基于BP神经网络的机械运输设备管理与回报率分析

Management of Mechanical Transport Equipment and Analysis of Return Rate by Using BP Neural Network

基于中冶宝钢某厂的24个月的机械运输设备运行数据,首先利用单层隐含层的BP神经网络预测下月回报率,提出了对模型的输入层、隐含层及其节点数的改进方案,保证模型拟合程度高的同时也确保训练的高效和快速收敛。经过改进模型的训练,仅用了150次左右的训练就达到期望误差0.0005。将影响回报率的设备,选取数据进一步筛选,并利用MATLAB进行占比分析,使工程设备的回报率逐年上升。结合运输设备回报率的波动性,解决高利用高风险率机械运输设备的设定标准,通过双高设备出现问题的因素分析提出以“检”代“修”、以“修”代“换”的机械设备管理要求。

Based on the single mechanical transport equipment report of MCC,by using BP neural network with a single hidden layer from MATLAB,the company’s return rate in next month is forecast.Propose a scheme that improve the input layer,the hidden layer and its node to ensure the high efficiency of training and the speed of convergence rate.After using the improved model,with only 150 training,the expected error reach 0.0005.Subsequently,further screening the selected data for equipment that affects the rate of return,combining with the fluctuant return rate of transportation equipment to set a screening standard for high utilization rate and high risk rate.Analyzing the problems arising from the emergence of dual high equipment to put forward the equipment management requires that replace‘repair’with‘inspection’and replace‘change’with‘repair’.