CONTROL OF WIP AND REDUCTION OF LEAD TIME IN A FOOD PACKAGING COMPANY by MASAUET~TS INTITT OF TECHNOLOGY KEVAN YONG CAI CHIM NOV 042010 B.Eng., Mechanical Engineering (2009) Nanyang Technological University, Singapore LI BRARI ES SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING IN PARTIAL FOR THE DEGREE OF REQUIREMENTS THE OF FULFILLMENT ARCHIVES MASTER OF ENGINEERING IN MANUFACTURING AT THE MASSACHUSETTS INSTITUTE OF TECHNOLOGY SEPTEMBER 2010 C20 10 Massachusetts Institute of Technology. All rights reserved. Signature of Author: ( / Department of Mechanical Engineering August 18, 2010 Certified by: Stephen C. Graves Abraham J. Siegel Professor of Management Science Department of Mechanical Enginng and Engineering Systems TJhesi Sunervisor Accepted Dy: David E. Hardt Ralph E. and Eloise F. Cross Professor of Mechanical Engineering Chairman, Committee on Graduate Students CONTROL OF WIP AND REDUCTION OF LEAD TIME IN A FOOD PACKAGING COMPANY by KEVAN YONG CAI CHIM Submitted to the Department of Mechanical Engineering on August 18, 2010 in partial fulfilment of the requirements for Degree of Master of Engineering in Manufacturing Abstract High inventory holding costs, strain on warehouse capacity and the competition of lead time are concerns of a food packaging manufacturer. In this work, causes of high WIP were identified and three approaches were developed to reduce and control better the WIP. We propose to split the production line into dedicated lines, to change the push production system to a pull production system, and to align the process times to the takt time. A simulation model of the proposed production line was analyzed and was verified by a simulation model of the current production line. Three configurations were tested for production line A and production line B. We found that the proposed production line reduced the WIP by 70% and saved $990,000 annually in inventory holding costs. In addition, the lead time was reduced from 6.4 days to 1.9 days and the strain on the warehouse capacity was eliminated. Thesis Supervisor: Stephen C. Graves Title: Abraham J. Siegel Professor of Management Science Table of Contents C hapter 1: Introduction ...................................................................................... ......--- 1 1.1 Company background ......................................................................................... 1 1.2 Com pany products ................................................................................................ 1 1.3 M arkets and customers ......................................................................................... 3 1.4 CA S operations ..................................................................................................... 5 1.4.1 D esign departm ent .................................................................................. 6 1.4.2 Planning departm ent ............................................................................... 6 1.4.2.1 M aterials planning .................................................................................. 6 1.4.2.2 Production planning ................................................................................ 7 Production departm ent ............................................................................. 8 1.4.3.1 Pre-press process....................................................................................... 9 1.4.3.2 Printing process....................................................................................... 10 1.4.3.3 Lam inating process ................................................................................ 11 1.4.3.4 Slitting process....................................................................................... 12 1.4.3.5 D octoring process .................................................................................. 13 1.4.3.6 Palletizing process ................................................................................. 13 1.4.4 Storage and w arehousing ...................................................................... 13 1.4.5 Purchasing departm ent........................................................................... 15 1.4.3 Chapter 2: Problem statement......................................16 2.1 Project m otivation.............................................................................................. ........ ................................... 16 16 2.1.1 H igh inventory holding cost......... 2.1.2 Exceed w arehouse capacity .................................................................... 17 2.1.3 Competition of lead tim e ........................................................................ 18 Causes of high W IP ........................................................................................... 20 2.2.1 Complexity............................................................................................ 20 2.2.2 Local autonom y of production rate........................................................ 20 2.2.1 Em phasis on O EE .................................................................................. 21 2.2.1 Push production system ........................................................................ 21 2.2 2.3 Project objective................................................................................................ 22 C hapter 3: Literature review ....................................................................................... 23 3.1 Little's law ............................................................................................................. 23 3.2 Push and pull production system ...................................................................... 24 3.2.1 M aterial and inform ation flow ............................................................... 24 3.2.2 Control of W IP and throughput ............................................................. 25 3.3 K anban system ....................................................................................................... 25 3.4 Conw ip system ....................................................................................................... 26 3.4.1 N umber of cards.................................................................................... 27 3.4.2 Stress level of w orkers........................................................................... 27 Simulation.......................................................................................................... 28 C hapter 4: M ethodology.............................................................................................. 29 3.5 4.1 Problem solving phase ........................................................................................... 29 4.2 D esign proposed production line ...................................................................... 30 Changes to current production line ........................................................ 30 4.2.1.1 Dedicated production lines ............................... 30 4.2.1.2 Pull production system ........................................................................... 33 4.2.1.3 Takt time ................................................................................................ 34 4.2.2 Selection of dem and.............................................................................. 36 4.2.3 Products allocation................................................................................. 38 4.2.4 Sequence of production... 4.2.5 W ork station stoppage........................................................................... 44 Sim ulation m odeling......................................................................................... 48 4.3.1 Proposed production line ...................................................................... 48 4.3.2 Current production line ......................................................................... 51 4.2.1 4.3 4.4 .......................... .......... 42 Im provem ent analysis ....................................................................................... 53 4.4.1 Conversion betw een rolls and reels ........................................................ 53 4.4.2 Inventory holding costs......................................................................... 54 4.4.3 Lead tim e ................................................................................................ 54 C hapter 5: R esults and discussion............................................................................. 55 5.1 Current production line ....................................................................................... 55 5.2 Line A perform ance ............................................................................................... 56 5.2.1 Throughput and card num ber.................................................................. 56 5.2.2 W IP inventory ....................................................................................... 57 5.3 Line B perform ance ..................................................................... .................. 59 5.3.1 Throughput and card num ber................................................................ 59 5.3.2 W IP inventory....................................................................................... 60 5.4 Inventory level and holding costs ...................................................................... 63 5.5 W arehouse capacity ........................................................................................... 65 5.6 Lead tim e ............................................................................................................... 67 Chapter 6: R ecom mendations and conclusion.......................................................... 68 C hapter 7: Future opportunities ............................................................................... 70 R eferences.........................................................................................................................71 Table of Figures Figure 1.1: Major Products of CAS....................................................... I Figure 1.2: The different layers of a package.......................................... 2 Figure 1.3: Markets served by Company A............................................. 3 Figure 1.4: Order Flow Diagram.......................................................... 5 Figure 1.5: Block planning system........................................................ 8 Figure 1.6: An overview of the manufacturing processes............................. 9 Figure 1.7: Cliche used for printing....................................................... 9 Figure 1.8: M ounted sleeves............................................................... 10 Figure 1.9: The Printing Process.......................................................... 10 Figure 1.10: The Laminating Process..................................................... I1 Figure 1.11: Process of slitting station................................................... 12 Figure 1.12: Capacity of warehouse...................................................... 14 Figure 2.1: Inventory holding costs for 2009............................................ 16 Figure 2.2: Warehouse capacity for printed, laminated and doctor WIP............ 17 Figure 2.3: Lead time....................................................................... 18 Figure 2.4: Comparison of lead time with other factories............................. 19 Figure 2.5: Mixed flow production........................................................ 20 Figure 3.1: Queuing system............................................................... 23 Figure 3.2: Material and information flow in a push system and pull system...... 25 Figure 3.3: Kanban system................................................................. 26 Figure 3.4: CONWIP system............................................................... 26 Figure 3.5: Steps in a simulation study.................................................. 28 Figure 4.1: Phases of problem solving.................................................... 29 Figure 4.2: Dedicated production lines.................................................. 32 Figure 4.3: Pull production system with card return................................... 34 Figure 4.4: Simulation model of proposed production line........................... 48 Figure 4.5: Simulation model of current production line.............................. 51 Figure 4.6: Slitting of a roll into reels..................................................... 53 Figure 5.1: Throughput trend of line A.................................................. 56 Figure 5.2: Throughput trend of line B.................................................. 59 Figure 5.3: Printed WIP with L22 process time = 14minutes ............... Figure 5.4: Printed WIP with L22 process time = 12minutes ............... Figure 5.5: Inventory cost of proposed production line..... Figure 5.6: Capacity of printed and laminated WIP......... .......... .......... Figure 5.7: Capacity of doctor WIP..................................... Figure 5.8: Lead time of proposed production line.......... ........... List of Tables Table 1.1: Shipping schedules............................................................. 4 Table 4.1: Work station capabilities...................................................... 30 Table 4.2: Takt time of work stations .................................................... 