Design Synthesis via HPD Methodology and Software
(Updated from the abstract of an early talk given to MIT's CIPD** & Engineering Faculty)
For Electro-Mechanical product development, almost all characteristics that need be considered have Variability; thus Design Synthesis IS the activity of ultimately making ("best") decisions on the Nominals & Tolerances of "parts" and operating characteristics that would enable meeting product performance and cost requirements.
But there is a huge gap between engineering curricula and how engineering is practiced. The former enables a deterministic Design, while the latter is ~ 80% on addressing Variability for any clean-sheet Design.
Achieving Fit & Function, against constraints and Variability, is engineering's objective; achieving Function is by far more difficult & time-consuming. Functional Tolerancing is one part of that work content; the other related part is finding the "best" (or adequate) design/operating set point (defined by the Nominals) – in the sense of having robustness and latitude. Existing methodologies, tools and techniques for these are rudimentary, inadequate, and (some) even erroneous (e.g., even for Sensitivity/Contribution Analysis!).
Initiated at Xerox after a major ‘90s study concluded that the Corporation must address this deficiency, HPD was to enable engineers to appropriately, coherently, and pervasively deal with Variability during product synthesis. The HPD Methodology and the 2 comprehensive & user-friendly Software Suites, HPD_VA and HPD_Opt (which enable executing the methodology), comprise a breakthrough capability for Design Engineering:
- The Methodology guides how to: structure the Stochastic problem; "model;" conduct experiments; analyze; optimize; approximate; trade off within and across entities; treat local and global problems; . . . ("Model" includes "a Total Stochastic Model")
- HPD_VA is for the Umbrella area of Variability Analysis (which includes conducting Tolerance Analysis & Sensitivity/Contribution Analysis and computing Failure Probabilities and resulting distributions, . . .)
- HPD_Opt is for finding Operating Windows and optimizing for Nominals, comprehending all Variability and target/constraints
HPD rigorizes, yet simplifies, and goes far beyond what exists. As importantly, HPD enables unifying, simplifying, and clarifying work processes, techniques, and terminologies, thus additionally impacting engineering productivity! HPD should become the 21st Century work process for Design Engineering.
HPD applications (including to a diverse set of projects other than Design Engineering) have shown its great power, handling many complex hitherto unsolvable problems, and have brought much revelation!
2019 Question: Shouldn't Engineering curricula include a course based on HPD since (i) the new mathematically rigorous methods incorporated are novel to academia, yet can be understood by students, (ii) its concepts are so unifying, and (iii) it greatly simplifies the understanding of dealing with Variability which is so important for engineering a product?