Self-learning neuro-fuzzy control system and its application on HTST heating in aseptic processing
Abstract
The purpose of the present study was to develop a self-learning neuro-fuzzy logic control system (SNFCS) to provide a model-free-method for designing an intelligent control system with self-learning ability for food processing. This system was applied to temperature control of a high-temperature-short-time (HTST) heat exchanger in aseptic processing by a computer simulation and pilot scale experiments. The time/temperature profile in aseptic processing is critical to the quality and sterility of aseptically processed foods. Better temperature control will improve product quality and reduce waste. The heating process becomes complicated because of unsteady inlet temperature of raw product, varying properties, changing flow rate, steam hysteresis, etc. The developed SNFCS is a hierarchy based control system consisting of three levels: primary control, learning, and supervising level. The primary control level is designed to directly control the process by applying several neuro-fuzzy decision systems (NFDS). The performance of fuzzy system is strongly affected by fuzzy rules and the membership functions of linguistic variables derived from experienced operators. Therefore, in the learning level the process performance measurement and adapting procedures are to tailor the knowledge base of control system to fit the new process given in the beginning, and to adapt to the process disturbances in-line. The supervisory control is integrated in the supervising level with functions such as alarming, fault diagnosis, start-up and shut-down procedures and higher level information. The results of the computer simulation and experiments showed that the developed SNFCS can tailor its control knowledge to fit new process conditions. SNFCS was more robust to noises and achieved better control than the conventional PID controllers. Besides, it was able to improve the HTST heating of the aseptic processing by adapting to the disturbances such as variation in raw product temperature, time-changing parameters in the process, varying product flow rate, and steam shut-off.
Degree
Ph.D.
Advisors
Singh, Purdue University.
Subject Area
Agricultural engineering|Artificial intelligence|Food science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.