Temperature control with a neural fuzzy inference network

Although multilayered backpropagation neural networks (BPNN's) have demonst rated high potential in the nonconventional branch of adaptive control, the ir long training time usually discourages their applications in industry. M oreover, when they are trained on line to adapt to plant variations, the ov er-tuned phenomenon usually occurs. To overcome the weakness of the BPNN, i n this paper we propose a neural fuzzy inference network (NFIN) suitable fo r adaptive control of practical plant systems in general and for adaptive t emperature control of a water bath system in particular. The NFIN is inhere ntly a modified Takagi-Sugeno-Kang (TSK)-type fuzzy rule-based model posses sing a neural network's learning ability. In contrast to the general adapti ve neural fuzzy networks, where the rules should be decided in advance befo re parameter learning is performed, there are no rules initially in the NFI N. The rules in the NFIN are created and adapted as on-line learning procee ds via simultaneous structure and parameter identification. The NFIN has be en applied to a practical water bath temperature-control system. As compare d to the BPNN under the same training procedure, the simulated results show that not only can the NFIN greatly reduce the training time and avoid the over-tuned phenomenon, but the NFIN also has perfect regulation ability. Th e performance of the NFIN is also compared to that of the traditional PID c ontroller and; fuzzy logic controller (FLC) on the water bath temperature-c ontrol system. The three control schemes are compared through experimental studies with respect to set-points regulation, ramp-points tracking, and th e influence of unknown impulse noise and large parameter variation in the t emperature-control system. It is found that the proposed NFIN control schem e has the best control performance of the three control schemes.