The Prediction of Pile Foundation Buried Depth Based on BP Neural Network Optimized by Quantum Particle Swarm Optimization

Due to the fluctuation of bearing stratum and distinct properties soil layer, buried depth pile foundation will differ from each other as well. In practical construction, since designed length is not definitely consistent with actual length, masses piles be required cut off or supplemented, resulting in huge cost waste potential safety hazards. Accordingly, prediction great significance construction engineering. this paper, a nonlinear model based on coordinates was established by BP neural network predict samples evaluated, consequence which indicated that easily trapped local extreme value, error reached 31%. Afterwards, QPSO algorithm proposed optimize weights thresholds network, showed minimum QPSO-BP merely 9.4% predicting 2.9% foundation. Besides, paper compared three robust models referred FWA-BP, PSO-BP, statistical tests (RMSE, MAE, MAPE). The accuracy highest, demonstrated superiority