36 Table 4.3: Demand of first three months in 2010....................................... 36 Table 4.4: Demand breakdown of March 2010......................................... 37 Table 4.5: Maximum rate of work stations............................................. 38 Table 4.6: Products allocation based on initial assignment .......................... 39 Table 4.7: Products allocation after demand adjustments ............................ 41 Table 4.8: Setup costs of laminators ................................... 42 Table 4.9: Sequence of production ...................................................... 43 Table 4.10: Breakdown of work stations ............................................... 44 Table 4.11: Rest time of work stations ................................................ 45 Table 4.12: Short stop of work stations ................................................ 45 Table 4.13: Setup time of work stations................................................ 46 Table 4.14: Reel inspection time of slitters .......................................... 47 Table 4.15: Process time of work stations in proposed model ..................... 50 Table 4.16: Percentage of rolls processed by each work station .................. 52 ....... 52 Table 4.18: Inventory holding costs of three types of WIP ......................... 54 ........... 55 Table 5.2: Error in the current production line model .............................. 55 Table 5.3: Throughput mean and standard deviation in line A ..................... 57 Table 5.4: Average printed WIP in line A ............................................. 58 Table 5.5: Average laminated WIP in line A .......................................... 58 Table 5.6: Average doctor WIP in line A .............................................. 58 Table 5.7: Throughput mean and standard deviation in line B ..................... 60 Table 5.8: Average printed WIP in line B ............................................. 60 Table 5.9: Average laminated WIP in line B ........................................ 60 Table 5.10: Average doctor WIP in line B ............................................ 61 Table 5.11: Inventory reduction .......................................................... 63 Table 4.17: Process time used in current production line model ........ Table 5.1: Performance of the current production line model ....... Table 5.12: Inventory holding cost savings ............................................. 64 CHAPTER 1: INTRODUCTION 1.1 Company Background Company A, Singapore (CAS) is a multinational food processing and packaging company of Swedish origin. Founded in 1951, it is one of the largest manufacturers in the food processing and packaging industry. Company A provides integrated processing, packaging and distribution lines as well as plant solutions for liquid food manufacturing. Today, the business spans more than 150 countries with 43 packaging material production plants worldwide. CAS was established in 1982. CAS focused on manufacturing finished packaging material for customers in 19 countries. CAS and Company A, Pune, in India are the production plants in the South and Southeast Asia Cluster. In year 2007, CAS received the Manufacturing Excellence Award (MAXA) for overall excellence in innovations, operations and sustainability as well as its World Class Manufacturing (WCM) approach to ensure operational improvement and downtime minimization. Due to the increase in complexity of Company A's supply material plants worldwide and increase in both complexity and number of products, Company A has been focusing on improving their supply chain and production efficiency. 1.2 Company Products Company A is one of the world's major packaging providers. It offers a wide range of packaging products, filling machines, processing equipment, distribution equipment and services. Figure 1.1 shows the major products of Company A. Figure 1.1: Major Products of CAS. ........... .. CAS produces carton packs, also known as CA packs. CA packs are used for food items like milk, juice and soy products. Designing service is also available for customers. Each CA pack is made of 6 layers of materials, including aluminum, paper and polyethylene, to prevent spoilage of the content. The base material for each package is paper. It provides structure and support to each package. After the design is printed onto the paper, it will be coated with a layer of aluminum foil, which makes the pack aseptic and preserve the flavor of the content. Four layers of polyethylene will also be coated onto the paper. The outside layer prevents damage from moisture; the adhesive layer between the paper and aluminum foil provide structure support and two protective innermost layers seals the liquid content. The layers of the package and their respective functions are shown in Figure 1.2. 1.PE - protects against outside moisture & enhance appearance 2 Paper - for stability and strength 5 & 2 3.PE - adhesion layer 4. Aluminium foil - oxygen, flavour and light barrier 5.PE - adhesion layer 6. PE - seals the liquid Figure 1.2: The different layers of a package. There are 10 classes of package available for CA packs. They are Brik, Brik Aseptic, Prisma Aseptic, Gemina Aseptic, Fino Aseptic, Classic Aseptic, Wedge Aseptic, Rex, Top and Recart. Different sizes are offered for each class of package, while the polyethylene layers are different for different carton content. Currently, different products are distinguished by the system code, size code, quality code and design code. The system code defines the class of the package, which describes whether the carton is aseptic, refrigerated or ambient. Different classes require different creasing in the printing stage. System codes also have a suffix indicating the content of ..... ...... the carton (juice or milk), which would affect the laminating stage as different contents require different polyethylene formulations. The size codes indicate the volume of liquid contained by the package and its shape (slim, base, square). Thus it describes two attributes and affects the printing, slitting process and sorting on the laminator. Products with the same size code would have same overall width and therefore, same number of webs. Quality codes determine the type, thickness in grams per square meter and the brand of paper used. Lastly, design codes describe a single attribute that is the design of the product. Compared with other factories of company A, CAS offers a large range of different CA pack products. Currently, around 20 different products of different package classes, carton sizes and polyethylene formulations are able to be produced in CAS. 1.3 Markets and Customers Positioned in Singapore, CAS efficiently serves customers in the South and South East Asia cluster. There are also customers from Europe and the Middle East. In total, CAS ships its finished products to customers in 19 different countries, as shown in Figure 1.3. Figure 1.3: Markets served by Company A. ............... The customers of CAS currently need to place their orders through a market company. The market company has a sales office in the customers' respective country. They take orders from beverage producing companies then receive and distribute the finished products to the customers. The Thailand market is the largest by volume, followed by Malaysia, Indonesia and Vietnam. Most of the products are ocean freight to the customers. At the moment, only the Malaysia market is being served by truck freight. For most shipping routes, the containers are only picked up from Singapore ports twice a week and the shipping dates are fixed. For example, for the case of the Thailand market, finish goods are shipped out on every Tuesday and Saturday. For the Malaysia market, the freight truck delivers orders from CAS everyday. The details are documented in Table 1.1 below. Table 1.1: Shipping schedules. usrana hina ng Kong ndonesia nu, bun on, Thu laysia epal akistan us, Sat Hiippines hu, Sun audi Arabia pan Korean Zealand tanka Mon, Fri Daily on aiwan land let Nam ue, Thu Mon ue, Sat ue ri Mon, Wed ue, Sat Mon, Wed, Thu 1.4 CAS operations The core corporate functions of CAS are the design, production, planning and purchasing departments. There is also a market company operating in CAS' premises and this is an independent entity from CAS. The market company is responsible for order management and customer service. CAS' warehouse and delivery operations are outsourced to a third party logistic company. Figure 1.4 shows the steps at which an order is processed. *Goods are being delivered out of the warehouse everyday Figure 1.4: Order Flow Diagram. 1.4.1 Design Department After the customers' designs have been submitted to CAS's design department, the design department reviews and adapts these designs to suit CAS's production systems. When faced with difficulty, the customers do receive assistance from the design department in designing the carton. Once the design is confirmed, the design is broken down according to the component colors. The process colors are Cyan (C), magenta (M), yellow (Y) and black or key (K). Special or spot colors may also be used to obtain specific shades of color. The number of spot colors can vary from none to seven. A sales order can only be made once the designed is confirmed. 1.4.2 Planning Department The planning department at CAS is responsible for materials planning and production planning. 1.4.2.1 Materials Planning Materials Requirement Planning does the ordering of the raw materials needed for production. The base materials ordered are paper, polyethylene and aluminium foil with many types of variants in terms of grade and size. The purchasing department is responsible for acquiring the additional materials such as water based inks, pallets and tapes that are used for production as they are relatively low volume and low cost. Company A International (CAI) is the parent body of CAS. CAI issues the annual global forecasts for number of packs and marketing directives. CAI's Global Supply places blanket orders on the basis of the annual forecast for each of the converting factories with the suppliers in order to obtain economies of scale and to pool the variation in demand. The converting factories then place the actual orders with the suppliers to withdraw from the blanket order placed initially. In addition, monthly forecasts are also issued and updated regularly. As the lead time of raw materials is very long, the ordering is done well in advance. The ordering is done on a weekly basis as this time period coincides with the frequency of dispatch. A continuous 6 review method is used to determine the ordering quantities. The re-order point is set at approximately 40% of the monthly demand while the order up-to point is around 60% of the monthly demand. 1.4.2.2 Production Planning The production system of CAS is make-to-order. The production schedule is drafted only upon receipt of a production order from the sales department. The scheduling is done on the SAP based P2 system and the current production lead time is around 12 days. Planning is based on the delivery due date. CAS uses three core work stations for their processes. They are the printer, laminator, and slitter. On each of the three work stations, the orders are grouped together based on certain criterions to minimize setups. The grouping for the printer is done on the basis of size and shape. The criterion for the laminator is the overall width of the roll. Lastly, the slitter orders are arranged based on pack width. A block scheduling system is used to plan the production schedule. In this collaborative planning, the planning department generates a weekly production schedule with blocks according to width of the paper rolls. This is to reduce the number of setups at the laminator. Customer orders are then fitted into the blocks. The latest order date for the customers is 4 days before the production cycle starts. The production cycle starts on every Monday. Thus, the customers must place their orders by Thursday of the previous week. The estimated delivery date is 3 days after the end of the production cycle. Therefore, the products would be ready on the Wednesday after the production cycle. The customers would be able to place orders many weeks earlier. However, when the orders are placed too early, the orders would be kept in the system and be produced in the subsequent production cycles. Figure 1.5 shows the block planning. However, some customers tend to place last minute orders, which would create disruptions to the planned production schedule. These last minute orders are rush orders which lower equipment efficiency. These last-minute orders are urgent orders that are placed within 1 to 3 days before the start of the production cycle in which it needs to be produced in. Also, when the current production schedule has been completed ahead of 7 time, the planning department would also bring forward some orders to fill up the empty block in the block schedule. By doing this, the equipment efficiency is improved but advanced production will also result in higher work in process (WIP) and finished goods inventory. 4 days 7 days M Latest order date T W TH F 3 days S Production cycle S Estimated delivery date Figure 1.5: Block planning system. 