Abstract: This plenary talk discusses real-time optimization (RTO) strategies for improving process performance in the presence of uncertainty in the form of plant-model mismatch, drifts and disturbances. RTO typically uses a plant model to compute optimal inputs. In the presence of uncertainty, selected model parameters can be estimated and the updated model used for optimization. Although very intuitive, this two-step approach suffers from the fact that the model is almost invariably “inadequate”, which prevents from reaching the plant optimum. Other approaches have been developed in the last two decades to overcome this difficulty. Recently, a generic formalization of these ad hoc fixes has been proposed under the label modifier adaptation. The basic idea is to leave the model parameters unchanged but to use plant measurements to “appropriately” modify the optimization problem. The modifier-adaptation approach will be presented and compared to the two-step approach, in particular with regard to model adequacy.

We will then go beyond this comparison and discuss different ways of using plant measurements for process improvement in the presence of uncertainty. There are many questions to be addressed: (i) what can be done off-line prior to process operation, and what should be performed in real time, (ii) how much of the optimization effort is model-based and how much is data-driven, (iii) what to measure, what to adapt and how to adapt? Two case studies concerned with kite power and semi-batch reactors will illustrate the applicability of the various approaches.

Keywords: Process and Control Monitoring, Process Applications
Abstract: Alarms indicate abnormal operation of the process plants and alarm floods constitute specific abnormal episodes that cannot be handled safely by the operators. In that regard, online alarm flood classification based on a bank of past historical episodes provides support on how to handle ongoing alarm sequences. This paper introduces a new approach based on alarm coactivations that is appropriate for the analysis of ongoing sequences. The method shows improvements when compared to an established sequence alignment approach for abnormal episode analysis of a gas oil separation plant.

Keywords: Optimization and Scheduling, Model-based Control
Abstract: In this paper, we consider the decomposition of scenario-based model predictive control problem. Scenario MPC explicitly considers the concept of recourse by representing the evolution of uncertainty by a discrete scenario tree, which can result in large optimization problems. Due to the inherent nature of the scenario tree, the problem can be decomposed into each scenario. The different subproblems are only coupled via the non-anticipativity constraints which ensures that the first control input is the same for all the scenarios. This constraint is relaxed in the dual decomposition approaches, which may lead to infeasibility of the non-anticipativity constraints if the master problem does not converge within the required time. In this paper, we present an alternative approach using primal decomposition which ensures feasibility of the non-anticipativity constraints throughout the iterations. The proposed method is demonstrated using gas-lift optimization as case study.

Keywords: Big Data Analytics and Monitoring
Abstract: Recently, the combination of cointegration analysis (CA) and slow feature analysis (SFA), has been adopted for concurrent monitoring of operation condition and process dynamics for nonstationary dynamic processes subject to time variant conditions. By isolating long-term temporal equilibrium features and specific temporal slow features from steady-state information, the CA-SFA based monitoring scheme can well distinguish between the changes of operation conditions and real faults. Considering that the temporal variation can provide an indication of control performance changes, the CA-SFA algorithm is further exploited based on dissimilarity analysis of temporal distribution to explore its unique efficacy in control performance monitoring (CPM). Two attractive features of the proposed approach are noticed. First, it is compatible with various operation conditions simultaneously including multifarious steady states and dynamic switchings between different working points. Second, a new performance monitoring index is used to monitor the control performance by quantifying the distribution structure of temporal features against the benchmark from both fast and slow dynamics aspects. Case study on a chemical industrial scale multiphase flow experimental rig shows the feasibility of the new CPM method.

Keywords: Big Data Analytics and Monitoring
Abstract: Improvements in sensing, connectivity and computing technologies mean that industrial processes now generate a vast amount of data from a variety of disparate sources. Data may take a number of different forms, from different time-domain signals, sampled at different rates using various types of sensors, through to more disparate sources such as alarm and event logs. New process and condition monitoring techniques are needed to be developed to tackle the new challenges of big and heterogeneous data. Although there are a few publicly available benchmark studies, e.g. the Tennessee Eastman process plant (Ricker, 1995), a multiphase flow benchmark case for statistical process monitoring (Ruiz-Cárcel et al., 2015), they provide only standard process data. This work presents a benchmark case on an industrial scale multiphase flow facility. Various operational conditions were tested under normal operating modes as well as with seeded faults. Heterogeneous data was collected from various sources, including process data, alarm data and high frequency ultrasonic and pressure data. Two different fault detection algorithms are applied to the data, a multivariate PCA-enhanced Canonical Variate Analysis (CVA) and a probabilistic Bayesian method. This benchmark case study with data from disparate sources can be used for algorithm development and validation for fault detection, fault identification, fault classification, fault severity detection, monitoring of fault evolution and prognostics.

Keywords: Big Data Analytics and Monitoring, Modeling and Identification
Abstract: Spectroscopic techniques such as near-infrared spectroscopy have gained wide applications in the last few decades. As a result, various soft sensors have been developed to predict sample properties from the sample’s spectroscopic readings. Because the readings at different wavelengths are highly correlated, it has been shown that variable selection could significantly improve a soft sensor’s prediction performance and reduce the model complexity. Currently, almost all variable selection methods focus on how to select the variables (i.e., wavelengths or wavelength segments) that are strongly correlated with the dependent variable to improve the prediction performance. Although many successful applications have been reported, such variable selection methods do have their limitations, such as highly sensitive to the choice of training data, and poorer performance when testing on new samples. This is because the variables that are removed from model building may contain useful information about the sample property. To address this limitation, we propose a statistics pattern analysis (SPA) based variable selection method for chemometric modeling. Instead of selecting certain wavelengths or wavelength segments, the SPA-based method considers the whole spectrum which is divided into segments, and chooses different features over each spectrum segment to build the soft sensor. Two case studies are presented to demonstrate the performance of the SPA-based soft sensor and compared with a full partial least squares (PLS) model, and a synergy interval PLS (SiPLS) model.

