基于混沌时序-随机森林回归的堆石坝料加水量预测研究

堆石坝碾压施工过程中一般通过对坝料加水以提高压实质量。然而现阶段针对坝料加水量控制的研究中，存在无法精确给定加水量、未考虑坝料上坝后摊铺施工过程中气象因素对坝料含水率的影响等问题，难以保证坝料在碾压施工开始时满足坝料设计含水率的要求。为有效控制堆石坝坝料加水量，基于混沌时序和随机森林回归方法，建立考虑气象因素影响的堆石坝料含水率变化预测模型并精确给定加水量，主要包括：①利用混沌时序方法，在设定时间内进行气象信息短期预测；②建立基于随机森林回归方法的含水率变化预测模型，预测气象因素引起的坝料含水率非线性变化，并通过十折交叉验证检验了模型预测精度；③结合预测的含水率变化、坝料初始含水率、设计含水率和坝料质量对加水量进行精确计算。在碾压监控系统和运输监控系统的基础上，结合某实际工程，验证了该方法的有效性和准确性，并与现有加水方法进行了对比，为堆石坝坝料加水量的精细化控制提供了科学指导。

Prediction of rockfill dam material watering volume based on chaotic time series and random forest regression

Watering dam material in the compacting operation of rockfill dams has been a general practice to improve compaction quality. However, in the previous studies of the watering control, several questions are left unanswered, such as how to control the water volume accurately and how meteorological factors influence dam material moisture content during the construction process. Consequently, operation at the compaction starting stage is difficult to meet the requirement of design moisture content. To control the watering volume effectively, this study develops a model to predict the change of dam material moisture content considering meteorological factors, using chaotic time series and random forest regression to accurately control the watering volume. It focuses on three aspects: (1) using chaotic time series to make short-term prediction of meteorological information in a given time period; (2) developing a prediction model of moisture content change based on the random forest regression to predict the nonlinear variations in moisture content caused by meteorological factors, and validating the prediction accuracy using the ten-fold cross-verification; (3) accurately calculating the watering volume based on the predicted moisture content, the initial moisture content of dam material, the design moisture content, and the mass of dam material. Based on a practical project and its rolling and transport monitoring systems, our method is compared with the existing watering technique and verified in terms of effectiveness and accuracy. It is thought to be useful to science-based fine control on the watering volume of rockfill dam material.