基于BP神经网络的电控汽油机传感器故障调节技术研究

On-Board Sensor Fault Accommodation Using BP Neural Network in Electronic Control Gasoline Engine

应用人工智能神经网络,提出了一套针对电控汽油机主要传感器的故障调节策略。故障调节模块根据处于正常工作状态的传感器信号输入,针对故障传感器信号进行估计输出,控制发动机继续运转。最后,通过台架试验采集的信号,对故障调节神经网络进行了训练和验证。结果显示,大部分工况下进气压力传感器和节气门位置传感器的信号重构误差可控制在10%以内,表明了所提出的调节策略是切实有效的。

A fault accommodation strategy was presented in this paper for the sensors of electronic control gasoline engine based on artificial neural networks （ANNs）. In this strategy, fault accommodation module sent out the estimated value of the faulted sensor according to other sensors value at normal state to keep the engine running. Finally, ANNs was trained and the method was validated using bench experiments sampiing data. The results show that the error of estimated value of both throttle position sensor and manifold pressure sensor are less than 10%, and the strategy put forward in this paper is valid.