Keywords: Big Data Analytics and Monitoring
Abstract: The modern steel industry aims to produce high-quality products with higher product yield, lower costs, and lower energy consumption to meet market demands. To accomplish these goals, it is necessary to reduce or eliminate product defects. However, the relationship of operating conditions to the defect formation is not fully understood. There is increasing interest in developing models to monitor the quality and predict the number of defects in real time. Modeling and analyzing the defect count data is a very challenging problem because the defect count data exhibit the unique characteristics of non-negative integers, overdispersion, high skewed distribution, and excess zeros. To explicitly account for these unique characteristics, the present work develops an on-line quality monitoring and prediction system based on the hurdle regression model. The basic idea of the hurdle model is that a binomial model governs the binary outcome of the dependent variable being zero or positive. If the dependent variable takes a positive value, hurdle is crossed, and the conditional distribution of the positives can be modeled by a zero-truncated Poisson or negative binomial (NB) model. Compared to Poisson and NB models, the hurdle model is not only suitable for modeling discrete and non-negative integer data, but also sufficient for handling both overdispersion and excess zeros data. The effectiveness of the hurdle model was verified through its application to the real defect data of a steelmaking plant. The results have demonstrated that the hurdle NB model is superior to the Poisson, NB, hurdle Poisson, and PLS models in the prediction performance.

Keywords: Big Data Analytics and Monitoring
Abstract: In industrial processes, various types of faults often propagate from one unit to another along information and material flows. In severe cases, fault propagation can eventually affect the entire plant, leading to the reduction in product quality and productivity, and even causing damages. In order to avoid these issues, effective root cause diagnosis is desired because the correct identification of the sources of process abnormalities is critically important for restoring the system to its normal condition in a timely manner. In recent years, the data-driven causality analysis method, such as Granger causality (GC) test, has been adopted to identify the causes of process faults. However, the conventional pairwise GC only considers the causal relationship between a pair of time series. In multivariate cases, repeated pairwise analyses are often conducted, which yet often give over-complex and misleading results. To solve this problem, in this research, the multivariate GC technique, which measures the conditional dependence between time series, is utilized to construct the causal map between process variables. In addition, the obtained causal map is further simplified by finding its maximum spanning tree, facilitating the identification of the root cause. The feasibility of the proposed method is illustrated by case studies.

Keywords: Big Data Analytics and Monitoring, Process Applications
Abstract: Slow feature analysis has proven to be an effective process monitoring and fault diagnosis approach. By isolating temporal behaviors from steady-state variations in process data, slow feature analysis enables a concurrent monitoring of operating condition and process dynamics, based on which false alarms triggered by nominal operating condition deviations can be effectively removed. However, the present formulation of slow feature analysis only makes use of the first-order time difference of time series data, thereby falling short of addressing high-order dynamics in process operations. In this work, we propose a second-order formulation of slow feature analysis, and further develop a systematic framework for process monitoring and fault diagnosis, which can provide more meaningful information about process dynamics to assist decision-making of operators. Case studies on the Tennessee Eastman benchmark process are conducted to demonstrate the efficacy of the proposed method.

Keywords: Model-based Control, Optimization and Scheduling, Process Applications
Abstract: This paper proposes a novel formulation of economic MPC for nonlinear discrete-time systems that is able to drive the closed-loop system to the (unknown) optimal equilibrium, despite the presence of plant/model mismatch. The proposed algorithm takes advantage of: (i) an augmented system model which includes integrating disturbance states as commonly used in offset-free tracking MPC; (ii) a modifier-adaptation strategy to correct the asymptotic equilibrium reached by the closed-loop system. It is shown that, whenever convergence occurs, the reached equilibrium is the true optimal one achievable by the plant. An example of a CSTR is used to show the superior performance with respect to conventional economic MPC and a previously proposed offset-free MPC still based on a tracking cost. The implementation of this offset-free economic MPC requires knowledge of plant input-output steady-state map gradient, which is generally not available. To this aim a simple linear identification procedure is explored numerically for the CSTR example, showing that convergence to a neighborhood of the optimal equilibrium is possible.

Keywords: Optimization and Scheduling, Model-based Control
Abstract: We present a fast sensitivity-based nonlinear model predictive control (NMPC) algorithm, that can handle non-unique multipliers in the discretized dynamic optimization problem. Non-unique multipliers may arise, for example when path constraints are active for longer periods of the prediction horizon. This is a common situation in economic model predictive control. In such cases, the optimal nonlinear programming (NLP) solution often satisfies the Mangasarian-Fromovitz constraint qualification (MFCQ), which implies non-unique, but bounded multipliers. Consequently, any sensitivity-based fast NMPC scheme must allow for discontinuous jumps in the multipliers. In this paper, we apply a sensitivity-based path-following algorithm that allows multiplier jumps within the advance-step NMPC (asNMPC) framework. The path-following method consists of a corrector and a predictor step, which are computed by solving a system of linear equations, and a quadratic programming problem, respectively, and a multiplier jump step determined by the solution of a linear program. We demonstrate the proposed method on an economic NMPC case study with a CSTR.

Keywords: Model-based Control
Abstract: Backstepping controller (BS) and model predictive controller (MPC) have been widely used for many applications by virtue of their own merits. BS works even with non-minimum phase and finite-time escape and MPC can handle state and input constraints explicitly. Nevertheless, BS requires repeated differentiations of the virtual control, whereas high computational loads of MPC are obstacles to practical implementation. This study proposes a control strategy that combines BS and MPC for nonlinear systems in strict-feedback form. It is proven that the controller renders the closed-loop system asymptotically stable. The proposed MPC-BS requires less computational load than that of MPC, since it only optimizes the virtual input of the first step and computes the input by backstepping approach. The explosion of terms caused by the consecutive differentiation in BS approach is also addressed.

