基于分位数回归的升船机变形监控模型构建方法

为深入分析升船机变形影响因素，提出了一种基于分位数回归的升船机变形监控模型构建方法。该方法根据升船机的结构特点，将温度、前期上游水位均值等因素引入候选影响因子集，采用自适应弹性网络分位数回归对影响因子进行筛选，建立各分位数下的回归模型，并根据拟合的良好性和检验的有效性原则选出最优的升船机变形监控模型。实例验证结果表明：相对于常规的逐步回归模型，本文方法构建的最优模型的预测精度波动性小，具有较强的稳定性，同时具有良好的长期预测能力。

Construction method for ship lift deformation monitoring model based on quantile regression

To deeply analyze the influence factors of ship lift deformation, a method for building ship lift deformation monitoring model based on quantile regression is proposed. Based on the structural characteristics of the ship lift, the influence factors such as air temperature and mean value of upstream water level in early stages are introduced into the candidate influence factor set. Adaptive elastic net penalized quantile regression is used for choosing the influence factors of ship lift deformation and building the quantile regression model. The optimal model can be obtained based on the principle of good fitting and validity check. Example verification shows that compared with traditional stepwise regression model, the proposed optimized model has low volatility for prediction accuracy with strong stability, indication a good ability for long term prediction.