基于深度神经网络的微扰动压入式沉井顶压力智能预测方法

Intelligent Prediction Method for Micro-Disturbance Press-In Open Caisson Top Pressure Based on Deep Neural Network

摘要: 为了提高沉井施工精度，提出了一种基于深度神经网络的微扰动压入式沉井顶压力智能预测方法。该方法针对沉井下沉状态参数与油缸顶压力之间的强非线性、高维度复杂关系，建立深度神经网络预测模型，对微扰动压入式沉井的油缸顶压力进行智能预测。实验结果表明，在相同的作业条件下，与LSTM和LSSVR模型相比，本文建立的深度神经网络预测模型对油缸顶压力的预测精度更高，能有效修正沉井下压过程姿态偏差和提高沉井施工精度，为压入式沉井姿态控制提供了可行的施工保障。

Abstract: In order to improve the accuracy of open caisson construction, a deep neural network-based intelligent prediction method of top pressure of micro-disturbed press-in open caisson is proposed. Aiming at the strong nonlinear and high-dimensional complex relationship between open caisson sinking state parameters and cylinder top pressure, the method establishes a deep neural network prediction model to intelligently predict the cylinder top pressure of micro-disturbed press-in open caisson. The experimental results show that under the same operating conditions, the deep neural network prediction model established in this paper has a higher prediction accuracy of cylinder top pressure compared with LSTM and LSSVR models, which can effectively correct the attitude deviation of sinking process and improve the accuracy of sinking construction, and provide feasible construction guarantee for the attitude control of press-in open caisson.