ON-LINE MULTIVARIABLE SYSTEM IDENTIFICATION AND ADAPTIVE CONTROL OF INDUSTRIAL PROCESSES

IN THE PROPOSED STUDY, THE CANONICAL VARIATE ANALYSIS (CVA) METHOD OF SYSTEM IDENTIFICATION IS EVALUATED FOR IDENTIFICATION AND ADAPTIVE CONTROL OF INDUSTRIAL PROCESSES.CURRENTLY AVAILABLE ALGORITHMS FOR SYSTEM IDENTIFICATION AND ADAPTIVE CONTROL ARE NOT COMPLETELY RELIABLE FOR AUTOMATIC IMPLEMENTATION ON MICROCOMPUTERS IN REAL TIME. IN THE CVA APPROACH, THE ALGORITHMS ARE COMPUTATIONALLY STABLE AND NUMERICALLY ACCURATE AND CAN BE IMPLEMENTED ON INEXPENSIVE MICROCOMPUTERS. THE CVA METHOD AUTOMATICALLY DETERMINES THE DYNAMICAL STATE ORDER AND STRUCTURE OF THE PROCESS. THE PROPOSED PHASE I RESEARCH IS TO SHOW THE FEASIBILITY OF THE CVA METHOD ON INDUSTRIAL PROCESSES THAT ARE DIFFICULT TO IDENTIFY, AND TO EXTENDED CVA TO THE IDENTIFICATION OF NONLINEAR AND TIME VARYING SYSTEMS. THE COMPUTER SIMULATION SOFTWARE AND EXPERIMENTAL PROCESS FACILITIES OF THE UNIVERSITY OF CALIFORNIA AT SANTA BARBARA WILL BE USED FOR EVALUATION OF FEASIBILITY. IN PHASE II, THE METHODS AND ALGORITHMS WILL BE INVESTIGATEDIN MORE DETAIL FOR A WIDER RANGE OF PROCESSES, AND ALGORITHMS AND SOFTWARE WILL BE DEVELOPED FOR APPLICATION OF THE CVA METHOD TO A BROAD CLASS OF INDUSTRIAL PROCESSES. IN ADDITION, FULL-SCALE INDUSTRIAL PROCESS DATA FROM A MAJORMANUFACTURER WILL BE USED.