Performance analysis of multivariable processes

This thesis is concerned with the methods of process monitoring and control loop performance assessment. Statistical process control, as well as performance assessment method based on minimum variance benchmark are discussed. Various measures are developed to assist performing assessment for feedback and feedforward-feedback control systems. Effect of deviation of process delay estimate on the feedback performance measure is also investigated. In addition, a system identification based method for assessing the performance of multivariate closed-loop systems is proposed. The method uses a priori knowledge of the delays between different input-output pairs of the process and performs assessment on each individual loop separately. Results indicate that for cases with insignificant interaction between disturbances, this new technique could be used in place of existing multivariate techniques that require a priori knowledge of the unitary interactor matrix.; The practicality of the methods presented in this thesis are demonstrated using simulated, experimental and industrial data.