Keywords: Process Applications, Modeling and Identification, Big Data Analytics and Monitoring
Abstract: Online estimation of internal states and parameters is often required for process monitoring, control and fault diagnosis. The conventional approach to estimate drifting parameters is to artificially model them as a random walk process and estimate them simultaneously with the states. However, tuning of the random walk model is not a trivial exercise. Recently, Valluru et al. (2017) have developed a moving window based state and parameter estimator which assumes that the parameters change slowly and remain constant within the window. Also, in another development, a moving window based recursive filter, receding horizon nonlinear Kalman (RNK) filter has been proposed by Rengaswamy et al. (2013). In this work, a novel simultaneous state and parameter estimator is proposed by combining the window based parameter variation model with RNK filter formulation. The performance of the RNK based estimator is demonstrated by conducting simulation studies on the benchmark quadruple tank system and a CSTR system. The efficacy of RNK based estimator is compared with that of the conventional simultaneous EKF approach and Moving Horizon Estimator (MHE) based state and parameter approach. Analysis of the simulation results reveals that the proposed state and parameter estimation scheme is able to generate better estimation performance than that of the simultaneous EKF and closer to that of the MHE based parameter estimator with less computational efforts.

Keywords: Batch Process Modeling and Control, Model-based Control
Abstract: Economic model predictive control is a popular method to maximize the efficiency of a dynamic system. Often, however, uncertainties are present, which can quickly lead to lower performance and constraint violations. In this paper, a new approach is proposed that incorporates the square root Unscented Kalman filter directly into the optimal control problem to estimate the states and to propagate the mean and covariance of the states to consider noise from disturbances, parametric uncertainties and state estimation errors. The covariance is propagated up to a predefined “robust horizon” to limit open-loop covariances, and chance constraints are introduced to maintain feasibility. Often variables in chemical engineering are non-negative, which however can be violated by the Unscented Kalman filter leading to erroneous predictions. This problem is solved by log-transforming these variables to ensure consistency. The approach was verified and compared to a nominal nonlinear model predictive control algorithm on a semi-batch reactor case study with an economic objective via Monte Carlo simulations.

Keywords: Model-based Control, Batch Process Modeling and Control, Modeling and Identification
Abstract: Thermal runaways cause significant safety issues and financial loss for industrial batch reactors due to the disruption of normal operation. The intensification of processes is restricted, since control systems are not capable of detecting stability boundaries of the system and hence are overly conservative. For this purpose Lyapunov exponents are introduced as a stability criterion. It is shown that Lyapunov exponents can correctly predict the stability of batch reactor systems. This stability criterion is embedded in Model Predictive Control, which results in a novel control scheme. This scheme allows the controlled increase of the reaction temperature to achieve a target conversion in a reduced completion time of reaction.

Keywords: Process Applications, Optimization and Scheduling
Abstract: We analyze and exemplify how some classical control structures can be used for optimizing a chemical process directly. Some key structures in this context are split-range control and mid-ranging control. In applications, split-range control is primarily used to manage several manipulated variables (valves) affecting the same controlled variable, without explicit reference to a control objective, or optimization criterion. However, the same basic principle can be used for directly optimizing a process, or subprocess. In some cases that scheme provides a simpler and more clear cut solution than MPC. A similar observation applies to mid-ranging control (also known as input resetting control, or valve position control).

Keywords: Optimization and Scheduling, Process Applications, Energy Processes and Control
Abstract: Control structure design for the large-scale hydrodealkylation of toluene (HDA) process has been extensively studied. The systematic procedure based on self-optimizing control was successfully applied to the HDA process with a promising control configuration. Besides of the active constraints identified, the remaining self-optimizing controlled variables are selected as the mixer outlet methane mole fraction and the quencher outlet toluene mole fraction, which are nonetheless single measurements. In this paper, we consider control reconfiguration for the HDA process by selecting measurement combinations as CVs to improve the process optimality. To this end, the recently proposed global self-optimizing control (gSOC) approach is employed for CV selection and the partial bidirectional branch and bound algorithm (PB^3) is used for fast-screening measurement candidates. Numerical controlled variables are derived and the economic optimality is improved by the reconfiguration.

Keywords: Optimization and Scheduling, Process Applications, Model-based Control
Abstract: With "classical advanced control" we mean the control structures that are commonly used in industry for multivariable control. These have been in use for at least 50 years, but surprisingly there is little literature published on how to design such structures in a systematic manner. We present a design procedure to assure optimal operation when active constraint changes occur. In this paper, we focus on input saturation. We suggest to use a priority list of constraints as an important first step of the presented design procedure. We also discuss how to handle input saturation using split range control, valve position control (input resetting), and selectors. As a case study, we consider optimal operation and a priority list of constraints for a cooler with temperature and flow control, and evaluate alternative classical advanced control implementations that maintain optimality.

Keywords: Model-based Control, Optimization and Scheduling, Modeling and Identification
Abstract: This paper introduces an Interaction Measure named Prediction Error Index Array (PEIA), which can be applied both to linear and non-linear systems. The linear PEIA is constructed as an extension of previous results using the H2-norm. The non-linear PEIA is an extension for systems represented by Volterra series. Additionally, the paper gives an interpretation of both linear and nonlinear PEIA based on the prediction error of the structurally reduced model which results from the control configuration selection. Examples illustrate and compare the interaction measure with established methodologies, like the relative gain array, participation matrix, and Hankel Interaction Index array.

