Automated design of fuzzy non-destructive testing system by a genetic algorithms

This paper proposes an application of a genetic algorithms (GA) to fuzzy non-destructive testing (NDT) system. We are concerned here with the system which extracts embedded tubes in concrete from pulse-radar images. The system usually uses several fixed parameters to analyze the input data. The optimal values of the parameters depend on the environment (e.g. the permittivity of concrete, pulse-radar device, and so on). Therefore, the system often fails to extract embedded tubes when the fixed parameters are not optimal in the environment. This paper attempts to optimize the parameters by a simple GA in order to solve the problem. We applied the system with the GA to the two data sets obtained by various environments. As the result, the system with GA was able to adapt each environment.