INTEGRATION OF THE DESIGN AND MANUFACTURE HIGH-PRECISION CAST COMPONENTS OF by NORMAN CORNELIUS CLAMPITT III Bachelor of Science in Mechanical Engineering, Purdue University 1997 Submitted to the Department of Mechanical Engineering and the Sloan School of Management in partial fulfillment of the requirements for the degrees of Master of Science in Mechanical Engineering and Master of Science in Management In conjunction with the Leaders for Manufacturing Program at the Massachusetts Institute of Technology June, 2000 2000 Massachusetts Institute of Technology. All rights reserved. Signature of Author - I,,- Department of Mechanical Engineering Sloan School of Management May 5, 2000 Certified by zDafiiel Whitney, Thesis Supervisor Senior Research Scientist, Center for Technology, Policy and Industrial Development Certified by Roy Welsch, Thesis Supervisor Professor of Statistics and Management Accepted b Professor Ain A. Sonin, Chairman of the Graduate Committee Department of Mechanical Engineering Accepted by MargaretAndrfws, Executive Director of the Masters Program Sloan School of Management MASSACHUSETTS INSTITUTE OF TECHNOLOGY SEP 2 0 2000 LIBRARIES 'sit ABSTRACT Forcanptitwprductdewopment, the design omganization ofa mamzactrngcnpy requis k nodaigeof the suppliers'manufactumingpnessesand ofprnaples ofdesignformnfacurahility. This knoud& nust he coll&a, store, made accessible, andactiny utilizi. A ll of these knokdcge managemnt activies are cticalto the sucassful realizationofthelue ofthis knodaige. The purpose ofthe desis is to illustrate the imran ofmanagigknodaige in a supply camn and to enmend zuys to inpmw knodedgemanagmmt The illustrationsare drawnflrma sezwn-maih intemshippnpt sponsom& by A BB A L S TOM POWER, a manufactuerofpomer generationequipmet anda partnerofthe Leadersfor Manufacturingpgramat the MassadusettsInstituteof Tendokg The purpose of thepmt us to ident ays to build a knouWage base ofprmples and ridesfor impvving de manufacurahilityofpnuca designs, therby rnducingcost, tme-to-market, andnanrformanx. The intent uas to inpvw the mn c rm men ofthe design oganization,reducingthe depodoze upon a single supplierby givig it the poer to designparts to he manufacturableso that Axy amld be made by many suppliers. The result aus a set of newrmndatinsfor in7 ingpnluat deW/opMt and the effectiawess of knoalgemanagmnt in conanmmt ennering. The focus of th projctus e praisioncast turbinecanponmts ofdx A BB A L STOM POWER heavy duty gas turbieengines. Each ofdxse conpnaitswdeyges a series of pratins,often perfonralby difeent supplie. These parts repmsent a signficanttmhnical challmge to dx design team. Thrugh an analysis ofquality data and inteviewsz ith designer andsuppliers, a set ofmnmendatimszus made to the canpzyfor impovg the design oiganizatin'smanufacturingknowa/ge base. There am opportimiiesfor impoving thefeedickfian manufturmg by idnwvg all supplien ofa part early in its designand by motiwting suppliers to make acurateandon-time deliwnes. A nwner ofobsertionsuuremade uhi shouldhe includa in such a knowage base. Many of the tos afmmtly used by dx design teanformanagmgfieldexperenx know/a/ge can he adaptaito use in managmgkndge gainafiraniufacturingexpere . There ar opportuniies to enhane dxse toos to make dm nw acessibleandeffati. 7here arealso appotwnities to comwrt lessons leamed into design rides and to incoporatethese into design softure. Many of dese arimmndatioms mquire inwstment of resountes. This inustmet can hejustfud fthe result is awdiingthe need to re/eampastlessons. -3 - -4- ACKNOWLEDGEMENTS First I wish to thank my Lord and Savior Jesus Christ, through whom all things are possible. I thank my family and friends who prayed for me during my two years at MiT. I especially wish to thank my best friend Zhelinrentice Scott for her love and support throughout this project. I wish to acknowledge the Leaders for Manufacturing Program for its support of this work through its resources and support of this project.. I also wish to thank my advisors Daniel Whitney and Roy Welsch for their guidance. I thankfully acknowledge the people at ABB ALSTOM POWER and its suppliers who made it possible for me to learn about their processes and to collect the necessary information which forms the basis of this thesis. I hope that this thesis project provides a valuable contribution to the company. In particular, I wish to thank Prith Harasgama, Erhard Kreis, Alastair Clark, Miriam Park, Stephen Howarth, Christoph Nagler, Mark Richter, Christof Pfeiffer, Alex Beeck, Tim Strauss, Stefan Florjandi, Ernst Pauli, Nick Jovic, Bernard Robic, Diethelm Boese, Ivan Dim, Gordon Anderson, Markus Oehl, Martin Meyer ter Vehn, Stephen Bowes, Klaus Schneider, Edi Primoschitz, Gary Dewis, Mark Baker, Dean Westerman, and Dan Tasker. -5 - -6- TABLE OF CONTENTS AB STRA C T ................................................................................................................................................................ 3 SEC TION 1: OV ERVIEW ...................................................................................................................................... 13 1.1 INTRODUCTION .................................................................................................................................................. 13 1.2 PROBLEM STATEM ENT ....................................................................................................................................... 13 1.3 THESIS OBJECTIVES ........................................................................................................................................... 13 1.4 PROJECT OBJECTIVES ........................................................................................................................................ 13 1.5 A READER'S GUIDE TO THE THESIS .................................................................................................................... 15 1.6 SUM M ARY ......................................................................................................................................................... 16 SEC TION 2: BA CKG RO U N D ................................................................................................................................ 17 2.1 INTRODUCTION .................................................................................................................................................. 17 2.2 A BB A LSTO M PO WER ................................................................................................................................. 17 2.3 GAS TURBINES - TECHNICAL CHALLENGES ....................................................................................................... 17 2.4 THE PRECISION CAST PARTS SUPPLY CHAIN ....................................................................................................... 18 2.5 IGT INDUSTRY ANALYSIS ................................................................................................................................. 19 Competition - Strong ......................................................................................................................................... 20 Blade and vane suppliers - Strong .....................................................................................................................20 Customers - Strong............................................................................................................................................ 20 PotentialEntrants- Weak ...................................................................................................................................20 Substitutes - W eak ............................................................................................................................................. 21 Implications........................................................................................................................................................ 21 2.6 THE GT24/26 GAS TURBINE .............................................................................................................................. 21 2.7 THE PRODUCT DEVELOPM ENT PROCESS ............................. ................................................................................ 21 2.8 BENCHM ARKING CONCURRENT ENGINEERING .................................................................................................. 21 2.9 PRIOR PROJECT WORK ........................................................................................................................................22 2. 10 BACKGROUND LITERATURE ............................................................................................................................ 23 2.11 SUM M ARY ....................................................................................................................................................... 23 SECTION 3: THE BUSINESS CASE FOR IMPROVING FEEDBACK AND ORGANIZATIONAL LEA RN IN G ............................................................................................................................................................... 24 3.1 INTRODUCTION .................................................................................................................................................. 24 3.2 A M ODEL OF D IM LEARNING ........................................................................................................................... 24 3.3 SUM M ARY ......................................................................................................................................................... 26 SEC TION 4: A C C O UN T O F THE PRO JECT ..................................................................................................... 27 4.1 4.2 INTRODUCTION .................................................................................................................................................. 27 NONCONFORM ANCE REPORT (NCR) ANALYSIS ................................................................................................ 27 5.1 INTRODUCTION .................................................................................................................................................. 30 What N CRs tell................................................................................................................................................... 27 What they don't tell ............................................................................................................................................ 28 4.3 INTERVIEW S W ITH DESIGNERS ...........................................................................................................................28 4.4 INTERVIEW S W ITH SUPPLIERS ............................................................................................................................28 4.5 SUM M ARY ......................................................................................................................................................... 28 SECTION 5: RECOMMENDATIONS FOR DEALING WITH SUPPLIERS IN DEVELOPMENT ............. 30 5.2 COMM IT TO A CAPABLE SUPPLIER EARLY .......................................................................................................... Ch a lleng es ......................................................................................................................................................... 5.3 INVOLVE ENTIRE CHAIN IN CONCEPT PHASE ...................................................................................................... Example: Vane I Hook Feature......................................................................................................................... Other Examples .................................................................................................................................................. Challenges ......................................................................................................................................................... 5.4 ALIGN INCENTIVES ............................................................................................................................................ Paym ent based on quality .................................................................................................................................. Limited NC acceptanceperiodwith regular intermediatemilestones ............................................................... Establish a regulardelivery schedule................................................................................................................ Encourageaccuratecomm itm ents ..................................................................................................................... Challenges ......................................................................................................................................................... 5.5 SUMMARY ......................................................................................................................................................... 30 30 31 31 32 32 32 32 32 33 33 33 33 SECTION 6: RECOMMENDATIONS FOR DESIGNING PARTS ................................................................... 35 6.1 INTRODUCTION .................................................................................................................................................. 35 Ch alleng es ......................................................................................................................................................... 3 7 6.2 M AKE COOLING CHANNELS ACCESSIBLE ........................................................................................................... 38 Challenges......................................................................................................................................................... 39 6.3 PROVIDE FOR SLAVE DATUM FEATURES ........................................................................................................... 40 6.4 PROJECT PLANNING INTEGRATION ..................................................................................................................... 40 Ch allenges ......................................................................................................................................................... 4 1 6.5 UNDERSTAND IMPACT OF DECISIONS AND CHANGES ON THROUGHPUT TIME, COST AND QUALITY .................... 41 Ch alleng es ......................................................................................................................................................... 42 6.6 KEEP MANUFACTURER INFORMED EARLY ABOUT POSSIBLE CHANGES ............................................................... 42 Challenges ......................................................................................................................................................... 44 6.7 ACTIVELY INCORPORATE MANUFACTURING FEEDBACK .................................................................................... 44 6.8 SUMMARY ......................................................................................................................................................... 44 SECTION 7: RECOMMENDATIONS FOR LEARNING FROM EXPERIENCE (FEEDBACK TO DESIG N) .................................................................................................................................................................... 46 7.1 INTRODUCTION .................................................................................................................................................. 7.2 SUPPLIER VISITS ................................................................................................................................................ Challenges......................................................................................................................................................... 7.3 ANALYSIS OF QUALITY DATA ............................................................................................................................ NCR .................................................................................................................................................................... N CR SerialNumber Traceability....................................................................................................................... NCRF4PS, Knowledge-basedsystem ................................................................................................................ Challenges ......................................................................................................................................................... 7.4 PRODUCT SUPPORT TOOLS ................................................................................................................................ Experience Response System .............................................................................................................................. Problem History Files........................................................................................................................................ Project status and communicationfiles ............................................................................................................. Case Studies ....................................................................................................................................................... Challenges ......................................................................................................................................................... 7.5 INTRANET .......................................................................................................................................................... Challenges ......................................................................................................................................................... 7.6 STON ERULEO .................................................................................................................................................. Automate 2D Drawing Generation.................................................................................................................... IncorporateFeedback into Mfg ......................................................................................................................... Challenges ......................................................................................................................................................... 7.7 SUMMARY ......................................................................................................................................................... 