1.4.3 Production Department The Production Department performs the major manufacturing processes to produce the packing materials. The 3 major processes in CAS are printing, laminating and slitting. Before printing, a pre-press process has to be carried out, and after slitting, a doctoring process sometimes needs to be done. An overview of all the production processes is shown in figure 1.6. ... . PRODUCTION PROCESS WAM BASMDI SR~MDPAM PBOD L~IHNENG PALLETZEG PA Figure 1.6: An overview of the manufacturing processes. 1.4.3.1 Pre-Press Process This is the first stage in the production process. In the pre-press stage, the clich6s for printing are prepared from the negatives. The cliches are polymeric stamps with elevated portions for the areas to be printed. These clich6s are prepared on photopolymer plates through a process of controlled exposure to UV light. There will be a cliche prepared for each color used for printing. After which, the cliches are mounted onto a sleeve with a rotating spindle. The number of cliches mounted on one sleeve depends on the width of the individual pack and the paper roll. This corresponds to the number of webs. A cliche used for printing is shown in figure 1.7. Figure 1.8 shows the mounted sleeves. Figure 1.7: Cliche used for printing. ... Figure 1.8: Mounted sleeves. 1.4.3.2 Printing Process In the printing stage, the flexography method is used. This is a method of direct rotary printing that uses resilient relief image plates of photopolymer material. The design pattern on the cliches is reproduced onto the paper board by rotary contact of the paper roll with the stamp. Water based ink is used. The incoming paper roll is loaded on the unwinder, which opens it up and feeds it to the printing stations. An illustration is shown in Figure 1.9. Ink transferred Impression Cylinder Ink transferred from anilox to plate Doctor Chamber Figure 1.9: The Printing Process. ...... .... - ................. .. 0000000000VWOV There are seven stations on the printer. Each station holds the sleeves and the water based inks for one of the colors of the design. Depending on the design colors, some of the stations may be idle for a design as not all colors are used for every design. A colored image is formed by 4 process colors, Cyan (C), magenta (M), yellow (Y) and black or key (K). The different colors are then superimposed one over the other to get the complete final printed design. Fold creases for the produced cartons are also form during the printing process. The purpose of creasing is to enable proper folding of the pack during the filling stage at the customer site. The tool is used to form creases also punched the holes for straws. For routine printing, flexographic technology is used. For higher resolution designs, CAS uses offset printing, which is more expensive compared to flexography. 1.4.3.3 Laminating Process Laminating involves the coating of aluminum foil and polyethylene (PE) layers onto the printed paper. A roll is first unwound at the unwinder. It then goes through three stations for the coating process. The last step is to rewind the laminated paper into a roll. In the first station, a layer of aluminium foil is layered onto the printed paper. After which, PE film is coated in the inner surface of the packaging material to prevent contamination and leakage. The final station adds another layer of PE on the outer surface of the packaging material to protect the paper. This process is shown in figure 1.10. Polymer Laminator I Laminate layer Lwminhtor2 inside layer Laminator3 D6cor layer Figure 1.10: The Laminating Process. 1.4.3.4 Slitting Process The paper roll can have 4, 5, 6, 7, 8 or 9 webs (columns) depending on the product size. The slitting process cuts the entire roll into reels of a single pack width so that they can be fed into the filling machines at customer production plant. The rolls are unwound, slit using a row of knives and counter-knives and then rewound to form reels. The reels are then grouped into defective reels and non-defective reels. Defective reels are reels that consist of at least one defect and these reels need to go through the doctoring process to have the defects removed. Non-defective reels are reels which have no recorded defects throughout all the processes and they do not need to go through the doctoring process. A schematic diagram of slitting process is shown in figure 1.11. Figure 1.11: Process of slitting station. 1.4.3.5 Doctoring Process After the slitting process, the defective and non-defective reels are separated. The nondefective reels would be kept at the shop floor and the defective reels would be doctored. Doctoring or rework is the process of removing the packs with defects from the reels. Approximately 24% of the reels require doctoring. These defects are due to any of the upstream processes and they are removed collectively at this stage. There are 14 doctoring stations after the slitting process. Each doctoring work station would have an operator to find the defects and remove them using a machine. One reel could be doctored every 24 minutes on average by an operator. 1.4.3.6 Palletizing Process The doctored reels would join the good reels to be palletized. Palletizing is the process of stacking reels together on a pallet and wrapped with a plastic layer. There would be 5 reels on the pallet on average. The palletizing time is 8 minutes. The palletized reels would be transported to the warehouse and await delivery. These palletized reels are handled solely by a third party logistics company from this point onwards. 1.4.4 Storage and Warehousing CAS' in-house warehouse is shared among raw materials; work in process (WIP) and part of total finish good inventory (FGI). Currently, the in-house warehouse is managed by a third-party logistics company. The current daily FGI level is around 1000 rolls (converted from pallets), among which up to 600 rolls are stored in the internal warehouse. Each roll is approximately 5513m long and it takes up about 3 pallets. The current daily FGI level of about 1000 rolls is approximately equal to about 6 to 7 days of inventory as CAS produces approximately 130 to 150 rolls per day on average. The external warehouse is engaged when there is not enough space. The floor layout and capacity for each category of inventories are illustrated in figure 1.12. As we can see from the figure, the full capacity for WIP is approximately 500 rolls only. Yet currently, average WIP levels can reach over 1000 13 zzv v Zzzzzw ::':_ - - - - - - - . "- ----------- rolls, not including raw material rolls. The raw material rolls are stored in a huge container yard beside the warehouse and it can hold up to 800 rolls of raw material. The movement of raw material, WIP and FGI between the production floor and in-house warehouse is facilitated by the laser guided vehicles (LGVs). These vehicles can move a roll at a time and they are programmed to follow a specific route. Forklifts and clamp trucks are used for the movements of the rolls within the internal warehouse. Capacity 1800 FG Storage Area (Pallets) Paper Storage Area (Rolls) 700 100 Alum Foil Storage Area (Crates/Boxes) WIP Storage Area (Rolls) 500 70 Additional Storage Area (Pallets) Doctor Reels Storage Area (Pallets) 200 Malaysian Picking and Storage Area (Pallets) Total Warehouse Storage Space Figure 1.12: Capacity of warehouse. 100 1.4.5 Purchasing Department The purchasing department at CAS is responsible for the purchase of additional materials and indirect services. Examples of additional materials include inks, pallets, cores, straws etc. Indirect services mainly refer to equipment maintenance, electricity, and water utilities. The base materials comprise 60% of the total monetary value spent by the purchasing department, while additional materials and indirect services make up the remaining 40%. There are more than 10 suppliers for the additional materials and more than 500 providers for indirect services. The purchasing reviews all the suppliers regularly and will provide assistance when the suppliers are underperforming. The purchasing department has a well-established system to source for alternative suppliers. Hence, suppliers who consistently underperform will be substituted. .. .................... :. ............. ........ . .............. ............ ......... ..... .... - CHAPTER 2: PROBLEM STATEMENT 2.1 Project motivation 2.1.1 High inventory holding cost CAS has incurred high inventory holding cost. The inventory holding costs of each month in 2009 is shown in figure 2.1. The printed WIP holding costs is $315,305 for the year, while the laminated WIP holding costs is $570,372 for the year. The total WIP inventory holding costs is $885,677 for the year of 2009. CAS has a production of over 1.2 billion packs per month since March 2010. CAS' marketing company has projected that sales will grow at 12 percent annually. Therefore, CAS needs to produce over 1.35 billion packs by March 2011. As such, the inventory holding costs would be even higher in 2011. CAS would be able to reduce the capital tied up by the inventories by reducing the WIP. 120,000 N Laminated 100,000 * Printed 80,000 - 60,000 40,000 20,000 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2009 Figure 2.1: Inventory holding costs for 2009. 2.1.2 Exceed warehouse capacity CAS stores its WIP in the internal warehouse. Capacity has been allocated to different types of inventory. The capacity allocated for printed and laminated WIP is 500 rolls while the capacity allocated for the doctor WIP is 500 reels. Doctor WIP is the term CAS uses to refer to the WIP awaiting to be doctored. Figure 2.2 shows that the WIP has exceeded the capacity allocated for them. This has two impacts. Firstly, it forced CAS to store WIP at forklift lanes. As such, the forklift operators have a harder time to retrieve the rolls and it might cause safety issues. Secondly, CAS has to relocate some of the finished goods inventory to the external warehouse. The external warehouse incurs additional expenses for CAS. Printed and Laminated WP 1000 900 Doctor WIP 1800 -__ __ 1600 - 1400 1200 1000 7'A 600 -- , -0 600 50 400 400 200 300 0_FbJn Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2009 Jan Feb Mar Apr May Jun Jl c Jul Aug Sep Oc 2009 Figure 2.2: Warehouse capacity for printed, laminated and doctor WIP. o e Nov Dec -iiW ft im ! E M" m ........... 2.1.3 . ..... I... ................... . .. ....... ... Competition of lead time The lead time quoted to customers is 12 days from time of order placement until the order is shipped by CAS. (The lead time from point of shipment until the receipt by customer is outside of the control of CAS. The shipment is arranged between the customer and the logistics company.) The production lead time takes up about 10 of these 12 days. The remaining 2 days is due to the handling of the finished goods by the 3 rd party logistics company. This long lead time is mainly due to the long waiting time between processes. An illustration of the current lead time is shown below in figure 2.3. 1 day <0.5 day 4 days <0.5 day 3 days <1.5 days Figure 2.3: Lead time. There is both internal and external competition on the lead time. The internal competition comes from the comparison of CAS and other factories within the same company. Figure 2.4 shows the lead time quoted to customers by other factories. The factories in India and China are of concern to CAS as they are the neighbouring clusters. The lead time at India is only one day behind that of CAS. The lead time of 7 days in China is 60% of that in CAS. Thus, there is a strong internal competition. The external competition comes from company C. Company C is able to quote a lead time that is comparable to CAS. CAS has a strong desire to shorten its lead time to stay ahead in the competition. Others Others 17 Others 16 Others 15 Others 14 India 13 CAS 12 Others 12 Company C 9 Others 9 Others China 7 China 7 China 7 Others 7 0 2 4 6 Smaller number = shorter lead time 8 10 8 10 I I 12 14 16 Figure 2.4: Comparison of lead time with other factories. 18 20 2.2 Causes of high WIP 2.2.1 Complexity CAS is using a mixed flow production such that a product can flow to almost any station downstream (subjected to work station capability). This mixed flow production is shown below in figure 2.5. There is no clear route for the production. It is too complex for CAS to set an appropriate route at which the products are produced at each stage. The mix of orders for each week is different, and CAS has a fast pace environment. As such, it is difficult for the planners to determine a good routing plan. This difficulty is coupled with rush orders (i.e. orders made by customer after the weekly production cycle has started). The planned production route would be disrupted by these rush orders. S52 P13 D1 to D14 L21 S54 Pal Queue S53 P18 FGI L22 Pa2 S55 Figure 2.5: Mixed flow production. 2.2.2 Local autonomy of production rate There is no fixed rule to control the production rate of the individual work stations. Each process manager would be given a production schedule and sequence. However, they have the autonomy to choose the work station speed at each stage. As such, every stage is producing at different rates. The general rule is that the individual process must not starve the downstream process. Thus, the process managers will produce a suitable level of safety stock based on their judgment to prevent the starvation of the downstream processes. As there is no 'pacer' in the production line, the WIP tends to build up. When sufficient WIP has built up, the work stations are rested. 