基于自适应加权融合的汽车衡故障传感器预估

Estimation of faulty load cell in truck scale based on adaptive weighted fusion

传统汽车衡任意一路称重传感器发生故障都将导致称重系统失效。提出了一种基于自适应加权融合的汽车衡故障传感器预估方法:根据多路称重传感器输出相关性和相邻传感器输出比值的相关性,建立基于径向基函数神经网络(RBFNN)的故障传感器输出预估网络和传感器输出比值的预估网络,得到两组冗余估计信号,采用自适应加权融合方法完成冗余信号融合,获得故障传感器输出估计值。仿真实验与现场测试表明,采用该方法的故障传感器称重误差小于任何单个预估网络误差和算术平均值融合误差,任一传感器发生故障时的汽车衡整体称重误差≤0.5%,避免了称重系统失效。

Conventional struck scale is disabled when anyone load cell goes wrong in operation. Estimation for faulty load cell based on adaptive weighted fusion is proposed. According the internal correlation among load ceils and the internal correlation among ratios of load cell outputs, two independent predictive models for wrong load cell output and the ratio of the outputs are established based on radial basis function neural network （RBFNN） to get two-group redundant estimative signals; then adaptive weighted fusion method （AWFM） is used to carry out redundant signal fusion and obtain more accurate estimative output of the wrong load cell. The results of simulation experiment and field test show that using AWFM, the errors of faulty load cell are less than those using any predictive model and mean fusion method （MFM）. Using AWFM the total weighing error of the truck scale is less than 0.5% and the truck scale still can work when anyone load cell goes wrong.