Temperature decoupling control of double-level air flow field dynamic vacuum system based on neural network and prediction principle

Abstract

Double-level air flow field dynamic vacuum (DAFDV) system is a strong coupling, large time-delay, and nonlinear multi-input–multi-output system. Decoupling and overcoming the impact of time-delay are two keys to obtain rapid, accurate and independent control for two air temperatures in two concatenate chambers of the DAFDV system. A predictive, self-tuning proportional-integral-derivative (PID) decoupling controller based on a modified output–input feedback (OIF) Elman neural model and multi-step prediction principle is proposed for the nonlinearity, time-lag, uncertainty and strong coupling characteristics of the system. A multi-step ahead prediction algorithm is presented for temperature prediction to eliminate the effects of time-delays. To avoid getting into a local optimization, an improved particle swarm optimization is applied to optimize the weights of the OIF Elman neural network during modeling. By using the modified OIF Elman neural network identifier, the DAFDV system is identified and the parameters of PID controller are tuned on-line. The experimental results for two typical cases indicate that the settling times are obviously shorten, steady-state performances are improved and more important is that one temperature no longer fluctuates along the other, which verify the proposed adaptive PID decoupling control is effective.

Introduction

Double-level air flow field dynamic vacuum system (DAFDV) is that the air flow temperatures in two concatenate chambers maintain in steady equilibrium states while maintaining a certain flow rate into and out of the chambers. DAFDV systems have great application value for the simulation of dynamic atmospheric environments, calibration of meteorological instruments and research of humidity generation technology based on two-temperature or two-temperature two-pressure principle (Wang, 2003, Kitano et al., 2008, Helmut, 2008). Air temperature control of single chamber has been widely studied by many scholars (Lawton and Patterson, 2000, Ghoumari et al., 2005, Thompson and Dexter, 2005, Qi and Deng, 2009, Li et al., 2011), and some findings have been extensively applied in industrial practices. However, researches on DAFDV systems have scarcely been represented in the open literature. This may be attributed to two major sources of difficulty. First, the simple and adequate dynamic relationships among the process variables are difficult to establish due to the nonlinearity, uncertainties and structure complexities. Second, the two air temperatures in the two concatenate chambers suffer impact from each other’s fluctuations. If the two air temperatures cannot be decoupled effectively, they would not only delay reach steady states, but also not accomplish independent control at all. Therefore, in order to achieve rapid, accurate and independent control, it is necessary to decouple these two variables. However, how the appropriate decoupling method is selected according to the characteristics of control object is a key problem.

The traditional decoupling ways to a multi-input multi-output (MIMO) system is mainly represented by modern frequency domain methods such as diagonal dominance matrix, relative gain analysis method, characteristic curve method, state variable method, inverse Nyquist array and so on (Wang, 2000). These methods, which are based on strict transfer functions or state spaces, play an important role in decoupling the linear time-invariant MIMO systems. However, these methods are difficult to achieve dynamic decoupling for nonlinear or uncertain or time-variant MIMO systems because accurate system models are difficult to develop for these systems. Thereby, these traditional decoupling methods are confined to a certain application scope.

With the development of decoupling control theory, a multitude of other decoupling methods such as adaptive decoupling, energy decoupling, disturbance decoupling, robust decoupling, fuzzy decoupling, neural networks decoupling, prediction decoupling, intelligent decoupling methods represented mainly by the fuzzy decoupling and the neural networks decoupling, have been proposed and applied in many control practices. The detailed introduction to these decoupling methods are summarized (Dong et al., 2011), this paper no longer repeats them.

Adaptive decoupling has merits in decoupling a system with many uncertain factors and can solve the system’s uncertainty to a certain extent. Multilayer neural networks have adaptive, self-learning, strong fault tolerance abilities and are universal approximators capable of approximating any nonlinear function to any desired degree of accuracy, making it a powerful tool for the decoupling control of nonlinear systems. The modified output–input feedback (OIF) Elman neural network, as a kind of recurrent neural network, is superior to the static neural network such as back-propagation (BP) and radial basis function (RBF) neural network on the dynamic characteristic (Wu et al., 2011a), and it is now extensively applied in the fields of system identification, nonlinear control and prediction control (Serhat et al., 2003, Qi et al., 2005, Gao and Wang, 2007). However, neural networks commonly require to be combined with other algorithms to realize decoupling control (Wu and Chai, 1997, Li, 2006).

Prediction is an effective means to control a time-delay system. Furthermore, the proportional-integral-derivative (PID) controller is widely used in many fields due to its simplicity and robustness (Kumar et al., 2007, Shi et al., 2008, Xu et al., 2008).

Based on the above discussions, an adaptive PID decoupling control method based on the modified OIF Elman neural network and prediction principle is proposed in this paper. Because the DAFDV system is a strong coupling, complex MIMO nonlinear system with large time-delay and uncertainty, which can hardly acquire satisfactory control performance and even cannot reach the steady state at all by the conventional PID controller with fixed parameters. By the identification function of the OIF Elman neural network, the PID parameters are tuned on-line. Thus, the couplings between the manipulated variables can be treated as corresponding exterior disturbances, so the proposed controller is used to eliminate disturbance and obtain desired control performance in different operating regions. The time-delay effects can be reduced or eliminated by the prediction. The main contribution of this study is to propose an effective decoupling control strategy, which can be applied to a real-time plant conveniently, for the strong coupling, large-time delay and nonlinear system make it is difficult to elaborate a mathematic model precise enough for the control.

The paper is structured in the following way. In Section 2, the composition of DAFDV system is presented. In order to analyze the system properties conveniently, the models of the DAFDV system are qualitatively developed in Section 3. Section 4 is proposed the adaptive PID decoupling control method based on the modified OIF Elman neural network and prediction principle. In Section 5, the experimental results on a real-time DAFDV system are presented. Conclusions are drawn in Section 6.