基于BP神经网络的装载机自动铲装作业油耗预测研究  

Research on the prediction of fuel consumption of automatic shoveling operation of loader based on BP neural network

作　　者：谌炎辉 蒋恒[2] 周涛 郑特[2] CHEN Yanhui;JIANG Heng;ZHOU Tao;ZHENG Te(School of Electrical and Mechanical Engineering,Guangxi Vocational College of Water Resources and Electric Power,Nanning 545023,Guangxi,China;School of Mechanical and Automotive Engineering,Guangxi University of Science and Technology,Liuzhou 545006,Guangxi,China)

机构地区：[1]广西水利电力职业技术学院机电工程学院,广西南宁545023 [2]广西科技大学机械与汽车工程学院,广西柳州545006

出　　处：《中国工程机械学报》2024年第3期379-383,共5页Chinese Journal of Construction Machinery

基　　金：广西自然科学基金资助项目(2015GXNSFAA139271);广西科技开发计划资助项目(桂科能1598021-2)。

摘　　要：针对装载机自动铲装过程中作业能耗大、难以优化以及测试繁琐的问题,运用BP神经网络对作业油耗进行了研究。首先装载机根据规划出的7条平行铲装轨迹进行自动铲装测试试验,通过传感器获取相关试验数据作为BP神经网络的训练和验证样本;然后建立BP神经网络预测模型,对该模型的误差进行了验证;最后对该模型预测出的非训练铲装轨迹的作业油耗与测试试验的作业油耗进行了对比,以此验证BP神经网络模型的准确性。研究结果表明:BP神经网络预测模型的预测值与与验证样本的试验值的平均偏差仅为6.05 mL,平均误差仅为5.46%;预测平行铲装深度为400、600 mm,这两条非训练铲装轨迹的作业油耗时,其预测值与试验值的误差分别为2.30%~8.36%、2.67%~8.18%;因此,建立的BP神经模型能够较为准确地预测装载机自动铲装油耗,为作业油耗的优化提供依据。Aiming at the problems of high energy consumption,difficulty in optimization and tedious test in the process of automatic shovel loading of loader,the oil consumption in operation was studied by using BP neural network.Firstly,the loader carried out automatic shoveling test according to the planned 7 parallel shoveling tracks.The relevant test data were obtained by sensors as training and verification samples of BP neural network,and then the prediction model of BP neural network was established to verify the error of the model.Finally,the operating fuel consumption of the non-training shovel trajectory predicted by the model is compared with that of the test,so as to verify the accuracy of the BP neural network model.The results show that the average deviation between the predicted value of the BP neural network model and the experimental value of the verified sample is only 6.05 mL,and the average error is only 5.46%.When predicting the operation fuel consumption of parallel shoveling depth of 400 mm and 600 mm,the errors between the predicted value and the experimental value are between 2.30%-8.36% and 2.67%-8.18%,respectively.Therefore,the established BP neural model can predict the automatic shovel fuel consumption of loader more accurately,which can improve the optimization degree of operation fuel consumption.

关 键 词：BP神经网络 铲装轨迹 油耗 装载机 

分 类 号：TH243[机械工程—机械制造及自动化]