Dual‐mode control with neural network based inverse model for a steel pickling process

Abstract

This article describes a novel implementation of the dual-mode (DM) control utilizing a neural network inverse model on a multivariable process (a steel pickling process). This process is highly nonlinear with variableinteraction, and is multivariable in nature, hence an accurately nonlinear model is required to provide acceptable control. The requirement of a true analytical inverse can be avoided when neural network models are used; they have the ability to approximate both the forward and the inverse system dynamics. Various changes in the open-loop dynamics are performed before implementation of the inverse neural network modeling technique. DM control based on neural network inverse model strategy is used to design the controllers to control concentration and pH of the process, which is guaranteed to remove steady-state offset in the controlled variables to obtain the maximum reaction rate and to comply with limits imposed by legislation. The robustness of the proposed DM control is investigated with respect to changes in disturbances and model mismatch. Comparisons are also made with the conventional inverse neural network controller (NNDIC) and other conventional controller proportional-integral (PI). Simulation results show the superiority of the DM controller in the cases involving disturbance and model mismatch, while the conventional controller gives better results in the nominal case. 

Machine Learning for Process Control of (Bio)Chemical Processes

“…Inverse model-based methods directly model the inverse process dynamics to predict the desired control actions over a prediction horizon to drive the process to the desired reference state (cf. Figure 1.9 Case 2) [169]. The control performance is dependent on the accuracy of these inverse models.…”

“…The control performance is dependent on the accuracy of these inverse models. Machine learning can be used to obtain the inverse models and relies on the amount and accuracy of the data available to train [68,169]. Direct inverse control and internal model control are two common types of inverse control methods [67,68,145].…”

“…A direct inverse control structure, using neural networks, has been combined with a conventional proportionalintegral (PI) controller in [169] to control the concentration and pH in a steel pickling process. The PI controller was used to lower setpoint offsets and prevent oscillations of the control variables.…”

“…Internal model control adds a plant model to the direct inverse control strategy to improve the robustness of the controller, (cf. Figure 1.10) [65,67,169,145]. The plant model accounts for mismatches and irregularities that can occur due to equipment degradation and noise [67].…”

“…• learning modifiers of the modifier adaptation scheme via GPs [6,11,148] • hybrid modeling for RTO models [194] • plant models for model-based control [52,67,95,105,112,150,194] • inverse models to provide control actions [65,67,88,145] • observer for parameter and state estimation [19,48,52,53,119,141] • self-tuning PID controllers [7,79,85,92,30,132,169] • solving static optimization of RTO using RL [140] • set-point optimization using RL [75,40] • imitation (supervise) learning [2] • RL-based [137,4] • RL for tuning PID controller [44] FIGURE 1.14 Selected, non-extensive, overview of works on machine-learning-supported control. Note that there is no clear-cut classification, as in some cases it is difficult to classify the methods into these categories.…”