Level Control of Quadruple Tank System Based on Adaptive Inverse Evolutionary Neural Controller
- PDF / 2,254,706 Bytes
- 12 Pages / 594.77 x 793.026 pts Page_size
- 91 Downloads / 302 Views
ISSN:1598-6446 eISSN:2005-4092 http://www.springer.com/12555
Level Control of Quadruple Tank System Based on Adaptive Inverse Evolutionary Neural Controller Nguyen Ngoc Son Abstract: This article proposes an adaptive inverse evolutionary neural (AIEN) controller for liquid level control of the quadruple tank system. Firstly, an inverse evolutionary neural model (IEN) that is utilized for offline identifying a dynamics of quadruple tank system, provides a feed-forward control signal from the reference liquid level. In which, the evolutionary neural model is a 3-layers neural network that is optimized by a hybrid method of modified differential evolution and backpropagation algorithm. Then, a hybrid feedforward and PID feedback control is realized to eliminate the steady-state error. Finally, to solve an uncertainty and disturbance characteristic, an adaptive law is proposed to adopt online in its operation. Simulation and real-time control experimental results demonstrated the feasibility and effectiveness of the proposed approach for the quadruple-tank system. Keywords: Adaptive feedforward controller, adaptive neural control, differential evolution, level control.
1.
INTRODUCTION
Liquid level control is widely used in many industries, such as petrochemical plants, water treatment processes, paper making and so on. To improve the performance, some studies have been introduced such as an adaptive PID control with a parallel feedforward compensator [1], a sliding mode control [2], a backstepping control [3], a linearized feedback control [4], robust H∞ observer based stabilization [5] are proposed for level tank control. Those approaches require pre-knowledge of the system. However, in practice, a common tank system consists of tanks, pumps, piping, valves, sensors. During operation, depending on requirements, valves can change the aperture, or the liquid type in the tanks can change. Moreover, the parameters such as the size and discharge coefficient of the valves, the power supply of the pump in some cases may be still unknown. This leads to a nonlinear characteristic of the system that can vary. Therefore, it is not easy to obtain an accurate mathematical model of a tank system. To overcome the above problem, some identification approaches which deal with the problem of how to estimate a nonlinear characteristic using measured input and output signals, have been studied for the liquid level control. Bououden et al. [6] proposed a predictive control based on adaptive fuzzy model and ant colony optimization (ACO) for nonlinear process systems. Where an adaptive fuzzy identification was introduced to identify the
nonlinear process system parameters. Wang et al. [7] used a radial basis function neural network as the prediction function due to its approximation ability. Then, a predictive control method was proposed to guarantee the system stability and compensate for the network-induced delays. Cetin et al. [8] introduced a model predictive control based on neural models that determine a function approximation. T
Data Loading...
