Intermittent production process fault detection method based on Markov chain and spectral clustering

Abstract

The invention relates to a production process fault detection method based on Markov chain spectral clustering, and the method mainly comprises the steps: offline model building and online detection, the offline model building comprises the steps: carrying out the sub-stage division of training sample data, building a detection model through principal component analysis, and determining the control limits of T2 and SPE statistics; the online detection mainly comprises the steps of collecting new sample data, calculating a detection index value of the new sample data, and determining whether to give an alarm by judging whether the detection index value exceeds a control limit; according to the method, stage division is carried out on batch of the intermittent process, the problems of data interruption and unequal length of data are effectively avoided, and the influence of outlier data and noise on clustering and fault detection is effectively reduced.

Description

technical field [0001] The invention relates to a fault diagnosis technology, in particular to a detection method based on Markov chain and spectral clustering applied to fault diagnosis in batch production process. Background technique [0002] Process industry, also known as process industry, refers to the modern manufacturing industry that processes and manufactures process material products. The processing of products in the process industry deals primarily with continuous or intermittent material and energy flows. Process industries involve chemical reactions, separations and mixing, etc. Common petrochemical, crude oil smelting, pharmaceutical, chemical, metallurgy, papermaking, etc. in industrial production all belong to the category of process industrial production. [0003] Thanks to the development of control science and industrial technology, the production level of the process industry is getting higher and higher. At the same time, the production process, prod...

AI Technical Summary

Problems solved by technology

[0008] Aiming at the problem that the fault detection effect is not good due to intermittent data usually occurring in the production process, the present invention proposes a production process fault detection method based on Markov chain spectrum clustering, which divides batches of intermittent processes into stages, effectively It avoids the problem of data discontinuity and data unequal length, and effectively reduces the impact of outlier data and noise on clustering and fault detection

Images