Abstract:During deep well filling, the distance of the filling slurry transportation pipeline is long, and the rheological characteristics of the slurry transportation are complex. It is difficult to grasp the normal operation status of the filling pipeline in real time, and it cannot provide comparative data of the normal status of the pipeline for pipeline fault warning. It is also difficult to make quick judgments when pipeline faults occur. To predict the operation status of pipelines, the gold mine filling pipeline is taken as the research object, and industrial data such as pipeline pressure is collected. The original dateset is preprocessed by combining the improved SMOTE algorithm and mixed sampling to improve the random forest algorithm. A pipeline key point pressure prediction model is established, and the model fitting goodness can reach 0.978, with high accuracy, which can be well applied to predict the pressure change law of filling pipeline key points and provide a basis for fault warning.