Performance Analysis of Cascade Tank System Using Deep Learning Controller

Abstract

The conventional proportional–integral–derivative (PID) controller is used in the majority of process industries. Although the most commonly used, the classic PID controller has several drawbacks like variable performance for the non-linear system, multivariable control design is not straightforward; the response is influenced by dead time, no constraints involvement, etc. The field of process control systems has grown quickly, and other controllers have been developed that try to overcome the weakness of PID controllers. Advances in artificial neural networks, specifically deep learning, have widened the application domain of process control systems. In this paper, a cascaded tank system, which is a benchmark problem, has been simulated. The input-output data of the plant has been generated and used to train a deep-learning controller using backpropagation. Several measures, such as time response, frequency response, and signal statics performance indices, are used to evaluate the outcomes of the proposed controller. The proposed model performs better on every assessment criterion than the traditional controller.