Model-Based Robust Design

Robust design encompasses the theories and methodologies that make performance measures (responses, critical times, energy, etc.) invariant to uncertainties in design variables (environmental conditions, manufacturing processes, material dimensions and properties, etc.). In this course you will learn how robust design methods and mathematical models of engineering systems help find improved designs at a lower cost. Topics include: the building of simple, efficient, meta-models through computer experiments to replace traditional mechanistic models. Performance measures and their design specifications. Sensitivity and importance analysis to select design variables. Second moment methods using Taylor series to provide parameter (mean) design. Probabilistic methods combined with manufacturing and scrap costs to perform simultaneous parameter design and tolerance allocation. Desirability functions and loss functions to deal with multiple competing performance measures. Integrated design by constrained optimization. Examples come from industrial processes, as well as hydraulic, electrical and mechatronic systems. Mathematics required: Total derivatives, Matrix calculus, Kronecker product, singular value decomposition. Matlab serves as the computing tool. Course notes are available.