基于改进的RBF神经网络车辆动态称重研究

Research on Vehicle Weigh-in-Motion by Improved RBF Neural Network

针对道路交通的复杂性,提出利用改进的径向基函数神经网络算法处理车辆的动态称重数据。该算法使用粒子群寻优的方式确定RBF神经网络中心,通过惯性权重因子控制寻优速度,将车辆的动态称重重量、车速、车长、轴数作为辅助神经网络的输入向量预测真实车重。训练结果表明,车辆速度与误差率呈正相关关系,改进的RBF神经网络可以明显提升动态称重数据的精度,在处理高速数据时,改进的RBF算法优化效果更好,在实际应用中具有重大意义。

Aiming at the complexity of road traffic,an improved RBF neural network algorithm was proposed to process the weigh-in-motion data of vehicles.This algorithm determined the center of RBF neural network by using a particle swarm optimization search,controlled the speed of merit seeking by inertial weighting factors.Weight,speed,length,and number of axles of the vehicle were used as input vectors for neural network to predict the real vehicle weight.The training results show that vehicle speed is positively correlated with error rate.The improved RBF algorithm can significantly improve the accuracy of weigh-in-motion data,and have better effect when dealing with high-speed data.This research has certain significance in practical applications.