46 46 46 47 47 47 47 48 48 48 48 48 48 49 49 49 49 49 50 50 50 SECTION 8: CON CLUSIO NS ................................................................................................................................ 51 8.1 INTRODUCTION .................................................................................................................................................. 51 51 ....... - - - - - --............................. ..... 8.2 D EALING W ITH SUPPLIERS....................................................................... 51 ... ................................................ 8.3 D ESIGN ING PA RTS ...................................................................................... ... -......................... 51 8.4 LEARNING FROM EXPERIENCE........................................................................................... . ---.. ....................... 51 8.5 O PPORTUNITIES FOR FUTURE W ORK........................................................................... -- - REFERENCES ................................................................................................................................---..............--- 53 APPENDIX A: CONCURRENT ENGINEERING PROCESS ............................................................................ 54 APPENDIX B: RELEVANT LITERATURE ........................................................................................................ 57 DF M .............................................................................--. ............ . ----..-...-. ---. ----........................................... .............----............... .......................................................................... MANUFACTURING TECHNOLOGIES -----....................... ........... SUPPLY CHAIN M A NAGEM ENT ..................................................................................... ORGANIZATIONAL LEARNING ................................................................................................................................. 57 57 57 58 APPENDIX C: CAUSAL LOOP DIAGRAMS...................................................................................................... 60 APPENDIX D: A MODEL OF DFM LEARNING AND ITS COST IMPLICATIONS ................. 61 M OD EL FORM U LATION ......................................................................................................... . ---------------........... 61 APPENDIX E: MACHINING COST ESTIMATION MODELS ........................................................................ ... .........----------...................................... LASER D RILLING .................................................................................... ELECTRIC D ISCHARGE M ACHINING....................................................................................................................... .... -------.---- .--------- . ELECTRO-CHEM ICAL M ACHINING............................................................................. 76 76 76 77 APPENDIX F: OVERLAPPING OPTIMIZATION MODEL ............................................................................. 78 APPENDIX G: DEFINITIONS.....................................................................................................------............ 79 - 9- - 10- LIST OF FIGURES ......14 Figure 1 The concurrent engineering process......................................................................... Figure 2 Concurrent engineering with feedback to a knowledge base ...................................................... 15 Figure 3 Partial assembly tree showing the role of precision cast parts in a power plant.......................... 19 Figure 4 A supply chain map for a typical precision cast part showing material flow. ............................. 19 Figure 5 A causal loop diagram of the DFM learning process. ................................................................ 25 Figure 6 The learning process and associated costs. ............................................................................. 26 Figure 7 V ane 1 hook feature ....................................................................................... ...... . .----.... 31 Figure 8 Key Characteristic flowdown for Vane 1 airflow example. ....................................................... 36 Figure 9 Effect of core shift on size of trailing edge cooling air exit holes. ............................................ 37 Figure 10 Cost of solution curves for three basic types of sensitivity analysis......................................... 38 Figure 11 Simplified blade showing cooling air passages and exit holes ................................................. 39 Figure 12 Slave datum feature.................................................................................................................. 40 Figure 13 A simplified financial model of the costs and revenues of a single engine project..................... 42 Figure 14 Recommended overlap strategy given estimates of evolution and sensitivity............................ 43 Figure 15 Types of overlapping and the resulting performance tradeoffs............................................... 44 Figure 16 Concurrent Engineering Timeline.............................................................. ...... 56 Figure 17 Horizontal modular structure vs. vertical integral................................................................... 58 Figure 18 The double helix of oscillation in industry structure and product architecture.......................... 58 Figure 19 The basic components of a causal loop diagram................................................................... 60 Figure 20 Causal loop diagram of complete DFM learning model.............................................................. 61 - 11 - - 12- SECTION 1: OVERVIEW 1.1 INTRODUCTION This section explains the aim of the thesis, showing the importance of knowledge management in product development. Then follows a description of the purpose of the project from which examples are drawn is also described. The following guide to the thesis explains the structure of the remainder of this document. 1.2 PROBLEM STATEMENT In manufacturing companies, there is a great deal of opportunity to save money and improve customer satisfaction through improvements in knowledge management during product development. This is especially true for companies whose manufacturing is outsourced, but holds for companies in which the manufacturing organization is internal as well. In order to improve, companies require an understanding of the problems in knowledge management and potential solutions. 1.3 THESIS OBJECTIVES The purpose of the thesis is to illustrate the importance of managing knowledge in a supply chain and to recommend ways to improve knowledge management. The illustrations are drawn from a seven-month .internship project sponsored by ABB ALSTOM POWER, a partner of the Leaders for Manufacturing program at MIT. 1.4 PROJECT OBJECTIVES The purpose of the project was to identify ways to build a knowledge base of principles and rules for improving the manufacturability of product designs, thereby reducing cost, time-to-market, and nonconformance. The intent was to improve the manufacturing competence of the design organization, reducing the dependence upon a single supplier by giving it the power to design parts to be manufacturable so that they could be made by many suppliers. Concurrent engineering is the simultaneous development of a product and the manufacturing process used to produce it.1 This process shortens the time to market and requires the cooperation of the designer and the manufacturer. Another intended benefit of cooperation is the improvement of the manufacturability of the design through feedback from the manufacturer during development, reducing production costs. Figure 1 is a simplified illustration of this feedback. 1Note: Fine (1998, 124) builds upon this idea by introducing a third concurrent activity: supply chain design. See Appendix A. - 13 - Concurrent -- Engineering Itgerti Ieration Design Specifications Suggestions for Manufacturability Improvement Manfatuin Figue 1 The conammt engineringprocess The building of a knowledge base of design principles does not eliminate the need for cooperation with manufacturing in the concurrent process. Rather, the knowledge base strengthens the concurrent process by eliminating redundant feedback. The knowledge base helps the design organization to retain lessons learned from experience in concurrent development and to avoid going through the process of learning previous lessons over again. - 14 - DFM Design Rules Concurrent Engineering Iteration Design Specifications Suggestions for Manufacturability ImprovementMau ctrn Figum 2 Conammt engmremg withfeerback to a knouliedge base Feedback from manufacturing is used to refine the design during development. The addition of a knowledge base in Figure 2 increases the effectiveness of the process by retaining experience and eliminating the need to re-learn old lessons through new experience feedback. Such a knowledge base exists in any concurrent engineering effort in which the people involved have prior experience. The recommendations in this report are not meant to replace the knowledge and experience of the people involved, but to support them and to improve the degree to which this knowledge is shared throughout the design organization. 1.5 A READER'S GUIDE TO THE THESIS The rest of this thesis is organized as follows: Section 2 gives the necessary background information to put the analysis and conclusions in proper context. The Company's history and current state of affairs are presented in a brief form, as relevant to the project. The industry of interest is analyzed from the perspective of the Company. A description of the product of interest and a brief history of its development are included to support the following sections. The original project upon which this project was based is described to cover prior related work done at the company. Section 3 builds the business case for improving knowledge management. A simple model of the system dynamics of learning from experience is used to show the cost impact of learning capabilities. Section 4 describes the project itself, how the information was collected and what was discovered. Section 5 recommends ways to improve the way the company works with suppliers in concurrent engineering and foreseen challenges in the implementation of these recommendations. - 15 - Section 6 suggests ways to improve the way parts are designed to reduce their cost, throughput time and defect rate. Section 7 recommends ways to improve knowledge management in product development. Tools are described that will allow knowledge to more effectively be captured, organized, accessed, and used in future development projects. Section 8 draws general conclusions from the specific lessons learned in the process of making the preceding recommendations. A literature review, machining cost estimation models, and table of definitions are included as appendices to clarify references within the text with more detail. 1.6 SUMMARY This thesis will illustrate the importance of knowledge management to product development through examples observed during the internship project at ABB ALSTOM POWER The remainder of the report will establish the framework and context, motivate the need for improved organizational learning, recommend ways to improve the design process and learning processes, and draw conclusions for the general case. - 16 - SECTION 2: BACKGROUND 2.1 INTRODUCTION The following background information explains the context of the project from which examples are drawn. A brief history of the sponsoring company is followed by a description of the product of interest and the supply chain that produces the components that will be the focus of the thesis. An analysis of the state of the industry explains the relative strength of the different players whose interests are at stake. A history of the design itself leads into the product development process and a comparison of the state of concurrent engineering within the company with that of its competitors. A description of the preceding project places the project of interest in its proper context. The section concludes with a reference to relevant literature and a brief summary of the background. 2.2 ABB ALSTOM POWER In 1987, ASEA AB of Vdsteris, Sweden, and Brown Boveri Ltd. of Baden, Switzerland, announced plans to merge and form Asea Brown Boveri Ltd. (ABB), headquartered in Zurich, Switzerland. The technical lead center for Power Generation remained in Baden. Over the next few years, ABB made a series of acquisitions, joint ventures, and divestitures. Just prior to the joint venture described below, ABB consisted of the following businesses: Power Generation Power Transmission Power Distribution Automation Oil, Gas and Petrochemicals Building Technologies Financial Services On March 23, 1999, ABB and ALSTOM formed ABB ALSTOM POWER, a 50-50 joint venture consisting of ABB Power Generation (Excluding Nuclear) and ALSTOM Energy (Excluding Heavy Duty Gas Turbines), employing 58,000 people in over 100 countries. As the formation of the joint venture did little to affect ABB's gas turbine product line, which is the focus of this thesis and project, the ALSTOM background is not included here. Furthermore in April 2000 Alstom acquired the remaining share (50%) from ABB and the present company is designated Alstom Power, a wholly owned subsidiary of Alstom. Hereafter the term used is AAP. 2.3 GAS TURBINES - TECHNICAL CHALLENGES Gas turbines are heat engines used to power jet aircraft and electric generators, among other things. Industrial gas turbines (IGTs) used to drive electric power plants are attractive to power producers because of their efficiency and the fact that some are fueled by natural gas, some by fuel oil, and some by either interchangeably. Energy can be recovered from the exhaust gases to generate steam and power steam turbines in what is called a "combined cycle" power plant. Combined-cyde plants achieve up to 60% fuel efficiency. The basic principle behind gas turbines involves a compressor, a combustion chamber, and one or 17 - more turbines. Both the compressor and each turbine are rotating stages of airfoils. The compressor draws in air and compresses it to the proper inlet pressure and temperature for combustion. Fuel, mixed with air, is burned in the combustion area. The expanding hot gas drives each turbine, which drives the compressor and/or the load, which is a generator in the case of an electric power plant. The frequency of the power generated is determined by the rotating speed of the engine. Some markets use 60-hertz power (e.g. North America), while others use 50-hertz (e.g. Europe and Asia). Over time IGT customers demand higher efficiency, while environmental regulations require lower NOx emissions. Both of these requirements are met by increasing combustion temperatures. However, this begins to stretch the limits of durability of the engine components, particularly the burners, heat shields, and turbine blades (rotating airfoils) and vanes (stationary airfoils). At one time, turbine blades were machined from forged parts. At that time the company (then BBC) did all the work in-house. Later, when higher combustion temperatures required parts with substantially higher temperature resistance, turbine blades and vanes were investment cast in high-melting temperature alloys (superalloys). As temperature and stress resistance requirements increased, new technologies were employed. Directionally solidified castings have grain boundaries in only one direction, increasing strength along that axis. Single crystal parts have no grain boundaries and are the strongest (and most expensive) investment castings available. This process is significantly more complex and requires a very high level of technical expertise and experience to be found in only a few places in the world. Both AAP and its competitors depend upon foundries that specialized in this type of casting. Even with these exceptional material properties, turbine parts would not survive the hot gas conditions without additional cooling, since these temperatures can be above the melting point of the metal. Internal channels in the parts conduct cooling air, which is diverted from the compressor. Many tiny holes on the hot surfaces of turbine components allow this air to escape and form a protective film over the parts. Many turbine components are cast with these internal cooling channels, requiring ceramic cores, an additional technical challenge and expense. For additional protection, many parts are also covered with a ceramic coating. Although the parts are cast near net shape, some machining operations are still necessary to produce mating surfaces and film-cooling holes. The types of materials used require relatively expensive nontraditional machining techniques such as electric discharge machining (EDM), electro-chemical machining (ECM), and laser beam machining (LBM) in addition to the more traditional methods of grinding and milling. 2.4 THE PRECISION CAST PARTS SUPPLY CHAIN For clarity, Figure 3 shows the basic role of precision cast parts in the end product. These are the parts of the turbine that are designed and manufactured to withstand the hot gas temperatures. The figure contains only a portion of the total number of turbine components. Precision cast parts consist of blades, vanes, burners, and heat shields, among other types. These parts are then assembled into the turbine portion of the housing and rotor of the engine or thermal block. The thermal block is only one piece of an entire power plant. - 18 - Burners Blades Precision Cast Parts Vanes Shields Therma LargeBlocks P Castigs GenertorsPower Plant Structure Figure 3 Partialassanboytree shown the role ofprecsion castparns ina power plant Figure 4 illustrates the supply chain of a typical precision cast part. Parts begin life as a casting. Some parts are equiax castings, while a few are either directionally solidified or single crystal castings. These castings are then machined to achieve the desired final shape. Many parts are coated for additional protection. Some machining takes place before coating (Machining I) and some after (Machining II). These processes are often, but not always, performed by separate manufacturers at geographically dispersed sites. Casting Coating Machining 1 Machining II 10Assembly (ABB) Figure 4 A supply chain mapfor a typicalprecision castpart showing materiaflow. 