USE OF GENETIC ALGORITHMS IN THERMAL-PROPERTY ESTIMATION - PART I - EXPERIMENTAL-DESIGN OPTIMIZATION

This two-part study is on the use of genetic algorithms (GAs) to desig n experiments and develop estimation methodologies for the determinati on of thermal properties; Part I is focused on the development of an i mproved GA and the implementation of this algorithm to optimize experi mental designs for the estimation of thermal properties, while Part II is directed toward the use of this algorithm in the estimation of the rmal properties. In Part I the methodology used in the improved GA, ca lled the extended elitist generic algorithm (EEGA), is presented, and results from two optimization test problems are compared with those ob tained previously from a basic elitist genetic algorithm (BEGA) and a parametric study. GAs are based on the genetic and selection mechanism s of nature, and the EEGA improves on the BEGA by enhancing the Darwin ian principle of the ''survival of the fittest.'' In the test problems , several key parameters in two experimental designs used for the simu ltaneous estimation of thermal properties were optimized. The results from these two test problems indicated that the computational efficien cy of the EEGA was much higher and the results were slightly better th an those of either the BEGA or the parametric study. Overall, genetic algorithms were found to be wed suited for use in the optimization of experimental designs for the estimation of thermal properties.