矿用卡车单位油耗的神经网络预测

Neural network prediction of unit fuel consumption of mining truck

利用TRAINCGF算法构建矿用卡车外部环境参数与卡车单位燃油消耗的BP神经网络预测模型。模型的输入信息为阶段产量、平均运距、平均高差、故障率、道路质量、司机操作、天气状况和日常维护,输出信息为单位油耗。BP神经网络模型为8—12—1结构,动量因子和学习因子分别为0.7和0.5。模型测试结果表明,相对误差最大值为4.5237%,相对拟合率值为0.9513,模型精度较高。该模型可为卡车油耗考核和油库进油提供参考。

Back--Propagation artificial neural network model with TRAINCGF algorithm is used to predict ex ternal environment parameters and unit fuel consumption of mining truck. In this model, production in a certain stage, average haul distance, average elevation difference, failure rate, road quality, driver operation level, weather conditions and daily maintenance are employed as input message while unit fuel consumption was employed as output message. The optimal network architecture is considered to be 8--12--1 with momentum factor chosen to be 0.7 and learning factor chosen to be 0.5, respectively. The test result shows that the biggest relative tolerance is 4. 5237%, and the correlation coefficient is 0. 9513, and the model has good accuracy. The result strongly indicates that the ANN model is a tool to provide reference for fuel consumption assessment and oil-taking.