Estimating the crashworthiness performances of crushboxes using artificial neural network

Abstract Studies on the development of energy absorbing systems that minimize vehicle chassis damage in traffic accidents are increasing day by day. Many designs have been made in the studies on crushboxes used to absorb the energy released in the event of an accident. These design works are quite costly and take a long time. In this study, to design crushboxes faster and more economically was estimated using artificial neural network. The input layer of the artificial neural network model consists of three different materials, thicknesses (between 0.8 and 2.2 mm) and three different initial speeds. In the artificial neural network model, 42 different models were created by changing the different training functions (training, trainlm and trainrp), transfer functions (tansig and logsig) and the number of neurons in the hidden layer (between 9 and 33). R 2 and root mean square error (RMSE) methods were used to evaluate the efficiency of artificial neural network models. The training function was found to be highly accurate (R 2 : 0.99999 and root mean square error: 0.314727E‐05) when the training function was “trainlm” and the number of neurons in the hidden layer was 33. The training and testing results of the artificial neural network model show that artificial neural networks can be used to estimate the specific energy absorption/energy/peak crush force value of crushboxes.