Design for SixSigma(DFSS) & Design for Reliability(DFR) 六西格玛设计和可靠性设计六西格玛设计和可靠性设计 Intro BE July 2010 Page 2 The Journey 1998 Seagate adopts Six Sigma defect reduction, cost savings 2001 DFSS in Product & Process Development 1999 Lean in Manufacturing & Supply Chain Intro BE July 2010 Page 3 DFSS in the Beginning Iterative Use of historical requests Test and re- test Short term estimates Isolated CTQ optimization Predictive Requirements hierarchy Model building Long term estimates System optimization Initial Approach: Top down Educate the masses in design centers - “DFSS Certified” DFSS Foundation 2 weeks of Statistics DFSS Project Systems Engineering 3 days Train the suppliers and factory BrB/BB/MBBs in DFSS Intro BE July 2010 Page 4 Allows us to set “need- based” requirements for CTQs and to evaluate our capability to meet those requirements. What Is Design for Six Sigma? Design for Six Sigma (DFSS): Is a process that focuses on predictive product design. Emphasizes the use of statistical methods to predict product quality early in the design process. Is a complement to good engineering/decision making practices. Intro BE July 2010 Page 5 Six Sigma Improvement Methodology 1.A high level Business need is identified(CTQ gap) 2.Does a Current Business Process/Product exist to address the gap 3.Are the Processes/Products that support your key outputs optimized but still not capable of meeting customer requirements? 4.Is the solution or part of the solution a new process, product, or service. 5.Does the capability of one or more KPIV need to be improved to optimize KPOV? Identify Design Optimize Define Analyze Improve Control Measure Validate 1 A 2 NO YES 4 3 5 NO YES YES NO YES NO A Intro BE July 2010 Page 6 Breakthrough Feasibility Point Tollgate Exception Review Perform Tradeoffs to Ensure that All CTQs Are Met OK Not OK Not OK Not OKValidate Optimize Design Identify OK Translate Into Critical To Quality (CTQ) Measures and Key Process/Product Output Variable (KPOV) Limits Formulate Designs/Concepts/Solutions Evaluate Designs For Each Top Level CTQ, Identify Key Product/Process Input Variables (KPIVs) Identify Customer Requirements Develop Transfer Functions Between Key Input and Output Variables Assess Process Capability to Achieve Critical Design Parameters and Meet CTQ Limits Optimize Design DFSS Scoring Determine Tolerances Test & Validation Assess Performance, Failure Modes, Reliability and Risks Validate The Measurement Systems Not OK Exception Review OK Perform Tradeoffs to Ensure that All CTQs Are Met Not OK Statistical Design Intro BE July 2010 Page 7 Six Sigma and Design for Six Sigma Design for Six SigmaDesign for Six Sigma andand “Standard” Six Sigma“Standard” Six Sigma work together!work together! Defects Lower Spec Limit Upper Spec Limit Design for Six Sigma DMAIC Six Sigma Focus on reducing variation around the mean Design robust products so that specs can be loosened Intro BE July 2010 Page 8 Design Evolution Evolving Design requirements Design rework Build and test performance assessment Performance and manufacturability after product is designed Quality is “tested in” FROMTO Disciplined CTQ flow- down Controlled design parameters Performance modeled and simulated Design for robust performance and manufacturability REACTIVEPREDICTIVE Intro BE July 2010 Page 9 Key Elements Systems relationships Transfer Functions, KPIV & KPOV Statistical Design: Meeting not only target but address variations in design Identify, Design, Optimize, & Verify (IDOV) Intro BE July 2010 Page 10 Systems Engineering -Flowdown System CTQs Subsystem CTQs Sub- assembly CTQs Components CTQs Process CTQs QFD/FMEA Intro BE July 2010 Page 11 Systems View Of a Hard Disc Drive Customer CTQs Elec/Interface Servo- Mech RSS- H/M 38 CTQs 111 Subsystem CTQs Component CTQs . . . Motor/Base HDA Encl. Media Head HGA HSA Channel/Preamp Mech Servo Firmware ASIC 120 Factory CTQs Cert/Test Assembly/Test 7 CTQs Process CTQs Intro BE July 2010 Page 12 Transfer Function What is a Transfer Function? f(X1,X2, Xn) X1 X2 X3 Xn Y It is a relationship of the CTQ (Y) to the key input variables (Xs). It is not necessarily as rigorous as a process model. It is key to predicting product performance before building prototypes. Intro BE July 2010 Page 13 Getting to the y = f(x1, x2) “All models are wrong, some are useful.” -George Box Physical Models -dedicated experts Explore design space run simulations with DOE Model management process Statistical Models DOE, Regression, Response Surface, etc Parametric data analysis especially for reliability MSA Intro BE July 2010 Page 14 Flowdown/Flowup Process Identify Customer CTQs. Translate into System CTQs. Identify Measurement for each system CTQ. Determine Specifications for each system CTQ (Y). Identify Transfer Function Y=f(x1,x2,xn) System Trade off mean/variance requirements to x1,x2,xnto best meet system CTQ need. Use transfer function and