2.5 IGT INDUSTRY ANALYSIS The following analysis is based upon Porter's 5 forces framework (Oster 1994, 31). The industry of interest is the manufacture of industrial gas turbines for use in power generation. This includes producers of both single- and combined-cycle plants that incorporate gas turbines. This analysis is presented as motivation for the rest of the thesis and as relevant background information that places the product of interest in its proper context. - 19 - COMPETITION - STRONG ABB-ALSTOM POWER is currently third in its share of the GT/CC market, behind the General Electric Company (GE) and Siemens-Westinghouse. It distinguishes itself by offering complete power plants in turnkey or "well to wire" packages to customers. BLADE AND VANE SUPPLIERS - STRONG Some of the suppliers themselves have a great deal of power. As mentioned earlier, there are very few casting companies with the technical expertise and capability to produce the level of technology required by the precision cast components. IGT make up a smaller portion of their business than aircraft engines. For example, a press release from Howmet International, Inc. 2 indicated that 50% of their revenues were from aircraft (25% new and 25% after-market), while 35% was from IGT. Therefore, the aircraft engine companies, including GE, have somewhat more power than AAP in dealing with these suppliers. This technology is not simple or cheap to develop. The existing experts have the experience of dealing with many customers, including the relatively large-volume aircraft industry. An IGT manufacturer would find it difficult, to say the least, to develop this technology in-house. Therefore, IGT manufacturers depend on these few suppliers. Howmet, the world's leading supplier for IGTs, is particularly powerful, serving all of the major IGT manufacturers. Howmet has the unique position of being the leading supplier of directionally solidified and single crystal castings. Howmet also has the majority market share in aircraft engine airfoils. CUSTOMERS - STRONG IGT customers have traditionally been government-supported utilities. However, in the wake of deregulation an increasing environmental regulation in the electric power industry, there is an increasing number of independent power producers (IPP). These IPPs are profit-driven, and somewhat more cost-conscious than the former customers are (Schimmoller 1999). They are able to demand guarantees of performance and ontime delivery. Since a late delivery or high fuel cost reduces the net present value of operations, these customers demand price penalties as compensation. Currently, IPPs purchase approximately the same quantity of gas turbine-based power plants (in megawatts) as purchased by utilities. POTENTIAL ENTRANTS- WEAK There are several significant barriers to entry. A reliable design requires a great deal of technical expertise and experience. It also requires an extensive supply chain with a great deal of capacity and experience. There are smaller firms that have the engineering capability to develop parts of the engine, but lack the manufacturing capacity. Although these things can be bought, the market share of all remaining manufacturers after the top three is less than that of AAP, "Howmet Forecast Holds Steady After Boeing Announcement". (January 1, 1999). Online. Internet. Available http://www.howmet.com/corp/news.nsf/(NR+by+Number)/1998-062 2 - 20 - SUBSTITUTES - WEAK Because of tighter restriction on emissions, power producers are driven to cleaner forms of power generation. At the present time, gas turbines fueled by natural gas are the cleanest form of power generation based on the burning of fossil fuels. Hydroelectric power is, of course, cleaner, but is not available everywhere or in sufficient quantities to supply the world's energy demands. Environmental laws are aimed at reducing the amount of nuclear power. Other alternative energy sources such as solar and wind power have yet to prove economically feasible. Therefore, there is little threat to the demand for gas turbine power generation equipment. IMPLICATIONS The conclusion from this analysis of the gas turbine industry is that one can restrict the focus of any analysis to a struggle between component suppliers, gas turbine customers, and three major equipment producers. At the present, there is little gained by including potential entrants or substitute technologies in a model of the system. However, among component suppliers, turbine customers and equipment producers, there is sufficient competition for value extraction and control of the supply chain to warrant the discussion presented in this thesis. 2.6 THE GT24/26 GAS TURBINE By the end of the eighties, BBC had decided to get out of the gas turbine business. A large part of the engineering staff responsible for turbine design was released. After the merger with Asea, the company decided to keep gas turbines as the focus of the power generation business. To remain competitive, the company needed a new gas turbine product, but lacked the engineering knowledge and design capacity to develop the turbine, an important and technically challenging part of the whole engine described above. Therefore, the company jointly developed the turbine with external contractors, while rebuilding the capability internally. The two products based on this design were called GT24 (60Hz) and GT26 (50Hz). The 50 Hz design is basically the 60 Hz design scaled by a factor of 1.2 The focus of this thesis is the precision cast turbine components of the upgrade (B version) design. 2.7 THE PRODUCT DEVELOPMENT PROCESS During the development of a new engine, a set of basic cost and performance targets are transformed into stable, serial production and delivery of complete engines. The basic process begins with the specification of these targets. Design teams develop the various major parts of the engine. The design team works with the manufacturing suppliers to develop and validate the production processes. A complete collection of components for one engine is called an "engine set". Experience gained from engine sets produced during development (preserial sets) is used to improve and stabilize the manufacturing processes. The stable production processes are reviewed in an Initial Sample Inspection (ISI). This milestone represents the end of the product development process and the beginning of serial production. After ISI, the manufacturer has sufficient specification to independently determine whether parts are acceptable. A more detailed explanation of the concurrent engineering process can be found in Appendix A. 2.8 BENCHMARKING CONCURRENT ENGINEERING Several engineers within the design organization had formerly worked for Pratt and Whitney (Hereafter referred to as Pratt). They were able to comment on their perceptions of the differences between concurrent engineering at AAP and at Pratt. Based on their impressions it seems that there is somewhat less communication between the designers and the manufacturers at AAP than at Pratt. At AAP, theieis more resistance to choosing a supplier early, mostly from the supply chain management organization, not from design. There is a significantly higher level of education among AAP engineers, but somewhat less 21 - experience. A former Pratt employee commented that both organizations were understaffed, but AAP wa; more so. At AAP component owners are responsible for a set of parts from the concept phase through serial production. At Pratt, this function is divided among design engineers and project engineers, who are responsible for parts in production. Another former Pratt engineer found it difficult at first to work in the AAP environment, suggesting that the new employee training was not as good as that of Pratt. Some individual engineering managers, however, make the effort to look after new engineers, pairing them with mentors until they are ready to work independently. This method, though not a standard company-wide practice, has the added benefit of helping to maintain and distribute the organization's knowledge and experience by sharing it with new members. More than one former Pratt engineer noted AAP's strength in documentation, which seemed to take up a larger part of an engineer's time at AAP than at Pratt, though one person questioned its effectiveness. Culturally, one engineer observed that one had to be more aggressive at AAP to get things done, but that people were receptive to ideas for improvement. 2.9 PRIOR PROJECT WORK Prior to the start of this project, a similar effort was focused on integrating the design and manufacture of large castings, mainly casings. These castings, usually one per engine and weighing several tons each, were studied to find common problems in manufacture that could be reduced or eliminated by changes in design. An analysis of Nonconformance Reports (NCR) led to the identification of problematic features in the castings. A Nonconformance Report is a type of quality documentation that allows the issuing organization (supplier) to report parts that do not conform to the specification. The customer (AAP) can then give an engineering disposition for the parts. The parts can then receive one of the following dispositions: accept without modification accept and identify for special future treatment rework scrap The NCR system streamlines this process by handling the information electronically and keeping a record of the incident. The locations of casting defects were gathered from the NCRs and summarized in a 2-D histogram. This showed the areas that caused the most defects. This information was useful to the casting supplier in adjusting the design of their tooling in order to reduce the amount of defects and, consequently, the time and cost required to produce a casting. The initial focus of this project was to conduct a similar analysis on precision cast parts (i.e. blades, vanes, burners, heat shields, etc.). It soon became clear, however, that a sufficient quantity of data of the type analyzed for large castings did not exist in the NCR records for precision castings. The improvements realized in the large casting project were possible because the factory recorded a great deal of casting defect data in the NCRs, but did not know what the trends were in the defect locations. Once this was pointed out to them, they were able to make corrections. As described below, such casting defect data and other localized defect data was scarce among NCRs of precision cast parts, requiring further investigation. - 22 - 2.10 BACKGROUND LITERATURE A description of relevant existing work can be found in Appendix B. All of the principles mentioned in this thesis upon which the recommendations were based are illustrated by examples observed during the project. However, many of these same principles can be found in the literature. The same was true of the prior project involving large castings. This helps to illustrate the importance of managing knowledge within the design organization and across the supply chain, as discussed later. 2.11 SUMMARY The rest of the thesis focuses on the precision cast turbine components of ABB ALSTOM POWER's GT24 and GT26 engines. These components represent a major technical challenge requiring a great deal of expertise to achieve both performance and reliability. The company faced a major challenge when it decided to compete in the industrial gas turbine business. One of these challenges was to redesign the turbine with a relatively new engineering team. Opportunities for improvement in design for manufacturability led the project management to initiate an effort to correct manufacturability problems in large structural castings. The success of this effort led to the project, which serves as the basis for the remainder of the thesis: to improve the manufacturability of precision cast parts and to develop ways to improve the design organization's ability to learn from experience in manufacturing. - 23 - SECTION 3: THE BUSINESS CASE FOR IMPROVING FEEDBACK AND ORGANIZATIONAL LEARNING 3.1 INTRODUCTION The aim of this section is to show the value of organizational learning within the firm, particularly the extent to which the design organization learns the expertise to develop more manufacturable products. Building a knowledge base requires the investment of resources, some of which are described in following sections. This section builds a framework for justifying that investment. A sinple model of the learning process is used to illustrate its nature as a dynamic system of elements and to show the importance of all of these elements of learning from experience. 3.2 A MODEL OF DFM LEARNING Senge (1994) emphasizes the need for learning organizations to adopt a "systems" perspective. Senge suggests that many organizations are limited in their ability to learn because they perceive every effect as having a single cause and fail to see the effect that businesses, teams, and individuals are parts of larger systems. The process of learning from experience can be thought of as a system of competencies working together iteratively. Figure 5 illustrates the process by which the design organization learns design for manufacturability from experience. For an explanation of casual loop diagrams, see Appendix C. The stock labeled "DFM" is an abstract measure of the design organization's ability to design manufacturable products. The diagram shows how the DFM knowledge is gained from experience. The "Manufacturability" of designs is determined by the team's "DFM" knowledge and "Knowledge Management Effectiveness", or ability to access and utilize its knowledge. The "Manufacturability Gap" is the amount by which the design falls short of perfect or "100% Manufacturability." Depending upon the "Quality of Communication with Suppliers", some portion of the information needed for improvement is fed back to the design team as "DFM feedback." The "retention fraction" is a measure of the portion of the feedback that is converted into useful DFM knowledge. - 24 - retention fraction Retained learnin + g DFM feedback.s.a DFM Learning loop DFM Learning rate Quality confmunicatio nith suppliers Manufacturabilit _y Gap Manufacturabilit yZ/+ Knowledge Management Effectiveness I00o Manufacturabl e Figure 5 A causal loop diagram of the DFM learnmingprocess. Three aspects of the organization contribute to the effectiveness of the learning process. These are "Quality of Communication with Suppliers," "Retention Fraction," and "Knowledge Management Effectiveness," representing the organization's ability to obtain, preserve, and use feedback from manufacturing. Each of these capabilities requires an investment of resources. An effective structure must be built and maintained for receiving feedback from manufacturing. Systems and people must be dedicated to organizing and storing this information in an accessible form. The degree to which these systems are utilized determines their value to the bottom line. Figure 6 suggests some ways in which learning impacts cost. A product that is more difficult to manufacture (large "Manufacturability Gap") will consume more engineering resources redesigning parts and processes. It will cause more scrap and rework in the process of making these improvements and will cause significant delays in delivery. The penalties for late delivery can be severe. The "Cost per Project" is simply the sum of "engineering cost" "scrap/ rework cost" and "delivery penalties." Appendix D contains the details and formulation of a more complete form of the same model. Simulation will show what may already be clear to the reader: the greater the organization's ability to obtain, store and use feedback from manufacturing, the lower projects will cost. Because of the model's assumptions, "Quality of Communication with Suppliers," "Retention Fraction," and "Knowledge Management Effectiveness," each have an equal impact on cost. The model is not suitable to weigh the relative importance of the three, but instead illustrates the fact that all three aspects of the system are critical to effective learning. It is relatively straightforward to implement a reporting system that requires manufacturers to record problems that occur in production. It is equally straightforward to record this information for future use, as is currently done by the NCR database. It is quite another matter to make use of such a database. The usefulness of a knowledge base depends largely on the way it is organized and recorded. While it is true that designers must be motivated to use past knowledge, motivation is not enough. The knowledge must be accessible. The recommendations in section 7 contain suggestions for improving the accessibility of DFM knowledge. - 25 - retention Retained fraction + learning± DFM feedback + DFM Learning DFM + -- Quality of communication with suppliers loop Manufacturability rate rae, .- Manufacturabili Knowledge Management Effectiveness Gap y- 100% Knowledge anIUac Ir e, % +± Scrap/ Delivery Penalties Rework cost Cost per Project Engineerin cost Figure 6 The learningprocess and associatedcosts. 