Thermal Control System Design Platform For The Development And Application

The simulation of control system has been a crucial step of designing the control system. The thermal controlled members are complex and have been given a high expectation in safety and economy; therefore we can't design the control system in thermal object directly but by simulating the system to understand the principle usually. At the same time, there are more advantages in object-oriented programming than traditional procedure-oriented programming about thermal control system design software, which will give the facility now and in the future.This paper adopts the ideas and methods of object-oriented programming to develop thermal control design software in Java by analyzing the characteristics of thermal control system. With the typical element as basic operated unit, the software can draw and dispose the block diagram rapidly and conveniently. Through exploring the signal flow chart and Mason equation, this paper proposes a new computer solution about how to get the transfer function of SISO control system by block diagram and accomplishes the deduction of the formula and the discretization of the transfer function and so on.This paper exploits a new module related artificial neural network model building about double input single output thermal object in gray box model and single neuron self-adaptive prediction control about SISO thermal object, and applies which to the model building and control simulation of heat exchangers. This paper deduces the universal gray box model-building algorithm based on operating point to forward network and improved Elman network. These two algorithms are easy to be understood and programmed and have strong applicability since they are expressed by matrix operations. This paper also provides a new synthetically self- adaptive step-size change adjustment method, thus avoiding the local minimal point and assure the convergence. By simulation test, this paper defines the optimal parameter range during ANN's train and single neuron self-adaptive prediction control. Compared with the traditional PID control method, the single neuron control system gets better control quality, which indicates the latter is feasible in practice. This paper also points out measures to make further improvement in the control system.