Keywords: Model-based Control
Abstract: In this study, a modified version of Hankel Interaction Index Array (HIIA) for control configuration selection is presented which can overcome some of its shortcomings, like e.g. scaling dependency, or not relating to closed loop system properties. Inspired by the relative gain array approach, the HIIA is reformulated in the relative gain thinking by considering the effect of closing loops. The ratio of the Hankel norm of the subsystems in closed and open loop are used to state a modified version of HIIA, which has improved characteristics compared to the original HIIA. Properties of the modified HIIA are discussed and benchmarked with established methods on three example cases.

Keywords: Process Applications
Abstract: Self-optimizing control focuses on minimizing loss for processes in the presence of disturbances by holding selected controlled variables at constant set-points. A measurement combination can be found, using the Null-space method, which further reduces the loss. Since self-optimizing control focuses on the steady-state operation, little attention has been put on the dynamic performance when selecting measurement combinations. In this work, an iterative LMI approach is combined with the sparsity promoting weighted l1-norm, to find a measurement subset together with PI controllers for the Null-space method. The measurement combination and the controllers are designed such that, the dynamic response is improved when the process is facing disturbances. The proposed method is illustrated on a Petlyuk column case study.

Keywords: Big Data Analytics and Monitoring, Batch Process Modeling and Control, Process Applications
Abstract: Data collected from Industry 4.0 scenarios present a variety of data structures, reflecting the evolution of industrial processes, measurement systems and IT infrastructures (“variety” is actually one of the 4 V’s of Big Data, meaning that its existence is widely recognized). Data analytics platforms must adapt to this context and keep the pace of its evolution, in order to continue providing effective solutions to practitioners for dealing with the large data resources now available. In this context, one prevalent feature of industrial data has been largely overlooked: their multiresolution nature. The multiresolution nature of data is directly connected to their granularity in the time domain, an aspect that induces inner dependencies that current frameworks cannot address in a consistent and rigorous way. Furthermore, multiresolution has been often mistaken as a simple multirate scenario, where in fact the meaning of the observations is completely different. In this paper, we highlight such differences and discuss current multiresolution frameworks for effectively handling industrial data sets.

Keywords: Big Data Analytics and Monitoring
Abstract: Process monitoring is considered to be one of the most important problems in process systems engineering, which can be benefited significantly from deep learning techniques. In this paper, deep neural networks are applied to the problem of fault detection and classification to illustrate their capability. First, the fault detection and classification problems are formulated as neural network based classification problems. Then, neural networks are trained to perform fault detection, and the effects of two hyperparameters (number of hidden layers and number of neurons in the last hidden layer) and data augmentation on the performance of neural networks are examined. Fault classification problem is also tackled using neural networks with data augmentation. Finally, the results obtained from deep neural networks are compared with other data-driven methods to illustrate the advantages of deep neural networks.

Keywords: Big Data Analytics and Monitoring, Modeling and Identification
Abstract: In this paper, a novel dynamic-inner canonical correlation analysis (DiCCA) algorithm is proposed to model high dimensional dynamic data. DiCCA extracts latent variables with descending dynamics, which are referred to as principal time series. Since DiCCA requires the principal time series to have maximal predictability, the most important dynamic features in the data are guaranteed to be extracted first. Therefore, the lower dimensional principal time series are able to provide good representation of the dynamic features, leading to the ease of interpretation and visualization. A case study on Eastman oscillating data demonstrates the effectiveness of the proposed method.

Keywords: Big Data Analytics and Monitoring, Process Applications, Modeling and Identification
Abstract: Ethylene cracking furnace is a key unit in the ethylene production process, whose operating condition is subject to changes with fluctuation in feed condition, the aging of equipment, and other possible disturbances. To promptly and correctly identify the root causes of process changes, an online process monitoring system based on Principal Component Analysis (PCA) and Signed Directed Graph (SDG) method is proposed for multiple operational conditions monitoring and fault detection. Active process adjustments or passive fluctuations are first differentiated, then the root cause is isolated by SDG reasoning based on the contribution percentage of principal variables in PCA. The True Positive Rate (TPR) of fault detection is 98%, and False Alarm Rate (FAR) is 1.56%. The root causes of fault match very well with operation records.

Keywords: Big Data Analytics and Monitoring, Process Applications
Abstract: In this work, a new algorithm is developed to identify stochastic faults in the Tennessee Eastman (TE) process, which integrates Ensemble Empirical Mode Decomposition (EEMD), Principal Component Analysis (PCA), Cumulative Sum (CUSUM), and half-normal probability plot to detect three particular faults that could not be properly detected with previously reported techniques. This algorithm includes three steps: measurements pre-filtering, sensitivity analysis, and fault detection. Measured variables are first decomposed into different scales using the EEMD-based PCA for extracting fault signatures, from which a subset of variables that are sensitive to faults are selected with the half-normal probability plot. Based on the specific variables, CUSUM-based statistics are further used for improved fault detection. The algorithm developed in this work can successfully identify three particular faults with small time delay.

Keywords: Big Data Analytics and Monitoring, Batch Process Modeling and Control, Process Applications
Abstract: In the processing industries, operating conditions often change to meet the requirements of the market and customers. To cope with the difficulty of on-line quality prediction for such multi-grade processes widely operated in process industries, a novel common and special feature extraction method is proposed for modeling multi-grade processes. A common feature extraction algorithm is proposed to determine the common directions shared by different grades of these processes. After extracting the common features, a partial least-squares modelling algorithm is used to extract the special directions for each grade, respectively. Hence, product quality prediction can be simply conducted by integrating the common and special parts of each grade for model building. A numerical case and an industrial polyethylene process are used to demonstrate the effectiveness and advantage of the proposed method.

