Search

The aerodynamic behavior of the fans array of the direct air-cooled condenser is different from that of a single fan because of the interaction effect between fans and the influence of wind. In this paper, a scale-down fan array experimental platform with measurement of the airflow and ambient wind speed is established to study the fan array partition schemes and cluster effect. The divisional cluster factors are obtained by experiment and generalized by random forest regression. The dynamic models of the condenser backpressure are introduced by the moving-boundary method. To achieve transient performance, a divisional control strategy for the direct air-cooled condenser is realized by MPC-PID cascade control. The simulation result shows that the fan power consumption is reduced by 14.7% compared to the existing control strategy. Furthermore, the wind speed disturbance is quickly suppressed cascade control.

Biobased slurry processes and large-scale continuous chemical production share many similarities. These concern not only process technologies and equipment but also staffing (positions/educations) and plant operations, including methods and technologies for planning, scheduling, and control. Process Systems Engineering (PSE) approaches have been applied successfully in large-scale chemical production to optimize processes of high complexity. By replacing rules-of-thumb and oversimplified calculations with rigorous quantification and automation, PSE increases the profitability of a wide range of processes. Such improvement potential also exists in the bioindustry, which traditionally comprises many small-scale and batch plants deployed in food or pharmaceutical production. The article elucidates the most important operational challenges of such processes. It brings together academic research and industrial perspectives by contrasting a careful revision of literature with a survey of personnel working in industrial biobased production settings in Denmark. This analysis allows showing the existence of a gap between academic research focus and industrial need, for instance, concerning production processes with humans-in-the-loop. Furthermore, differences between chemical and biobased processes are pointed out, highlighting some limitations of applying PSE to biobased processes. A deeper understanding of these limitations is fundamental to assessing risks when identifying process improvement potential. Finally, the article suggests a possible roadmap for the development of frameworks for the optimization of industrial biobased processes based on existing works in chemical processing.

Plantwide control (PWC) is an inherently complex task, which is why systematic guidelines and step-wise procedures are necessary to identify proper control pairings. Several different methodologies for the development of PWC structures have been proposed over the years, trying to balance the field between formal methods, practicality, and heuristics. The Integrated Framework of Simulation and Heuristics (IFSH) by Konda et al. (2005) utilizes simulators to support the heuristic decisions, thereby reducing the dependence of experience and prior understanding of the process behavior. In this article, the framework is successfully applied to the Forced Circulation Evaporator process by Newell and Lee (1989). A step-wise application of the methodology provides the basis for a discussion of how the methodology benefits from the utilization of simulations and how it differs from previously developed control structures for this evaporator process.

Graphical abstract, Production-scale fermentation processes in industrial biotechnology experience gradients in process variables, such as dissolved gases, pH and substrate concentrations, which can potentially affect the production organism and therefore the yield and profitability of the processes. However, the extent of the heterogeneity is unclear, as it is currently a challenge at large scale to obtain representative measurements from different zones of the reactor volume. Computational fluid dynamics (CFD) models have proven to be a valuable tool for better understanding the environment inside bioreactors. Without detailed measurements to support the CFD predictions, the validity of CFD models is debatable. A promising technology to obtain such measurements from different zones in the bioreactors are flow-following sensor devices, whose development has recently benefitted from advancements in microelectronics and sensor technology. This paper presents the state of the art within flow-following sensor device technology and addresses how the technology can be used in large-scale bioreactors to improve the understanding of the process itself and to test the validity of detailed computational models of the bioreactors in the future.

