Multi-objective optimization of active vibration control systems by means of evolutionary algorithms and accompanying real-time simulationDonnerstag (23.05.2019) 12:10 - 12:30 Uhr Bestandteil von:
There is a continuing trend towards digitalization in the product design and development process. Numerical simulation tools speed up the development process and minimize the need for costly experimental evaluation. Faced with increasingly detailed multi-domain models and the ability to integrate results from multiple software tools, there is a need for automatized multi-parameter optimization.
In recent years, various evolutionary approaches for parameter optimization have been presented and implemented also in open source software libraries like the Distributed Evolutionary Algorithms in Python (DEAP) library. However, the evolutionary tools, such as mutation and crossover, are usually not specifically used and adapted for parameter variations in mechatronic applications in order to run a multi-objective optimization.
This paper illustrates the multi-objective optimization of an active vibration control system in an active suspension application applying a fuzzy control system. A Python-Matlab interface is set up and the core functions (i.e. evolution criteria, fitness evaluation and selection of individuals) of the genetic algorithm are adapted to take into account the requirements available in an early stage of the development process. Based upon a numerical simulation model comprising the car body, the active suspension system and the chassis, a real-time simulation hardware is set up in order to allow for a rapid evaluation of the overall system behavior. As the genes are directly connected to a physical meaningful parameter, this allows for a direct evaluation of a given parameter set using the real-time simulation hardware. The paper highlights the simultaneous optimization of the hardware requirements, control performance and estimated power consumption of the active control system by applying a multi-objective cost function.