Process fault detection based on continuous hidden Markov model

With the increase of the scale and complexity of the industrial process, the requirements for process safety and reliability are further improved. In order to detect the equipment failure accurately and timely, a fault detection method based on continuous hidden Markov model (CHMM) is proposed. The principal component analysis (PCA) method is used to extract the characteristic data of the process variable data, and the dynamic data is effectively tracked by the variable moving window. A new implementation statistic is proposed based on the conditional probability as the quantization index of the fault detection and the real — time threshold. CHMM fault detection. The test results of this method to Tennessee-Eastman (TE) chemical process show that the fault detection effect based on CHMM is better than PCA-based fault detection, indicating that CHMM-based fault detection method can more accurately detect faults.