基于深度卷积-长短期记忆神经网络的整车道路载荷预测

Prediction of Vehicle Road Load Based on Deep Convolution Neutral Network-Long-Short Term Memory

针对传统道路载荷获取方法耗费周期长、成本高,且无法高效应用于整车参数变化后的新车型的问题,利用现有多款车型的载荷数据库,通过建立可确定整车结构参数、运行工况与道路载荷之间关联关系的深度卷积-长短期记忆神经网络(DCNN-LSTM)模型,提出了基于数据驱动的整车轮心载荷预测方法。对比试验结果表明,该方法预测的整车轮心载荷与试验场采集数据非常接近,有利于逐步取消路谱采集试验并极大地提高整车耐久性分析的效率。

In view of traditional vehicle road load acquisition method which is costly and time-consuming,and is unconvenient for new vehicle models with changed configuration parameters,a data-driven wheel center load predication method is proposed by using load database of multiple vehicle models,and establishing a deep-learning model named Deep Convolution Neutral Network-Long-Short Term Memory(DCNN-LSTM)to build the relationship between vehicle road load and parameters including designed configuration and driving conditions.The comparison with test results indicates that the wheel center load predicted by this method is very close to the data collected from proving ground tests,conducive to gradually cancelling road spectrum acquisition test and improving the efficiency of vehicle durability analysis greatly.