2.2.3 Emphasis on OEE The overall equipment effectiveness (OEE) is an important key performance index (KPI) for the production employees as their bonuses are tied to this KPI. As such, every process has an incentive to process the rolls that will bring the OEE to the target OEE. This builds up unnecessary WIP. 2.2.4 Push production system CAS is currently using a pure push production system. The system ignores the real production shop-floor situation. Thus, the upstream work station continues to produce even when the downstream work station is having a breakdown or a long setup change. This causes the WIP to build up. The laminator is currently the bottleneck of the entire system and it is estimated to have a capacity of 100 rolls per day. CAS would estimate the capacities of all work stations to plan the production for the push system. However, it is difficult to estimate the capacity as the actual capacity of the work stations would be altered by variations. Thus, if the actual capacity of the laminator drops and the actual capacity of the printer increase, there would be more WIP between these 2 stages. 2.3 Project Objective The author aims to improve the production system in CAS by the following means: 1) Control the WIP to reduce capital tied up by inventories and to eliminate the problem of warehouse capacity. 2) Reduce the lead time to keep CAS competitive in the industry. 3) Reduce the complexity of the production system to have better visibility and control and improve cooperation between processes. CHAPTER 3: LITERATURE REVIEW 3.1 Little's law Lead time and inventory level are two major concerns in a production system. A discreteparts production system can be viewed as a queuing system where the items are queuing to be processed by a number of steps. Little's law offers an understanding on the queuing system. A simple queuing system is shown in figure 3.1. The work items arrive at the system where there is one server. The rest of the work items that are not served immediately have to wait in a queue for their turn to be processed. When a work item is processed by the server, it departs from the system. The Little's law relates the average number of items in a queuing system, the average arrival rate for the work items and the average queue time of work items by Eq. (3.1) [1] Eq. (3.1) L = XW Where L=average number of work items in queuing system A=average arrival rate of work items W=average queue time of work items Departure Arrival Queue Server ) Figure 3.1: Queuing system [1]. Little's law is general and could be applied in different systems. For example, Little's law could be applied in a production system where it relates the work items' flow time, throughput and the level of WIP by Eq. (3.2) [1] WIP = TH x FT Eq. (3.2) Where WIP=average number of work items between start and end of production system TH=throughput rate of work items (which is assumed to match the arrival rate) FT=average flow time of work items Little's law applied at the production system is similar to its original form. However, there is a distinct difference. The original Little's law is using the arrival rate of work items. On the other hand, Spearman et. al. used the departure rate (throughput) of the system [2]. It is more meaningful to use throughput as it is often a measure of the system performance. 3.2 Push and pull production system In a push system, a production release date is calculated by taking into account the time planned for production, shipping and other operations. Once these orders are released into the production system, they are pushed to the end of the system. As such, an upstream process would produce work items without considering the situation at the downstream process. This might build up the WIP unknowingly. Thus, in a push system, the throughput is controlled and the WIP (and flow time) is used as a performance measure [3]. In a pull system, the work item cannot move downstream without authorization. The work item would be pulled downstream when the downstream process gives a signal to the upstream. Thus, in a pull system, the WIP is controlled and the throughput is a performance measure [3]. 3.2.1 Material and information flow In the push system, the work items flow is in the same direction as the information flow [4]. The information in this case could be a master production schedule where the work items are produced according to predefined due dates. In the pull production system, the work items flow is in the opposite direction as the information flow. The information flow in this case would be a signal from the downstream process that the work item is ready to be produced downstream. Figure 3.2 show the information flow in the push and pull system. ............. ... .. ......... - Work items flow Push System M 7) ( Information flow Work items flow Pull System Information flow Figure 3.2: Material and information flow in a push system and pull system [4]. 3.2.2 Control of WIP and throughput Spearman et al. discussed that a pull system is relatively easier to control as compared to a push system [5]. WIP would be easier to control than throughput in practice as WIP could be observed in a straightforward manner. In addition, capacity estimation of the plant is required when throughput is the parameter to be controlled as in the push system. However, the capacity of the plant is difficult to estimate as it is subjected to many variations. These variations includes worker efficiency, machine breakdown and setup time. The orders could exceed the perceived capacity and cause the WIP to build up. The building up of WIP would be aggravated when seeking high utilization is of importance. 3.3 Kanban system Kanban system is a type of pull production system where there is a set of kanban cards circulating between each pair of buffer and the upstream machine [6]. These kanban cards are localized and not transferred along the entire line. When a work item has become a finished good, the kanban card is returned to the immediate upstream machine to signal that it has the authority to process one more work item. This relationship is the same throughout the line. Figure 3.3 illustrates a kanban system. Thus, the demand propagates up the production line. The machines will stop processing when the buffer is full. On the other hand, they will continue to process when the buffer is not full. ............................ Although kanban helps to reduce the WIP, it cannot be used under some environments. Kanban would not be useful when [5]: 1) The setup time is long 2) The scrap loss is high 3) The demand has large and unpredictable demand M - M B -* B M -*G Figure 3.3: Kanban system [6]. 3.4 CONWIP system CONWIP system is different from kanban system. There is only one set of cards that circulates around the entire line [6]. When a work item has become a finished good, the card is returned to the first machine to signal that it has the authority to process one more work item. The new work item would then be processed down the line like a push system. As the card is always transferred to the first machine, the WIP in the system should always be a constant number, assuming that the first station is never starved. A CONWIP system is shown in figure 3.4. M FrB 34 M W B Figure 3.4: CONWIP system [6]. Msytm 3.4.1 Number of cards CONWIP is easier to control and less complex than kanban as there is only one set of cards as compared to having one set of cards between each pair of adjacent processes in the kanban system [3]. There are more set of cards to be determined on the line in the case of the kanban system. Since the number of cards in the system might need continuous adjustments over time, a kanban system is harder to control and maintain. However, the number of cards used in a CONWIP system can be more than that in a kanban system. Thus, the WIP level is likely to be higher in a CONWIP system. This implies that the cycle time is higher for the case in CONWIP system. 3.4.2 Stress level of workers Spearman et. al. has also pointed out that a CONWIP system is less stressful to the production workers as compared to the kanban system [3]. There would be more pacing stress in the kanban system as the upstream workers need to replenish the void in the system to prevent the starvation of downstream machine. On the other hand, the workers in the CONWIP system experience less pacing stress as CONWIP acts as a push system at any machines other than the first one. 3.5 Simulation It is possible to obtain analytic solution when the problem of interest is simple. However, many real life problems are complex. Simulation is useful in problems that are too complex to be evaluated analytically. It is more cost effective to carry out simulations to determine the suitability of the proposed plan as compared to experimentation with the real system. The simulation allows the user to have an insight on the performance of the plan. Simulation models could be static or dynamic. A dynamic model would be more applicable here as the production line evolves over time. Discrete event simulation is a model where it is dynamic, discrete and stochastic. Computer package that provides discrete event simulation includes Arena, ProModel and Simul8. Averill et. al. has proposed the following steps in a simulation modeling that would help in modeling a system [7]. These steps are shown in figure 3.5. Figure 3.5: Steps in a simulation study [7]. CHAPTER 4: METHODOLOGY 4.1 Problem solving phase The author proposed 6 problem solving phases to meet the 3 objectives stated in chapter 2 of the thesis. The phases are shown in figure 4.1. Phase 1 introduces the production line proposed by the author. Different aspects of the proposed production line are discussed in this phase. Phase 2 involves the building of a simulation model of the proposed production line. In phase 3, a model of the current production line is created to act as a form of verification. In phase 4, three alternative configurations are tested for each line to find a suitable system for CAS. An improvement analysis is done in phase 5 to measure the performance of the proposed system. In phase 6, the author proposes the production line and makes recommendations to CAS. Pha,-se I Des(in prIopiosed pr-oductOio P ase 2 Veif1ication1 using( Model Of current piroduction1 line Pl ase 4 Collnpare- alterna~tive system1 conlfigur-ations Pha'.se 5 lIiimos etet anial ,s s Phai~se 6 P -opose sOtolto Figure 4.1: Phases of problem solving. ................. ONERM 4.2. Design proposed production line 4.2.1 Changes to current production line -, - The author proposed to make three major changes to the current production line to meet the objectives outlined in chapter 2. These changes are: e Use of dedicated production lines * Change to pull production system * Align production to takt time 4.2.1.1 Dedicated production lines The current production system is a mixed flow line where the products can flow to almost any downstream work stations. However, some work stations are unable to process all of the products. The products that could be processed by each work station are shown in table 4.1. A shaded box means that that product could be processed by that work station. For example, L22 could process both juice and milk products while L21 could only process juice products. The planner in CAS has to plan the route of the products during each production cycle to ensure that there are no clashes between the product routings and the work station capabilities. Table 4.1: Work station capabilities. Products Juice 567 Juice 350 P13 P18 L21 L22 S52 S53 S54 S55 ............................. In order to reduce the complexity of the current production line, the author proposes to split the entire line into two dedicated production lines A and B as shown in figure 4.2. The allocation of products between the two lines is explained in greater detail in section 4.2.3. There are several motivations to use dedicated production lines over the mixed flow production lines: e The lines are simpler to manage by CAS. * There is more visibility of the products flow and decisions are made faster. e Any rush orders inserted into line A does not affect line B and vice versa. e Planning lead time during each production cycle is reduced because planning should be simpler. Figure 4.2: Dedicated production lines. One slitter (S54) is assigned to line A while three slitters are assigned to line B. This is due to two reasons. Firstly, the demand assigned to line A is less than that of line B. Secondly, only 5 and 9 web products are assigned to line A while 5, 7, 8 and 9 web products are assigned to line B. The setup time for changing of webs at the slitters is 60 minutes. Since there is more type of webs on line B, frequent setups are expected and the capacity of the slitter is taken up by these setups. This situation would be aggravated when rush orders of different webs are inserted after the production cycle has started. The number of setups is reduced by dedicating slitters to slit only products of 1 or 2 web types. S54 and S55 are assigned to slit 5 and 9 webs products. S53 is dedicated to slit 7 webs products and S52 is dedicated to slit 8 webs products. The advantage of the assignment on S54 and S55 is that when either S54 or S55 experience unexpected and prolonged downtime, products could be switched from one slitter to the other without disrupting other products on either line A or line B. There are 14 doctoring work stations in CAS. 4 Doctoring work stations are assigned to line A while 7 doctoring work stations are assigned to line B. This is because line B has a higher demand and