The use of genetic programming (GP) in nonlinear system identification enables the automated search for mathematical models that are evolved by an evolutionary process using the principles of selection, crossover and mutation. Due to the stochastic element that is intrinsic to any evolutionary process, GP cannot guarantee the generation of similar or even equal models in each GP process execution; still, if there is a physical model underlying to the data that are analyzed, then GP is expected to find these structures and produce somehow similar results. In this paper we define a function for measuring the syntactic similarity of mathematical models represented as structure trees; using this similarity function we compare the results produced by GP techniques for a data set representing measurement data of a BMW Diesel engine.
|Number of pages||8|
|Journal||Topics on Chaotic Systems - Selected Papers from CHAOS 2008 International Conference|
|Publication status||Published - Mar 2009|
- Nonlinear system identification
- genetic programming
- modeling of mechatronical systems
- results similarity estimation