基于ARIMA的盾构机液压推进系统数据预测方法研究

液压推进系统是盾构机的关键构成，承担着盾构机姿态控制、纠偏和同步前进等重要功能，以推进系统的运行数据为基础，精准预测数据的变化是分析、预测和避免盾构机产生安全问题的重要手段。基于随机时序分析法(Autoregressive Integrated Moving Average model, ARIMA)对盾构机液压推进系统数据进行预测研究。首先利用相关性分析方法，获得了与盾构机液压推进系统推进过程相关性较高的数据类别为掘进速度，基于该数据进行了自相关性的分析；之后，基于ARIMA方法，建立了盾构机液压推进系统ARIMA模型，并利用该模型进行了平稳性分析与贝叶斯信息准则；最后，基于优化模型分析比较了基于K-means的循环神经网络(Recurrent Neural Network, RNN)预测方法以及线性回归预测方法对数据预测的效果。研究表明，ARIMA模型下的线性回归方法能很好的预测盾构机液压推进系统数据变化趋势及异常数据预测，对盾构机的故障诊断及预测有重要的意义。

Research on Data Prediction Method of Shield Machine Hydraulic Propulsion System Based on ARIMA

The hydraulic propulsion system is the key component of the shield machine system, which undertakes significant functions in attitude control, rectification and synchronous advancement. To predict and avoid shield machine safety issues, accurate prediction of data changes based on the operation data of the propulsion is an important means to analyze. In this paper, In this paper, the data of the hydraulic propulsion system of the shield machine is predicted and studied based on the stochastic time series analysis method (Autoregressive Integrated Moving Average model, ARIMA). The high correlation data of the system category is the driving speed, and the autocorrelation analysis is carried out based on the data. Moreover, the ARIMA model of the hydraulic propulsion system is established, then the stationarity analysis and Bayesian Information Criterions (BIC). In the end, based on the analysis of the ARIMA model, the effect of the K-means plus RNN network prediction method and the linear regression prediction method on data prediction are compared. The results show that the linear regression method with the ARIMA model can well predict the data trend change and abnormality prediction of the hydraulic propulsion system of the shield machine, which has great significance to the fault diagnosis and prediction of shield machines.