this is explained in the later sections. Thus, the defects on line B are higher and more doctoring work stations are required. The remaining 3 doctoring work stations act as backup for either of the line. 4.2.1.2 Pull production system The current production system is a pure push production system. Orders are released into the production line through P13 or P18 when the printers have finished processing the previous roll. The current production system does not consider the status of the downstream work stations. Orders are still released to the printers into the production line even when the downstream work stations are stopped for a prolonged period of time. A CONWIP pull production system is able to react to the actual situation in the production line and allow CAS to have a better control over the WIP. There are several advantages of CONWIP production system over the current push system. * Orders are not released into the system when the WIP has reached a predefined level using CONWIP cards. * The total WIP in the production system stays relatively constant. * The lead time to customers could be better estimated when the WIP stays at a near constant level. Figure 4.3 shows a schematic of the pull production system proposed by the author. There are separate queues in each of the dedicated lines just before the printers. An order is released into the production line when there are available cards at the entry point. As CAS has a large production floor area, a physical card is not appropriate to be attached to each of the roll. CAS has initiated a concurrent project where radio frequency identification is used to track the rolls in the production line. When a roll leaves the end of the line (palletizer), a signal is sent back to the queuing area and authorizes that a new roll can be released into the production line. The products in line A go through the following path: P13 -+ L21 -> S54 -> D1 to D7 (only defective products) -> Pal and the products in line B go through the following path: P18 -+ L22 -* S53 or S52 or S55 -* D8 to D14 (only defective products) -> Pa2 .................... ...... .............. ................. ... ....... ... - ..... 11-,..... Conwip card 53 7 B web D5 to D11 P18 L22 S52we 8 & 9web Conwip card Figure 4.3: Pull production system with card return. 4.2.1.3 Takt time The current production line utilizes all the work stations at their maximum rate to increase their OEE. When a sufficient WIP buffer has been built up, the work stations are made to rest. The author proposes that the work stations rate should be aligned to the takt time of the demand during each production cycle. This implies that the work stations do not process the rolls at their maximum rate. When the demand is high, all work stations operate at a high rate and vice versa. The takt time for a given week for line A and line B is found using Eq. (4.1) Takt time (mins/roll) (mins)-Time loss (mins) = Total time in production per week(roll) Demand cycle Eq. (4.1) Net avaliable time (mins) Demand per week(roll) The total time in each production cycle is the same. Each production cycle comprises 7 days or 10,080 minutes. The time loss comprises planned maintenance, setups and other types of stoppages that can be anticipated or expected. The time loss is estimated from past data in CAS internal P2 system. The net available time for each work station is different as their time loss is different. The demand in the denominator of Eq. (4.1) is the demand in a week (one production cycle) expressed in rolls. Rush orders are not considered in the models due to two reasons. Firstly, the percentage of rush orders is small. Secondly, CAS is currently working on a customer order management project to minimize the occurrence of rush orders. The doctoring work stations and palletizers do not have any planned maintenance, setups and other types of stoppages. Thus, their net available time is the total time in each production cycle. The number of doctoring work stations required on each line is different. 24% of the rolls on each line contain defects and require doctoring. The number of doctoring work stations is determined by Eq. (4.2) Number of doctors = Process time (mins) + Number of doctors = Process time (mins) + Net avaliable time (mins) Demand require doctoring per week(rolls) Eq. (4.2) Net avaliable time (mins) Demand per week(rolls)x Defects (%) The takt time for each of the work station is shown in table 4.2. The takt time required for the doctoring work stations are much slower than the rest of the work stations as only 24% of the rolls on each line contain defects and are directed to the doctoring work stations. A doctoring work station takes 392 mins (standard deviation of 72) to rework a roll. It is found that line A requires 4 doctoring work stations and line B requires 7 doctoring work stations. Thus, 11 doctoring work stations are required for the entire production line. This is the same as the number of doctoring work stations in the current production line. This leaves 3 doctoring work stations unused and assigned as backup. This assignment of the doctoring work stations is based on the assumption that the percentage of defects on each line remains at 24% regardless of the assignment of products on each line. CAS has to do further analysis to determine whether there is a correlation between product type and the percentage of defects. Table 4.2: Takt time of work stations. Line A Rate required (mins/roll) Capacity (mins/roll) Work station Net avaliable time (mins per week) P13 L21 6,498 9,491 15.8 22.9 9.2 12.8 S54 (5 & 9 web) 5,523 13.3 6.8 Doctoring (4) 10,080 101.3 392 per station Pal 10,080 24.3 8.3 Work station Net avaliable time (mins per week) Rate required (mins/roll) Capacity (mins/roll) P18 L22 S52 (8 web) S53 (7 web) S55 (5 & 9 web) Doctoring (7) Pa2 6,408 9,545 7,269 7,623 7,078 10,080 10,080 9.3 13.9 10.6 11.1 10.3 61.1 14.7 9.2 9.2 5.4 7.5 4.5 392 per station 8.3 B Line 4.2.2 Selection of demand Table 4.3 shows the demand of the first three months in 2010. CAS has estimated that the future demand would increase by 12% a year. The demand of March 2010 is 4461 rolls and this is the highest among the three months. Thus, the demand of March would reflect the future demand better. As such, the author chose the demand of March for the modeling of the production system. Table 4.3: Demand of first three months in 2010. Week Weekly demand (rolls) Monthly total (rolls) 1 2 3 4 747 1019 987 771 3524 March February January Month 1 2 3 4 1 2 3 4 1003 798 1505 674 900 1142 1177 1242 3980 4461 The author excluded products that have infrequent and low demand in order to simplify the problem. These products account for 1.2% of the total demand, so it would not have a big impact on the accuracy of the model. The demand breakdown of all the products used in the modeling is shown in table 4.4. The products are differentiated by two codes. The lamination code is juice or milk. The size codes are numbers such as 465 and 350. The lamination and size code together would form a product, for example Juice 465. Table 4.4: Demand breakdown of March 2010. Products Juice 465 Wk 1 Wk 2 Wk 3 368 359 March 1079 181 489 Juice 350 89 Juice 811 31 Juice 813 95 243 258 173 769 Juice 466 Juice 565 Juice 460 Juice 560 Juice 585 13 291 7 80 86 98 131 17 15 83 48 136 81 32 57 172 67 94 71 216 730 172 221 240 Juice 600 8 2 39 49 Milk 465 Milk 460 Milk 702 87 78 8 37 16 28 8 163 134 24 Milk 811 9 1242 4404 Total (rolls) 882 132 Wk 4 352 87 58 89 11 32 16 20 1142 1138 ___ 29 4.2.3 Products allocation The capacity on each line is determined to ensure that the weekly demand does not exceed the capacity of line A and line B. The bottleneck (based on work station rate) of each line is at the laminating stage. Table 4.5 illustrates the maximum rate of each work station. The maximum rate on line A is 430m/min while the maximum rate on line B is 600m/min. The doctoring work stations and palletizers have a higher capacity than the printers, laminators and slitters. Table 4.5: Maximum rate of work stations. Line B Line A Work station Maximum rate (rn/min) 600 P13 430 L21 1000 S54 Work station P18 L22 S52 S53 S55 Maximum rate (m/min) 600 600 1000 800 1200 CAS has estimated that the maximum daily output of L21 and L22 is 80 rolls and 110 rolls respectively. The total time in a day is 1,440 minutes and the average length of a roll is 5513m. The processing time, based on the maximum capacity of 430m/min (L21) and 600m/min (L22), is 12.8 minutes per roll and 9.2 minutes per roll, respectively. CAS estimates that there is a capacity loss of 30%. This capacity loss includes setups, trial runs, break downs, short stops, rush orders and others. Thus, the available time per day is 1,008 minutes (70% of 1,440 minutes). The maximum daily output is found using Eq. (4.3) Maximum daily output = Eq. (4.3) Available time per day (mins) Fastest processing time (mins) 1,008 Maximum daily output of L21 or line A =128 12.8 78.75 ~ 80 rolls per day 1,008 Maximum daily output of L22 or line B = 1 9.2 109.57 - 110 rolls per day Thus, the weekly capacity of the two lines A and B are 560 rolls per week and 770 rolls per week respectively. As such, the demand allocated to each production line must not exceed this capacity. The initial assignment of products to each line is based on the maximization of the number of flying setups on line B. The setup time and setup cost is minimized by maximizing the number of flying setups on line B. Table 4.6 illustrates the demand breakdown of every product and their allocation on line A and line B based on the initial assignment. The demand on line B is within the capacity of the line. However, the demand on line A has exceeded the capacity of the line in three of the weeks. The capacity of line A is 560 rolls a production cycle, but the demand for week 2, 3 and 4 is more than 700 rolls. Table 4.6: Products allocation based on initial assignment. Line Products Juice 465 Wk 1 Wk 2 368 Line A Wk 3 359 Wk 4 352 March 1079 769 95 243 258 173 Juice 350 89 132 87 181 A Total 184 Juice 813 Products Juice 466 Juice 565 Wk 1 Wk 2 13 98 B Wk 3 48 Wk 4 57 March 216 730 291 131 136 172 489 Juice 460 7 17 81 67 172 2337 Juice 560 80 15 32 94 221 Juice 585 86 83 71 240 Juice 600 8 2 39 49 Juice 811 31 Milk 465 87 37 11 28 163 Milk 460 78 16 32 8 134 Milk 702 8 Milk 811 89 58 16 9 24 20 29 B Total 698 399 434 536 2067 A & B Total 882 1142 1138 1242 4404 One of the challenges in using dedicated production lines is the allocation of products on each line. We observe that the demand on each line, based on the initial assignment, is more than the capacity of the line. Therefore, CAS has to do products reallocation between the lines during some of the production cycles to keep the demand on each line within the capacity. The author has established three simple rules to determine the allocation of products 1) All milk products are assigned to line B. Only L22 is capable of processing milk products. Thus, all milk products must be on line B. 2) Juice products that are capable of having flying setups are assigned to line B. Only L22 is capable of doing flying setups (for width difference smaller than 20mm). Setup time and cost is minimized by assigning products with width difference smaller than 20mm to line B. 3) When the demand exceeds the capacity on line A, some products are moved from line A to line B and vice versa. Both line A and line B have been assigned with 5 and 9 webs products. Only 5 or 9 webs products would be moved from line A to line B when the capacity of the line is exceeded and vice versa. Products with other types of webs on line B are not moved as this would be very disruptive to the setup planning for the slitters. The routing of the products to the slitters is not affected since S54 and S55 is assigned to slit the 5 and 9 web products. Table 4.7 illustrates the product allocation on each line after demand adjustments. Some of the products are moved from line A to line B. The author chose to move Juice 813 to line B as moving other products do not help to keep the demand within capacity. Juice 811 is moved from line B to line A to increase the demand allocated on line A as flying setup could not be done on Juice 811 alone on line B. Demand fluctuations of each product are a hurdle for the dedicated production line as the products might have to go through reallocations during some of the production cycles. However, it is less complex for CAS to do product reallocation during some of the production cycles than to plan the route for every product in each production cycle. Table 4.7: Products allocation after demand adjustments. Line A Products Wk 1 Juice 465 Juice 350 89 Juice 811 31 A Total 120 Wk 2 Wk 3 Wk 4 March 368 359 352 1079 132 87 181 489 58 500 B Line 504 533 Products Wk 1 Wk 2 Wk 3 Wk 4 March Juice 813 95 243 258 173 769 Juice 466 13 98 48 57 216 89 Juice 565 291 131 136 172 730 1657 Juice 460 7 17 81 67 172 Juice 560 80 15 32 94 221 Juice 585 86 83 71 240 Juice 600 8 2 39 49 Milk 465 87 37 11 28 163 Milk 460 78 16 32 8 134 Milk 702 8 Milk 811 16 9 24 20 29 B Total 762 642 634 709 2747 A & B Total 882 1142 1138 1242 4404 4.2.4 Sequence of production The production planning is done on a weekly basis in CAS. The orders are confirmed several days before the start of each production cycle. Each production cycle starts on a Monday and the cycle is 7 days. These orders are divided to each line according to the assignment given in section 4.2.3 and then sequenced on each line. The proposed production uses the first in first out (FIFO) rule on both lines. The production is sequenced solely based on the setup costs of laminators as the setup costs on other work stations are negligible compared to that of the laminators. Setup is required at the laminators when there is a product family change or a paper width change. The PE continues to flow down as waste during these setups and this waste is called drooling waste. CAS estimates that the drooling waste costs $49 per minute. The cost of drooling waste for different types of lamination setup is illustrated in table 4.8. L22 is capable of doing flying setup when the paper width change is less than 20mm from wide to narrow. Flying setup allows the setup to take place with minimum loss of speed. Thus, the setup time is approximately zero minutes. Table 4.8: Setup costs of laminators. setup time Drooling ($) Work station Type of setup Average (mins) L21, L22 40 L21, L22 Product family (juice to milk; milk to juice) Width (narrow to wide) 30 1,470 L21, L22 L21 L22 Width (wide to narrow, >20mm) Width (wide to narrow, <20mm) Width (wide to narrow, <20mm) 20 20 0 (flying setup) 980 980 0 waste 1,960 The author determined the sequence of the production on each line using three steps: * Separate the products into juice product family and milk product family. It is indifferent in the order of the two groups. This step is only applicable to line B as all the products on line A are juice products. This is the first step as the setup time between product families is the longest. Thus, the setup between product families is minimized. wV : -..................... .