基于神经网络自适应滤波器的故障检测与诊断

在稳态过程的故障检测和诊断中 ,有大量反映故障状态的数据 ,在用测量方法来获取这些数据的情况下 ,异常数据能被快速、有效的检测出来就显得非常重要。用自适应滤波器无疑是一个可行的方法 ,但采用这种方法进行故障检测和诊断时 ,有计算速度慢 ,难以跟踪输入信号变化的缺点。据此 ,本文提出了用神经网络自适应滤波器来完成故障检测和诊断的方法 ,它具有速度快的特点。如能用硬件完成 ,并调整好参数 ,检测速度是极短的 (2× 10 -10 s) ,能很好地完成故障检测和诊断任务。

Neural Networks for Adaptive Filter and Fault Diagnosis

On fault detection and diagnosis of stable process (such as chemical industries), there are a lot of measurable data. Sometime, the data often associate with extraordinary conditions. In this case, it is important to detect those unusual date by faster and more accurate methods. A new scheme of adaptive filter based on neural networks is proposed, and it can be applied to detect damage date in complex system. Compared with tradition filter, the network has faster calculation speed, and can be used to detect fault on line. Simulation shows that the damage data can be reliably detected.