An approach to simulation-based parameter and structure optimization of MATLAB/Simulink models using evolutionary algorithms

In engineering, a broad range of environments exist for modeling and simulation with integrated parameter optimization. The established techniques only optimize model parameter values, the model structure is considered to be fixed. As system performance is optimized, one may have to redesign the model structure. The redesign is done manually by an analyst. The suboptimal combination of automatic parameter optimization and manual structural changes leads to an optimization task that is prone to error. This paper details an approach that provides optimization through automatic reconfiguration of both the model structure and model parameters. An optimization method that uses an evolutionary algorithm is supported by a model management method. This method is based on the system entity structure/model base framework. The admissible model structures and their associated model parameter sets are specified using the system entity structure ontology. Basic dynamic model components are organized in a model base. In addition to this, new algorithms are introduced. These map knowledge coded in the system entity structure to a set of numerical (structure) parameters, and also perform this mapping in reverse. In this manner a combined structure and parameter optimization problem is derived. Since both methods – evolutionary algorithm and model management – work together concurrently, different system configurations can be evaluated automatically. The objective is to provide an optimal solution; a model optimized for both parameter and structure.