Optimization of Roll forming process using the integration between Genetic Algorithm and Hill climbing with neural network

Abstract

Knowledge-Based Neural Network (KBNN) model is one of the most useful methods which is used to predict every single variability to perform the parameters on data of the Roll forming (RF) process. It is true that the quality of product and the parameters in RF process depend on the reliability of the training in KBNN. To achieve this, the new novel of the optimal algorithm including integration between Genetic Algorithm (GA) and Hill climbing Algorithm (HCB) was proposed to train the KBNN model. Initially, the GA is applied to find the local optimal region, then, the HCB will detect the best location area in which the training error of the KBNN model is less than 8%. In addition, the Finite Element Analysis (FEA) results of the high fidelity FE model were used to obtain the trained data set of the KBNN model. From simulation results, it can be concluded that the efficiency of the proposed method is higher than that of the conventional methods in optimization of the RF process.