Design of neural networks by using genetic algorithm for the prediction of immersed CBR index

The most important parameter of soil for the conception of flexible pavements is the California Bearing Ratio after immersion (CBRimm). This parameter is determined from laboratory testing, which requires skilled workforce and time. Based on parameters simply measured like Maximum Dry Density (MDD), Optimum Moisture Content (OMC), Liquid Limit (LL), Plastic Limit (PL) and the fine fraction passing at 0.08 mm and 2 mm (F 0.08 mm, F 2mm) we proposed a neuro-genetic model to predict the index CBRimm The aim to use the genetic algorithm is to evolve at the same time: The determination of the artificial neural network architecture, transfer function and the optimization of synaptic weights. Using a neuro-genetic approach helps to increase neural network performance and it gave us a minimal average absolute error.

کلیدواژه ها:

CBRimm ، Compacting ، Prediction ، Artificial neural network ، Genetic algorithm