Research On Nonlinearitv Detection In Process Control Loops

Oscillation is one of the most serious problem encountered in process control. The presence of oscillations may result in increased consumption of energy and raw materials, fluctuations of product quality, and potential risk. Among the factors which cause oscillation, nonlinearity in the control loop is more common, especially valve stiction. Recent nonlinearity detection techniques usually assume that there is no periodic disturbance in the control loop. However, disturbances that propagate throughout a plant can have such a impact that even though a well performed loop would suffer from oscillation. In order to cope with the nonlinearity detecting problems in oscillation diagnosis, following issues are addressed in the thesis.1. A nonlinearity detection process for oscillation diagnosis in industrial control loops in proposed. The technique aims at dealing with identifing the nonlinear root cause when there is a plant-wide oscillation which makes nonlinearity detection a more complex issue. The whole process consists of not only nonlinearity detecting methods, but also the locating methods for nonlinear oscillation source and detecting methods for specific nonlinear characteristics (stiction detecting method, etc).2. A effective nonlinearity detection method based on Modified bicepstrum is proposed. An new higher order statistical tool is developed to overcome the disadvantage of the traditional method using bicepstrum. Compared with the current nonparametric method for detecting nonlinearity in the control loop, method in this paper performs better in dealing with data of small size and weak nonlinearity. Besides, the nonlinear index proposed in this method can be used to measure the degree of nonlinearity in the oscillation signal within a control loop. Relying on the fact that the harmonic in the disturbance would be weakened when it propagates through any linear process, a nonlinearity-ranking method based on the Modified bcp index is proposed to locate the root cause of oscillation.3. A robust nonparametric causal analysis method for locating the nonlinear root cause of oscillation is proposed. It is based on the phase slope of cross spectrum and a PSI index to estimate the direction of information flux between the variables in the process. PSI method is applicable to detecting the way through which oscillation is propagated and then locating the root source. This measure is insensitive to irrelevant component between the variables and gives meaningful results even though the cross spectrum is nonlinear.