3.3 SUMMARY Products that are difficult to manufacture will cost more. Manufacturability can be improved when the design organization learns from experience. The model presented in this section shows that organizational learning depends on investment in the organizations capability to obtain, record, and use feedback from the manufacturer. The recommendations in section 7 suggest ways to improve these capabilities. The following section describes the methods by which the origins of these recommendations were discovered. - 26 - SECTION 4: ACCOUNT OF THE PROJECT 4.1 INTRODUCTION As explained in section 1.4, the aim of the project was to develop ways to build a knowledge base of principles and rules for improving the manufacturability of product design. The project plan consisted of data collection and analysis, leading to a set of recommendations and, to the extent permitted by time, implementation of these recommendations. The following list describes the basic tasks, some of which were carried out in parallel: Selection of pilot components Analyze Nonconformance Reports Understand and analyze The design process Technical requirements of the components Manufacturing processes Recommend Improvements in dealing with suppliers General design principles Feedback tools 4.2 NONCONFORMANCE REPORT (NCR) ANALYSIS The project began with an analysis of the NCRs generated by suppliers in the production of precision cast parts. Specifically, two pilot components were chosen for the analysis, the first stage low-pressure turbine blade and vane. The defects raised in the NCRs were collected and analyzed for trends. The quantity of defects identified by geometric location was very small, however, and not useful for a statistical analysis of the kind previously performed for large castings. Instead, the defects were categorized by type. Those types that occurred more frequently were investigated further as described below. WHAT NCRS TELL Except in the cases where the obligation to report nonconformance had been waived, NCRs describe the conditions under which parts were out of tolerance and either accepted, reworked or scrapped. At a macro level, this allows one to break down the total amount of scrap and rework by engine, part, process, supplier, or any of a number of other fields. For example, one problem that was mentioned in a number of NCRs was a lack of material on datum surfaces, listed as a casting defect. Upon further investigation, it was discovered that the machining supplier was using, as a slave datum, a surface that was designed to be left as cast. This problem is described in detail in section 5. The NCRs indicated that there was a problem, but it required further investigation to actually determine what the problem was. - 27 - WHAT THEY DONT TELL A process may run perfectly, without producing nonconforming parts, and still be excessively slow, costly, or altogether redundant. The NCR will not indicate that a part could be made cheaper or more quickly by another method without compromising customer requirements. NCRs don't give a great deal of insight into the root causes of problems. Rather, they indicate places or operations where problems may have occurred. It is then necessary to follow up these indications with further investigation. The NCR database was difficult to analyze because of inconsistency in use and the limited number of fields available. The same form is used for any kind of defect found in a part as a result of a manufacturing operation. It is, therefore, not particularly useful for analyzing the locations of specific types of defects, statistically, nor is it useful for analyzing a large quantity of exceeded tolerance errors. Contrary to the specification for filling out NCRs, some reports included descriptions of more than one kind of problem per report. Also, the defect related information and serial number list was often included only in the form of a digitized picture file. It is not possible to search for information in picture files, since the database is only capable of searching for alphanumeric text. The reports actually had many fields for the manufacturer to complete. However, for a defect analysis of any detail, the most important information was not organized into fields, but was entered as open text. This made any kind of statistical analysis of the data (for example tolerance deviations, grouped by feature) a very manual and time-consuming process. Some recommendations in the following section resulted from this analysis. Some deal with the use of the NCR database itself. 4.3 INTERVIEWS WITH DESIGNERS While analyzing the NCRs, it was possible to interview the engineers who had designed the parts. These interviews gave a great deal of insight into the concurrent engineering process used, what went well, and what could have been improved. This insight contributed to the benchmarking analysis and the recommendations in the following sections. Designers also shed light on the performance issues that drive the technical requirements imposed upon the design. 4.4 INTERVIEWS WITH SUPPLIERS Defects raised in NCRs led to inquiry into their causes by interviewing people at the factories that made the parts. Although some problems that arose in manufacturing could be observed from the NCR record, these problems and more all came out in the interviews. Unlike the large-casting project, the manufacturing engineers at the suppliers were already aware of the features that tended to raise NCRs. More importantly, they often knew the root causes and were able to explain the circumstances surrounding them. They were also able to describe the concurrent engineering process from their perspective, to evaluate it, and to offer suggestions for improvement. The supplier and designer interviews were the major sources of information used to analyze the most recent iteration of concurrent engineering, and to develop the following recommendations for improving the process. These recommendations fell into three categories: the supplier relationship, design for manufacturability (DFM) principles, and tools for developing a knowledge base of DFM principles to improve future iterations of concurrent engineering. 4.5 SUMMARY The majority of the project consisted of data collection and investigation into the causes of problems in manufacturing development of the GT24 and GT26 B version precision cast turbine components. It was determined that the existing system for reporting manufacturing problems was inadequate as a tool for organizational learning. Defect data was collected and stored, but was not easily accessible for use in avoiding - 28 - the same problems in future design projects. Through interviews with designers and process development engineers, new insights were gained as to the origins of manufacturing defects and problems in development. The result of these investigations was a series of recommendations for improving design, which should be incorporated into the organization's knowledge base of design for manufacturability, and ways in which the organization can improve the feedback it receives and the way it uses this feedback in future designs. The following recommendations were presented to members of the supply chain management and design organizations. The recommendations are grouped into three categories, dealing with suppliers, designing parts, and learning from manufacturing experience. The background and significance of each recommendation are included for clarity. - 29 - SECTION 5: RECOMMENDATIONS FOR DEALING WITH SUPPLIERS IN DEVELOPMENT 5.1 INTRODUCTION It was expressed more than once by leaders in the design organization that the issues addressed by the following recommendations should be separate from an analysis of design and the concurrent development process because they relate to the supply chain. However, each of these recommendations represents an improvement strictly in terms of quality of product development and the building of a manufacturing knowledge base within the design organization. In addition, Fine (1998) suggests that the design of the supply chain is an integral part of a competitive concurrent engineering effort and is a critical competence to maintain within the design organization, rather than an issue to be handled by a separate supply chain management group. 5.2 COMMIT TO A CAPABLE SUPPLIER EARLY In the environment of Concurrent Engineering, it is important to make decisions as early as possible in order to minimize costly rework. In order to reap the benefits of concurrent development, suppliers must be identified early so they can guide the design from a manufacturing perspective. Before suppliers will invest in the development of a new product significantly, they will require some level of assurance that they are not just helping to develop a product for someone else to produce. This is especially true for operations like machining, which could be done by any of a large number of suppliers. Traditionally, contractors were selected by competitive bidding after a design was complete. While this has the advantage of minimizing part of the manufacturing cost (process inefficiencies and suppliers' profit margins), there are already costs designed into the part which even the most efficient manufacturing operation cannot remove. These built-in costs are the motivation for integrating design and manufacturing. Through cooperative development, designs are more manufacturable and inherently less expensive to produce. This does not preclude a second supplier for each part. Sufficient commitment could be made by guaranteeing a minimum percentage of the business to a partnering supplier, subject to minimum quality standards. This agreement could be used to form the necessary partner relationship with a supplier to develop the product concurrently. In fact, it is unwise to depend upon a single source for many reasons. A company that does so is exposed to the risk of that supplier's ability to deliver, which could be affected by the actions of a competitor. It is also subject to the supplier's monopoly power, especially if, in cooperative development, the design is optimized for that supplier at a significant cost advantage over other suppliers. Careful management and engineering judgement must be exercised to avoid this condition. CHALLENGES The organization seems to be moving in this direction already. The Engineering organization is naturally willing to do this, since it benefits them at no obvious cost. There is more likely to be resistance from the supply chain and logistics organization, which would like to remain flexible to switch suppliers. The middle ground must be established in which a primary supplier can be identified for the concurrent development, with some allowance for doing business with a second supplier. The agreement can guarantee, under reasonable circumstances, a certain percentage of the business to the primary supplier. - 30 - 5.3 INVOLVE ENTIRE CHAIN IN CONCEPT PHASE The principle of early supplier involvement applies to not only the casting supplier, but downstream suppliers as well. In this context, the term downstream refers to suppliers of processes that follow casting. All of these processes are upstream relative to AAP. If the development of upstream processes is carried out without the input of suppliers of downstream operations, there is a greater opportunity for inherent costs to subsequent stages to be built into the design of the early stages. For example, if only the casting supplier is involved in the casting design, there is no opportunity for machining suppliers to suggest ways in which the casting could be changed to reduce machining cost. The same applies to any downstream supplier. Therefore, it is best to involve the entire supply chain from the beginning, to save the costs of tooling rework and to minimize the total manufacturing cost of the product. This is an important part of adding the Supply Chain Design dimension to traditional concurrent engineering (product and process design). Fine (1998, 124) calls this broader view three-dimensional concurrent engineering (3 DCE). Hook feature Figure 7 Vane 1 hook feature EXAMPLE: VANE 1 HOOK FEATURE One problem that might have been prevented by the insight of subsequent suppliers was the clearance in the trailing edge hook feature of Vane 1. There are many dimensions to control in casting a turbine component. Although computer numerically controlled coordinate measuring machines are used for this purpose, the most important dimensions must be emphasized to ensure control. The hook was designed to be cast with clearance for assembly into the stator. The only part of the cast feature that was designed to be machined was a pad on the underside of the hook. The geometry of this feature was not constrained by strength requirements, but by assembly features. However, the cast surface inside the hook of the actual parts interfered with mating parts. As a result, an EDM operation was added to remove the excess material and permit assembly of the Vanes into the stator, adding about CHF 75,000 to the annual cost of machining, in addition to extra development and production delays. It was noted by the machining supplier that this could 31 - have been prevented if suppliers of subsequent operations had been involved in the casting development. It is possible that a team member from assembly would have ensured that the casting supplier control this dimension in the interest of ensuring the possibility of assembly. OTHER EXAMPLES A problem with laser-drilled cooling holes, described in detail in section 5, arose in machining, but was caused by the casting design. The best method for stopping the laser beam is to block it with strips of polytetrafluoroethylene (PTFE). The surfaces of many precision cast turbine parts are covered with blinddepth holes for cooling air. Similarly, the need for excess material on surfaces to be used as slave datum features must be included in the casting design, also described later. Another example was the unsuitability of the cast surface for brazing, as designed. This led to an additional machining step to erode a surface for brazing, which, in turn complicated the brazing process by adding a step to remove the recast layer formed by EDM. CHALLENGES The engineering organization is starting to adopt this practice as well. New development projects are taking steps to involve machining, coating, and assembly suppliers in the early stages of development. 5.4 ALIGN INCENTIVES PAYMENT BASED ON QUALITY If a supplier gets paid equally, regardless of part quality, there is little incentive to make parts to satisfy the design requirements. For example, a part that has only half the intended lifetime is worth much less than a part made to specification. Quite a few parts raised in NCRs were accepted with the restriction that the parts be marked as having reduced lifetime. However, since suppliers were paid the same for parts with a reduced lifetime, there was less incentive to make the effort to improve the process that delivered poor quality, and little reason to recommend design improvements. Suppliers are often too small to absorb the complete risk of defective parts. However it is only necessary to provide a sufficient financial incentive to motivate behavior which is compatible with the customer's goals. This suggests an arrangement where minimum quality standards are established consistent with a predicted target lifetime. Price penalties should be set for acceptance of parts with defects leading to a reduced predicted lifetime. The same must apply for scrap. If the manufacturer cannot bear the full risk of scrapping parts, they must be accountable for some portion of it. In the ramp-up phase, poor quality in ECM drilling of the trailing edge of Blade 1 resulted in approximately IF 390,000 in scrap castings. It is likely that the supplier would have been inclined to correct the problem earlier if required to pay a portion of the scrap cost. LIMITED NC ACCEPTANCE PERIOD WITH REGULAR INTERMEDIATE MILESTONES If nonconforming parts are to be accepted for a limited time during development, that time schedule should be made clear to the suppliers and should include regular events requiring improved quality. Nonconforming parts should never be accepted without clear plans of corrective action and time commitments by the responsible supplier. - 32 - ESTABLISH A REGULAR DELIVERY SCHEDULE In order to ensure the timely resolution of quality issues, it is important to require regular deliveries from each supplier. This brings manufacturability problems to the surface more frequently, reducing the delay in correction and the resulting accumulation of bad parts in the system. It also creates the incentive to solve them either through process changes or through changes in specifications. Every type of part required from a supplier must be delivered to a regular schedule. Otherwise, it is tempting to deliver an excess of parts that are less costly and to delay the production of more problematic parts. Ultimately, regular delivery of the required parts is compatible with a lean production system and is in the best interests of both the customer and the supplier. ENCOURAGE ACCURATE COMMITMENTS Many of the problems encountered in the development phase result from the fact that suppliers over-commit in capacity planning in order to ensure full utilization. This is analogous to the airline industry's practice of over-booking flights to ensure full loading of aircraft. This practice continues to satisfy customers if they are sufficiently compensated to wait for a later flight when their original flight is full. However, if there is no penalty for underestimation of capacity, the supplier's dominant strategy is to over-commit. In the present situation where AAP's customers defer the risk of late delivery by requiring penalties for lateness, AAP is obligated to accept as many parts as possible. This lowers the incentive for suppliers to improve quality as mentioned above. It also drives them to maximize production efforts at the expense of quality improvement. There are several examples of quality problems that took months to resolve because AAP was desperate for parts and was forced to accept a lower standard. For instance, many engine sets of blades were accepted with excessive airflow measurements, although this causes problems later. The parts have to be mixed with lower-flowing parts to avoid excessive losses and efficiency problems. This could be addressed in more than one way. AAP could purchase capacity, rather than ask for commitments to produce a certain number of parts. This, however, depends on the supplier to efficiently utilize the capacity to produce the right parts. Alternatively, the supplier could be required to pay price penalties for late deliveries of promised parts. This would force the supplier to reassess risk in light of the cost of over-comnitment. C-ALLENGES Alignment of incentives will be more difficult to achieve than the other recommendations in this section since it requires careful contract design. This is not helped by the fact that the organization seems to resist suggestions to change the contracts. The basic elements of the new contracts should include price penalties for parts accepted with limited lifetimes or other quality deficiency and for late delivery of promised parts. This cannot be introduced without also including a penalty for excessive scrap. If the supplier is not accountable for scrap and only adds a small part of the total value of the part, then he will be inclined to simply write off a part as scrap, rather than address quality problems. 