Machine Learning-Based Driving Style Identification of Truck Drivers in Open-Pit Mines

Abstract

The significance in constructing a driving style identification model for open-pit mine truck drivers is to reduce diesel consumption and improve training. First, we developed a driving behavior and mining truck condition monitoring system for an open-pit mine. Under heavy-load and no-load conditions of a mining truck, based on the same experimental truck and haulage road, the data of driving behavior and truck status of different drivers were collected. The driving style characteristic parameters of mining trucks under heavy-load and no-load conditions were constructed through Pearson correlation analysis. Using a k-means clustering algorithm, driving style can be divided into three types: normal type, soft type, and aggressive type, and we verified the validity of this driving style classification with a box plot. On this basis, the parameters of random forest, k-nearest neighbor, support vector machine, and neural network models were optimized and the accuracy was compared through a cross-validation grid search, and then a driving style identification model based on the random forest method was finally proposed. Driving style parameter weight values were obtained based on the Gini coefficient. Last, the fuel consumption characteristics of different driving styles were calculated. The results show that the driving style identification models based on random forest can effectively identify different driving styles when the mining truck is operating under heavy load and no load, and the overall accuracy of the model is 95.39% and 90.74% respectively. The fuel consumption of the aggressive driving style was the largest and was 10% higher than the average fuel consumption. The research results provide data support and new ideas for operation training and fuel-saving driving of mining trucks in open-pit mines.