- ::::: " - - --- Sort the products within each group by their width with the widest width product at the top. Thus, within each product family, the product with the widest width is processed first and the product with the narrowest width is processed last. This is because the setup time for wide to narrow is shorter than the setup time from narrow to wide. * When the product width is the same (e.g. Juice 350 and Juice 811), the products with the same webs are placed together to minimize the setup time at the slitters. The sequence of production is shown in table 4.9. The sequence of production line A is Juice 465, followed by Juice 350 and Juice 811. Juice 350 and Juice 811 have the same width of 1468mm. Juice 350 is produced first followed by Juice 811 so as to reduce the number of web setups on S54 on line A. CAS is required to do 1 web setup during every production cycle to change the web number from 9 to 5. Table 4.9: Sequence of production. Line B Line A Products Juice 813 Products [Width (mm) 1574 Juice 465 webs Sequence 1468 9 2 Juice 466 1468 5 3 Juice 565 Juice 460 Juice 560 Juice 585 Juice 600 Milk 465 Milk 460 Milk 702 Milk 811 Juice 350 Juice 811 Width (mm) 1533 webs Sequence 1 5 2 8 8 3 7 4 7 7 7 1574 1506 1506 1468 9 7 7 5 5 6 7 8 9 10 11 There are some products on line B that are highlighted. Products with the same highlighted color indicate that flying setup is possible on that group of products. It is found that there is no web setup time on loss on any of the slitters on line B. Although 43 S55 on line B is assigned to produce 5 web and 9 web products, time is not lost to do the web setup of 60 minutes. After the production of Juice 813 (5 webs), the subsequent products are 8 and 7 web products. Thus, S55 can do the web setup during this idle time. After the production of Milk 465, the subsequent products are 7 web products. Thus, S55 can do the web setup during this idle time. After the production of Milk 811 (5 webs), the subsequent product is also a 5 webs product (Juice 813). Thus, no web setup is required. Therefore, there is no web setup of 60 minutes on any of the slitters on line B and there is no disruption to the flow from L22 to the slitters. 4.2.5 Work station stoppage There are several types of stoppages that would cause a work station in CAS to stop production. They are categorized and explained in the following sections. These stoppage data is extracted from CAS internal P2 system and analyzed using the Minitab software to determine a suitable distribution. These stoppages are used in the simulation models of the proposed production line and the current production line. Breakdown The printers, laminators and slitters experience breakdowns. However, the doctoring work stations and the palletizers are well maintained by CAS and they do not breakdown. The mean time between failures (MTBF) and average mean time to repair (MTTR) in minutes for January to April 2010 are shown in table 4.10. The slitters are more reliable than the printers and the laminators. Table 4.10: Breakdown of work stations. Work station MTBF (mins) P13 P18 L21 L22 S52 S53 S54 S55 13,964 12,570 13,011 11,185 15,982 18,548 19,403 18,001 MTTR (mins) 61 63 57 65 11 9.5 9 6.5 Rest time Work stations are made to rest due to 2 reasons. Firstly, they are rested during planned maintenance. CAS does not have a fixed schedule for planned maintenance and the planned maintenance time is very short. For example, the S54 has maintenance of 12 minutes in the entire month of March. Secondly, they are made to rest when sufficient WIP is built up. The sufficient level of the WIP is based on the judgment of the process managers. Thus, there are no strict rules to determine the level of WIP. The rest of the work stations are shown in table 4.11. Table 4.11: Rest time of work stations. Work station Total time (mins) Mean (mins) Frequency of rest periods over month of March P13 P18 L21 L22 S52 S53 S54 S55 1,616 404 4,316 1,079 13,254 6,627 4,212 1,053 13,680 4,560 19,260 6,420 14,772 7,386 12,024 4,008 4 4 2 4 3 3 2 3 Short stop Short stops occur when the work station did not breakdown but other activities caused the production to stop. These activities could be formation of air bubbles, roll breakage and misalignments. Short stop only occurs at the printers and laminators. The other work stations do not experience short stops. Short stops occur at random. Table 4.12 shows the distribution and frequency of the short stops. Table 4.12: Short stop of work stations. Work station Mean (mins) per stop Standard deviation Distribution Frequency of short stops over month of March P13 P18 L21 L22 17.7 13.3 Lognormal 17.8 17.4 Lognormal 33.9 26.7 Lognormal 39.1 31 Lognormal 146 172 41 81 Setup Table 4.13 shows the setup time of the work stations. Printer setups occur when there is a change of design or paper width. Laminator setups occur when the there is a change in product family or paper width. The setup time of the printers and laminators varies substantially even for the same type of setup. Thus, the author used the actual setup time of printers and laminators in March for the simulation models. This distribution is more reflective of the actual setup than the usage of a fixed setup time. Slitter setups occur when there is a change in number of webs or paper stiffness. There is no actual data on the web change setup as the slitters are assigned to slit only a single web in the current production system. The web and paper stiffness change setup time is modeled as deterministic whenever there is a change in the web and paper stiffness respectively. Table 4.13: Setup time of work stations. Work station Mean (mins) Slitters P13 P18 Laminators Slitters 17.6 15.1 Refer to table 4.5 60 15 0 Deterministic Paper stiffness Standard deviation 11 9.3 0 0 Distribution Lognormal Lognormal Deterministic Frequency 654 678 Deterministic Product family or width change Web Reel inspection Reel inspection only occurs at the slitters. The rolls are slit into smaller reels at the slitters. The average roll length is 5513m. The reel length is 2595m. Thus, a single roll could be slit into 2.12 sets of reels. Every time a set of rolls has been slit, the operators would need to check the reels and separate the defective and non defective reels before sending them to the doctoring work stations or palletizers. Table 4.14 shows the reel inspection time at the slitters. These are actual data captured from CAS P2 system. The time in the table is for the inspection for 1 roll and this inspection is required for every roll. Thus, it is modeled for every roll that is processed by the slitter. Table 4.14: Reel inspection time of slitters. Work S52 S53 station S54 10.4 9.7 9.8 Mean (mins) 3.3 3.2 Standard deviation 3.1 Normal Normal Normal Distribution S55 10.6 3.1 Normal 4.3 Simulation modeling 4.3.1 Proposed production line The modeling of the proposed production line is done using the SIMUL8 simulation software. Figure 4.4 shows the screenshot of the proposed production line A and line B in the software interface. There are 3 objectives for simulating the proposed production line. Firstly, it is used to test out the feasibility of the route of products. Secondly, it is used to determine a suitable configuration of the process times of the work stations in each line. Thirdly, it is used to determine the card number (or level of WIP) of each line to meet the demand of 1,657 rolls in line A and 2,747 rolls in line B. The demand used in both lines is the March 2010 demand as explained in section 4.2.3. The simulation time is 40,320 minutes (4 production cycles, equal to four weeks). The simulation time is chosen to get stable results. We found that a simulation time of 10,080 (1 production cycle) is not long enough to determine stable results. However, a simulation time of 40,320 is able to achieve stable results. Juice465 Juice 350 Juice811 Line A co LineAiorder P135h Juice813 Juice466 Juice565 0 0 0 stop P13 Doctor1 (4) L21 shortstop L21 1111111S53 Juice460 Juice560 Juice585 S54reelinspection 354 (5 and9 webs) Pa r 1 Pa 1 ine A FGI Paroute2 Pa2 Line8FGl reelinspectionS53 (7webs) c LineB order E Juice 600 Milk 65 Milk460 card post P18sh~rtstop a]~~ Pit L22shortstop L22 4 S52reelinspection 0 S52(8webs) ' 0 0 0 t n lweis reel inspection S53ou S53 Milk 702 Mile811 0 0 Figure 4.4: Simulation model of proposed production line. A D c2(7) The author made several assumptions in the simulation model of the proposed production line. These assumptions are " The travel time between work stations are zero. This is reasonable as the travel is very short as compared to the other events such as processing time and stoppages. " The raw materials required at every work station are always available. CAS kept about 2 days of raw materials in the internal warehouse, so the chance of starving the work stations is low. * The production is modeled for the whole month of March. So, the demand for the entire month is constant across the 4 weeks. * There are no rush orders. There is no record of rush orders and the proportion of rush orders is likely to be small. Thus, it is not important to model it. e The entire roll is sent to the doctoring work stations when defects are found. In the actual production, only reels with defects are sent to the doctoring work stations. However, as the author used percentage (24%) to define the proportion of defects present, it does not affect the accuracy whether it is modeled as a roll or a reel. The takt time required for each work station is calculated in section 4.1.2.3. The author set the process time of each work stations to equal the takt time. The work stations are set equal to the takt time to remove the imbalances in the production line. Thus, the work stations all work at the same rate and do not produce ahead and build up inventory, as is seen in the current production line. The process time of the work stations in each line are shown in table 4.15. The card number (WIP) is varied in each of the configuration with increment of 50 cards. Each line starts off with WIP in the laminating and slitting queue. This is reasonable as each production cycle does not start with zero WIP. If 100 cards are used, the laminating queue starts with 50 cards while the slitter queue starts with 50 cards. Table 4.15: Process time of work stations in proposed model. Line A work station Number of station Process time (mins) P13 1 L21 1 16 23 S54 Doctor 1 4 392 Pal 1 24 Line B work station Number of station P18 1 Process time (mins) 9 L22 1 14 S52 1 10 S53 S55 1 11 1 10 Doctor 7 392 Pa2 1 14 13 There are 2 tools in SIMUL8 to model the proposed production line. The first tool is the visual logic, which is the language used in the software. The author used this visual logic to route the cards from the palletizers to the 'CONWIP card post'. When a roll leaves the line through the palletizer, visual logic increases the number of cards in the 'CONWIP card post' by 1. An order only enters the system when there is both an order and a card at the 'CONWIP card post'. The visual logic used for the routing of the cards in both lines is Pal Work Complete logic Add Work to Queue Products, Line A CONWIP card post Pa2 Work Complete logic Add Work to Queue Products, Line B CONWIP card post The second tool is called label. The author attached labels to each roll to direct the roll to the correct slitter based on the number of webs the roll has. A roll with 7 webs is labeled as '1', a roll with 8 webs is labeled as '2' while a roll with 5 or 9 webs is labeled as '3'. These labels are read when the roll leaves L22 and then directed to the intended slitter. 