5.5 SUMMARY After discussions with suppliers and designers about the causes of problems discovered in the NCR analysis, it was determined that there are a number of opportunities to improve the DFM learning process by changing the way the company deals with its suppliers. To fully realize the value of this partnership, all the suppliers that will be performing operations on a part should be involved from the beginning in the design of that part. 33 - In addition to the obvious commercial implications, accurate and timely delivery of parts is also essential to getting effective manufacturing feedback and to building a knowledge base of DFM principles. The next section describes some of the principles that can be learned from this project. - 34 - SECTION 6: RECOMMENDATIONS FOR DESIGNING PARTS 6.1 INTRODUCTION After an analysis of the version B design, the following suggestions were made to improve future designs. These suggestions represent the type of knowledge that must be available to all designers of such components. Each recommendation is the result of an investigation to some problem that caused nonconformance or delays in product delivery, both of which result in unnecessary cost to the company. MINIMIZE SENSITIVITY OF KEY CHARACTERISTICS TO PROCESS VARIATION Every manufacturing process results in some part-to-part variation. This variation can affect how nearly a part meets the requirements of the customer. With any design, there are measurable characteristics of a finished part that are critical to satisfying a customer need. However, the relationship between variations in the manufacturing process and variations in these characteristics is defined by the design. Thornton (1999) defines key characteristics (KCs) as ... the pmro&t, sub-assenbly, part, andpwxess featus that siiflcanty inpa thefinal cost, perform x, or safety ofa pdw dxn the KG zwyflrn naninaL Specialconbvn shoLd be applied to those KG zbem the cost ofzriatonjustifies the cost ofcntmmL - 35 - Cost ofC Customer Requirement OperationT Engine Efficiency Time Before Reconditioning/ Replacement Cooling Air Used Surface Temperature Cross-Sectional Area of Trailing Edge Holes Product KCs Subsystem KCs Part KCs Core Shape Core Position Process KCs Figure8 Key CharacteristicfloudozenforVane 1 airflow example. A good example is the airflow of a cooled vane with holes cast in the trailing edge, which is critical to the efficiency of the engine and the lifetime of the vane and downstream turbine stages. Figure 8 shows a portion of the KC flowdown for this part. The airflow is largely defined by the geometry of the cooling passages, which are produced by the core. At the trailing edge, where cooling air exits these passages, the cross-section of the core determines the size of the holes. The core position tends to vary from one part to the next. If the edges of the core are not parallel, the cross-section at the trailing edge will vary with core position. Therefore, the relationship between core position and exit hole size is determined by the angle between the core edges, a design parameter. Figure 9 shows a cutaway view of the cored exit holes of a typical vane. The edges of one hole are extended to show that they are not parallel. In this case, since making the edges parallel has no negative effect, it makes sense to do so. - 36 - Exit Hole Height Nonparalle edges External View Core Position 1 Core Position 2 Figure 9 Effect of core shift on size of trailing edge cooling air exit holes. The problem of the sensitivity of key characteristics to manufacturing variation is analogous to the impact of product failure on customer requirements, such as safety or cost. Therefore a systematic approach to evaluating the sensitivity of a design to manufacturing variation would resemble a Failure Modes and Effects Analysis (FMEA), a common design tool (http://www.fmeca.com/Default.htm). The modes of variation are analogous to modes of failure. The manufacturing process can be systematically analyzed for variation modes that are significant and difficult to control. For each of these modes, the design can be analyzed for the effect of each variation mode on important top-level performance characteristics. Decisions can then be made to minimize this impact. Thornton (1999) proposes a model for quantifying the effectiveness of monitoring a particular KC. This effectiveness is analogous to the Risk Priority Number in FMEA. Since it is often economically infeasible to monitor every tolerance, this model allows the designer to prioritize tolerances and to select those that are cost-effective to monitor. CHALLENGES The difficulty of evaluating the sensitivity of a design to manufacturing variation depends greatly on the level of detail and complexity of the model used. Figure 10 illustrates, for three solution types, the relationship between the complexity of the analysis (the number of variables considered) and the cost of evaluating the sensitivity of the design to variations in manufacturing. As complexity increases, there is a point where each method becomes prohibitively expensive. Basic intuition can be applied quickly to simple problems involving only a few varying parameters and quality characteristics. However, this problem quickly becomes daunting as the numbers of varying parameters and quality characteristics are increased. Mnemonic lists help structure the analysis and indicate which parameters are important, so the sensitivity can be evaluated systematically. But this method, too, becomes infeasible when the problem becomes very complex. This argues for the use of mathematical models that can be analyzed by computer when the complexity is large. - 37 - Models nemonic lists rtuition 0 1 Complexity of problem (number of variables) Figure10 Cost of solution cursfor thr basic types of sensitivity analysis However, if the project is constrained such that the basic initial cost of modeling is too high, steps must be taken to simplify the problem so that one of the other two methods can be used. For example, a few variation parameters and quality characteristics that are known to be problematic can be collected into lists and analyzed manually at a lower cost. This may not anticipate all the problems, but may maximize the chance of catching them given the time constraint. There are many business analogies to this situation. In business strategy, the analysis of industries could easily be monstrously complex. Mnemonic lists are used to help a strategists systematically consider the important factors in any industry (e.g. Porter's 5 forces, Oster 1998, 31), without overlooking an obvious player. The same is true in marketing. The "4 Ps and 3 Cs" mnemonic helps one address all the important issues of successfully marketing a product or service. Similarly, there is an initial investment (hence, the higher base cost) in developing such a list from experience. However, once the list is established, it can be used to avoid overlooking important opportunities for variation in quality characteristics due to manufacturing. Using the database tools described below (e.g. an Intranet) this list can be maintained and updated with time and experience. As previously mentioned, the model proposed by Thornton attempts to prioritize tolerances by the effectiveness of monitoring them, thereby reducing the complexity of the monitoring task. If there is sufficient time to use it, this model represents a more mathematically systematic approach to simplifying the problem. 6.2 MAKE COOLING CHANNELS ACCESSIBLE One example of considering the needs of subsequent operations in the design and development of earlier processes is the accessibility of the cooling channels. Figure 11 shows a simplified blade with cored cooling channels. The cooling holes must be drilled from the outside into the channel, but must not continue on through the blade. In the figure, the laser would be drilling more or less into the page to form the cooling holes shown in the external view. Note from the cutaway view that these holes must not continue through the rest of the part. When drilling cooling holes with a laser, it is necessary to stop the beam from striking the back wall. This can be done with wax, but is more reliably done with strips of PTFE (Teflon). However, the use of PTFE strips requires easy access to the channel into which the holes are to be drilled. Serpentine cooling channels, similar to those shown in the cutaway view of Figure 11, achieve more effective cooling than simple, straight channels. These cooling channel shapes are often used to cool parts that are - 38 - exposed to very high temperatures. However, these shapes can make insertion of PTFE strips difficult or impossible. If prior operations leave no direct opening into the right portion of the channel, wax must be used instead. Unfortunately wax is not as reliable, since it tends to melt during the laser drilling operation. This is the type of problem that would be identified early if the supplier of the laser drilling operation were involved in the design of prior operations. Although this problem is well known to designers, a laser-drilling expert would be able to provide intuition on the additional cost of drilling into blind channels. One option might be to leave one end of the blade open until after the laser drilling operation. In any case, there are usually small holes in the end of the blade left by pieces of the core used for support during casting. These small holes must be plugged or covered at some stage. If these holes could be left large enough for PTFE strips, the remaining problem would be to find a way to block them afterwards. The cost of the additional hole closing operations would have to be weighed against the cost of drilling into blind channels without PTFE. ble Pori ~ion, Ina ccessa Cooling Holes 21 n( C )red Coo lin Chan nels Cutaway View External view Figure 11 Simplified blade showing cooling airpassagesand exit holes CHALLENGES The need for protecting internal walls from penetration when drilling laser holes has been well known for some time (Corfe 37). Quality data from the laser drilling blades with inaccessible channels can be used to estimate the cost, in additional scrap, of eliminating the possibility of protecting a surface with PTFE. This - 39 - cost can then be weighed against the potential benefit of blind channel designs in the future. The challenge here is to estimate this cost and to incorporate it into the decision-making processes of designers. 6.3 PROVIDE FOR SLAVE DATUM FEATURES Another example of considering machining needs in the casting design is the need to fixture parts off secondary or slave datum features. All dimensions on a cast blade are referenced to six points on the cast surface of the airfoil. However, it is not possible bear the loads of machining operations such as grinding on these six points. Therefore, datum surfaces must be machined by fixturing the parts on the six locating points. Then these new surfaces can be used to fixture the part for subsequent machining operations. Surfaces used as slave datum features ME- Figure 12 Slave datumfeature In the case of the first low-pressure turbine stage vane, the machining supplier used surfaces with excess material for machining these slave datum features. However, the casting was not designed to always have excess material on these surfaces, since the parts could have insufficient material for machining these features and still remain within the drawing tolerance. Although the situation was corrected by altering the wax die, many parts were scrapped in the interim and written off as casting defects. This could have been prevented if the machining supplier had specified the need for material in these areas during the casting design. It is likely that this request would have been made if the machining supplier had been more directly involved during the casting design. 6.4 PROJECT PLANNING INTEGRATION To prevent costly delays, design information must be delivered and used according to schedule. Delays along any point in the critical path lead to delays in finished product delivery. Suppliers expressed concern that they had not received certain pieces of information early enough. For example, a machining supplier suggested that it had received the details of airflow testing late. This resulted in late delivery of the first sets of parts to be tested. - 40 - The fact that suppliers recognize this as a problem emphasizes the need for integrating the project management of the suppliers' process development with that of the overall product development (Fine 1998, 186). This would allow designers to estimate the impact of a design change on tooling cost and delay of project completion. CHALLENGES The challenge in managing the project in the larger sense is to collect all the relevant activities and their associated costs, estimated durations, and dependencies from the suppliers. At the present, the supply chain organization has difficulty determining even the current state of work in progress. Suppliers must be convinced of the need for increased information sharing to permit project planning to be integrated. 6.5 UNDERSTAND IMPACT OF DECISIONS AND CHANGES ON THROUGHPUT TIME, COST AND QUALITY In order to maximize the present value of future earnings, the impact of design decisions on throughput time, development time, manufacturing cost, and product quality must be understood. This suggests that a heuristic cost model of the development process is required. Although the effectiveness of such a model would increase with its accuracy and precision, there are practical limits to its complexity. Design-related factors that affect the cost of producing engines include the elements of Table 1. Table 1 Elements of a cost model Manufacturing Cost Work-in-process, a function of throughput time Tooling and machinery cost Material cost Scrap rate Labor, rework Other Costs Late Penalty, a function of total throughput time for early sets Performance penalty, a function of efficiency and power output Replacement cost, a function of predicted lifetime Overhead, including design The basic formula for NPV is the discounted revenue stream minus the discounted cost stream, including the initial one-time investments (engineering time, tooling, and other development costs). Figure 13 shows the significance of several cost, revenue and timing components. The model assumes an initial development cost and costs associated with the production of each unit sold. Penalties and scrap costs are reduced with time and learning. The throughput time is defined as the time between the initial input of raw materials and the delivery of the product. Revenue from the sale is assumed to coincide with delivery. - 41 - Throughput Time - < Revenue from sales Penalties Ak T 2o Time to Market -ime U Design Aoin Scrap Allowance anufacturing cost aterial cost Figure 13 A simplifiefmncial mldel of the custs and rwues ofa single ngineprct Engineers must balance performance and manufacturing cost. Since the design fixes these costs, it makes sense to have the necessary information available to help them make informed design decisions. Examples of manufacturing cost estimations relating the physical properties of the product to process costs are included in Appendix E. CHALLENGES Though it seems simple, finding the relevant cost of a component can be difficult. The relevant cost depends largely upon which decision is being made (Fine 1998, 237). Designers have developed the tools to determine the effect of a design decision on performance and lifetime. The effects of the same decision on throughput time and the manufacturing costs of Table 1 are more elusive. Designers require information about the supplier's processes, something suppliers often see as proprietary. Again, suppliers must be assured a mutual benefit from the partnership in order to become more involved in the design process. Appendix E shows how the relative cost of drilling holes by laser and EDM can be estimated. If these cost functions are included in the design review software, it seems more likely that they will be used, since such calculations take significant time to formulate and the underlying costs take time to gather. If many of these cost calculations were standardized and automated, designers should be more inclined to use them as decision support tools. 6.6 KEEP MANUFACTURER INFORMED EARLY ABOUT POSSIBLE CHANGES Concurrent engineering inevitably requires design iterations and changes to the design during manufacturing development. The challenge is to minimize the negative impact of these changes on manufacturing cost, quality and time-to-market. Ideally, the manufacturer would like to have a subset of the total design that can be considered fixed, in order to begin development and tooling while the rest of the design is finished. Alastair Clark, an AAP engineer, suggested a means of communicating the certainty of permanence of design features. In addition to nominal geometry and tolerance information, the design documentation could also include an estimate of the likelihood that each dimension will change. This need not be a precise probability, but a categorical estimate of which features are essentially fixed and which are likely to undergo further - 42 - iteration. Such information could be communicated through color-coded features, indicating how certain the designer is that the feature will remain the same in the final design. Krishnan (1993) develops a model for minimizing the total duration of overlapping (concurrent) activities in the face of uncertainty in the value of evolving (unfinished) design parameters and "sensitive" process development activities, the cost of which increases with changes in these parameters (e.g. costs incurred by changes in dimensions after tooling has begun). Krishnan (1992) also recommends ways to improve the transfer of information from design to process development through timing, content, and preliminary information. The formulation of Krishnan's model is listed in Appendix F. Design evolution is a measure of certainty of the value of a particular dimension. Assuming that, at any given point during the design phase, a maximum and minimum value of the dimension can be specified. The evolution is defined as the fraction of the original gap between minimum and maximum values, which has been eliminated by raising the minimum and/or lowering the maximum. Therefore, evolution is 0 at the start of design and 1 at the end. Sensitivity is the degree to which the duration of the downstream activity is lengthened by a change in the design variable. Since some of the parameters may be difficult to obtain quantitatively, Krishnan suggests a qualitative method for making overlapping decisions. If evolution and sensitivity can each be classified as high or low, the chart in Figure 14 can be used to decide what kind of overlapping strategy should be used. Degree of upstream evolution Degree of upstream evolution Slow Evolution Case Process Time ast Evolution Case Process Time Increase in Downstream Duration Low Sensitivity Case Design Change Increase in Downstream Duration High Sensitivity Case Distributive Overlapping Iterative Overlapping LL A2 Ap Design Change Figure 14 Recamnmd Divisive Overlapping or No Overlapping Precipitative Overlapping otwrlap strategygrvm estimates ofezalution and sensitivity As with most engineering and management decisions, the effects must be weighed and balanced. Figure 15 compares the effects of each overlapping strategy on quality, effort, and lead time. - 43 - Degree of upstream evolution Degree of upstream evolution Slow Evolution Case Increase in Downstream Duration Low Sensitivity Case Process Time ast Evolution Case L- Process Time No Iterative Overlapping No quality loss Increase in Effort Smaller lead time Distributive Overlapping Some quality loss Increase in effort Much smaller lead time Divisive Overlapping or No Overlapping No tradeoff or one of the other three types Precipitative Overlapping Some quality loss No increase in effort Smaller lead time Design Change Increase in Downstream Duration High Sensitivity Case Design Change Figure 15 Types of owrlappingand the resultmgperfomanxe tradeoffs CHALLENGES Preliminary information release seems more difficult to implement since it represents additional effort on the part of the designers. In the immediate time horizon, simple certainty information can be communicated directly. For example, if the airfoil shape is considered fixed, the designer can simply say so and allow the developers of the drilling processes to begin programming and fixture designing based on the airfoil geometry. In the future, this would be a prime candidate for incorporation into the CAD software. The challenge at that point would be getting the designers and manufacturing engineers to use it. This seems to be a matter of illustrating its value with one of many examples of cost and delay resulting from poor communication which would be corrected by the use of such tools. 