Hybrid Genetic Algorithm ― Rule-based System Solution for Optimizing SMT Placement Machines

摘要

Analyses of surface mount technology (SMT) assembly lines have shown that automated placement machines are often the bottleneck. Thus, optimizing the setup and placement sequence is a key issue for optimizing SMT production line throughput. In the research literature, genetic algorithms (GAs) have been proposed as a solution to this problem. However, exclusively using GAs does not take advantage of existing domain knowledge. In this paper, we present a new hybrid genetic algorithm ― rule-based system (GA-RBS). Domain knowledge, represented in the form of a rule-based system, enables the GA-based optimizer to find better solutions in a shorter computation time. Data mining techniques have been applied to automatically discover underlying knowledge from observed data, e.g., the simulated cycle times achieved with various settings of execution/iteration parameters of the genetic algorithms for the optimizer. Experimental results with a set of actual production programs indicate the potential for significant performance improvement by automating the control and tuning of genetic algorithms using a rule-based system.