Concepts NREC公司多学科优化能力介绍.pptx

MULTIDISCIPLINARY OPTIMIZATION (MDO);Copyright Statement;How does optimization help you? Optimization approaches Traditional single objective optimization Simulated Annealing (1960’s) More recent search strategies Multiple objective optimization – Pareto front Parallel optimization Robust design optimization Optimizers supported by Concepts NREC Optimization case studies – pump and compressor ;How does optimization help YOU?;Manual Process: Limited Picture;Optimization Process: Better Road Map;Optimization in Integrated Agile Engineering Design System?;TurboOPT II?;Optimization Approaches;Any variable/parameter accessible via Concepts NREC’s application interface (+ a few extras) User must determine which inputs can vary, within what range Care must be taken to choose a reasonable range on geometrical inputs (Bezier points/thickness curves) for grid generation Manufacturability limits Application (PbCFD) may crash with certain inputs!;Maximize efficiency, head rise; minimize maximum stress Design constraints Max lean 45 degrees, impeller tip deflection 0.1 mm Structural constraints Modal frequency 12 kHz Manufacturability constraints Minimum hub separation 0.025 mm Solver constraints (complete, converged) Completed iterations 1000, mass error 1 e-4 kg/s Noise constraints (improbable answers) EFF TT 0.98 Oscillations in CFD calculations (unsteady flow);Multiple objectives summed (scaled/weighted) Minimize objectives * (1) Maximize objectives * (-1) Constraint violations calculated (Lower/upper bound scaled/weighted) ;Penalty function from all violations Optimizer seeks to minimize sum of (1) objective + (4) penalty All objectives, all constraints merged to single function to be minimized ;Local Search;Global Search;Simulated Annealing;Effect of Temperature on SA Search;Deterministic (systematic/geometrical) Cutting plane Stochastic (random variables) Simulated annealing Monte Carlo Heuristics and metaheuristic (mimics) Evolutionary (genetic) Swarm-based Re