6.7 ACTIVELY INCORPORATE MANUFACTURING FEEDBACK The feedback of manufacturing experience into design requires that the documentation of lessons learned be made available in a convenient format for reference. It also requires that designers actively seek this knowledge in the development of new designs. The most useful guidelines are worthless unless followed. 6.8 SUMMARY Many of these principles of design for manufacturability are well known, not only by the general engineering community, but by many engineers at AAP as well. The fact that some of these problems arose suggests the need for increasing the degree to which such knowledge is distributed among and available to all members of 44 - the design team. The company is improving the manufacturing knowledge of the engineering group. This must continue in order for the company to remain competitive. The next section describes ways the company can go about improving the distribution and accessibility of knowledge. - 45 - SECTION 7: RECOMMENDATIONS FOR LEARNING FROM EXPERIENCE (FEEDBACK TO DESIGN) 7.1 INTRODUCTION The previous section described important principles to be included in a knowledge base of design for manufacturability. This section suggests ways the company can build this knowledge base and make it accessible to the designers. Particular emphasis is given to the knowledge gained from experience in product development. This is how individuals learn best, by experience. It is important, however, that the organization as a whole learn from the experience of individuals. As illustrated by the lessons learned from the B version design in the previous section, it is not enough for a few individuals to learn from experience. There is a great deal of savings to be gained if this experience is available to all designers. This reduces the number of times the same lesson must be learned through trial and error. The following suggestions include methods of building the knowledge of individuals, as well as codifying and organizing knowledge for the entire group. To help build the intuition and expertise of the individual designers, it is suggested that they visit the suppliers and share experiences with one another. New lessons can be learned through analysis of quality data collected by the manufacturers. Databases of experience can be built for access through the internal computer network Some suggestions for building and arranging these databases are given as well. 7.2 SUPPLIER VISITS It is important for designers to actually have some experience in each of the manufacturing processes used to make their parts. It makes sense to send the designers to the factories where the parts are made before they start designing, as well as during development. This allows them to gain intuition in designing for manufacturability and creates the opportunity for them to identify operations that are unnecessary or overly complex and expensive for the design intent. It also requires that suppliers be chosen early as partners in development and that they be willing to have such visitors. For example, Alex Beeck, an AAP component owner, observed that a deburring operation at AETC BMF, a machining supplier, involved extreme care due to a misinterpretation of the specification, which required the manufacturer just to break the edge. This was intended to make the operation easier since the quality of the edge was not very critical. However, it was interpreted to mean that the operator must just break the edge, but avoid cutting any deeper into the part. The operation required careful attention and consumed much more time than necessary. When Alex, a designer with a thorough understanding of the design requirements, walked through the operation from start to finish, he was able to observe the error in communication and make a correction. In the future, this could be avoided by specifying a tolerance on the edge radius. In visiting suppliers, designers should not neglect assembly. There is a common perception that manufacturing ends before assembly begins. However, many manufacturing problems become immediately apparent at assembly. Geometrical problems show up when parts don't fit. Assembly is starved when parts are late. Unlike the aircraft engines industry, there is little opportunity to test land-based gas turbines prior to service because of their size and cost. This makes assembly even more important since it is an opportunity to observe some of the problems of the design before the engine is turned over to the customer. CHALLENGES This is done to some extent already, but should include the entire chain, not just the casting. It is tempting to concentrate on casting and leave out machining and coating because of the relative cost of casting due to - 46 - material cost. However, mistakes that scrap castings in machining cost as much or more than mistakes that occur in casting. This focus on material cost at casting also neglects to take into account the relative throughput times of casting, machining and coating. As noted earlier, delays in power plant delivery incur expensive penalties. A delay in machining or coating represents a delay in final delivery of the part and of the entire project if the part is on the critical path. 7.3 ANALYSIS OF QUALITY DATA There is an opportunity to learn from manufacturing experience by analyzing quality data collected over many parts. Patterns can be found that indicate potential problems. In some cases this suggests a change in the manufacturing process. Alternatively, it may make sense to change the design specification to make the part more manufacturable. NCR The NCRs of two components were analyzed for patterns in defects. For each component, the defects were classified by the operation in which they occurred. For Vane 1, some of these defects could be eliminated by eliminating the operation that produced them. After an investigation of causes of the defects, it was revealed that the operations were, in fact, not necessary. Some were operations added to correct casting defects. When the casting defects were corrected, the machining operations were redundant. This was the case for the hook relief and the brazing surfaces on the platforms. NCR SERIAL NUMBER TRACEABILITY In order to apply a similar analysis to the NCRs, one must be able to search NCRs by serial number to find correlation between different types of nonconformance that occurred on the same part. At the start of the project, it was not possible to search for every NCR raised by a particular serial number. NCRs generally, though not always, had the correct serial numbers listed with their associated defects. However, these serial numbers were often included in a form that was readable, but not searchable. For example, many serial number lists were included as an attached picture or spreadsheet file. The NCR database did not have the capability to search for numbers in such attachments. A method of indexing the NCR records by serial number was developed and recommended for all subsequent NCRs. The method was adopted and is in the process of being implemented. This capability has the added benefit of permitting a search for all NCRs related to each of the parts in an engine, which may become necessary in the case of an engine failure and resulting liability investigation. As the number of NCRs raised using the new method grows, and as old NCRs are converted to the new format, this type of analysis will be possible, revealing interactions between different kinds of defects. NCRFAPS, KNOWLEDGE-BASED SYSTEM The Non Conformance Report Fast Answering Process Software, being developed by Bernard Robic, an AAP engineer, presents a significantly improved platform for analyzing quality data. Currently applied to the answering of cooling airflow NCRs, the software accepts data measured by suppliers and generates a disposition for each affected part. Error proofing is performed at several levels to ensure that the information is correct before it is accepted. The information for each part is then stored in a database in an easily searchable form. This allows each manufacturer's airflow quality to be analyzed for process stability, currently a manually intensive process, prone to error. - 47 - CHALLENGES This should be easier since the implementation of serial number traceability in the NCR database. As the NCRFAPS system is used, process control data will be much easier to access and analyze. The measurements will be stored in a more structured way. Error checking will ensure cleaner data. Once NCRFAPS is demonstrated in airflow, it should be easier to implement in other areas of quality control. This will not happen by itself, though. A champion must fight the natural tendency to use a different tool in every department by promoting the value of NCRFAPS to other areas of the design department. 7.4 PRODUCT SUPPORT TOOLS The Product Support group has defined a process for incorporating field experience into design. This consists of several tools: Experience Response System (ERS), Read-only project status and operating data, Problem History Files (PHF), Design Office, and case studies. This structure would be well suited for manufacturing feedback as well. To avoid the proliferation of information tools, it seems best to try to use existing structures to build the manufacturing knowledge base. When a new kind of tool is required for manufacturing, it makes sense to integrate it with product support in order to maintain uniformity and compatibility. EXPERIENCE RESPONSE SYSTEM The Experience Response system is a database of documentation of field failures. It contains documents of various forms (including faxes, memos, e-mail messages, and NCRs) that describe the problems and the subsequent actions and contain links to more detailed reports. These documents can be sorted by component, project, system and other fields and are.indexed for full-text searchability. PROBLEM HISTORY FILES Similar to the ERS, Problem History Files contain a description of a problem in the field and the action taken to solve it. The PHF form provides a structured approval path for each problem. The records can be sorted by owner, GT type, approver, and status and are indexed for full-text searchability. PROJECT STATUS AND COMMUNICATION FILES Project information and the status of problems currently being solved are stored on network servers to which most users have read-only access. The Design Office database contains documentation of the proposal and approval of official design changes resulting from field failures. The ERS, PHF, and Design Office databases allow easy storage of information describing field experience. However, designers suggested that relevant information is still difficult to access. This system seems appropriate for the documentation of manufacturing problems and solutions, but can only be effective if the process of storing and retrieving the information is made smoother. CASE STUDIES The Product Support Team conducts case studies with large groups as learning exercises. Field failures are presented to the participants, who suggest solutions. These suggestions are compared with the actual implemented solutions and discussed. For those who participate in these case study workshops, they seem to be relatively effective. The participants, including design engineers, get a good understanding of the problem 48 - of interest. When a large number of people are thinking about the same problem, they generate solutions and ideas that had not occurred to the few who dealt with the original problem. These also serve as an exercise for introducing new employees to the kinds of problems faced in design. While case studies cannot be relied upon to communicate every type of lesson learned from experience, they are good for making people aware of major problems that occurred in the field and how these problems were solved. They seem, therefore, appropriate for communicating and studying problems faced in manufacturing and development as well. CHALLENGES These tools were recommended because they are already used for field feedback and should, therefore, be relatively easy to implement in manufacturing feedback. The main challenge is to see whether or not the information is used. This will depend a great deal on how easily relevant information can be accessed from the database. There is great potential to make both the manufacturing feedback and the field feedback more powerful tools by improving the user interface. This would require a commitment of resources from management to organize the experience data and configure the search tools in a way that maximizes the usefulness to designers. This includes the time of a designer that uses or could use the tools and the time of a programmer. 7.5 INTRANET Within the company's intranet, a structure is being developed for publishing turbine design information internally. The structure contains both public and private sections. It allows the user to find test data, lessons learned, design standards, and other relevant design information through a hierarchical taxonomy. This facilitates knowledge transfer from the receiving end, allowing the user to quickly find and "pull" relevant information, rather than wading through volumes of documents or waiting for the relevant memo to be circulated. This intranet structure would be an ideal interface for integrating the Product Support tools and analogous manufacturing feedback tools. CHALLENGES Similarly, this tool will simply require the investment of a designer's time and a programmer's time to develop a user-friendly system to allow designers to pull a wide range of relevant information about the project, design standards, field and manufacturing experience, etc. It only makes sense to incorporate search capability of the product support database into this intranet portal interface. 7.6 STONER ULE® AAP is currently working with a consulting firm, CADFEM, to develop tools to build knowledge into the design software (CATIA) with a package called STONEruLe@. An object-oriented programming language, STONEie@ provides an interface between the user and CATIA, applying design rules to the user input to automate some design tasks. The STONErue@ package was already used successfully to automate much of the design of vane carriers. It is doubtful that much of blading design could be reduced to rules for automation. However, there is a great opportunity to use the tool to incorporate rules learned from field and manufacturing experience into the CAD system to alert the designer of potential problems. AUTOMATE 2D DRAWING GENERATION The STONEne@ package is currently being programmed to automatically generate 2D drawings based on 3D models of precision cast parts. - 49 - INCORPORATE FEEDBACK INTO MEG In its "design review" mode, the STONEnie@ software can give the user advice on design features based on pre-programmed rules. When a new rule is learned from manufacturing, it can be programmed into the software for use by all designers. Although the name "stone rules" implies permanence, it is the possibility of adding to or changing the rules over time that makes the software a powerful tool for knowledge transfer. CHALLENGES Again, this will require the time of a programmer and part of the time of one or more design engineers. 7.7 SUMMARY In order to improve DFM learning, these recommendations for building the design organization's manufacturing knowledge base were presented to the company. For ease of implementation, acceptance, and maintainability, many of these are simply modifications of existing tools used by the design team. The development of these tools will still require the investment of resources. It may be difficult to quantifiably justify this investment, since one cannot predict what problems will be avoided in the future. However, if these tools can prevent some of the major problems experienced in the past from recurrmng, the investment will be quickly recovered. The following section summarizes the lessons learned and concludes the thesis. - 50 - SECTION 8: CONCLUSIONS 8.1 INTRODUCTION The project's recommendations achieve both aims of the thesis: to illustrate the importance of knowledge management in concurrent engineering and to recommend ways to improve it. The recommendations for dealing with suppliers illustrate the importance of knowledge management and hint at the consequences of failing to do it well. The mere existence of issues like those that brought up the recommendations for design calls for improvements in knowledge management. These issues are common to many manufacturing companies, not just the sponsoring organization. Finally, the suggestions for improving the feedback from manufacturing offer ways to improve the management of knowledge within the design organization as well as with manufacturers. 8.2 DEALING WITH SUPPLIERS In committing to suppliers early, the company reaps the value of incorporating the manufacturers'knowledge into the design (alternatively, avoiding the cost fixed in the design by excluding this input). At the same time, careful management is required to avoid becoming dependent upon a single source. Involving the whole chain means taking advantage of the knowledge of ewry supplier. The value of this is illustrated by several examples given: the hook feature of Vane 1, blocking the laser, and EDM of the brazing surface. Aligning incentives shows the synergy of enforcing quality standards, project management, inventory management, and improving manufacturing feedback. 