The normalization PCA model and its application under the periodic non-steady conditions

The principal component analysis method is usually used for fault detection under the steady conditions, however, when system works under the non-steady conditions, the false alarm rate and the missing alarm rate, tested by the T² control limit, are so high. The main reason for this situation is that the sampled data is accord with normal distribution under the steady conditions, whereas the data mostly does not satisfy normal distribution under the non-steady conditions, but T² control limit can only detection fault effectively for the data conforming to normal distribution. So this paper firstly analyzes two characteristics of the periodic non-steady conditions and mathematical theory and Q-Q figures have verified the characteristics, then a normalization PCA (NPCA) is proposed. Finally it is applied to the system of the motor cyclical process and alleviating the load, the test result is good and can verify the effectiveness of the model.