论文部分内容阅读
传感器故障诊断在化工生产中有着重要地位。该文以小波变换与神经网络方法为基础,提出了一种传感器故障诊断的方法。该方法能够有效区分传感器故障造成的信号变化与过程本身正常波动造成的信号变化,同时在训练神经网络时只需要系统正常状态下的样本,克服了传感器故障样本稀少的困难。此外,该方法可以在传感器发生故障后估计出正常的模拟信号。实验证明,该方法能够有效完成故障诊断,并可以判断出传感器的故障类型。
Sensor fault diagnosis in chemical production has an important position. Based on wavelet transform and neural network, a method of sensor fault diagnosis is proposed. The method can effectively distinguish the signal changes caused by the signal changes caused by the sensor failure and the normal fluctuation of the process itself, and only need the samples in the normal state of the system when training the neural network, thereby overcoming the problem of sparse sensor failure samples. In addition, this method estimates normal analog signals after a sensor failure. Experiments show that this method can effectively diagnose the fault and judge the fault type of the sensor.