8.3 DESIGNING PARTS Minimizing sensitivity to variation requires the engineering organization to have knowledge of the manufacturer's process variables. Providing for slave datum features requires knowledge of the features to be used, or at least the issues that drive the choices of these surfaces. Timely information delivery and informing the manufacturer early both require knowledge of the manufacturer's timeline, cost information, and sensitivity to changes and integration of the project management of the manufacturing process development with that of the product design. Supporting design decisions with cost models requires bringing together cost information from many sources to the engineering organization. Data from one product introduction must be analyzed to predict the dynamics of subsequent development efforts. Cost and process information must also be collected from manufacturers. This requires transferring knowledge from sources that traditionally have had complete and careful control of it. The political barriers to obtaining such information must not be overlooked as this will likely be perceived as a threat to the former knowledge brokers' power. 8.4 LEARNING FROM EXPERIENCE When designers visit suppliers and before, during, and after the design project, they have the opportunity to gain valuable, relevant knowledge first-hand. Analyzing quality data allows the organization to transform volumes of numbers into valuable lessons. The product support tools and the intranet help the whole organization to learn from its individual members. CAD software automates the process of pulling relevant knowledge into the design process and maps it to appropriate geometric features. 8.5 OPPORTUNITIES FOR FUTURE WORK There are a number of potential projects for future interns resulting from the recommendations of this project. There are still problems to be solved in implementing these suggestions. There are opportunities for future interns to: - 51 - * Facilitate the building of a set of rules for manufacturing that can be programmed into the STONEruk-@ software and used in the design review mode to flag potential manufacturing problems. This would require working with suppliers and designers to collect a set of rules and working with a consultant to program them into the software. * Support the development of the Intranet design manual by working with designers to develop the most valuable structure and content and by designing the process by which it can be maintained. This may include incorporating the existing Product Support tools and modifying them to include manufacturing feedback. * Develop a cost model of individual manufacturing operations to help the turbine design department understand the impact of design decisions on TPT, Cost, and Quality. This model could also be incorporated into the design review mode of the STONEruLe® software. - 52 - REFERENCES ASM Handbook. (1988) Materials Park, Ohio: ASM International. Boothroyd, G. and P. Dewhurst (1983). Design for Assembly: A Designer's Handbook. Department of Mechanical Engineering, University of Massachusetts at Amherst. Bralla, J. G. (Ed.) (1998). Design for Manufacturability Handbook, 2nd Ed. New York: McGraw Hill. Corfe, A. G. (1983) "Laser drilling of aero- engine components." Proceedings of the 1s International Conference on Lasers in Manufacturing. 31-40. Fine, CH. (1998). Clockspeed: Winning Industry Control in the Age of Temporary Advantage. Reading, Massachusetts: Perseus Books. Kalpakjian, S. (1995). Manufacturing Engineering and Technology, 3rd Ed. New York: Addison-Wesley Publishing Co. Krishnan V. (1992). "Overlapping product development activities by analysis of information transfer practice." Working paper # 3478-92 MS. MIT Sloan School of Management. Krishnan V. (1993). "A model-based framework to overlap product development activities." Working paper # 3635-93 MSA. MIT Sloan School of Management. Oster, S. M. (1994). Modern Competitive Analysis. Oxford University Press, New York Powell, J. (1989). "The influence of material thickness on the efficiency of laser cutting and welding." Proceedings of the 6t International Conference on Lasers in Manufacturing. 215-221. Schimmoller, B. K. (1999). "Balancing compliance with competition." Power Engineering. 103(10): 22-28. Senge, Peter M. (1994). The Fifth Discipline: the Art and Practice of the Learning Organization. Currency Doubleday, New York. Springborn, R. K. (Ed.) (1967). Non-traditional Machining Processes. Dearborn, Michigan: American Society of Tool and Manufacturing Engineers. Thornton, A. C. (1999). "A Mathematical Framework for the Key Characteristic Process." Research in Engineering Design. 11:145-157. - 53 - APPENDIX A: CONCURRENT ENGINEERING PROCESS Figure 16 shows the sequence of events in the product development process. Note that many design and process development activities occur concurrently in order to minimize time to market. The following table briefly describes each of the tasks listed in Figure 16. See the figure for the sequence of events. Note that the timeline is not to scale. Task Description Component Specifications and Targets Project managers create the project plan, including performance and cost targets. Preliminary Design Project managers, and design team leaders, conduct a feasibility study, and assess potential problems, suppliers and delivery schedule. Review Preliminary Design The project targets are reviewed and converted into boundary conditions and targets for the conceptual design. Conceptual Design, Manufacturing Documentation The design team leaders work with the proposed suppliers to estimate feasibility, cost, and delivery schedule. Concept Design Review Project managers approve the design plan, including boundary conditions. Concurrent Design Phase A The design team works with suppliers to develop the component design specifications. Concurrent Design Phase B A continuation of Phase A with input from Concurrent Manufacturing Phase Al. Design Review The project managers review and release the technical drawings and casting and core dies. Production Documents, Phase A The design team works with suppliers to develop production specifications and test procedures for each phase of manufacturing (casting, coating, machining, and assembly). Production Document Review Project managers review the result of production process development and release pre-serial production. Component / Design Validation Production parts are tested for conformance to the design intent. Production Documents, Phase B Production documentation is finalized. Product Review The results of product development are evaluated and the design is frozen. - 54 - Manufacturing Preparation The supply chain management team coordinates the plan for routing parts through the supply chain. Suppliers provide cost and schedule estimates and expertise to support the manufacturability of the design. Supplier Selection Orders are placed with suppliers. A delivery schedule is established. Concurrent Manufacturing, Phase Al Suppliers provide input to the design team and a draft of production routing. Concurrent Manufacturing, Phase A2 Suppliers continue to support design and update routing and schedule. Concurrent Manufacturing, Phase A3 Tools are designed and manufactured. The manufacturing and testing processes are documented. Preseries Production Release Production documentation is reviewed and preserial production is approved. Concurrent Manufacturing, Phase B 1 The supplier provides preserial sets and updates to production documentation as yields improve. Concurrent Manufacturing, Phase B2 The production process is stabilized and documentation is finalized. Initial Sample Inspection Quality managers approve production processes. Tolerances are fixed. Transfer Serial Order Execution The result of the development process is documented. Transfer Document The design department relinquishes responsibility for production. - 55 - Task Component Spec and Targets I Preliminary Design Review Preliminary Design Conceptual Design, Mfg. Doc. 7 Concept Design Review Concurrent Design Phase A Concurrent Design Phase B Design Review K.____ K____ Production Doc. Phase A Production Doc. Review Component Design Validation Production Doc. Phase B Product Review Manufacturing Preparation I Supplier Selection Concurrent Mfg. Phase Al Concurrent Mfg. Phase A2 Concurrent Mfg. Phase A3 Preseries Production Release * Concurrent Mfg. Phase B1 Concurrent Mfg. Phase B2 Initial Sample Inspection Transfer to Serial Order Execution Transfer Document Figum 16 Conament Enginering Tmeline I APPENDIX B: RELEVANT LITERATURE Many of the recommended design principles and supply chain concepts are based upon existing research. This section describes several sources of knowledge on the topics relevant to this thesis. The reader is encouraged to consult these sources for more information on Design for Manufacturability, Supply Chain Management, and Organizational Learning, since all of these are useful resources for anyone engaged in concurrentengineering and product development. DFM Boothroyd and Dewhurst (1983) did a great deal of groundbreaking work in Design for Assembly. Their research resulted in several basic principles for designing parts to simplify the assembly operation. They also developed tools for evaluating the ease of assembly of a particular design. There are similar tools for evaluating the ease of manufacturing a design. Separate sets of principles are written for different manufacturing processes, some of which are included in Appendix E. For example, in designing a part to be cast, it is generally best to avoid sharp corners, abrupt changes in thickness, and undercuts, and to let cross-sectional area decrease with increasing distance from the gates, so that material will not be restricted before it has filled the cavity. The Design for Manufacturability Handbook (Bralla, 1998) is a collection of such principles in a single text. MANUFACTURING TECHNOLOGIES A basic understanding of the processes used to manufacture precision cast parts was obtained from general handbooks. Kalpakjian (1995) briefly describes most of the relevant operations in sufficient detail to allow the reader to get an appreciation for the physics involved. The text includes articles on investment casting (p. 306), hot isostatic pressing (p. 510), electrical-discharge machining (p. 836), electro-chemical machining (p. 832), laser drilling (p. 840), and coating (p. 996). A more thorough treatment of each can be found in the ASM Handbook (1988). SUPPLY CHAIN MANAGEMENT Fine (1998) emphasizes the importance of supply chain design as a core competence and the only truly sustainable source of competitive advantage. He examines companies in "high clockspeed" industries (frequent organizational changes, new product introductions, etc.) to gain insight into the principles of supply chain management for industries of all clockspeeds. He draws the analogy to fruitflies, which are studied in genetics because of their short life cycle. Several generations of fruitflies can be observed in the laboratory setting. The observations made can be generalized to species with life cycles that are much too long to observe. By observing such "fruitfly" industries as personal computers and multimedia entertainment, Fine makes several observations about the nature of changes in industries over time. One important tendency is for industries to oscillate between a horizontal structure and vertical integration along a path Fine calls a doublehelix. Figure 17 illustrates the difference between vertical and horizontal industries and between integral and modular product architecture. Figure 18 (Fine 1998, p. 49) lists the forces that drive industries back and forth between vertical and horizontal structure. - 57 - 1 Firm A Component 2 Firm D Component Firm B Firm C Firm A Firm B Firm C Fir m E Component 3 Component 4 Firm Firm Firm Firm F G H I Firm K Firm L Firm J Fir m M Horizontal Industry, Modular Product Vertical Industry, Integral Product Figure 17 Horizontal modular structure s. wrtical integral Niche competitors Integral product, vertical industry Modular Product, horizontal industry Technical adcances High dimensional complexity Supplier market power Organizational rigidities Pressure to disintegrate Pressure to integrate ----- Proprietary system profitability Figure 18 The double helix of oscillation in industry structureand product architectum Fine asserts that, because industries and products are constantly changing, competitive advantage is temporary. At best, the only sustainable advantage is having the ability to manage knowledge across the supply chain. This meta-competency of supply chain design must be integrated with product and process design in what Fine calls "Three-dimensional concurrent engineering" (3DCE). ORGANIZATIONAL LEARNING Senge (1994) builds a framework for learning organizations based on the experience of several companies. He suggests that the fundamental requirement for learning organizations is a change in thinking. Rather than seeing oneself as a powerless victim of circumstances, one should see the part one plays in the larger system, of which one is an element. Senge outlines five "disciplines" for an effective learning organization: 1) Personal Mastery - commitment of the individuals to a group goal and self-improvement 2) Mental Models - assumptions, world view, learning to recognize one's own, and sharing models with others. 3) Shared Vision - translating individual vision into shared vision, not dictating. 4) Team Learning - a team's ability to improve performance by working together over time and to achieve a higher level of competence than any individual could 58 - 5) Systems thinking - understanding interactions and system dynamics Of particular interest are the basic principles of system dynamics. Senge lists several system archetypes, such as balancing loops with delay, limits to growth, shifting the burden, eroding goals, escalation, tragedy of the commons, etc., which can be found in many business systems. The learning model in Section 3 is a very simple system, which includes growth and balancing loops. Appendix C explains the symbols used in the model's diagrams. Appendix D lists the details of the model. - 59 - APPENDIX C: CAUSAL LOOP DIAGRAMS Complex dynamic systems can be modeled using only a few simple components and establishing causal relationships between them. Figure 19 shows a simple causal loop diagram and its basic elements. Variables can be constants or functions of other variables. Arcs between variables indicate that one variable is a function of the other, a causal relationship. A plus or minus symbol near the head of the arrow indicates the relationship between the two variables. If the sign is positive, then as one variable increases, the other tends to increase as well. If the sign is negative, then as one tends to increase, the other tends to decrease. Variable c Source F71w in Sink k Flow out Figure 19 The basic componmts ofa causal loop diagran There are two special types of variables: stocks and flows. A flow is a variable that happens to feed or drain a stock. Flows can be constants or functions of other variables. A stock is strictly a function of those flows that feed or drain it. Formally, a stock is the integral of the sum of incoming flows minus the sum of outgoing flows. A bathtub could be modeled as a stock of water where the faucet allows water to flow in at a certain rate and the drain allows water to flow out at another rate. Sources and sinks provide a simple way to limit the scope of the model. Each has limitless capacity and basically serves to terminate a flow without specifying an additional stock from which the flow draws or into which the flow drains. Such systems can be modeled after the relationships between variables are formulated and initial conditions are specified. The system model presented in this thesis was simulated using VensimPLE32 Version 3.OD1. The software performs a discrete-time simulation by calculating the values of each variable once for every time increment. The values of any variables can then be plotted to allow the user to observe trends. - 60 - APPENDIX D: A MODEL OF DFM LEARNING AND ITS COST IMPLICATIONS Figure 20 shows all the variables needed to run the simulation used to draw the conclusions in section 3. The following formulation explains the meaning of each variable and some assumptions. ntion ction Initial DFM Retained in itial SUTA In itial Product Tech + Projects per month Cost of Capital Vf learning + feed back 4-communication with suppliers D DFM Learning loop DFM + Learning rate future payment Quality of + Manufa cturability Gap Manufacturability Knowledge Management Effectiveness Rate of Ra fA advaning 1000/1 Worst case scrap cost Manufact urable Tech Development per eng.-hr. Worsst case Eng. hrs. for + D FM + <Projects p er month> techology Rate of advancing state of the--. art - + echnology g ap + Scrap/ Rework cost + Wors t case performance penalty <Time> <TIME STEP> + Delivery Penalties Engineering Time improving M Total DFM + Engineering Engineering + Time time -V developing technology + Engineerin ngineering + cost Cost per hu + + + Worst cas delivery dela penalty Technology slI + N-, cost stream of + + Cost per Project + + Performance + penalties Figure 20 Causal loop diagramof complete DFM leaming mcdel MODEL FORMULATION (01) "100% Manufacturable"= 1 Units: Dimensionless The maximum value of manufacturability. (02) Cost of Capital= 0 .1/12 Units: Dimensionless Interest rate that determines the opportunity cost of capital. Cost per Project= Delivery Penalties+Engineering cost+Performance penalties+"Scrap/ Rework cost" Units: Dollars The average cost per new product (03) - 61 - (04) Delivery Penalties= Manufacturability Gap*Worst case delivery delay penalty Units: Dollars Penalty incurred per project for late delivery. (05) DFM= INTEG ( Learning rate, Initial DFM) Units: Dimensionless Learned ability of engineering to design for manufacturability. (06) DFM feedback= Manufacturability Gap*Quality of communication with suppliers Units: Dimensionless Information received by engineering from manufacturing on the manufacturability of the design. (07) Engineering cost= Engineering Cost per hour*Total Engineering Time Units: Dollars The total engineering cost per project. (08) Engineering Cost per hour= 100 Units: Dollars/hr Cost of one hour of one engineer's time. (09) Engineering time developing technology= 5120 Units: hr Engineer-hours (10) spent developing product technology per project Engineering Time improving DFM= Manufacturability Gap*"Worst case Eng. hrs. for DFM" Units: hr Amount of engineering time required to make design manufacturable. (11) FINAL TIME = 100 Units: Month The final time for the simulation. (12) Initial DFM= 0 Units: Dimensionless Value of DFM at time=0 - 62 - (13) Initial Product Tech= 1 Units: Dimensionless Value of Product Tech. at time=0 (14) Initial SOTA= 1 Units: Dimensionless Value of SOTA at time=0 (15) INITIAL TIME Units: Month = 0 The initial time for the simulation. (16) Knowledge Management Effectiveness= 0.75 Units: Dimensionless The fraction of relevant DFM knowledge that is successfully retreived and applied to new designs. (17) Learning rate= Retained learning-Technology slip Units: 1/Month The increase in the ability of the design organization to produce manufacturable designs. (18) Manufacturability= DFM*Knowledge Management Effectiveness Units: Dimensionless The ease with which designs can be manufactured. (19) Manufacturability Gap= "100% Manufacturable" -Manufacturability Units: Dimensionless The amount by which manufacturability could improve. (20) NPV of cost stream= INTEG ( NPV of future payment, 0) Units: Dollars The total net present value of all future costs within the time horizon. NPV of future payment= Cost per Project*Projects per month/(1+Cost of (Time/TIME STEP) Capital)^ Units: Dollars/Month (21) - 63 - (22) Performance penalties= Technology gap*Worst case performance penalty Units: Dollars The cost of poor product performance due to lost sales and performance-based price penatlies. (23) Projects per month= 0.15 Units: 1/Month Number of new projects started per month, assumed two per year. (24) Quality of communication with suppliers= 0.75 Units: Dimensionless Fraction of design-related manufacturing problems communicated by supplier. (25) Rate of advancing state of the art= 0.1 Units: 1/Month The percent of DFM knowledge that becomes obsolete due to advances in product technology. (26) Rate (27) Retained learning= DFM feedback*retention fraction Units: 1/Month of advancing techology= Engineering time developing technology*"Tech Development per eng.