4.3.2 Current production line The modeling of the current production line is also done by SIMUL8 software. Figure 4.5 shows the screenshot of the current production line in the software interface. The objective of simulating the current production line is to verify that the method of modeling the proposed production line is credible. The printed WIP (WIP between printers and laminators), laminated WIP (WIP between laminators and slitters) and the doctor WIP (WIP waiting to be doctored) in the model is compared with the actual WIP in March for verification. The demand is 4404 rolls and the simulation time is 40,320 minutes (4 production cycles). 052 reelinspecaionS52 rest S52(5 and8 webs) 0 0 Juice465Juice350 Juice811 0 Juice813 Juice466 Juice565 Pl3shortstop Pl3rest P13 L21 shodstop L21rest L2e Dodor(11) o s Pal Or~rs Pr g L ro ngi- a ute FG Pa 2 Juce40 Jue c6 5 Juice585 P18ish itstop P18rest P8 L22 sh stop L22rest 02 54 teel ispecton S54 est S54(9 webs)' 4r Juice600 Milk465 Milk460 9 vb routinr 0S55 Milk702 Mik~l1 S4 (9wes reenspection esS55 (9 wet) Figure 4.5: Simulation model of current production line. The assumptions made in this model are the same as that in the model of the proposed production line. In the current model, S52 is assigned to slit 5 and 8 webs products, S53 is assigned to slit 7 web products and S54 and S55 are assigned to slit 9 web products. 11 doctoring work stations are used in the current model. The current production line is modeled to have WIP at the laminator and slitter queues at the start of the simulation. There are 396 rolls in the laminator queue and 564 rolls in the slitter queue on the last day before the start of the March production. Therefore, the author used the same number for modeling. ..... The number of rolls processed by each work station is extracted from CAS internal P2 system. The length of the products produced by each work station in March is shown in table 4.16. The percentage of the rolls going to each work stations is calculated using Eq. (4.4) % of rolls = Length processed at work station(m) Total length produced at same type of work station(m) Eq. (4.4) The percentage of rolls processed by the doctoring work stations is extracted directly from CAS internal P2 system. The process time for each of work station is shown in table 4.17. The data is also extracted from CAS internal P2 system. Table 4.16: Percentage of rolls processed by each work station. Work station P13 P18 L21 L22 Doctors Processed length(m) 15,280,000 13,237,000 11,787,000 19,371,000 % of rolls Rolls 53 47 37 63 2334 2070 1629 2775 24 1052 Work station S52 S53 S54 S55 Processed length(m) 7,121,000 5,679,000 6,829,000 8,264,000 % of rolls 25.5 20.3 24.6 29.6 Table 4.17: Process time used in current production line model. Work station P13 P18 L21 L22 S52 IS53 IS53 Run speed (m/min) 546 519 410 548 1,000 800 800 Process time (min/roll) 10.1 10.6 13.4 10.1 5.5 Work station S54 Run speed (m/min) 1,000 S55 1 200 Doctoring (11) Pal Pa2 Process time (min/roll) 5.5 4.6 392 8.3 8.3 Rolls 1123 894 1083 1303 4.4 Improvement analysis 4.4.1 Conversion between rolls and reels Figure 4.6 illustrates the splitting of a single roll into smaller reels. A roll is equivalent to 16.5 reels. The weighted average length of a roll is 5513m and all reels have the same length of 2595m. Thus, a roll is split into 2 sets of reels of normal length (2595m) and a third set of reel of shorter length (323m on average). The shorter reel is combined with other shorter reels and delivered to the customers like the reels of normal length. As such, a roll is equivalent to 2.12 sets of reels on average. For each set of reels, there are several webs (5, 7, 8 or 9). The weighted average number of webs is 7.8 webs. Therefore, the number of reels in a roll is found using equation Eq. (4.5) Number of reel = Number of webs x Length of roll ()(4.5) Length of reel (in)Eq(45 Number of set of reel Number of webs 2595m 5513m Figure 4.6: Slitting of a roll into reels. 4.4.2 Inventory holding costs CAS estimated the inventory holding costs for the three types of inventory. Table 4.18 shows the inventory holding costs for the printed WIP, laminated WIP and the doctor WIP. The inventory holding costs of each month is found using equation Eq. (4.6) Inventory holding cost($/month) = Number of rolls x holding cost ($/roll/month) Eq. (4.6) Table 4.18: Inventory holding costs of three types of WIP. Inventory type Inventory holding cost per roll per month ($) Printed WIP Laminated WIP Doctor WIP 4.4.3 95 116 130 Lead time The lead time is calculated using the Little's Law. The total WIP is calculated by adding the printed WIP, laminated WIP and the doctor WIP. The lead time is then found using Eq. (4.7) Lead time (days) = Total WIP in production line(rolls) Throughput (rolls/day) Eq. (4.7) CHAPTER 5: RESULTS AND DISCUSSION 5.1 Current production line The purpose of building the simulation model of the current production line is to verify that the method of modeling the production line is credible. Table 5.1 shows the average printed WIP, laminated WIP, doctor WIP and their standard deviation from 20 simulation runs. The throughput of the model is 4,526.6 rolls. This throughput is reasonable as the author has excluded a small percentage of the products. Thus, the throughput is higher than 4,404 rolls. The average printed WIP is 310.7 rolls while the average laminated WIP is 618.2 rolls. The average doctor reel WIP 861.0 reels. A roll is equivalent to 16.5 reels. Table 5.1: Performance of the current production line model. Simulation Object Throughput (4 weeks) Printed WIP Laminated WIP Doctor WIP (reels) Average (rolls) Standard deviation 35.8 4,526.6 25.0 310.7 7.0 617.0 85.8 861.0 The results from the model of the current production line are compared with the actual inventory level in March 2010 to find the error of the model. Table 5.2 shows the inventory level of the model and the actual production line as well as the percentage error. It is found that the model has underestimated the printed WIP and the laminated WIP and has overestimated the doctor WIP. However, the percentage error of the model is low. The highest percentage error is only 6.8%. Thus, the method of modeling the current production line represents the actual production line well. It is expected that the modeling of the proposed production line has a good prediction of the actual performance since the method of modeling is the same. Table 5.2: Error in the current production line model. Model (roll) March inventory (roll) Error (%) Inventory type 2.5 318.7 310.7 Printed WIP 6.3 658.5 617.0 Laminated WIP 6.8 806.3 Doctor WIP (reel) 861.0 5.2 Line A performance 5.2.1 Throughput and card number The number of cards has been varied to find the number of cards that gives the production line A a throughput of more than 1657 rolls in a four-week period (four production cycles). Figure 5.1 illustrates the throughput trend when the card number is increased at interval of 50 cards while table 5.3 shows the throughput mean and standard deviation of the 20 simulations. We found that the throughput is increased when the card number is increased. However, there is a critical point beyond which a higher card number does not help to increase the throughput of the production line. The throughput reached its highest throughput of 1714 rolls at the card number of 250 and the increment of card number to 300 does not help to increase the throughput. 1,720 1,710 1,700 0 .21,0 . 0 1,680 1,670 1,660 1,650 50 100 150 200 250 300 Card number Figure 5.1: Throughput trend of line A. It is also found that the increment of throughput increment diminished as the card number is increased. Thus, holding more inventories in CAS does not give the same throughout increment. Holding more inventories would only cause CAS to incur unnecessary inventory holding costs and tie up the capital. We found that line A is able to give a throughput of more than 1657 rolls when the card number is 100. A card number of 50 is not able to achieve the target throughput. Thus, the use of 50 cards is not feasible. If the work stations are made to work at a faster rate, an even lower card number could be achieved for the same throughput requirement. However, a lower card number is not realistic as unexpected variations might cause the work stations to be starved during production. Table 5.3: Throughput mean and standard deviation in line A. Card number 50 100 150 200 250 300 5.2.2 Mean (rolls) Standard deviation 11.8 1,655.4 12.0 1,678.2 12.6 1,703.5 10.4 1,713.0 10.1 1,714.4 10.2 1,714.0 WIP inventory The next performance measure of the configurations is the average printed, laminated and doctor WIP. The 3 types of WIP are shown in table 5.4, 5.5 and 5.6. A shaded box means that the particular card number is not feasible as the throughput is below 1657 rolls. These non feasible card numbers are not discussed in the rest of the discussion. We found that the minimum average printed WIP is 63.5 rolls and it is achieved when 100 cards are used. The average printed, laminated and doctor WIP increases when the card number is increased. The minimum average laminated WIP is 15.7 rolls while the minimum average doctor WIP is 142.5 rolls. 11I'll- 111,__- _R, w"aLQij% I , - Table 5.4: Average printed WIP in line A. Card number 50 100 150 200 250 300 Mean (rolls) Standard deviation 4.5 27.2 6.1 63.5 8.0 94.5 9.5 121.2 9.7 146.9 9.8 172.3 Table 5.5: Average laminated WIP in line A. Card number 50 100 150 200 250 300 Mean (rolls) 5.3 15.7 31.9 53.4 77.6 102.2 Standard deviation 1.2 3.3 6.0 8.5 9.3 9.3 Table 5.6: Average doctor WIP in line A. Card number Mean (reels) 110.4 50 142.5 100 164.6 150 174.7 200 177.2 250 178.0 300 Standard deviation 63.4 79.4 89.0 90.6 90.2 90.5 ............................................... .... ..................................... ..... ... .. . 5.3 Line B performance 5.3.1 Throughput and card number The number of cards has been varied to find the number of cards that gives the production line B a throughput of more than 2747 rolls in a four-week period. Figure 5.2 illustrates the throughput trend when the card number is increased at interval of 50 cards while table 5.7 shows the throughput mean and standard deviation of the 20 simulations. We also observe in production line B that the increment of throughput decreased when the card number is higher. Thus, a high card number has diminishing effects on the throughput increment of the line. 2,900 2,800 2,700 U1 -5 : 0 2,600 2,500 ,00 2,300 2,200 50 100 150 200 250 300 Card number Figure 5.2: Throughput trend of line B. We found that the minimum card number required to achieve the throughput of 2747 rolls is 200 cards. We note that 150 cards are able to achieve a throughput of nearly 2747 rolls, but it does not reach the target. Thus, card numbers 50 to 150 are not feasible. Table 5.7: Throughput mean and standard deviation in line B. Card number 50 100 150 200 250 300 5.3.2 Mean (rolls) Standard deviation 15.7 2,222.9 15.9 2,644.5 13.4 2,722.4 14.0 2,753.9 14.9 2,785.2 18.4 2,785.2 WIP inventory The average printed, laminated and doctor WIP are shown in table 5.8, 5.9 and 5.10. A shaded box means that the particular card number is not feasible as the throughput is below 2747 rolls. Thus, they are ignored in the rest of the discussion. We found that the minimum average printed WIP is 109.4 rolls when 200 cards are used. The minimum average laminated WIP is 59.9 rolls while the average doctor WIP is 108.6 reels. Table 5.8: Average printed WIP in line B. Card number Mean (rolls) Standard deviation 50 4.0 0.4 100 150 200 250 300 21.5 62.1 109.4 155.1 201.0 2.2 3.6 3.5 4.2 4.2 Table 5.9: Average laminated WIP in line B. Card number Mean (rolls) Standard deviation 0.9 34.3 50 100 54.5 1.7 150 200 250 300 58.3 59.9 61.8 63.2 1.8 1.9 1.8 1.8 Table 5.10: Average doctor WIP in line B. Card number Mean (reels) 16.9 50 Standard deviation 6.3 100 77.2 34.4 150 99.1 45.5 200 250 300 108.6 125.1 118.1 49.4 61.9 25.0 The laminators are the only work stations in CAS that must be in operation continuously. It is expensive to starve the laminators. When the laminator is starved, it has to be shut down and it is required to be restarted (cold start) to process the roll. It costs about $9,000 each time a cost start is required. The process time of L22 is 14 minutes in the proposed production system. The average length of a roll is 5513m. Thus, the work station rate is 393 m/minute. Figure 5.3 depicts the printed WIP during the simulation when L22 process time is 14 minutes (aligned to takt time) and 200 cards are used. It is observed that the minimum printed WIP is 71 rolls. Therefore, L22 is not starved at all times. Figure 5.3: Printed WIP with L22 process time = 14minutes. CAS should take precautions to prevent local autonomy of the laminator workers to change the work station rate and stray away from the intended takt time. Figure 5.4 shows an illustration of the printed WIP during the simulation when L22 process time is 12 minutes (not aligned to takt time) and 200 cards are used. A process time of 12 minutes corresponds to work station rate 460 m/minute. We observe in figure 5.4 that the printed WIP reached zero about 7 times, which would force a shutdown of the printer. If L22 were starved 7 times, the cold start costs are about $63,000. Figure 5.4: Printed WIP with L22 process time = 12minutes. 