Developing Automatic Multi-Objective Optimization Methods for Complex Actuators

• Main Authors: CHIS, R., VINTAN, L.
• Format: Article
• Language: English
• Published: Stefan cel Mare University of Suceava 2017-11-01
• Series: Advances in Electrical and Computer Engineering
• Subjects: actuators; computer aided engineering; machine learning; pareto optimization; response surface methodology
• Online Access: http://dx.doi.org/10.4316/AECE.2017.04011
• ISSN: 1582-7445; 1844-7600

This paper presents the analysis and multiobjective optimization of a magnetic actuator. By varying just 8 parameters of the magnetic actuator’s model the design space grows to more than 6 million configurations. Much more, the 8 objectives that must be optimized are conflicting and generate a huge objectives space, too. To cope with this complexity, we use advanced heuristic methods for Automatic Design Space Exploration. FADSE tool is one Automatic Design Space Exploration framework including different state of the art multi-objective meta-heuristics for solving NP-hard problems, which we used for the analysis and optimization of the COMSOL and MATLAB model of the magnetic actuator. We show that using a state of the art genetic multi-objective algorithm, response surface modelling methods and some machine learning techniques, the timing complexity of the design space exploration can be reduced, while still taking into consideration objective constraints so that various Pareto optimal configurations can be found. Using our developed approach, we were able to decrease the simulation time by at least a factor of 10, compared to a run that does all the simulations, while keeping prediction errors to around 1%.