A cloud-based large-scale capable multi-objective optimization framework for model-based engine control unit (ECU) transient calibration
2022-01-0488
03/29/2022
- Event
- Content
- For advanced powertrain solutions, there are increasing number of powertrain components and related ECU control set point variables to be calibrated in order to fulfill stringent emission regulations while maintaining customer satisfaction and product desirability. This imposes challenges to engineering deliverable time and resources. Utilizing combinations of simulation and optimization algorithms have become an inevitable trend. However, most intellectually patented simulation software are black box based, and thus difficult to be utilized with highly efficient deterministic optimization algorithms, many of which require explicit mathematical formulations. Additionally, surrogated based optimization and heuristic algorithm may be efficient to solve black box problems but they tend to either infeasible to deal with or converge slowly on large scale problems. In this study, the authors proposed a large scale optimization solver framework called "weighted-optimization framework with NSGA-II (WOF-NSGA-II) ", coupling the weighted optimization framework(WOF) with Non-dominated Sorting Genetic Algorithm (NSGA-II), this in-house developed optimization solver framework (WOF-NSGA-II) solved a large scale engine transient ECU calibration problem with 3150 control set point parameters. The simulation models involved an engine model developed in GT-Power-xRT and ECU control model developed in Mathworks Simulink. The in-house optimization solver was implemented in Python, and GT-automation enables the simulations and the solver to execute on Amazon Web Services (AWS Cloud). This solution provides large scale capable computational resources while it maintains on-demand flexibility. This work proposed workflow greatly shortened the simulation deliverable time and provided reliable solutions for multi-objective model based engine calibration and achieved good results to obtain engine BSFC and BSNOx trade-off pareto.
- Citation
- Wei, Y., and Sun, Y., "A cloud-based large-scale capable multi-objective optimization framework for model-based engine control unit (ECU) transient calibration ," SAE Technical Paper 2022-01-0488, 2022, .