Douglas Van Bossuyt PhD Qualifier June 11, 2009 OSU Triple Play Research Interests • Collaborative design • Complex system design • Design for cultures • Psychology • Business management • Overview of Design Trade Studies • Methods, Tools, Groups Who Perform, Examples • Overview of Risk and Uncertainty • Risk quantification and mitigation tools • Uncertainty assessment methods • Uncertainty mitigation tools • • • • Possible ways to incorporate risk and uncertainty into Trade Studies Future areas of research Expected contributions of overall research Research plan Complex systems are here to stay Every complex system design tries to maximize system utility • System utility metrics: ROI, system integrity, public perception of project, etc… Subsystems optimized to achieve high overall system utility Design parameters (mass, power, volume, cost, heat dissipation, etc.) used to define subsystem parameters that determine system-level utility Parameters are traded between subsystems to optimize design in Trade Studies Risk and uncertainty of systems is another important factor in complex system design Definitions • Risk: probability of event X impact of event Sometimes more narrowly means probability of catastrophic event X impact of event • Uncertainty: caused by variability and doubt in the status of an output that is either predictable or unpredictable, or caused by an unknown process or device • Reliability: Ability of a device to perform as intended over a given period of time Definitions (Con’t) • Robustness: Ability of a device to continue to properly function under changes in input variables. • Design Margins: Quantify the influence of uncertainties in the design process. Often a high and low bounding around a central parameter quantification. Methods developed from this research will improve system utility and integrity Improved utility and integrity uses resources more efficiently and produces more desirable results • Overview of Design Trade Studies • Methods, Tools, Groups Who Perform, Examples • Overview of Risk and Uncertainty • Risk quantification and mitigation tools • Uncertainty assessment methods • Uncertainty mitigation tools • • • • Possible ways to incorporate risk and uncertainty into Trade Studies Future areas of research Expected contributions of overall research Research plan Trade Studies in complex systems design Both design and decision tool Trade Studies attempt to find maximum system utility System utility defined by many metrics: • Cost • Return on Investment • System Reliability Multi-step process to perform a Trade Study • Understand system goals, objectives and constraints • • • • • • (Eg: Functional requirements) Develop alternative conceptual design solutions Evaluate alternatives based on system utility Use mathematical models where appropriate to determine system utility Rank alternatives according to their system utility Remove less desirable alternatives Either refine and continue to eliminate alternatives or select most desirable alternative Image Source: NASA Systems Engineering Handbook Trade Studies search for max system utility Many mathematical ways to find max utility Modern software packages available to find optimum design points • ICEMaker: Used by many Collaborative Design Centers to find optimum designs • Advanced Trade Space Visualization (ATSV): Used to graphically view and explore optimum design points ATSV Screenshots Image Source: https://webhosting.its.psu.edu/atsv/webfiles/glyphscatter /WebStart_files/image003.jpg Software (Con’t) • ModelCenter: Integrates capabilities of ATSV with ability to link together many different types of programs Many other types of software available that help perform Trade Studies Image Source: Jensen, et. al.: ME 519 Class Project Many CDCs exist in government organizations, academia, and industry Original is Team-X housed at NASA JPL • Helped NASA reduce time to finish Trade Studies from 3-9 months to 2-3 days • Reduced cost by a factor of five Other NASA facilities with CDCs: Langley Research Center, Goddard, Johnson Space Center Image Source: http://jplteamx.jpl.nasa.gov/images/teamx/team.jpg European Space Agency uses CDCs and Trade Studies Boeing, Aerospace Corporation, TRW, and other aerospace companies use Trade Studies Several academic institutions also use trade studies Many examples in literature of Trade Studies Most come from Team-X but some also from academic institutions Very few from private industry • Due to proprietary information, etc For those interested, long list of Trade Studies is available • Overview of Design Trade Studies • Methods, Tools, Groups Who Perform, Examples • Overview of Risk and Uncertainty • Risk quantification and mitigation tools • Uncertainty assessment methods • Uncertainty mitigation tools • • • • Possible ways to incorporate risk and uncertainty into Trade Studies Future areas of research Expected contributions of overall research Research plan Overview of methods to account for risk Overview of uncertainty and how to account for it in design process Note: many methods not reviewed here due to space and time constraints Many methods and tools available Some used in practice, some only in academia Practice • RBD, Databases, FMEA/FMECA , ETA, FTA, PRA, QRA Theory • FFDM, FFIP, RED, HiPHOPS, RUBIC, FFA Used for understanding fault tolerance Energy, information, or material flow through block diagram Image source: http://www.itemsoft.com/rbd.shtml Contain failure and reliability data on systems, subsystems, components, and processes Proprietary and industry-specific High amount of front-end work to have worthwhile database Often used in oil, automotive, and aerospace industries FMEA used to examine: • Potential failures modes • Effects of failures • Severity of the effects • Potential causes of the failures • Probability or