摘要
基于BP神经网络非线性求解模型,结合冲击试验结果得到铁路罐式集装箱冲击试验参数与冲击响应谱的映射关系,预测了集装箱冲击时应采取的试验速度以及获得的冲击响应谱。通过实例,统计了不同罐式集装箱试验时的试验台座质量、冲击速度、冲击响应谱峰值和罐式集装箱的自重、容积、额定质量、实际试验质量,并分别建立了冲击速度和冲击响应谱的预测模型;结合BP神经网络进行了训练学习,当训练步数分别为18179步和6231步时,网络模型达到误差要求,基于这2种模型得到了既能够使铁路罐式集装箱满足冲击标准要求又能使其受到最小破坏的冲击速度,并与真实值进行了对比分析,分析结果显示,2种BP神经网络模型的输出样本与真实值十分接近,冲击速度模型预测最小误差为0.08%,冲击响应谱模型预测最小误差为1.39%,表明模型具有良好的精度。
Based on the nonlinear solution model of BP neural network and combined with the test data,the mapping relationship between the impact test parameters of railway tank container and the impact response spectrum is obtained,and the mapping relationship is used to predict the test speed that should be taken when the container is impacted,as well as the impact response spectrum to be obtained.In this paper,the test bench mass,impact velocity and peak value of impact response spectrum of different railway tank container tests are counted,as well as the self weight,volume,rated mass and actual test mass of railway tank container.The prediction models of impact velocity and impact response spectrum are established respectively;Combined with BP neural network for training and learning,when the training steps number are 18179 and 6231 respectively,the network model meets the error requirements.Based on the two models,the impact velocity that can make the railway tank container meet the impact standard requirements but also make it suffer minimized damage is obtained,and the real value is compared and analyzed with it.The analysis results show that the output samples of the two BP neural network models are very close to the real value,and the minimum prediction error of impact velocity model is 0.08%,the minimum prediction error of impact response spectrum model is 1.39%,which indicates that the model has good accuracy.
作者
王玉伟
阎锋
刘鑫
WANG Yuwei;YAN Feng;LIU Xin(CRRC Qingdao Sifang Rolling Stock Research Institute Co.,Ltd.,Qingdao 266031,China)
出处
《铁道车辆》
2022年第6期88-93,共6页
Rolling Stock