Controller Oscillation Behavior Detection Method And Application

The continuous development of industrial processes towards larger scale and higher integration brings challenges to the PID control loop performance detection and diagnosis,and the problem of poor performance in process control often appears.The control loop with poor performance will interfere with the long-term operation of the process,increase the consumption of raw materials and energy,decrease the quality of industrial products,and even affect the safety of industrial processes.An important manifestation of the control performance decline is the oscillation behavior of the control loop.Due to the coupling and integration of industrial processes,one single oscillation may be widely propagated throughout the system,which increasing the risk of instability.Therefore,it is important to detect and diagnose the oscillation behavior of the control loop.This thesis focuses on the detection,diagnosis and health management of control loop oscillation in industrial process.The main research work is as follows:1)An improved single-frequency oscillation detection method based on autocorrelation function(ACF)is proposed.Compared with the traditional ACF method,this method adopts all the autocorrelation coefficients of the control loop,and makes better use of data information during the performance calculation.The proposed method has advantages in anti-noise ability and improves the accuracy and reliability of frequency vibration detection in industrial data with high noise2)An improved method based on autocorrelation function(ACF)+empirical wavelet transform(EWT)is proposed for performance detection problem of multi-frequency oscillations with different frequencies and amplitudes,which are formed by the superposition of multiple oscillation sources.EWT is used to decompose the multi-frequency oscillation signal and reconstruct every component within a specific frequency range.For each component,detection is realized by the improved ACF method proposed before.This method can provide the information of all oscillation components.The frequency interval of the most significant oscillation component in the loop and its period can be given by the oscillation energy ratio index,which provides a priority recommendation for operators to carry out loop maintenance in the later stage.This method is designed to decrease the influence of the hybrid of multiple frequency oscillations on the detection and diagnosis of oscillations,and can effectively improve the accuracy of multi-frequency oscillation detection.3)Aiming at PID loop health management in industrial process,one control loop health management software is designed and developed,which providing the performance evaluation,the result of detection and diagnosis for the control loop.Six-dimensional evaluation index that include operation rate,saturation rate,adjustment index,oscillation index,stiction index and nonlinear index are calculated for each monitored control loop,then the loop performance rating which are excellent,good,medium and poor is provided.One oscillation diagnosis procedure is also proposed to determine the different causes of oscillation,which are set value oscillation,valve stiction and external disturbance.This procedure can help on-site maintenance personnel to accurately judge the oscillation causes and timely maintain control loop performance.4)The health management software and the oscillation detection and diagnosis method are applied to two practical industrial processes.One is an air separation unit with 8 control loops,and the other is a chemical process unit with 25 control loops.The detail analysis for all monitored loops are provided,then the traceability cause diagnosis for some oscillation loops are carried out,which verifies the effectiveness,reliability and practicability of the proposed method.