Keywords: Model-based Control, Process Applications, Batch Process Modeling and Control
Abstract: Plant-model mismatch and estimation errors are critical issues in the practical implementation of Nonlinear Model Predictive Control (NMPC). To address these challenges, we formulate a robust output feedback NMPC scheme that is real-time implementable and provides robust constraint satisfaction in the presence of parametric and additive uncertainties, and estimation errors. The robustness is achieved by combining the tube-based and the multi-stage NMPC approaches. Two controllers are used in the proposed framework: a primary controller with tightened constraints that optimizes a given objective and an ancillary controller that tracks the trajectories provided by the primary controller. Unlike standard tube-based NMPC, the primary controller predicts different state trajectories for different realizations of the most important uncertainties using the multi-stage NMPC framework. The synergy between the two approaches leads to a better trade-off between optimality and complexity. The advantages of the proposed approach are demonstrated for an industrial-scale fed-batch polymerization reactor example.

Keywords: Model-based Control, Optimization and Scheduling, Process Applications
Abstract: We present a new method for adaptively updating nonlinear model predictive control(NMPC) horizon lengths online via nonlinear programming (NLP) sensitivity calculations. This approach depends on approximation of the infinite horizon problem via selection of terminal conditions, and therefore calculation of non-conservative terminal conditions is key. For this, we also present a new method for calculating terminal regions and costs based on the quasi-innite horizon framework that extends to large-scale nonlinear systems. This is accomplished via bounds found through simulations under LQR control. We show that the resulting controller is asymptotically stable. Finally, we demonstrate this approach on a quad-tank system. Simulation results reveal that the proposed approach is able to achieve significant reduction in the average computation time without much loss in the performance with reference to fixed horizon NMPC.

Keywords: Model-based Control, Big Data Analytics and Monitoring, Modeling and Identification
Abstract: Model predictive control (MPC) is a popular control strategy that computes control actions by solving an optimization problem in real-time. Uncertainty and nonlinearity of a process, and the non-convexity of the resulting optimization problem can make online implementation of MPC nontrivial. Consequently, MPC is most often used in processes where the time constants are large and/or high-performance computing support is available. We propose a deep neural network (DNN) controller architecture to reduce the computational cost of implementing an MPC. This is done by training a DNN controller on simulated input-output data from a well-designed MPC. The online implementation of a DNN controller does not require solving an optimization problem. Once the DNN is trained, the MPC is fully replaced with the DNN controller. The benefits of this approach are illustrated through a simulated example.

Keywords: Modeling and Identification, Model-based Control
Abstract: This paper proposes a parallelizable real-time algorithm for integrated experiment- design model predictive control (MPC). Integrated experiment design MPC is needed if a system is not observable at a tracking reference and needs to be excited on purpose in order to be able to estimate the system’s states and parameters. The contribution of this paper is a real-time model predictive control (MPC) algorithm using two processors. On the first processor an extended Kalman filter as well as a parametric certainty-equivalent MPC controller are implemented, which can provide immediate feedback at high sampling rates. On the second processor, an optimal experiment design (OED) problems are solved in parallel in order to perturb the certainty-equivalent MPC control loop improving the accuracy of the state estimator at a lower sampling rate. We show that this framework can achieve optimal tradeoffs between OED and control objectives. The approach is applied to a biochemical process in order to illustrate that the proposed controller can achieve superior control performance when compared to certainty- equivalent MPC.

Keywords: Model-based Control
Abstract: A parameterized network topology based distributed model predictive control (DMPC) framework is proposed in this work, it is mainly applied in system reconfiguration and sensor fault-tolerant control. The Lyapunov stability condition for DMPC with a parameterized network topology is derived. Regarding to system reconfiguration, the parameterized network topology is served as the explicit reconfiguration model. Furthermore, for fault-tolerant control with sensor bias, the parameterized network topology is used to compensate the sensor fault, and a residual generator is designed by states of predictor and consider a time varying threshold for fault detection. The proposed approach is able to handle the system reconfiguration and fault-tolerant control without backup controllers or controllers redesign, and there is no need of the information of fault because of the using of predictor.

Keywords: Model-based Control, Process Applications, Optimization and Scheduling
Abstract: Simulated moving bed (SMB) processes have been applied to petrochemical, pharmaceutical, and fine chemical industries to separate products in high purity and yield since it was introduced in 1960s. Owing to process complexity and operational sensitivity, the control and dynamic optimization of the SMB process are still challenging issues. In this work, a simplified process model with linear isotherms was introduced to estimate process states of conventional four-zone SMB chromatography. In a simulation study, a controller can estimate current process states and find the optimal operating conditions ‘switch by switch’ up to moderately nonlinear ranges of the competitive Langmuir isotherms. Furthermore, the controller works for the process with system void volumes and delayed feedback information.

Keywords: Model-based Control
Abstract: An engineer always has to make assumptions about the system boundary. In this paper, the impact of neglected dependencies of manipulated variables on the disturbance variables as example for said assumptions is investigated in the context of self-optimizing control. The feedback through dependent disturbances influences both the optimal operating point and the combination of measurements. As a case study, we consider an ammonia synthesis reactor with a simplified model for the ammonia separation and the recycle. The disturbance dependency changes the optimal selection matrices through the recycle. However, we find that it is possible to neglect the recycle in the selection of the controlled variables for this example if the setpoint is adjusted.

Keywords: Process Applications, Energy Processes and Control, Modeling and Identification
Abstract: Integrated process and control design approach for cyclic distillation columns is proposed. The design methodology is based on application of simple graphical design approaches, known from simpler conventional distillation columns. Here, a driving force approach and McCabe-Thiele type analysis is combined. It is demonstrated, through closed-loop and open-loop analysis, that operating the column at the largest available driving force results in an optimal design in terms of controllability and operability. The performance of a cyclic distillation column designed to operate at the maximum driving force is compared to alternative sub-optimal designs. The results suggest that operation at the largest driving force is less sensitive to disturbances in the feed and inherently has the ability to efficiently reject disturbances.

