Journal of Process Control最新文献

Velocity control proves to be an effective and a more easily implementable actuation than boundary and distributed actuations for hyperbolic distributed parameter systems. However, the design of velocity control for these systems, following the late lumping approach, i.e., using the partial differential equations model, poses a challenging problem in control engineering. Noticeably, the velocity controller faces a control singularity issue, resulting in a loss of controllability that renders the controller impractical. In this paper, we demonstrate that the zeroing dynamics method is a viable alternative design approach for velocity control of hyperbolic distributed parameter systems following the late lumping approach. Thus, employing the partial differential equations model, a velocity state feedback forcing output tracking is developed based on the zeroing dynamic method. Furthermore, to address the control singularity problem, the zeroing gradient method is combined with the zeroing method to design a state feedback that achieves output tracking even when a singularity occurs. The tracking error convergence is demonstrated for both developed state feedbacks. The effectiveness of these design approaches is clearly demonstrated in the case of a steam-jacketed heat exchanger and a non-isothermal plug flow reactor.

It is common for the working conditions to change with time in actual industrial processes. However, the transition modes of complex industrial processes under different working conditions often have various degrees of dynamic nonstationarity, which makes the traditional process monitoring model based on the stationarity assumption ineffective. In this paper, a Recursive Slow Feature Analysis method based on Stability Factor Partitioning (SFP-RSFA) is proposed for fine process monitoring of transition modes under dynamic nonstationarity characteristics. First, we calculate the stability factor according to the different stationarity characteristics of the production process variables. Then, K-means clustering is carried out according to the stability factor of each variable, and the stability factor of the cluster center is mapped to the interval [0,1] as the smoothing coefficient of the exponential weighted moving average (EWMA), which is applied to each data subblock respectively to highlight the steady-state and dynamic characteristics of the monitoring data subblock. In the online monitoring stage, the monitored data are fed into the subblock recursive slow feature analysis (RSFA) monitoring model. Finally, a comprehensive statistic method is proposed to integrate the subblock monitoring statistics. The Tennessee Eastman (TE) process and actual cement clinker production process were tested and compared with existing RPCA, RCA and RSFA methods. The effectiveness and superiority of the proposed method in the problem of nonstationary transition mode process monitoring are verified.

Primal–dual feedback-optimizing control is a simple yet powerful approach to optimally handle active constraint changes at steady state. It is composed of two layers: Constraint control in the upper master layer and unconstrained optimization or gradient control in the layer below. The master constraint controllers operate on a slow time scale by updating the dual variables (Lagrange multipliers). This can result in too slow control of the constraints, for example, for hard constraints that cannot be violated dynamically. To address this issue, we propose introducing a third fast override constraint control layer. Additionally, to optimally coordinate the constraint handling between the master and override layers, we need to introduce auxiliary constraints for the master controllers. A gas-lift oil production optimization case study demonstrates the power of the proposed scheme.

Mixed integer Model Predictive Control (MPC) problems arise in the operation of systems where discrete and continuous decisions must be taken simultaneously to compensate for disturbances. The efficient solution of mixed integer MPC problems requires the computationally efficient online solution of mixed integer optimization problems, which are generally difficult to solve. In this paper, we propose a machine learning-based branch and check Generalized Benders Decomposition algorithm for the efficient solution of such problems. We use machine learning to approximate the effect of the complicating variables on the subproblem by approximating the Benders cuts without solving the subproblem, therefore, alleviating the need to solve the subproblem multiple times. The proposed approach is applied to a mixed integer economic MPC case study on the operation of chemical processes. We show that the proposed algorithm always finds feasible solutions to the optimization problem, given that the mixed integer MPC problem is feasible, and leads to a significant reduction in solution time (up to 97% or $50\times$) while incurring small error (in the order of 1%) compared to the application of standard and accelerated Generalized Benders Decomposition.

This manuscript presents a discussion on the algebraic state update step performed during recursive filtering of the differential–algebraic equation (DAE) systems. Existing DAE state estimation approaches follow a two-step state update procedure at each sampling instant. In particular, they first estimate the differential states using the Kalman update, and then update algebraic states by explicitly solving the algebraic equations. Specifically, for the case of DAE systems involving linear algebraic equations though the differential equations are nonlinear, we show that when appropriately initialized, this two-step state update procedure is not needed. It can instead be replaced with a one-step state update procedure that computes the differential and algebraic state estimates simultaneously through the Kalman update. The satisfaction of algebraic equations is guaranteed by this one-step update without it being explicitly enforced. Towards this end, we show that the error covariance matrix of augmented states, when properly initialized, satisfies a null-space property after prediction and update step at each sampling instant. This property ensures that the state estimates obtained using the proposed one-step update approach, satisfy the algebraic equations. This holds for both analytical linearization based extended Kalman filtering and statistical linearization based sigma-point filtering approaches. We also propose a heuristic-based update procedure for state estimation of DAE systems that involve nonlinear algebraic equations. This procedure draws out inferences from the case of DAE systems involving linear algebraic equations and is based on the analysis of algebraic equations residuals obtained from the updated differential and algebraic state estimates with a one-step state update. The efficacy of the proposed state update procedures is demonstrated by performing simulation studies on a benchmark drum boiler system case study. Results demonstrate that the proposed update procedures satisfactorily estimate the differential and algebraic states of a DAE system when compared to the traditional two-step update procedure.

