Looking into the Future of Design for Six Sigma (DFSS) • Description of past deployments • Comparison and observations • Suggestions for the future © Statistical Design Institute, LLC. All Rights Reserved. Jesse Peplinski January 16, 2012 Six Sigma Versus DFSS Define the Design Problem Capture the Voice of the Customer Identify Critical Requirements • You can use a flexible Improve or Create New Design approach to let each design problem dictate which process is followed • Use DFSS as a rigorous method for creating a design to satisfy multiple requirements © Statistical Design Institute, LLC. All Rights Reserved. Create Design Concept Measure the Requirements Build Math Models Analyze the Root Causes Optimize the Design Improve the Design Validate the Design Control the Root Causes MAIC data-driven method for design improvements DFSS • Use Six Sigma (MAIC) as a Select Approach Improve Existing Design Page 2 What is a “Deployment”? • A company-specific attempt to inject Six Sigma and/or DFSS into its culture and daily activities • Typically a customized mixture of: – Training classes with tailored content – Structure for projects and “belt” certification – Supporting software tools – Strategic communication by management and leadership • Scope of implementation can vary widely – All employees vs. targeted teams – Local vs. global © Statistical Design Institute, LLC. All Rights Reserved. Page 3 Past DFSS Deployments Company Description Status Automotive 1 • Global deployment • Mandatory training for all engineers • Projects and certifications Low level of activity Automotive 2 • Local deployment • Training and tools for selected experts based on role or skills Continued success Defense 1 • Emphasis on black belts and projects • DFSS as an afterthought to six sigma Low level of activity Defense 2 • Leadership evolved a design process intertwined with DFSS tools Continuing activity © Statistical Design Institute, LLC. All Rights Reserved. Page 4 Past DFSS Deployments Company Description Status Electronics 1 • Global deployment, mandatory training • Projects and certifications Low level of activity Electronics 2 • Local deployment for product teams • DFSS tools folded into an internal process excellence program Steady continuing activity Healthcare 1 • Global deployment with projects and certification • Significant backlash and years of inactivity Quiet resurgence through design reviews Healthcare 2 • DFSS integrated into development process • Emphasis on providing DFSS tools Continued activity © Statistical Design Institute, LLC. All Rights Reserved. Page 5 Observations • Pendulum swing – Larger, top-down deployments often end up with lower levels of long-term practice. • Backlash against projects and certification – Long-term health of deployment correlated with selective, low-key implementation • Challenge of demonstrating DFSS savings – Heroes get visibility for fixing mistakes; cost avoidance is difficult to recognize. • Tools stand the test of time – Six Sigma: Gage R&R, SOP’s, DOE, process control – DFSS: QFD, Pugh Matrix, Monte Carlo, Optimization © Statistical Design Institute, LLC. All Rights Reserved. Page 6 Suggestions for the Future • Design for Six Sigma: – DFSS tools fit naturally within a systems engineering group. (If you don’t have a systems engineering group, consider starting one.) – In addition, DFSS tools should be leveraged by your key participants in design reviews. (Principals, architects, etc.) – DFSS success hinges on modeling and simulation capability. Be prepared for resistance. • Six Sigma: – Let DMAIC flow naturally from leadership asking questions and demanding answers with data • Let plans for training and employee reward be driven by the forces above. (Not vice-versa.) © Statistical Design Institute, LLC. All Rights Reserved. Page 7 How does DFSS fit within Systems Engineering? Product Development Process Best Practice SE/DFSS Enablers & Tools Voice of the Customer Quality Function Deployment Exploration S E & D F S S Conceptual Design TRIZ & Design Selection Identify Critical Requirements Failure Modes & Effects Analysis Physics and First Principles Create Design Concept • First – use the Tools Build Models to Detail Design Design Verification Initial Production Final Production © Statistical Design Institute, LLC. All Rights Reserved. DOE and Regression Statistical Allocation and Monte Carlo support Sensitivity the Analysis Cost and Reliability Analysis Process Optimize the Design • Allocate Variability • Analyze Variability • Optimize Variability Validate the Design Multi-Objective Optimization FMEA & Fault Tree Analysis Test Effectiveness Analysis Design that best meets all requirements SE/DFSS Process Scorecards Page 8 Modeling and Analysis within DFSS Require that this be done everywhere, and if it isn’t, explain why not! Understanding Requirements, Specifications, & Capabilities Applying Models & Analyses Non-Compliance refers to any condition that results in Defects or Off-Spec conditions A B C D E Product Model (equation, simulation, workbook, hardware, etc.) Predicting Probability of Non-Compliance Y PNC “Noncompliant” LL “Compliant” T “Noncompliant” UL The fundamental metric is the Probability of Non-Compliance (PNC) © Statistical Design Institute, LLC. All Rights Reserved. Page 9 Modeling: Easier than It May Appear Key Design Parameters (X’s) Gather Design Parameter Information Can equations be developed? Yes Fast, Accurate Math Model No Yes Critical Requirements (Y’s) Identify Existing Models A simulation of sufficient accuracy exists? Yes No Simulation computes very quickly? No Best Design Alternative(s) Historical data exists? Yes Perform Regression Analysis No Create New Models © Statistical Design Institute, LLC. All Rights Reserved. No Prototypes exist? Yes Perform a Design of Experiments Page 10 Six Sigma Examples • What can we do to improve our process yield? Our goal is to get solid answers: ~ ~ It starts with hard problems: • How can we reduce • How can we increase the throughput of our call center? will reduce operating temperatures by 11 °C. ~ ~ increase sales volume? supplier B will improve yields by 8%. • This power supply redesign operating temperatures and fix our thermal issues? • What can we do to • Switching from supplier A to • A $50 rebate would increase sales by 15%. • Adding two more operators will increase throughput by 100 calls per day. How do we bridge the gap with high levels of confidence based on solid evidence? © Statistical Design Institute, LLC. All Rights Reserved. Page 11 Guiding Questions Answer these questions to bridge the gap: 1. What is our current state? – Product or process performance in measurable terms (Y’s) If we can’t measure it, we don’t know where we are. 2. What is our desired state? – How much improvement is needed in our measurable Y’s? If we can’t measure it, we can never know if we get there. 3. How good are our measurement systems? – If we measure the same thing twice, do we get the same answer? – If we made a process improvement, could we detect it? 4. What data do we need to collect? – Responses (Y’s) and Parameters (potential X’s) – How much data? Time period? Shifts? – Existing data? Or new data collection effort? © Statistical Design Institute, LLC. All Rights Reserved. Page 12 Guiding Questions Continued 5. If the Y is plotted versus the X’s, is there evidence of correlation (patterns) for some of the X’s? Which ones? – May begin to indicate the significant drivers for improvement 6. Is there statistical evidence that the Y changes when some X’s change? Which ones? – Type of analysis used (t-Test, F-Test, ANOVA, etc.) – Confidence level 7. What changes in the X’s are needed to achieve the desired state? Implement Six Sigma as a process for answering these questions. © Statistical Design Institute, LLC. All Rights Reserved. Page 13 Thank you… Questions? Contact: jpeplinski@stat-design.com © Statistical Design Institute, LLC. All Rights Reserved. Page 14