Keywords: Model-based Control, Modeling and Identification, Process Applications
Abstract: This paper proposes the use of the extent decomposition on homogeneous reaction systems for control purposes. The decomposition results in a Linear Parameter-Varying (LPV) representation, upon which parametric feedback and feedforward strategies are developed. In the first part of the paper, three different ways to obtain the Extent-Based LPV (ELPV) representation of the system are proposed. The representation is advantageous since the physical meaning of all the variables are kept and it has a Jordan type of structure which is used to establish controllability conditions. In the second part, a general parametric feedback and feedforward control laws are proposed for the ELPV system. The nonlinear state-parameter dependence is first considered in the feedback term. This allows to convert the original ELPV system into a Linear Time Invariant (LTI) system, which is used to design optimal control laws for reference tracking. Finally, the performance of the control strategy for ELPV system is illustrated in simulation and compared with a controller based on a constant-parameter LTI model (ELTI).

Keywords: Model-based Control, Batch Process Modeling and Control, Energy Processes and Control
Abstract: This article provides a comparison of control strategies for an economically attractive vapor-recompressed batch distillation (VRBD) column. A VRBD column exhibits nonlinear dynamics with strong inter-stream interactions due to energy-integration, which can give rise to difficulties in controlling the operation using traditional linear controllers. In this paper, we propose and compare model-based control strategies such as linear quadratic regulator (LQR), linear model predictive control (LMPC) and Globally linearizing controller (GLC) for a VRBD column to achieve desired distillate purity with minimum energy consumption. The effectiveness of these control strategies is illustrated using a simulation case study of benzene/toluene separation in a VRBD column.

Keywords: Process Applications, Energy Processes and Control, Modeling and Identification
Abstract: For the process of biogas pressurized water scrubbing, design and simulation of plant-level dynamic control system is researched based on the optimum structure and parameters of the steady state in this paper. According to the process requirements of the biogas pressurized water scrubbing pilot-scale plant, a steady state simulation system is established using Aspen Plus. The simulation research under different operation conditions is carried out to obtain the optimal operating parameters. Then plant-level dynamic control schemes and the dynamic simulation system are also designed to ensure the efficiency of biogas purification. The system’s performance is tested under introducing biogas inlet flow disturbance and intake flow component disturbance. The simulation results show that the system has good dynamic response performance, such as the removal rate of CO2 is greater than 99.8% under various disturbances.

Keywords: Process Applications
Abstract: The parametric identification problem for industrial crude distillation unit (CDU) is considered. We take the a priori knowledge of the process into account by using a system of constraints for parameters of soft sensors models. The identification problem is transformed into a constrained optimization problem, which we solved using the active set method. The static and dynamic soft sensors are evaluated for industrial CDU located at JSC “Gazprom neftekhim Salavat” refinery. It was found that the model performed better when we used the proposed constrained optimization approach for identification instead of robust regression methods.

Keywords: Model-based Control, Modeling and Identification, Process Applications
Abstract: In this paper, we consider state estimation of wastewater treatment plants based on model approximation. A wastewater treatment plant described by the Benchmark Simulation Model No.1 (BSM1) is used. We use the proper orthogonal decomposition approach with re-identification of output equations to obtain a reduced-order model for the original system and then use the reduced-order model in state estimation. An approach on how to determine an appropriate minimum measurement set is also proposed. A continuous-discrete extended Kalman filtering algorithm is used to design an estimator based on the reduced-order model with re-identified output equations. The estimator gives good state estimates for the actual process.

Keywords: Process Applications
Abstract: Poly(oxymethylene) dimethyl ethers (OME) are widely used in the reduction of soot formation in diesel engines, and also as physical solvents for absorbing carbon dioxide from natural gas. In this work, a new production process of OME is designed and rigorously studied. In the conventional process, multiple process sections are required to first separately produce trioxane and methylal, and then to react of these two intermediates to produce OME. In this paper, the plant-wide process via a much simply direct synthesis of OME from formaldehyde and methanol in aqueous solutions is newly developed. The overall process includes an upstream section and a downstream product purification section. After the steady-state simulation of the process is established, the process flowsheet is optimized by minimizing total annual cost (TAC). The dynamics and control of the proposed OME production process is also investigated in this study. A proper control strategy is developed for the optimized process to reject the process disturbances. Copyright © 2018 IFAC

Keywords: Model-based Control
Abstract: A production system which produces plenty of items in many steps can be modelled as a continuous flow problem governed by a nonlinear and nonlocal hyperbolic partial differential equation. One of important ways to adjust the output such process is by manipulating the start rate. This paper considers the control and regulation by proportional-integral (PI) controllers for the continuum production systems. In the considered system, the input and output are located on the boundary. In particular, the closed-loop stabilization of the linearized model with the designed PI-controller is proved using the method of spectral analysis and the Lyapunov theory. Numerical results demonstrate successful tracking for step inputs in the demand rate.

Keywords: Model-based Control, Modeling and Identification, Process Applications
Abstract: In this paper, the authors have proposed an ensemble Kalman filter based stochastic model predictive control algorithm to determine the optimal control policy at every sampling time instant for a constrained stochastic linear system. To determine the optimal control policy for the system affected by random disturbances and measurements corrupted by random noise, the authors have minimized the uncertain objective function, subject to uncertain state and output constraints and deterministic input constraints using the quantile based scenario analysis approach. In this work, ensemble Kalman filter is being employed, to generate a recursive estimate of states of the constrained stochastic linear system. The number of scenarios is considered to be equivalent to that of number of sample points in the ensemble Kalman filter. Each scenario is viewed as one realization of the process noise, measurement noise over the prediction horizon as well as the ith particle of the state estimate at the beginning of the prediction horizon generated by the ensemble Kalman filter. Simulation studies have been carried out to assess the efficacy of the proposed control scheme on the simulated model of the constrained single-input and single-output linear stochastic system.

