Genetic Optimization of Multidimensional Technological Process Reliability

A technological process (TP) is considered multidimensional if a number of defects of diverse types occur, and they are detected and corrected simultaneously within the process execution [4, 5]. Quality of the TP is estimated by the probability of output zero-defects as well as by the probabilities of zero-defects for each of the defect types. The tasks of TP-optimization involve the choice of such a process structure that will provide the necessary output level of product quality given some certain cost limits [5]. The typical example of such optimization tasks is optimal choice of multiplicity of control-retrofit procedures in a TP. This particular optimization problem is studied in this article. An initial perspective would view this problem as one that may be solved by using known mathematical programming methods. However, taking into account that defects of many diverse types increase the dimensionality of the state space. It means that using classical mathematical programming techniques becomes impractical. Therefore, in this article, the task of TP optimization is solved by using genetic algorithms (GA) [2, 3], which allow to find nearly global optimal solution, and additionally do not require much mathematical backgrounds about optimization. Principal