A method for detecting abnormal behavior of on-site process behavior in the manufacturing industry

Abstract

The invention relates to a detection method for on-site process abnormal behavior in the manufacturing industry. It mainly conducts research on the detection method based on the behavior characteristic knowledge base for the detection of the abnormal process in the manufacturing industry. The invention builds a random forest anomaly detection model based on the random forest algorithm. The specific implementation method is to carry out PCA (Principal Component Analysis) method for dimensionality reduction on the process data of the industrial site, use the integrated rule tree model for feature selection, and use the random forest algorithm to classify the process data. Each path of the random forest corresponds to a rule, which is very explanatory, greatly improves the classification accuracy, and can handle a large number of input variables. Even if the process data contains missing values, the classification results can still achieve high accuracy. The real-time emergency decision-making framework that integrates the knowledge base of behavioral characteristics and unknown behavioral scenarios provides early warning of abnormal behaviors by predicting the behavioral detection of industrial site process data.

Description

technical field [0001] The invention relates to an abnormal behavior detection method for on-site process behavior in the manufacturing industry. By constructing a knowledge base based on industrial on-site behavior characteristics, the correlation between industrial control safety detection technology and on-site behavior in the industry is realized, and the abnormal behavior detection alarm is more representative of the industry. On-site behavior characteristics belong to the field of industrial control network information security. Background technique [0002] At present, the research on abnormal behavior detection technology based on industry characteristic knowledge base is relatively blank at home and abroad. At present, industrial control protocols are added on the basis of traditional IT security protection products. Now the technology used for abnormal detection is based on support vector machines. Anomaly detection, its data space and time costs are relatively lar...

AI Technical Summary

Problems solved by technology

[0002] At present, the research on abnormal behavior detection technology based on industry characteristic knowledge base is relatively blank at home and abroad. At present, industrial control protocols are added on the basis of traditional IT security protection products. Now the technology used for abnormal detection is based on support vector machines. Anomaly detection, its data space and time costs are relatively large, and requires a large amount of storage space. Anomaly detection based on the naive Bayesian method, when the number of features is large and the correlation between features is large, the classification effect is not good

The prediction result of anomaly detection based on decision tree algorithm is unstable and the variance is large

Images