טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
M.Sc Thesis
M.Sc StudentShneiderman Dmitry
SubjectProperties of the Quadruple-Tank Process with Multiple
Dead Times and its Control via Auto-Tuned
Decentralized Dead-Time Compensator
DepartmentDepartment of Mechanical Engineering
Supervisor Professor Emeritus Zalman Palmor


Abstract

The presence of dead times in many engineering processes makes their control more difficult and imposes limitation on the performance. One possible way of coping with dead times is to use the Dead Time Compensator (DTC).

This research is the third and the final stage of a project whose purpose was to develop a full auto-tuner for decentralized DTCs for Multiple Input Multiple Output (MIMO) systems. One of the objectives of the research was the implementation of the auto-tuner and to test its performance under practical conditions on a laboratory process. The auto-tuner was implemented on a DSP (DS1104 R&D Controller Board) of dSPACE and on MATLAB.

In order to test the performance of the auto-tuner under practical conditions a laboratory MIMO process containing dead times was constructed as part of this research. The laboratory setup is based upon a process, called the Quadruple-Tank Process (QTP) which does not contain dead times. The latter process was found to be ideally suited to illustrate the performance limitations of MIMO systems in closed loop due to the right half plane zero. The new setup is called the Quadruple-Tank Process with Dead-Times (QTPDT).

Based upon the QTPDT developed in this research, a thorough investigation of the influence of the dead times on the process’ properties was carried out. Interestingly, it was found that the effect of the dead times on process’ zeros depends on the difference between the dead times on the diagonal and those on the off-diagonal of the QTPDT transfer matrix.

Experiments with the laboratory QTPDT were performed in the research. The experiments demonstrate that the auto-tuner is effective and robust also in difficult to control cases. The performance of the DTC was compared both experimentally and via simulations with that of conventional decentralized controllers. The comparison shows that the performance, achieved by the auto-tuned DTC, is significantly better than that of the conventional controllers.