基于GA-PSO-BP的车辆动态称重算法研究

Research on vehicle dynamic weighing algorithm based on GA-PSO-BP

针对车辆动态称重系统测量精度较低的问题,提出基于遗传算法(genetic algorithms)和粒子群算法(particle swarm optimization)混合优化的BP神经网络(back propagation neural network)动态称重模型。首先利用小波变换对信号进行去噪滤波与信号重构。通过GA-PSO算法迭代寻优神经网络的权阈值,以滤波重构的动态车重、车速、轴数作为网络输入,拟合动态车重与静态车重间的非线性关系,可以使称重误差控制在1%以内。实验结果表明该模型精度高,可应用于工程实践。

Aiming at the problem that the measurement accuracy of on-board dynamic weighing system is low,a dynamic weighing model based on BP neural network optimized by GA-PSO is proposed.The wavelet transform was used to pre-process noise of vehicle weighing signal and signal reconstruction.GA-PSO was used to optimize the weight and thresholds of the traditional BP neural network,the speed,axle number and dynamic vehicle weight are used as inputs to fitting the nonlinear relationship between dynamic vehicle weight and static vehicle weight,which can control the weighing error within 1%.The experimental results show that model has high accuracy and can be applied to engineering practice.