potential probability of failure • Current detection methods of failure • Detectability of failure • Recommendations to mitigate cause or effects of failure FMEA also can be used to assign a Risk Priority Number RPN = Severity x Occurrence x Detection • Severity of each failure is rated • Likelihood of each occurrence is rated • Likelihood of prior detection is rated FMECA is an extension of FMEA. Adds criticality analysis to FMEA. Mode Criticality = Expected Failures X Mode Ratio of Unreliability X Probability of Loss Item Criticality = SUM of Mode CriticalitieS Image Source: http://www.weibull.com/basics/fmea_fig1.htm ETA is visual representation of failure events and mitigating events in a system Used in safety system analysis Starting point is failure event Subsequent levels show additional failures and mitigations Image Source: http://www.event-tree.com/images/et_example.JPG FTA starts with failure at top-level and proceeds down to analyze all possible causes of failure Boolean operators and logic gates used Image Source: http://www.isograph-software.com/ftpoverdgc.htm PRA is used to quantify the risk of failure in a system Employs FTA, ETA, and other techniques as desired PRA quantifies risk by magnitude and likelihood of each possible failure PRA is essentially an umbrella for several other risk methods Used when quantitative assessment is not possible • Not enough time, money, expertise Relies on expert opinions Usually performed by interviewing key designers to determine their belief in the level of risk of a design FFDM used to investigate potential failure modes during conceptual design Uses failure databases to find failure rates of generic components Improves on FMEA and related techniques Image Sources: Stone, Tumer, Van Wie: The Function-Failure Design Method FFIP estimates potential failures and their propagation paths through systems Three components to FFIP: • Graphical system model • Behavioral simulation • Reasoning scheme called Function Failure Logic Image Sources: Kurtoglu and Tumer: A Graph-Based Fault Identification and Propagation Framework for Functional Design of Complex Systems An extension of FFDM Quantifies risks identified in FFDM Automated process for combining historical risk data with new system architectures Uses fever charts to show risks Displays riskiest failure states Image Source: Lough, Stone, Tumer: Implementation Procedures for the Risk in Early Design (RED) Method HiPHOPS uses elements of FMEA, FTA, and others to assess risk in systems Model of system is annotated with formalized logical component failure descriptions and expected effects This method is too complex to ever gain widespread adoption A continuous risk management tool Used to identify risk elements during conceptual design RUBIC continuously optimizes budgetary resources to mitigate risks Graphical tool helps find Pareto optimal sets of resource allocations Image Source: Mehr, Tumer: Risk-Based Decision-Making for Managing Resources During the Design of Complex Space Exploration Systems FFA captures physical system architecture including connections of energy, material, and data flows in a functional model Model contains sensor information, failure modes of each component, propagation effects of failure modes, and propagation timing Approach requires high level of detail in system before it is useful State Event Fault Tree Analysis Component Fault Tree Analysis Simulation-Based Probabilistic Risk Analysis Component Stress and Conceptual Strength Interference Theory Various Bayesian Network Analysis tools Many others All try to identify and quantify risk All good for identifying riskiest points in designs In practice, lists of failures versus failure paths methods Most theoretical tools trying to find subsystem and component interaction risks • Overview of Design Trade Studies • Methods, Tools, Groups Who Perform, Examples • Overview of Risk and Uncertainty • Risk quantification and mitigation tools • Uncertainty assessment methods • Uncertainty mitigation tools • • • • Possible ways to incorporate risk and uncertainty into Trade Studies Future areas of research Expected contributions of overall research Research plan Definitions of uncertainty Assessing System Uncertainty Mitigating Uncertainty Image Source: http://www.martin-koser.de/images/enjoy%20uncertainty.jpg Many different ways to define uncertainty and many different places for it to be found Easiest to think of uncertainty as being made of many different types and falling into two categories Categories: • Intrinsic: Caused by randomness in nature • Epistemic: Caused by lack of knowledge or data Several ways to assess uncertainties: • Probabilistic Methods • Bayesian Techniques 1st, 2nd, 3rd level Bayesian Analysis Bayesian Team Support • Stimulation Methods Monte Carlo Methods These methods quantify the behavior of model uncertainties as a result of random model design input variables This allows engineers to find variables that are the most sensitive to change • Engineers then concentrate on these variables Based on Bayesian statistics and probability Probability interpreted as a state of knowledge Bayesian probability assumes that posterior probability is proportional to prior probability All Bayesian analysis based on Baye’s Theorem First Level Bayesian Analysis: • Used for creating system success rate probabilities based on past success and failure data Second Level Bayesian Analysis: • Used for systems with no prior data but that are similar to existing systems. Existing system data used Third Level Bayesian Analysis: • Same as Second Level but with normalizing available comparable systems data. Makes best estimate of future system success rates Bayesian Team Support helps to solve Arrow’s Paradox • Arrow’s Paradox is not being able to rank order a group’s voting choices (eg: A-B-C-A) BTS assumes all information is “Uncertain, incomplete, inconsistent, and evolving” BTS implemented in Accord software package BTS helps groups make decisions robustly by showing where information needs to be improved, etc… Image Source: Ullman. Making Robust Decisions Many techniques to simulate systems Most work by simulating system design or input variables • Use random or semi-random numbers Simulation Methods useful for when direct analytic solution is not available Monte Carlo Methods Class of computational algorithms Models with a high degree of uncertainty in input variables use MCM Model outputs calculated using random or well-chosen semi-random input variables with many repetitions • Large dataset is created to adequately explore design space Many different algorithms available • Overview of Design Trade Studies • Methods, Tools, Groups Who Perform, Examples • Overview of Risk and Uncertainty • Risk quantification and mitigation tools • Uncertainty assessment methods • Uncertainty mitigation tools • • • • Possible ways to incorporate risk and uncertainty into Trade Studies Future areas of research Expected contributions of overall research Research plan Natural Uncertainties: • Hard to address without changing natural environment • Eg: dikes to hold back rivers, oceans; bridges over bodies of water; snow sheds for rail lines… Model Structure and Parameter Uncertainties: • Usually addressed by refining model and parameters Data Uncertainties • Hard to discover and diagnose. If found then: Inaccurate measurements: Better measurement tools Data gauging network problems: Improve networks Data handling and transcription errors: Fix process Alternative approach: Relax requirements so uncertainties are acceptable Computational Uncertainties • Faulty hardware: Very hard to find and diagnose. If found, replace bad hardware. • Faulty software: Also hard to find. Fix software as needed. (Truncation and rounding errors can be fixed with longer integer registers, etc…) Operational Uncertainties • Use Total Quality Management, Human Factors Engineering, etc to reduce this uncertainty Behavioral and Ambiguity Uncertainties • Various corrective behavioral techniques available • Overview of Design Trade Studies • Methods, Tools, Groups Who Perform, Examples • Overview of Risk and Uncertainty • Risk quantification and mitigation tools • Uncertainty assessment methods • Uncertainty mitigation tools • • • • Possible ways to incorporate risk and uncertainty into Trade Studies Future areas of research Expected contributions of overall research Research plan No examples beyond Thunnissen of risk and uncertainty being traded in Trade Studies Thunnissen proposed a method to do this This method is rudimentary and needs to be expanded Thunnissen also proposed a method of design margins for Trade Studies. Also needs to be implemented. Thunnissen was influenced by Antonsson and others • Grayscale System Reliability: Quantifies influence of partial failure states on system integrity and reveals tradeoffs between system reliability and cost • Method of Imprecision: Represents uncertainty and preference on 0-1 scale. System to trade attribute levels without allowing any attribute to go to zero performance. Some interest at Team-X to implement • Thunnisson’s methods • Bayesian Team Support • Risk methods Work conducted at JPL this summer to integrate some methods into Trade Studies using ModelCenter (replacing ICEMaker) • Overview of Design Trade Studies • Methods, Tools, Groups Who Perform, Examples • Overview of Risk and Uncertainty • Risk quantification and mitigation tools • Uncertainty assessment methods • Uncertainty mitigation tools • • • • Possible ways to incorporate risk and uncertainty into Trade Studies Future areas of research Expected contributions of overall research Research plan The valuation and perception of risk and uncertainty is not well accounted for in the complex system design process • It is researched in insurance industry, stock market, and elsewhere. Need to move this knowledge into a workable form for engineers. Cultural factors that affect risk and uncertainty need to be better understood by engineers • Several methods in business to quantify culture. Need to move and adapt into engineering. Method to maximize system utility and integrity via expanding Trade Studies • Risk methods • Uncertainty methods • Design Margins • Risk and uncertainty perception and valuation • Cultural methods Integration of method with ModelCenter May 09 Complete PhD qualifier. Research design Trade Studies and risk and uncertainty. Prepare for summer internship at JPL. June-August 09 Work with Steve Wall and Team-X at JPL. Prepare first example of design margins in Trade Studies using ModelCenter for Team-X. September 09 Submit paper to Journal of Engineering Design detailing work at JPL. Sept-Dec 09 Revise design margins in Trade Studies tool. Start work on integrating risk and uncertainty into Trade Studies. January 10 Submit conference papers to IDETC and others. Jan-March 10 Take final required coursework. Continue work on integrating risk and uncertainty into Trade Studies. April-May 10 PhD Prelim. Publish journal paper on integrating risk and uncertainty into Trade Studies. June-Aug 10 Work either at JPL, overseas, or another company to continue work. Test risk and uncertainty methods in an industrial setting. Begin investigating perception and valuation of risk, and cultural constructs of risk. Sept-Dec 10 Finish work on integrating risk and uncertainty into Trade Studies. Implement valuation and perception metrics, and cultural standards. January 11 Submit conference papers to IDETC, ICED and others. Jan-March 11 Submit journal paper. April-June 11 Prepare and defend dissertation. Take long vacation. Why so focused on aerospace? Why worthy of a PhD?