Keywords: Model-based Control
Abstract: This paper shows that for a class of nonlinear systems with a lower-triangular structure, the problem of semi-global asymptotic stabilization is solvable by sampled-data output feedback, without requiring restrictive conditions on the nonlinearities and unmeasurable states of the system, such as linear growth, output-dependent growth or homogeneous growth conditions as commonly assumed in the case of global output feedback stabilization. The main contribution is to point out that semi-global asymptotic rather than practical stabilizability of certain classes of nonlinear systems is still possible by sampled-data output feedback if a sampling time is small enough. A design method is also given for the construction of semi-globally stabilizing, sampled-data output feedback controllers.

Keywords: Model-based Control
Abstract: The problem of robust output regulation is studied for a class of parameter uncertain systems under unity output feedback control. The objective is tracking of the desired reference trajectory in the presence of the disturbance, both generated by a common exosystem. Because of the anti-stability of the exosystem and potentially unbounded reference trajectory, the output tracking error is employed as the measurement signal and used as the input to the feedback controller. The method of p-copy of the internal model is utilized to augment the plant dynamics. Assuming that the output regulation condition is satisfied for all the parameter uncertainties, it is shown that the problem of robust output regulation is equivalent to the problem of robust output stabilization. Furthermore for quadratically bounded parameter uncertainties, an application of the notion of the quadratic stability leads to H-infinity based robust control, and the maximum allowable uncertainty bound can be computed, below which the robust output regulation can be achieved.

Keywords: Modeling and Identification, Process Applications
Abstract: As a key stage in polyester production, polymerization process is difficult to model due to its complex reaction mechanism. As a result, online viscosity prediction in industrial polyester polymerization processes is not an easy task. An efficient data-driven prediction model is considered in this work. In order to solve the problem of low accuracy of the online viscosity measuring instrument and considerably time-consuming laboratory analysis, variables that are easily monitored during the polymerization process, i.e. temperature and pressure in the main reactor as well as the viscometer values, are selected to establish an Extreme Learning Machine (ELM) viscosity prediction model. A Just-in-time-based ELM model was established to predict the viscosity values under multi-mode operating and multi-standard production conditions. Consequently, without relying on the time-consuming laboratory analysis process, the PET viscosity can be predicted online. The industrial PET viscosity prediction results show the improved prediction performance of the proposed modeling approach in comparison with ELM and JPCR (Just-in-time principal component regression) approaches.

Keywords: Modeling and Identification, Big Data Analytics and Monitoring
Abstract: The Gaussian mixture models (GMM) is an effective tool for modeling processes with multiple operating modes that widely exist in industrial process systems. Traditional supervised version of GMM, namely the Gaussian mixture regression (GMR), for developing soft sensors merely relies on the labeled samples. However, labeled samples in the soft sensor application are usually very infrequent due to economical or technical limitations, which may lead the GMR to unreliable parameter estimation and finally poor performance for predicting the primary variable. To tackle this problem, a semisupervised GMM for regression purpose is proposed, where both labeled and unlabeled samples take effect, and the Gaussian parameters and regression coefficients are learned simultaneously based on the expectation-maximization algorithm. Two case studies are carried out using simulated dataset and real-life dataset collected from a primary reformer in an ammonia synthesis process, which demonstrates the effectiveness of the proposed method.

Keywords: Modeling and Identification, Optimization and Scheduling, Process Applications
Abstract: Abstract: The periodic operation of a non-isothermal continuous stirred tank reactor (CSTR) using inlet temperature modulation is investigated in this paper. The DC component of the CSTR output concentration is optimized by tuning the modulation parameters of the inlet temperature in order to achieve a maximum conversion using a Nonlinear Output Frequency Response Functions (NOFRFs) based approach. The results show that the new approach is fast and efficient in the analysis and design of the periodic operation of CSTR and can potentially be applied to conduct the optimal design of periodic operation of other chemical engineering processes.

Keywords: Modeling and Identification
Abstract: Online learning based Least Squares Support Vector Machine (LSSVM) can address the modeling problems of a time-varying process, which has a few advantages such as low training time and good general. Nevertheless, many of online learning algorithms cannot adapt the kernel parameters for the time-varying characteristic, so the inferred LSSVM models are low-accuracy. An online learning algorithm with time-varying kernels is proposed to improve online training accuracy of LSSVM model. The kernel parameters are optimized along with time-varying process using updating samples data. To achieve reliable performance during online optimization, we propose a controllable metaheuristic algorithm that adopts a contracted particle swarm optimization with an elaborate chaotic operator. The proposed modeling approach is utilized in the energy efficiency prediction of the electrical smelting process, and the experimental results show that the proposed online learning algorithm can both improve the accuracy of LSSVM model and ensure low online training time.

Keywords: Modeling and Identification, Optimization and Scheduling
Abstract: Kernel-based modeling methods have been used widely to estimate some difficulty-to-measure quality or efficient indices at different industrial applications. Least square support vector machine (LSSVM) is one of the popular ones. However, its learning parameters, i.e., kernel parameter and regularization parameter, are sensitive to the training data and the model’s prediction performance. Ensemble modeling method can improve the generalization performance and reliability of the soft measuring model. Aim at these problems, a new adaptive selective ensemble (SEN) LSSVM (SEN-LSSVM) algorithm is proposed by using multiple learning parameters. Candidate regularization parameters and candidate kernel parameters are used to construct many of candidate sub-sub-models based on LSSVM. These sub-sub-models based on the same kernel parameter are selected and combined as candidate SEN-sub-models by using branch and bound-based SEN (BBSEN). By employing BBSEN at the second time, these SEN-sub-models based on different kernel parameters are used to obtain the final soft measuring model. Thus, multiple kernel and regularization parameters are adaptive selected for building SEN-LSSVM model. UCI benchmark datasets and mechanical frequency spectral data are used to validate the effectiveness of this method.

