Using genetic programming techniques for inertia-free system identification tasks

The problem of identification of inertia-free objects is being investigated. The overall pattern of the process investi- gated is being described. As a research object, a stochastic inertia-free process of modeling has been chosen. A feature of the process under consideration is the fact that unmanaged but controlled variables influence the process under investigation. In addition, the process under investigation is affected by unmanaged and uncontrolled variables. The levels of priori information has been briefly considered and characterized. For each level of priori information the identification method has been described. Particular attention is paid to the levels of priori information under which the identification task in a “broad” sense needs to be addressed. As a method of identification, genetic programming is considered. Method of genetic programming has been chosen as a research object since this method is more commonly used in the identification problem. Despite the frequency with which this method is used, it is interesting to look at the results of this method under different conditions. As changing conditions, the object’s complexity and the change in the volume of the training sample were used. For the identification process, objects with different structures were selected. The dependence of the time of finding the structure of the object on the size of the training sample was investigated. As shown by studies, there is no clear correlation between the time of finding the structure of the object and the size of the training sample. In addition, relationship between the time of the structure and the “complexity” of the object was investigated. As a criterion of “complexity” of the object, the number of input variables was taken. The study showed correlation between some values; with the increase in the number of input variables, the time of finding the structure of the process also increased.