Research On Outlier Detection For Process Control System

With the rapid development of modern industry, the industry process control systems require higher performances on control precision, response speed, stability. Number of process models is also increasing, and the models are more and more complex. But most of the modeling methods often rely on data collected at the scene, requiring high accuracy of the modeling data, so accurate data is a prerequisite for an establishment of accurate model. Therefore, outlier detection for the process control system is particularly important.In this background, a series of detection methods about outlier in process control was provided based on characteristics of process data and structure of process control. The main contents are summarized as follows:Combined with structure of process control system and characteristics of outlier, definition of outlier was established. According to the relationship between wavelet transform and signal singularity, a detection method about outlier based on wavelet transform was proposed. The method can extract signal characteristic information effectively and detect outlier accurately.Because both outlier and catastrophe point are process data, it is necessary to establish definition of catastrophe point. According to scale propagation characteristics of wavelet transform maximal modulus, detection method about catastrophe point was proposed. Simulation illustrated validity of the method in paper.In view of the coupling characteristics of the multi variable system, a coupling-outlier detection method based on statistics theory was put forward in the background of three-phase electrode regulator system of arc furnace. The simulation results showed that the detection method has a good application effect and practical value.