Real-Time Process Control (A4)
Project PI: Thomas Edgar
PROJECT DESCRIPTION
In the process industries model-based control is rapidly eclipsing all other types of traditional advanced control methods. This project focuses on the application of this advanced control methodology using new algorithms implemented on an Intel platform operating as a real-time system. The model-based control algorithm uses a simultaneous solution and optimization approach to determine the open-loop optimal manipulated variable trajectory at each sampling instant. This scheme provides a framework to control processes described by different kinds of models, such as nonlinear ordinary differential/algebraic equations, partial differential equations, and artificial neural nets. Successive quadratic programming (SQP) is employed to compute the optimal control.
TECHNICAL CHALLENGE
Tests of this technique will be carried out on several pilot/lab facilities, including distillation columns and thin film deposition. This will permit an assessment of the capability of the Intel platform for carrying out real-time optimization calculations, with a sampling time of the order of several seconds or less. A real-time model requires simplification of the original theoretical model (nonlinear pde¹s), hence it will be necessary to investigate the trade-offs between model complexity and control performance.
IMPACT
Two of the leading automation suppliers currently beta-testing real-time control system architectures based on NT-Server are Fisher-Rosemount and National Instruments, both based in Austin, Texas. These companies, both members of the Texas-Wisconsin Modeling and Control Consortium (TWMCC), have expressed interest in this project. TWMCC is a consortium involving the University of Texas, the University of Wisconsin and twenty companies from a range of oil, chemical, microelectronics, paper, and service companies in the process industries. Another cooperative TWMCC project with major vendors of simulation software in the oil and chemical industry provides an opportunity to achieve a more rigorous assessment of the capabilities of Intel and Windows NT in a real-time optimization and control environment with commercial dynamic simulator packages. TWMCC members will receive detailed reports on the results of the Intel implementation of model-based control and simulator systems.
EQUIPMENT
The project is using four Intel Pentium Pro dual processor workstations. One will be used with an existing rapid thermal processing reactor for thin film deposition on a six-inch wafer, based on PECVD of silicon nitride. The second type of processing environment is the control of pilot distillation columns located in the Department of Chemical Engineering and in the Separations Research Program at the Pickle Research Campus. Both systems are close to industrial standards and should afford a realistic test of the computing environment. The third application is the use of the dynamic simulation tools HYSYS and DYNAPLUS in real-time decision-making in a simulation environment using a single workstation.
RESOURCES
The research will be carried out by faculty and graduate students from Chemical Engineering with expertise in process control, faculty from the College of Business working in the development and use of nonlinear optimization codes, and faculty from Computer Science working on real-time systems.
BENEFITS TO INTEL
This project will demonstrate to a range of industrial sectors the feasibility and performance of Intel platforms in a real-time operating environment and help identify the advantages of NT over traditionally used Unix systems. The complex optimization, control, and estimation algorithms employed in this research are computationally demanding in the content of the sampling time required (e.g., several seconds).