5.4 Inventory level and holding costs The inventory level in both line A and line B are compared with the actual inventory level in March 2010 to determine the cost savings for the proposed production lines. Table 5.11 shows the inventory of the proposed production line and the actual inventory in March 2010. It is evident that all the inventory types are lower for the proposed production line. The printed WIP is reduced by 46%, the laminated WIP is reduced by 88% and the doctor WIP is reduced by 68%. The total WIP is reduced by 74% from 1,026 rolls to 263 rolls. Table 5.11: Inventory reduction. Printed WIP (rolls) Laminated WIP (rolls) Doctor WIP (reels) Total WIP Line A Line B Proposed production line total Actual inventory level (March) % reduction 63.5 109.4 172.9 (65%) 318.7 (31%) 46 15.7 59.9 75.6 (29%) 658.5 (64%) 88 142.5 87.8 108.6 175.9 251.1 (6%) 263.7 806.3 (5%) 1,026.1 68 74 In the current production line, more WIP is stored as laminated WIP (64%) instead of printed WIP (31%). It is more expensive to store buffers downstream as more value has been added to the roll. Thus, the inventory holding costs is much higher in March. In the proposed production line, 65% of the inventory is stored as printed WIP while 29% is stored as laminated WIP. As such, the inventory holding costs is lower. Table 5.12 illustrates the inventory holding costs of the proposed production line and the current production line in March 2010. It is found that the inventory holding costs is $27,173 per month for the proposed production line while the inventory holding costs is $113,014 for the current production line. Thus, the inventory holding cost savings is 85,840 per month. This is a 76% reduction in inventory holding costs. Table 5.12: Inventory holding cost savings. Proposed Actual inventory Line A Line B production line total level (March) Savings Printed WIP ($) 6,033 10,393 16,426 30,275 13,849 Laminated WIP ($) Doctor WIP($) 1,821 1,123 6,948 856 Total 8,770 1,978 27,173 76,386 67,616 6,353 113,014 4,374 85,840 Figure 5.5 depicts the inventory holding costs of the proposed production line, the current production line from January to May 2010 and the average inventory holding costs of these 5 months. It is observed that the proposed production line has a lower inventory holding costs as compared to any of the months in 2010. The monthly average inventory holding costs in 2010 is $109,530. Therefore, the annual inventory holding costs of the current production line is estimated to be $1,314,360. The annual inventory holding costs of the proposed production line is expected to be $326,076. Thus, the annual inventory holding costs savings is $988,284. Figure 5.5: Inventory cost of proposed production line. - t -- -- -- - - - - - - - -Aw-A I '- - -- 5.5 Warehouse capacity The warehouse capacity allocated for the printed and laminated WIP is 500 rolls while the warehouse capacity allocated for the doctor WIP is 500 reels. Figure 5.6 illustrates the capacity allocated for the printed and laminated WIP and the actual WIP in the first 5 months of 2010. It is evident that all the 5 months in 2010 has exceeded the warehouse capacity allocated to the WIP. The WIP has exceeded the warehouse capacity by about 100% in February, March and April. As such, the warehouse has been piled up with too much inventory and the working lanes for forklift are forced to make way for these excess inventory. This high inventory caused the workers to take a longer time to excess the rolls. From the simulation, it is found that the proposed production line has an average of 248.5 rolls of WIP. This is 50% of the capacity allocated for it. Thus, the proposed production line WIP is kept within the capacity. During the actual production, the WIP would fluctuate. However, as the total card number used is 300 cards, the WIP in the production line would not exceed 300 rolls. Thus, the capacity allocated would not be exceeded at any point of time. 1200 1039 1000 - - - 97f -983 901 832 800 0 50 600 E 2 -500 z 400 200 0 Proposed Capacity Jan Feb Mar Apr 2010 Figure 5.6: Capacity of printed and laminated WIP. May The capacity allocated for the doctor WIP is 500 reels. Figure 5.7 shows the capacity allocated for the doctor WIP and the doctor WIP in the first 5 months of 2010. It is observed that the capacity allocated for the doctor has been exceeded in all the months. The doctor WIP in April is about 3 times of the capacity while the doctor WIP in May is nearly 4 times of the capacity. CAS has a general area allocated to store the doctor WIP. However, within that area, the reels could be placed anywhere. Thus, it is difficult for the workers to find the correct reel to be doctored when the number of doctor WIP is high. The workers mainly rely on their experience to navigate and find the correct reel. We found that the doctor WIP in the proposed production line is 251.1 reels. This is 50% of the capacity. Thus, WIP is kept within the capacity. In addition, the workers would be able to find the correct reel to be doctored faster as the number of doctor WIP reels is lower. Since the printed, laminated and doctor WIP utilizes about 50% of the capacity allocated to them, it gives CAS the opportunity to bring some of the inventories stored in the external warehouse back to the internal warehouse. This gives CAS the added advantage of reducing the high inventory costs in the external warehouse. 2000 1880 1800 1657 1600 1400 1200 0 1000 .0 E 806 800 600 -- 3 400 251.1 200 0 Proposed Capacity Jan Feb Mar 2010 Figure 5.7: Capacity of doctor WIP. Apr May 5.6 Lead time The production lead time quoted to customers is 12 days. The production lead time takes up about 10 days while the 3 rd party logistic company requires about 2 days to manage the finished goods inventory. Figure 5.8 shows the production lead time of the proposed production line and the production lead time in the first 3 months in 2010.The production lead time is about 6 to 7 days. The current production lead time is long as the WIP in the production line is high. A roll has to wait for all the previous rolls to be processed before it could be processed. Thus, the average time a roll spends in the production is long. From the simulation, the lead time of the proposed production line is found to be 1.9 days. This lead time of the proposed production line is the time from the release of a roll to the completion of a roll. It should be noted that this lead time does not include the waiting time of a roll to get a CONWIP card. This proposed production lead time is compared with current production lead time in March 2010. The lead time reduction is 4.5 days or 70%. This shorter lead time would make CAS more competitive within the company. The shortest lead time is quoted by the plant in China, and is 7days. The 1.9 days of production lead time together with the 2 days of lead time required by the 3 rd party logistic company makes the total lead time to be 3.9 days. Thus, this lead time is shorter than any of the other plants. As the lead time is shortened, it makes CAS more competitive to the external company C. 7.0 6.4 6.0 5.0 E 4.0 -U e3.0 2.0 1.0 0.0 Proposed )an Feb Mar 2010 Figure 5.8: Lead time of proposed production line. CHAPTER 6: RECOMMENDATIONS AND CONCLUSION This thesis uncovered the problems with the current production line. It is demonstrated that the production line could be improved by making the three major changes of using dedicated production lines, changing to pull production system and aligning the production rate to the takt time. It is found that the WIP in the production line is reduced by 70% from 1,026 rolls to 300 rolls when the proposed production line is used. This improvement helps CAS to save $988,284 annually and eliminates the capital tied up by the inventories. In addition, the strain on the internal warehouse capacity is removed as the printed WIP, laminated WIP and the doctor WIP takes up only about 50% of the total warehouse capacity. The reduction in WIP also alleviates the potential safety issues and time wasted in looking for the WIP in the internal warehouse. In addition, the production lead time is reduced from 6.4 days to 1.9 days. The reduction in production lead time puts CAS in a better position to fend off internal and external competition. There are several recommendations to CAS. Firstly, the author recommends switching only 5 and 9 webs products between lines when the demand has exceeded the capacity of the lines. This reduced the need of reassigning products to the slitters since S54 and S55 are both assigned to slit 5 and 9 web products. However, there is an opportunity for CAS to look for a more detailed rule for switching products across lines. Secondly, it is recommended that each slitter is assigned to slit only one or two type of products. This avoids the disruption of flow by the long web setup time. Thirdly, it is recommended that 4 doctoring work stations are assigned to line A while 7 doctoring work stations are assigned to line B. The rest of the work stations should remained as backup to either lines. These backups would help to buffer against any unexpected fluctuations of the percentage of defects on each line. CAS should do further analysis to determine if there is any correlation between product types and percentage of defects. Fourthly, the minimum WIP required for line A is 100 rolls while the minimum WIP for line B is 200 rolls. Thus, the total WIP required is 300 rolls. The author recommends that CAS does not reduce the total WIP to 300 rolls at once. The proposed production line is a major change to the way CAS has been operating. Since the warehouse capacity allocated to the printed and laminated is 500 rolls while the warehouse capacity allocated to the doctor WIP is 500 reels, CAS could first reduce the level of WIP to 500 rolls to eliminate the warehouse capacity problem. CAS could then reduce the WIP level gradually to 300 rolls. This would allow all the employees to have time to adapt to the changes and to build up their confidence in the proposed production line. Lastly, it is recommended that CAS should look into the way the KPI is structured. The OEE KPI is tied to the employees' bonuses. Thus, there would be resistance to the implementation of the proposed production line. The changes to the production line should be communicated carefully to the employees. In addition, the KPI should be altered to ensure that no one is penalized by moving towards the proposed production line. Without changes to the KPI, the employees would have a tendency to fall back to the old way of doing things. In conclusion, CAS is recommended to adopt the three major changes of using dedicated production lines, changing to pull production system and aligning the production rate to the takt time to control the WIP. CHAPTER 7: FUTURE OPPORTUNITIES The author identified two areas where future studies could be made. Firstly, it is mentioned that the cold start of the laminators costs $9,000 each time and it is expensive to CAS. Therefore, the laminators are made to run continuously. However, it costs CAS about $840 an hour to run the laminators. This cost includes the energy and the labor cost. Therefore, there is a trade off between running the laminators continuously and shutting down the laminators. This trade off is important when the demand on the line is low and the laminators have the chance to be switched off when the demand has been met. It is worthwhile to determine two things. Firstly, verifications should be done to determine whether the laminators should be shut down at times to achieve a lower cost. Secondly, the duration of each shut down should be determined. The second area is to reduce other types of inventory that CAS holds. This thesis is focused on the control of WIP in the internal warehouse. These WIP make up 19% of all the inventories in monetary value. The base materials make up 35% and the FGI make up another 35%. The remaining 11% of the inventories are the straws and packaging materials. Due to the high level of inventory, CAS internal warehouse could not accommodate all of the inventories. CAS stores an average of 2031 rolls of FGI in the internal warehouse and stores an average of 1720 rolls of FGI in the external warehouse in 2010. The author has identified a few causes of the high FGI. Firstly, the shipment to the customers is infrequent (twice a week). Therefore, the FGI has to wait in the warehouse for the shipment date. Secondly, some customers postponed the delivery date after the production has been completed. Thus, CAS stores these FGI in the warehouse out of goodwill. The storage of the FGI in the external warehouse is expensive. Thus, there is a huge opportunity to reduce the level of FGI in order to reduce the inventory costs. References [1] Dilip Chhajed and Timothy J. Lowe, Building intuition: insights from basic operations management models and principles, Operations research and management science (2008), p. 8 1-100 [2] Mark L. Spearman, David L. Woodruff and Wallace J. Hopp, CONWIP: a pull alternative to kanban, International journal of production research (1990), vol. 28, p. 879894 [3] Wallace J. Hopp and Mark L. Spearman, Factory physics, McGraw hill (2008) [4] M.C. Bonney, Zongmao Zhang, M.A. Head, C.C. Tien and R.J. Barson, Are push and pull systems really so different?, International journal of production research (1999), vol. 59, p. 53-64 [5] Mark L. Spearman and Michael A. Zazanis, Push and pull production systems: issues and comparisons, operations research (1992), vol. 40, p. 521-531 [6] Asbjoern M. Bonvik and Stanley B. Gerswin, Beyond kanban: creating and analyzing lean shop floor control policies, Manufacturing and service operations management conference (1996) [7] Averill M. Law and W. David Kelton, Simulation modeling and analysis, McGraw Hill (2000)