Research And Application Of Working Condition Recognition Method Based On Multi-Variable Trend Analysis

With the advancement of distributed control and model predictive control systems and the updating of numerous industrial devices,a wide variety of sensors are used in the fault diagnosis of equipment or systems.Since the amount of data generated in the process is very large,it is very complicated and difficult to extract useful information from it.Qualitative Trend Analysis(QTA)is a data-based working condition recognition method.Trend extraction and trend matching are important components of QTA.This method can obtain potential information from the data sequence and describe the process state,and is widely used in fault diagnosis,process state monitoring and other fields.However,most methods based on QTA use univariate trend characteristics to identify the working conditions.In complex industrial processes,the characteristics of the single variable are not sufficient to describe the process state,while the characteristics of the multi-variable can accurately describe the process state.Therefore,this paper studies the trend characteristics of multivariate.In this paper,the main objective is to realize the recognition of working conditions.The method of working conditions recognition based on multivariate trend analysis is studied in this paper,and the following results are obtained.1.Through the research,it is found that when the sliding window method is used to extract the trend of variables,the segmentation points are inaccurate,and the extraction efficiency is low.In this paper,the fixed window and fixed threshold in the original method are improved.The least square method is used to fit,and the dynamic threshold and dynamic initial window are added.A trend extraction method based on dynamic threshold and sliding window is proposed.The proposed method can accurately extract the trend characteristics of the variables.2.Fuzzy logic and knowledge base are introduced to match the trend and process state.A trend matching method based on fuzzy logic is proposed.The method uses the knowledge base and production rules to compare the trend of multiple variables with the working conditions and calculates the confidence between the trend and the working condition.Based on the proposed trend extraction method and trend matching method,a multivariate trend analysis method is proposed.3.In the research of froth flotation condition recognition,in view of the problem of small bubble segmentation and low efficiency in segmentation of froth images using OTSU method,a new froth image segmentation method based on pixel distribution characteristics is proposed,and the proposed method is applied to froth feature extraction in flotation conditions.4.In view of the difficulty of identifying froth flotation conditions,and the problem of froth flotation condition can't be recognized effectively only by froth image characteristics.It is found that the trend of pulp flow also affects the variation of flotation conditions.Therefore,this paper selects the trend of pulp flow and foam size and studies the method of foam flotation identification based on multivariate trend analysis.According to the characteristics of froth flotation conditions,the froth flotation conditions were divided into three categories: excellent,medium and poor.The knowledge base of flotation conditions was established,and a flotation condition recognition method based on multivariate trend analysis was proposed for the recognition of froth flotation condition.The proposed method can accurately identify various working conditions.