ESTIMATION OF TEMPERATURE-DEPENDANT THERMAL CONDUCTIVITY OF MATERIAL WITH GENETIC ALGORITHM

摘要

Estimation of temperature-dependant thermalconductivity of material is a typical inverse heatconduction problem. The temperature range canbe divided into several segments and thethermal conductivity is linearly interpolatedfrom the thermal conductivities on thesesegment points whose values are to be estimatedby Genetic Algorithm(GA) method from somemeasurement information. However, undersome circumstances such as the measurementnoise being significant, some unphysical resultsof significant oscillation may come out of theestimated result. So, in this paper, physicalconstraints that the thermal conductivity valueshould be greater than zero and not decreasewith the temperature rise are taken into theestimation method of GA with two strategies.One is to estimate the non-negative incrementalvalue of the thermal conductivity with thetemperature rise, the other is to implement thepenalty function in the estimation process. Fromthe numerical results of examples, it can befound that the constrained estimation methodsof both strategies are feasible and effective, andthe estimated values of both strategies areagreeable and generally better than theconstraint-free result. The estimated resultreflects the basic temperature-dependantcharacteristic of the material well and themethod is of good prospect in engineeringpractices.