Application of bacterial colony chemotaxis optimization algorithm and RBF neural network in thermal NDT/E for the identification of defect parameters

The identification of defect parameters in thermal non-destructive test and evaluation (NDT/E) was considered as a kind of inverse heat transfer problem (IHTP). However, it can be further considered as a shape optimization problem, and then a structure design optimization problem, and the design results should meet the surface temperature profile of the apparatus with defects. A bacterial colony chemotaxis (BCC) optimization algorithm and a radial basis function (RBF) neural network (NN) are applied to the thermal NDT/E for the identification of defects parameters. The RBFNN is a precise and convenient surrogate model for the time costly finite element computation, which obtains the surface temperature with different defect parameters. The BCC optimization algorithm is derivatively-free, and the convergence speed is fast. Then a simple but complete multi-disciplinary design optimization (MDO) framework is constructed for the sake of generality and flexibility. This method is applied to a simple verification case and the result is acceptable. The algorithm is also compared with the particle swarm optimization (PSO) algorithm, and the BCC algorithm can access the optimum with faster speed.