Classification of pile foundation integrity based on convolutional neural network

Pile integrity is a comprehensive qualitative indicator reflecting the relative change of pile section size, the compactness, and continuity of pile material. The evaluation of pile integrity is a systematic and comprehensive evaluation. However, manual detection has some defects, such as cost high, low efficiency, and strong subjectivity. In order to realize the automatic classification of pile integrity, this paper proposes a method of pile integrity classification and recognition based on convolutional neural network (CNN), including one input layer, four convolutional layers, four pooling layers, two fully connected layers, and an output layer. The stochastic gradient descent algorithm and overfitting prevention technology are used to improve the model. The experimental results show that the proposed model can effectively achieve the classification of pile integrity, with an average accuracy of 98.58% on the test set, and has good robustness and generalization performance, which overcomes the shortcomings of complex operation, high cost, and strong subjectivity of artificial extraction feature.