-hr."*Projects per month Units: 1/Month The percent increase in product technology level per unit time Amount of feedback actually retained for future product development cycles. (28) retention fraction= 0.75 Units: 1/Month Fraction of feedback actually retained for future product development cycles. (29) SAVEPER = 1 Units: Month The frequency with which output is stored. (30) "Scrap/ Rework cost"= Manufacturability Gap*Worst case scrap cost - 64 - Units: Dollars Cost of scrap and rework per project. (31) "Tech Development per eng.-hr."= 0.0001 Units: 1/hr The rate at which technology development proceeds per engineer-hour invested. (32) Technology gap= (Rate of advancing techology)/Rate of advancing Units: Dimensionless The amount by which the state of the state of the art-Rate of advancing state of the art company's product technology lags the art. Technology slip= DFM*Rate of advancing techology Units: 1/Month The amount by which the value of DFM knowledge is reduced because of increases in the level of product technology. (33) (34) = 1 Units: Month The time step for the simulation. TIME STEP Total Engineering Time= Engineering time developing technology+Engineering Time improving DFM Units: hr Total engineer-hours spent per project (35) (36) Worst case delivery delay penalty= 4e+007 Units: Dollars The penalty per project for delayed product delivery in the worst case, where the design is as difficult to manufacture as imaginable, arbitrary figure assumed. "Worst case Eng. hrs. for DFM"= 4000 Units: hr Engineering hours required to rework design to make it manufacturable in the worst case, where the design requires the (37) - 65 - most modification imaginable, assumed to be one fulltime engineer dedicated to modifications for two years. (38) Worst case performance penalty= le+007 Units: Dollars Cost of lagging the state of the art through lost revenues and performance price penalties in the worst case imaginable, arbitrary figure assumed. (39) Worst case scrap cost= le+007 Units: Dollars Cost of scrap and rework per project in the case of the worst manufacturability imaginable, arbitrary figure assumed. (01) "100% Manufacturable"= 1 Units: Dimensionless The maximum value of manufacturability. (02) Cost of Capital= 0 .1/12 Units: Dimensionless Interest rate that determines the opportunity cost of capital. (03) Cost per Project= Delivery Penalties+Engineering cost+Performance penalties+"Scrap/ Rework cost" Units: Dollars The average cost per new product (04) Delivery Penalties= Manufacturability Gap*Worst case delivery delay penalty Units: Dollars Penalty incurred per project for late delivery. (05) DFM= INTEG ( Learning rate, Initial DFM) Units: Dimensionless Learned ability of engineering to design for manufacturability. (06) DFM feedback= -66 - Manufacturability Gap*Quality of communication with suppliers Units: Dimensionless Information received by engineering from manufacturing on the manufacturability of the design. (07) Engineering cost= Engineering Cost per hour*Total Engineering Time Units: Dollars The total engineering cost per project. (08) Engineering Cost per hour= 100 Units: Dollars/hr Cost of one hour of one engineer's time. (09) Engineering time developing technology= 5120 Units: hr Engineer-hours spent developing product technology per project (10) Engineering Time improving DFM= Manufacturability Gap*"Worst case Eng. hrs. for DFM" Units: hr Amount of engineering time required to make design manufacturable. (11) FINAL TIME = 100 Units: Month The final time for the simulation. (12) Initial DFM= 0 Units: Dimensionless Value of DFM at time=0 (13) Initial Product Tech= 1 Units: Dimensionless Value of Product Tech. at time=0 (14) Initial SOTA= 1 Units: Dimensionless Value of SOTA at time=0 (15) INITIAL TIME = 0 Units: Month - 67 - The initial (16) time for the simulation. Knowledge Management Effectiveness= 0.75 Units: Dimensionless The fraction of relevant DFM knowledge that is successfully retreived and applied to new designs. (17) Learning rate= Retained learning-Technology slip Units: 1/Month The increase in the ability of the design organization to produce manufacturable designs. (18) Manufacturability= DFM*Knowledge Management Effectiveness Units: Dimensionless The ease with which designs can be manufactured. (19) Manufacturability Gap= "100% Manufacturable" -Manufacturability Units: Dimensionless The amount by which manufacturability could improve. (20) NPV of cost stream= INTEG ( NPV of future payment, 0) Units: Dollars The total net present value of all future costs within the time horizon. (21) NPV of future payment= Cost per Project*Projects per month/(l+Cost of Capital)^(Time/TIME STEP) Units: Dollars/Month (22) Performance penalties= Technology gap*Worst case performance penalty Units: Dollars The cost of poor product performance due to lost sales and performance-based price penatlies. (23) Projects per month= 0.15 Units: 1/Month Number of new projects started per month, assumed two per year. (24) Quality of communication with suppliers= - 68 - 0.75 Units: Dimensionless Fraction of design-related manufacturing problems communicated by supplier. (25) Rate of advancing state of the art= 0.1 Units: 1/Month The percent of DFM knowledge that becomes obsolete due to advances in product technology. (26) Rate of advancing techology= Engineering time developing technology*"Tech Development per eng.-hr."*Projects per month Units: 1/Month The percent increase in product technology level per unit time (27) Retained learning= DFM feedback*retention fraction Units: 1/Month Amount of feedback actually retained for future product development cycles. (28) retention fraction= 0.75 Units: 1/Month Fraction of feedback actually retained for future product development cycles. (29) = 1 SAVEPER Units: Month The frequency with which output is stored. (30) "Scrap/ Rework cost"= Manufacturability Gap*Worst case scrap cost Units: Dollars Cost of scrap and rework per project. (31) "Tech Development per eng.-hr."= 0.0001 Units: 1/hr The rate at which technology development proceeds per engineer-hour invested. Technology gap= (Rate of advancing state of the art-Rate of advancing techology)/Rate of advancing state of the art (32) Units: Dimensionless - 69 - The amount by which the company's product technology lags the state of the art. (33) Technology slip= DFM*Rate of advancing techology Units: 1/Month The amount by which the value of DFM knowledge is reduced because of increases in the level of product technology. (34) TIME STEP 1 Units: Month The time step for the simulation. (35) Total Engineering Time= Engineering time developing technology+Engineering Time improving DFM Units: hr Total engineer-hours spent per project (36) Worst case delivery delay penalty= 4e+007 Units: Dollars The penalty per project for delayed product delivery in the worst case, where the design is as difficult to manufacture as imaginable, arbitrary figure assumed. (37) "Worst case Eng. hrs. for DFM"= 4000 Units: hr Engineering hours required to rework design to make it manufacturable in the worst case, where the design requires the most modification imaginable, assumed to be one fulltime engineer dedicated to modifications for two years. (38) Worst case performance penalty= le+007 Units: Dollars Cost of lagging the state of the art through lost revenues and performance price penalties in the worst case imaginable, arbitrary figure assumed. (39) Worst case scrap cost= le+007 -70- Units: Dollars Cost of scrap and rework per project in the case of the worst manufacturability imaginable, arbitrary figure assumed. (01) "100% Manufacturable"= 1 Units: Dimensionless The maximum value of manufacturability. (02) Cost of Capital= 0.1/12 Units: Dimensionless Interest rate that determines the opportunity cost of capital. Cost per Project= Delivery Penalties+Engineering cost+Performance penalties+"Scrap/ Rework cost" Units: Dollars The average cost per new product (03) (04) Delivery Penalties= Manufacturability Gap*Worst case delivery delay penalty Units: Dollars Penalty incurred per project for late delivery. (05) DFM= INTEG ( Learning rate, Initial DFM) Units: Dimensionless Learned ability of engineering to design for manufacturability. (06) DFM feedback= Manufacturability Gap*Quality of communication with suppliers Units: Dimensionless Information received by engineering from manufacturing on the manufacturability of the design. (07) Engineering cost= Engineering Cost per hour*Total Engineering Time Units: Dollars The total engineering cost per project. (08) Engineering Cost per hour= 100 Units: Dollars/hr Cost of one hour of one engineer's time. -71 - (09) Engineering time developing technology= 5120 Units: hr Engineer-hours spent developing product technology per project (10) Engineering Time improving DFM= Manufacturability Gap*"Worst case Eng. hrs. for Units: hr Amount of engineering time required to make design manufacturable. (11) FINAL TIME = 100 Units: Month The final time for the simulation. (12) Initial DFM= DFM" 0 Units: Dimensionless Value of DFM at time=0 (13) Initial Product Tech= 1 Units: Dimensionless Value of Product Tech. at time=0 (14) Initial SOTA= 1 Units: Dimensionless Value of SOTA at time=0 (15) INITIAL TIME Units: Month = 0 The initial time for the simulation. (16) Knowledge Management Effectiveness= 0.75 Units: Dimensionless The fraction of relevant DFM knowledge that is successfully retreived and applied to new designs. (17) Learning rate= Retained learning-Technology slip Units: 1/Month The increase in the ability of the design organization to produce manufacturable designs. (18) Manufacturability= -72 - DFM*Knowledge Management Effectiveness Units: Dimensionless The ease with which designs can be manufactured. (19) Manufacturability Gap= "100% Manufacturable" -Manufacturability Units: Dimensionless The amount by which manufacturability could improve. (20) NPV of cost stream= INTEG ( NPV of future payment, 0) Units: Dollars The total net present value of all future costs within the time horizon. NPV of future payment= Cost per Project*Projects per month/(1+Cost of Capital)^(Time/TIME STEP) Units: Dollars/Month (21) (22) Performance penalties= Technology gap*Worst case performance penalty Units: Dollars The cost of poor product performance due to lost sales and performance-based price penatlies. (23) Projects per month= 0.15 Units: 1/Month Number of new projects started per month, assumed two per year. (24) Quality of communication with suppliers= 0.75 Units: Dimensionless Fraction of design-related manufacturing problems communicated by supplier. (25) Rate of advancing state of the art= 0.1 Units: 1/Month The percent of DFM knowledge that becomes obsolete due to advances in product technology. of advancing techology= Engineering time developing technology*"Tech Development per eng.-hr."*Projects per month Units: 1/Month (26) Rate - 73 - The percent increase in product technology level per unit (27) Retained learning= DFM feedback*retention fraction Units: 1/Month Amount of feedback actually retained for future product development cycles. (28) retention fraction= time 0.75 Units: 1/Month Fraction of feedback actually retained for future product development cycles. (29) SAVEPER = 1 Units: Month The frequency with which output is stored. (30) "Scrap/ Rework cost"= Manufacturability Gap*Worst case scrap cost Units: Dollars Cost of scrap and rework per project. (31) "Tech Development per eng.-hr."= 0.0001 Units: 1/hr The rate at which technology development proceeds per engineer-hour invested. (32) Technology gap= (Rate of advancing state of the art-Rate of advancing techology)/Rate of advancing state of the art Units: Dimensionless The amount by which the company's product technology lags the state of the art. (33) Technology slip= DFM*Rate of advancing techology Units: 1/Month The amount by which the value of DFM knowledge is reduced because of increases in the level of product technology. (34) TIME STEP = 1 Units: Month The time step for the simulation. (35) Total Engineering Time= -74- Engineering time developing technology+Engineering Time improving DFM Units: hr Total engineer-hours (36) Worst spent per project case delivery delay penalty= 4e+007 Units: Dollars The penalty per project for delayed product delivery in the worst case, where the design is as difficult to manufacture as imaginable, arbitrary figure assumed. (37) Eng. "Worst case hrs. DFM"= for 4000 Units: hr Engineering hours required to rework design to make it manufacturable in the worst case, where the design requires the most modification imaginable, assumed to be one fulltime engineer dedicated to modifications for two years. (38) Worst case performance penalty= le+007 Units: Dollars Cost of lagging the state of the art through lost revenues and performance price penalties in the worst case imaginable, arbitrary figure assumed. (39) Worst case scrap cost= le+007 Units: Dollars Cost of scrap and rework per project in the case of the worst manufacturability imaginable, arbitrary figure assumed. - 75 - APPENDIX E: MACHINING COST ESTIMATION MODELS The following formulae are included as examples of how manufacturing costs can be estimated in the design phase. While these calculations could be quite cumbersome if used manually, CAD software offers the possibility of automating them in design review mode. These tools can then be used to support design decisions. The basic cost function is the product of the duration of the operation and the cost per unit time of machine use (including labor and overhead), which can be estimated or based upon figures from the manufacturer. LASER DRILLING The following energy balance relates the cutting time t of a line of length I with a laser of power P (Powell 1989): (P - b)t(x/100)=Ectldk + tndk/2(A+B+C) where: b = laser power transmitted through the cut zone without interaction with the cut front. x = the absorptivity of the cut zone expressed as a percentage. Ecut = specific energy needed to melt and remove one unit volume of material from the cut zone. d = material thickness k = kerf width A = Conductive loss function. B = Radiative loss function. C = Convective loss function. These terms are clarified in the article. However, it is likely that information specific to the relevant equipment and material can be obtained from the manufacturer. ELECTRIC DISCHARGE MACHINING For both EDM and ECM, the basic formula for the machining time is t = (V/Rmr) where: V = Volume of material to be removed Rmr = material removal rate While formulae are presented here to help estimate the material removal rate, this value is best obtained from the manufacturer. - 76 - Springbom (1967, 111) uses the following to calculate material removal rate: Rw = 2.43Mw- 2 3 where: Rw Mw = = Average metal removal rate from workpiece (in3/amp-min x 104) Melting point of workpiece ( C) The removal rate Rm- is, therefore, the product of Rw and the current used. ELECTRO-CHEMICAL MACHINING Springborn (1967, 41) gives the following formula for ECM: s = [(N/n) x (1/d) x (1/95,500) x y] where: d = Density (g/cm3) y = Current efficiency I = Current N = Atomic weight of material N = Valence of material s = Specific removal rate (cm3/amp-sec.) Rm, = sI Again, this formula can be used to estimate the material removal rate, but it is best to get it from the manufacturer. - 77 - APPENDIX F: OVERLAPPING OPTIMIZATION MODEL Krishnan (1993) gives the following formulation to optimize the timing of design freezes, iteration start times and the number of iterations, given the evolution and sensitivity of the upstream and downstream activities. Minimize X Subject to tF tAf ti ti-I + to tAs tAs < ti tn X di1i <F = 1,2,... ,n 1 = 1,2,... ,n-1 tF = tn + di Ax(ti,ti) = (bin - ain) (ej-ei)/2 di = D(Ax(thr,ti)) where ti = Start time of ith iteration (decision variable) i = 0,1,2,... ,n d~i = Duration of ith iteration (variable) i = 0,1,2,... ,n do = Duration of planned iteration (input) tAs = tAf = Nominal Finish time of up stream activity (input) tF = Advance freeze time of exchanged information (decision variable) Start time of upstream activity (taken as origin of the time scale) = Product Development Lead Time n = Number of iterations subsequent to the planned iteration (decision variable) (D = The sensitivity funtion - 78 - APPENDIX G: DEFINITIONS Term Definition CE Concurrent Engineering DFM Design for manufacturability, a methodology or set of principles to help designers consider the impact of design decisions on the cost of manufacturing when developing products. ECM Electrochemical Machining, a method of material removal that operates by the same process as electroplating, but in reverse. EDM Electric Discharge Machining, a process of material removal using a dielectric medium and a shaped electrode to generate electric arcs that vaporize material locally GT Gas Turbine NCR Nonconformance Report PTFE Polytetrafluoroethylene, commonly called Teflon®, a registered trademark of DuPont NPV Net Present Value, the current value of a set of future transactions assuming that money loses value with time at an exponential rate, the cost of capital - 79 -