Stabilizing the Operation of Industrial Processes using Data Driven Techniques


  • M. A. A. Shoukat Choudhury Assistant Professor Department of Chemical Engineering BUET, Dhaka-1000, Bangladesh
  • Ian Alleyne



Stiction, Data Driven Techniques


Poor performance of a control loop is usually caused by poor controller tuning, presence of disturbances, control loop interactions and/or loop nonlinearities. The presence of nonlinearities in control loops is one of the main reasons for poor performance of a linear controller designed based on linear control theory. In a control loop, nonlinearities may appear either in the control instruments such as valves and positioners or in the process. Among the control valve nonlinearities stiction, deadband, deadzone, hysteresis and saturation are most common. A nonlinear system often produces a non-Gaussian and nonlinear time series. The test of Gaussianity or nonlinearity of a control loop variable serves as a useful diagnostic aid towards diagnosing the causes of poor performance of a control loop. Ttwo indices, the Non-Gaussianity Index (NGI) and the Non-Linearity Index (NLI), developed in [1] are used to detect the possible presence of nonlinearity in the loop. These indices together with specific patterns in the process output (pv) vs. the controller output (op) plot can be conveniently used to diagnose the causes of poor control loop performance thus ensuring smooth operation of the plant. The method has been successfully applied to many industrial data sets. One of the interesting case studies is presented in this paper. The results of the analysis were confirmed and the results after the troubleshooting was performed are also presented.

Keywords: Nonlinearities, stiction, performance monitoring, nonGaussianity, process industries, control valves

DOI = 10.3329/cerb.v13i1.2995

Chemical Engineering Research Bulletin 13 (2009) 29-38


Download data is not yet available.




How to Cite

Choudhury, M. A. A. . S., & Alleyne, I. (2009). Stabilizing the Operation of Industrial Processes using Data Driven Techniques. Chemical Engineering Research Bulletin, 13(1), 29–38.