In the presence of uncertainty, the optimum obtained based on a nominal identified model can neither provide any performance guarantee nor ensure that critical constraints are satisfied, which is crucial for e.g., bioprocess applications characterized by a high degree of complexity combined with costly experiments. Hence, uncertainty should be considered in the optimization and, furthermore, experiments designed to reduce the uncertainty most important for optimization. Herein, we propose a general framework that combines model-based robust optimization with optimal experiment design. The proposed framework can take advantage of prior knowledge in the form of a mechanistic model structure, and the importance of this is demonstrated by comparing it to more standard black-box models typically employed in learning. Through optimal experiment design, we repeatedly reduce the uncertainty most relevant for optimization so as to maximize the potential for improving the worst-case performance by balancing between exploration and exploitation. This makes the proposed method an efficient model-based robust optimization framework, especially in cases with limited experimental resources. The main part of the paper focuses on the case with modeling uncertainty that can be reduced with the availability of more experimental data. Towards the end of the paper, we consider extending the method to also include inherent uncertainty, such as input uncertainty and unmeasured disturbances. The effectiveness of the method is illustrated through a realistic simulation case study of medium optimization of Chinese hamster ovary cell cultivation in continuous monoclonal antibody production, where the metabolic network consists of 23 extracellular metabolites and 126 reactions.

This paper presents an approach for the dynamic speed drop compensation during threading in rolling processes. The feedforward control design exploits the differential flatness of the mechanical model and accelerates both the rolls and the drive train in a manner such that the acceleration torque is equal to the rolling torque during threading, while simultaneously maintaining the roll at the desired target speed. Ideally, this prevents the speed drop and enhances the quality and stability of the rolling process. The flatness-based feedforward trajectories are optimized in an online fashion to determine the optimal initial roll speed and duration of the acceleration process. An extensive experimental validation on a hot strip finishing mill shows superior performance in terms of various key performance indicators in comparison with a standard overspeed approach.

In minerals processing, the froth flotation is one of the widely used process that separates valuable mineral components from their associated gangue materials. The efficiency of this process relies on several factors, such as feed characteristics, particle size, pulp flow rate, pH, conditioning time, aeration, reagents system and many other affecting parameters. These processing parameters significantly impact the overall performance of the flotation process and influence the quality of the final concentrate. For instance, improper pulp flow and reagent dosing systems can result in metal loss and waste, particularly when dealing with frequently changing ore compositions. In this work, we established an Artificial Intelligence-based system which goal is to intelligently monitor flotation circuits and to recommend set-points for the process’s manipulated variables in order to achieve optimal performance.

The system has been developed and evaluated within an industrial flotation plant that processes complex Pb-Cu-Zn sulfide ores. Leveraging an Artificial Neural Network-based Mixture of Experts (MoEs) predictive model, the system accurately estimates the mineral grades in the final concentrate and tailing of the flotation circuit. Moreover, using a Genetic Algorithms-based optimization pipeline, the system recommends set-points for the manipulated variables of the process for a maximum recovery and optimal product quality.

The industrial validation of the predictive component demonstrated a 94% accuracy with a rapid 3s response time. Furthermore, the hypothetical simulation of the optimization component indicated a potential 5% increase in circuit recovery and a 4% increase of lead (Pb) grade in the circuit’s final concentrate. This developed system aims to enhance the control of froth flotation process, stabilize the product quality, and improve the overall economic benefits of production efficiency. This research contributes to the field of manufacturing systems by providing practical data-driven application for the advanced monitoring, optimization and control of industrial processes with a specific emphasis on the froth flotation process.

We study the optimal steady-state operation of processes where the active constraints change. The aim of this work is to eliminate or reduce the need for a real-time optimization layer, moving the optimization into the control layer by switching between appropriately selected controlled variables (CVs) in a simple way. The challenge is that the best CVs, or more precisely the reduced cost gradients associated with the unconstrained degrees of freedom, change with the active constraints. This work proposes a framework based on decentralized control that operates optimally in all active constraint regions, with region switching mediated by selectors. A key point is that the nullspace associated with the unconstrained cost gradient needs to be selected in accordance with the constraint directions so that selectors can be used. A main benefit is that the number of SISO controllers that need to be designed is only equal to the number of process inputs plus constraints. The main assumptions are that the unconstrained cost gradient is available online and that the number of constraints does not exceed the number of process inputs. The optimality and ease of implementation are illustrated in a simulated toy example with linear constraints and a quadratic cost function. In addition, the proposed framework is successfully applied to the nonlinear Williams–Otto reactor case study.

This paper proposes a cascade series control structure and design for two series processes represented by first-order plus dead-time FOPDT models. The proposed controller uses two series predictors, one for each process, and can deal with stable, unstable, or integrative processes. The design follows similar principles of the simplified filtered Smith Predictor (SFSP) for a single-loop dead-time system. Initially, the primary controller, composed of a set-point static gain and two feedback controllers, is tuned to achieve the desired set-point tracking. Then, the predictor filters are tuned to ensure stability, robustness, and disturbance attenuation. Different from standard cascaded SFSP, one of the feedback controllers, instead of only a static gain, includes a finite time integral. The main advantage of this approach is that disturbances generated in the primary or secondary process can be handled independently by the predictor filters, simplifying the tuning procedure and enhancing the overall control performance. Additionally, for the nominal case, the proposed cascaded controller allows obtaining an ideal set-point tracking and disturbance rejection. After the dead-time effect, the proposed cascade controller achieves the set point exponentially and rejects exponentially step-like disturbances where the user defines the time constants of the exponentials. Simulation results demonstrate the advantages of the proposed controller compared to other recently published approaches, mainly in the inner loop disturbance rejection, which is precisely what is expected from a series cascade controller.