Study Of Monitoring Methods For Process With Multiple Operation Modes And Its Application

With the development of our country’s economy and the improvement of people’s living standard, the society has high demand for the process safety and the product’s quality. Process monitoring, as one of the key technologies of the process system engineering, can detect process fault effectively, and related actions could be taken timely to avoid accident. With the computer and network technology being applied to the industrial, more and more data is collected into the process database which can be used for process monitoring. Data-driven process monitoring methods have been widely used in this decade and the methods of Multivariable Statistical Process Control have been a research hotspot.This paper mainly focues on the problem of process monitoring with multiple operation modes. Firstly, Fuzzy C Means (FCM) algorithm is used to divide the process modes and Support Vector Data Description (SVDD) method is used to monitor the process. Traditional FCM method easily sinks into local minimum, so a novel optimization method is proposed to replace the traditional iterative steps. The above two methods are applied for monitoring ethylene cracker and the monitoring result is discussed. Then, a multiple models local modeling method is developed according to the process input-output relation. Firstly, local modeling method is used to get the monitoring residual error. Then, the feature modeling samples which are extracted by Greedy method are used for building SVDD model. The application of monitoring Tennessee Eastman Process and ethylene cracker shows the algorithm’s efficiency. What’s more, multiple modes external analysis is proposed to remove the influence of multiple operation modes. Due to the problem of SVDD’s high computation complexity, a method called Kernel Possibilistic one-Mean clustering (KP1M) is used to monitor process with multiple operation modes. In the end, this paper introduces some technologies used in the process monitor software developing and process simulation platform.