ANALYSIS ULTIMATE BEARING CAPACITY ON BORED PILE WITH USING ARTIFICIAL NEURAL NETWORK

The issues that often arise within geotechnical engineering include uncertainty, complexity, and inaccuracies in planning. Therefore, this creates problems as relying on assumptions are the only way to determine parameters in design and construction. Recently, a new approach has emerged, inspired by the intelligence of the human brain, and it is called artificial neural network (ANN). This study aimed to utilize the ANN models with a back-propagation algorithm that feeds forward to predict the ultimate bearing capacity, namely NN_Qult. The total number of samples used are 375, and the input variables are d, Lp Le, A, K, f'c, Ntip, Nshaft, and P. According to Shahin (2001), the model was divided into two group: 2/3 training data and 1/3 validation data, processed in a modified ANN program. The prediction results of NN_Qult are then compared with the carrying capacity of pile driving analysis (PDA). It shows a good relationship, as evidenced by the value of R2> 0.8 and RMSE close to 0.1. The sensitivity analysis (AS) was also carried out to obtain the level of influence of the input compared to the output which are 12,367%; 10.255%; 14.576%; 8.323%; 15.870%; 5.154%; 8.218%; 14.314%; 10.923% respectively. The Le, Ntip and P variables are the most influenced of the dataset.