Research On Monitoring Methods About Transition Process In Multi-mode Industrial Process

Fault monitoring is one of the important measures to ensure the manufacture stably and securely in industrial production.When production condition changed or production indicator varied,the transition mode occurs.Because of its large data fluctuations,poor stability,and easy to break down,etc.,it has become an important area in monitoring field.In this paper,specific to the transition modes monitoring issues,weighting and local linearization are lead into monitoring area to establish the monitoring model and realize the monitoring of transition mode.The main researches are as follows:Firstly,due to the difficulty of dividing the transition mode accurately,a study of the k-means clustering and EM clustering is done to propose the corresponding improved clustering methods basing on the data relationship between the transition mode and the stable mode.Basing on the basic theory of the two clustering methods,the starting and ending points of transition mode and its data division can be realized.Secondly,the characteristics of nuclear weighted average function in nuclear methods are elaborated.Considering the inner connection of transition mode and stable mode,a monitoring model is established combining with the PCA basic theory,and the simulation analysis and verification are applied.Thirdly,combining the basic strategy of local linearization with the complicated variety of the transition mode,a new monitoring method about transition mode is established basing on its geometrical characteristics.The feasibility is verified by the simulation of TE process.Finally,taking the grinding and classification process as the background,and selecting the cyclone-related variables as the monitoring object,the field data can be obtained.The practicality of the WAF-PCA method and local linearization method is verified.