Keywords: Modeling and Identification, Big Data Analytics and Monitoring
Abstract: Train basic resistance is important for the design of the automatic train operation, which influences the efficiency, punctuality, stop precision, energy consumption, and the safety of the train. The multi-innovation theory is a novel concept which can improve the accuracy of parameter estimation and be used to modify the traditional recursive least squares algorithm. In this paper, we derive the regularization form of the multi-innovation least squares algorithm and apply it to the train basic resistance parameter estimation. The simulation results based on the Yizhuang Line of Beijing Subway indicate that, compared with traditional least squares algorithm, the multi-innovation least squares algorithm can provide higher estimation accuracy and robustness, and can be used for online identification.

Keywords: Modeling and Identification, Process Applications, Model-based Control
Abstract: This work proposes a passivity-based approach to deal with the output-tracking-error problem for a large class of nonlinear chemical processes including non-minimum phase systems. More precisely, in that framework, the system dynamics is firstly written into the relaxing (pseudo) port-Hamiltonian representation which does not necessarily require the positive semi- definite property of the damping matrix. Then, a reference trajectory associated with a certain structure passing through a desired equilibrium point (i.e., the set-point) is chosen so that the error dynamics can be globally asymptotically stabilized at the origin thanks to the assignment of an appropriate damping injection. This method is subsequently illustrated for a benchmark of multiple reactions systems, namely Van de Vusse reaction system. The numerical simulations show the applications of the proposed approach.

Keywords: Modeling and Identification
Abstract: Most regression approaches, such as principal component analysis (PCA), are based on an assumption that the process data follow a Gaussian distribution. However, the process data usually dissatisfy that assumption. Thus, the locally weighted standardization (LWS) method is employed for transforming data into an approximate Gaussian distribution. Furthermore, the LWS based subspace PCA ensemble modeling method is developed. The subspace PCA can select important variables in each subspace for ensemble modeling. As a result, the proposed method gives a weaker assumption constrain and a better regression performance. The effectiveness of this approach is testified by two study cases

Keywords: Modeling and Identification
Abstract: Abstract: Prediction of physical properties of crude oil plays a key role in the petroleum refining industry, therefore, it is of great significance to establish the prediction model of physical properties of crude oil. In this paper, we propose an ensemble random weights neural network based prediction model whose inputs are nuclear magnetic resonance (NMR) spectra and outputs are carbon residual and asphaltene of crude oil. The model uses random vector functional link (RVFL) networks as the basic components and employs the regularized negative correlation learning strategy to build neural network ensemble and the online method to learn the new data. The experiment using the practical data collected from a refinery is carried out and compared with the decorrelated neural network ensembles with random weights (DNNE), least squares support vector machine (LS-SVM), partial least squares regression (PLS) and multiple linear regression (MLR). The results indicate the effectiveness of the proposed approach.

Keywords: Modeling and Identification, Batch Process Modeling and Control, Process Applications
Abstract: A methodology is proposed for parameter ranking and parameter subset selection for ordinary differential equation (ODE) models with time delay, in which delay is treated as an unknown model parameter. The methodology builds on earlier algorithms for ranking model parameters in systems without time delay (Yao et al., 2003; Thompson et al., 2009) and for finding the optimum number of parameters for estimation (Wu et al., 2011; McLean and McAuley, 2012a). A polymerization reactor system for producing bio-source polyether is used to illustrate the effectiveness of the proposed method in comparison with prior results obtained by Cui et al. (2015) who neglected the time delay.

Keywords: Modeling and Identification, Process Applications, Optimization and Scheduling
Abstract: The state of health (SoH) of lithium-ion batteries and battery packs must be monitored effectively for abuse to prevent failure and accidents, and to prolong the useful lifetime of the batteries. Many studies have suggested that temperature and discharge/charge current rate are the primary factors causing battery aging. However, there is yet to be a concrete mathematical model to predict the battery SoH. In this study, we introduce two SoH prediction models: the decreasing battery V₀₊ model, the increasing CV charge capacity model. Additionally, we derive a simple thermal model for the cell based on variation of temperature data.

Keywords: Modeling and Identification, Model-based Control, Process Applications
Abstract: This paper deals with the control of simple process models with state-delays. A pre-compensator is used to cancel the delay. Due to the cancellation of some terms, the process as well as its undelayed part should be stable. Although the approach is general, in order to simplify the notation, second order systems are considered to describe the procedure. The proposed methodology is applied to compensate the delay in a recycled reactor as well as to control a pure state-delayed academic example.

Keywords: Process Applications, Optimization and Scheduling, Modeling and Identification
Abstract: This article deals with the development of a reduced model describing the temporal evolution of the fructo-oligosaccharide production by Aureobasidium pullulans and the optimal process control. First, a reduced model is derived from a detailed model reproducing with good accuracy the dynamics of the fructo-oligosaccharide production. The reduced model is obtained using maximum likelihood principal component analysis and parameter identification based on a weighted least squares criterion. Next, the fructo-oligosaccharide concentration at an a priori undetermined time is maximized using Pontryagin maximum principle. The methodology is analyzed based on experimental data from batch and fedbatch cultures, and results are compared with those obtained with another simple model available in the literature.