Simulated Chemical Reactor Research Articles

Asynchronous Quantized Control of Piecewise-Affine Systems

This technical note proposes a novel asynchronous control approach for discrete-time piecewise-affine (PWA) systems with logarithmic quantization of both multi-inputs and multi-state measurements. Since the actual system state and the quantized state may not always be in the same operating region due to quantization-induced uncertainties, the operating modes of the PWA system and the controller which depends on the quantized states may be asynchronous. Aiming at reducing the computational cost and the conservatism of the results, a mapping region-based algorithm is first proposed to determine the reachable dwelling regions for the quantized state. By using a convex combination model to approximate the quantization-induced uncertainties, a novel piecewise Lyapunov function taking into consideration the uncertainties is then proposed. It is shown that with the newly proposed Lyapunov function, the desired asynchronous controller can be obtained and the resulting closed-loop system is asymptotically stable. A simulated chemical reactor example is presented to illustrate the effectiveness and the superiority of the proposed asynchronous control approach.

Control invariant set enhanced safe reinforcement learning: Improved sampling efficiency, guaranteed stability and robustness

Reinforcement learning (RL) is an area of significant research interest, and safe RL in particular is attracting attention due to its ability to handle safety-driven constraints that are crucial for real-world applications. This work proposes a novel approach to RL training, called control invariant set (CIS) enhanced RL, which leverages the advantages of utilizing the explicit form of CIS to improve stability guarantees and sampling efficiency. Furthermore, the robustness of the proposed approach is investigated in the presence of uncertainty. The approach consists of two learning stages: offline and online. In the offline stage, CIS is incorporated into the reward design, initial state sampling, and state reset procedures. This incorporation of CIS facilitates improved sampling efficiency during the offline training process. In the online stage, RL is retrained whenever the predicted next step state is outside of the CIS, which serves as a stability criterion, by introducing a Safety Supervisor to examine the safety of the action and make necessary corrections. The stability analysis is conducted for both cases, with and without uncertainty. To evaluate the proposed approach, we apply it to a simulated chemical reactor. The results show a significant improvement in sampling efficiency during offline training and closed-loop stability guarantee in the online implementation, with and without uncertainty.

Incipient fault detection for nonlinear processes based on dynamic multi-block probability related kernel principal component analysis

In order to detect the incipient faults of nonlinear industrial processes effectively, this paper proposes an enhanced kernel principal component analysis (KPCA) method, called multi-block probability related KPCA method (DMPRKPCA). First of all, one probability related nonlinear statistical monitoring framework is constructed by combining KPCA with Kullback Leibler divergence (KLD), which measures the probability distribution changes caused by small shifts. Second, in view of the problem that the traditional KLD ignores the dynamic characteristic of process data, the dynamic KLD component is designed by applying the exponentially weighted moving average approach, which highlights the temporal data changes in the moving window. Third, considering that the holistic KLD component may submerge the local statistical changes, a multi-block modeling strategy is designed by dividing the whole KLD components into two sub-blocks corresponding to the mean and variance information, respectively. Case studies on one numerical system and the simulated chemical reactor demonstrate the superiority of the DMPRKPCA method over the conventional KPCA method.

Output modifier adaptation with filter-based constraints

Modifier adaptation (MA) and output modifier adaptation (MAy) are iterative model-based real-time optimization (RTO) algorithms that have the proven ability to drive plants to their optimal operating condition upon convergence despite disturbances and modeling uncertainty, provided the model at hand satisfies model adequacy conditions. But there is no guarantee that constraints are satisfied before convergence. In this article, an improvement of the formulation of MA and MAy is proposed that is proven to bring significant improvements w.r.t. these two limitations – model adequacy and feasibility of iterates. While standard MA or MAy suggests to perform optimization and filtering sequentially, it is proposed to integrate the input filtering stage in the modified model-based optimization problem by means of additional filter-based constraints. The corresponding approach, labeled “KMAy”, is (i) proven to preserve constraint qualification despite additional constraints, (ii) proven to preserve the property of MA methods to converge to the true plant optimal inputs, (iii) proven to significantly relax the model adequacy condition - leading it to be independent of the constraints of the optimization problem, (iv) shown to increase the chances of converging from the safe side of the plant constraints and (v) shown to support the choice of input filtering, instead of output or modifier filtering, if the input filter is appropriately chosen. A method for the automatic selection of the largest filter gain with the five aforementioned assets, while minimizing the filter-induced conservatism, is also proposed. The performances of KMAy with and without adaptive gain are successfully illustrated by means of the optimization of a benchmark simulated chemical reactor.

An Artificial Intelligence Software-Based Controller for Temperature Control of a Partially Simulated Chemical Reactor System

The processes that take place in industrial reactors are mainly highly nonlinear and, due to this, their controls have become a major factor in determining the aspect of process safety, quality, and productivity of these systems. For this reason, we have designed an inexpensive, practical reactor on a pilot plant scale to conduct such online control studies where the heat of reaction in the reactor is simulated by injecting steam into the reactor system. This paper highlights the development of the pilot plant with its software development to implement advanced control strategies using artificial intelligence approaches such as fuzzy logic and genetic algorithm. The online implementation results conclude that the Genetic Algorithm Model Based Controller (GAMBC) gave similar performance with the Fuzzy Logic Controller (FLC) for the set point tracking studies but in the load disturbance rejection studies, it was found that the FLC performed better. The results obtained show the usefulness and importance of these experimental studies to further understand advanced control applications for a typical industrial chemical process such as a reactor.

Dynamic process monitoring and fault diagnosis with qualitative models

Qualitative Modeling and Interpretation (QMI) and Qmimic are online monitoring and diagnosis systems which use multiple qualitative models of a plant to monitor noisy data streams and rapidly diagnose faults from observed dynamic behavior. Both systems continue monitoring after faults have occurred. QMI simulates normal and faulty plant behavior off-line using purely qualitative QSIM models, and uses plant data to select the correct model, yielding a diagnosis. Qmimic incrementally simulates online qualitative models which describe the current behavior of the plant, using plant data to constrain further predictions and select between the models. Although both systems are based on qualitative models of the plant, Qmimic also incorporates semi-quantitative data (quantitative ranges and bounding envelopes) into the qualitative simulation in order to achieve better predictions. QMI and Qmimic are described and compared in detail, and both are tested on a simulated chemical reactor. >