Due to the heterogeneous nature of large-scale fermentation processes they cannot be modelled as ideally mixed reactors, and therefore flow models are necessary to accurately represent the processes. Computational fluid dynamics (CFD) is used more and more to derive flow fields for the modelling of bioprocesses, but the computational demands associated with simulation of multiphase systems with biokinetics still limits their wide applicability. Hence, a demand for simpler flow models persists. In this study, an approach to develop data-based flow models in the form of compartment models is presented, which utilizes axial-flow rates obtained from flow-following sensor devices in combination with a proposed procedure for automatic zoning of volume. The approach requires little experimental effort and eliminates the necessity for computational determination of inter-compartmental flow rates and manual zoning. The concept has been demonstrated in a 580 L stirred vessel, of which models have been developed for two types of impellers with varying agitation intensities. The sensor device measurements were corroborated by CFD simulations, and the performance of the developed compartment models was evaluated by comparing predicted mixing times with experimentally determined mixing times. The data-based compartment models predicted the mixing times for all examined conditions with relative errors in the range of 3–27%. The deviations were ascribed to limitations in the flow-following behavior of the sensor devices, whose sizes were relatively large compared to the examined system. The approach provides a versatile and automated flow modelling platform which can be applied to large-scale bioreactors.

Operational complexity in batch-operated plants is large and companies struggle to produce at high equipment utilisation. Due to high fixed costs, capacity utilisation is an important operational directive. Many current plants are not in a state that allows the application of rigorous process systems engineering tools due to modelling challenges. This work proposes identifying the right engineering projects based on statistic evidence. In the case of incomplete process monitoring strategies in semi-automated facilities, these analyses are specifically challenging. The power of modern data processing tools in this context is shown at hand of a case study in an industrial pharmaceutical production. This includes the development of a recursive monitoring algorithm as well as plant performance evaluation based on established heuristics.

Numerical methods for simulation of linear and non-linear liquid chromatography processes are compared in terms of computational cost and quality of the solution using as few as possible degrees of freedom (discrete problem size). We apply, for the first time, the p-type (order enrichment) continuous Galerkin spectral method (p-CG) for spatial discretization with exponential decay of spatial errors (spectral convergence) for smooth problems. To stabilize the method for marginally resolved or very steep shock-waves typically encountered, we use a spectral filtering approach. This effectively removes aliasing driven instabilities while retaining the high-order accuracy of the numerical method. We benchmark the stabilized p-CG method against two state-of-the-art finite element methods, namely the second-order accurate h-type (element refinement) discontinuous Galerkin method using a total-variation-diminishing limiter (h-DG-TVD) and the arbitrary-order hp-type discontinuous Galerkin spectral element method (hp-DG). Not surprisingly, the three methods provide predictions of column elution profiles that converge to the same values. However, the computational costs differ considerably. Although second-order accurate shock-capturing methods are often adopted as the default implementation, the stable arbitrary high-order methods considered in this paper are found to significantly outperform these in terms of both computational cost and discrete problem size. The p-CG method performs as well as the hp-DG method in terms of computational cost, but it has superior convergence properties, even in the presence of strong shock-waves. Further, its implementation is much simpler.

The running state of auxiliary equipment in power plants directly affects the economy and safety of power production, thus a data-based fault early warning method for auxiliary equipment is proposed. A multivariate state estimation technique with improved process memory matrixes is used to build the nonparametric model describing the normal state of the equipment. Different from the traditional process memory matrix, the improved process memory matrix utilizes the probability density equal interval sampling method to greatly increase the sampling density of the typical operating state and effectively cover the common working space. Instead of the traditional residual threshold method, the sliding window similarity of states is adopted as the evaluation criteria for early warning, which can better evaluate the normality of multiple fault-related variables in the state. This study takes a 300MW CFB power plant in China as an example, the proposed method are compared with the time series, fuzzy logic and traditional MSET methods. The result shows that the proposed strategy can most effectively predict the faults.

A robust batch-to-batch optimization (RB2BO) is proposed for recipe optimization of batch processes in the presence of uncertainty. Robust optimization can result in performance loss due to its conservative nature, and measurement-based optimization approaches do not typically address the residual uncertainty after parameter identification. In the proposed approach, distribution and extreme-case scenarios of uncertain parameters are refined through Bayesian updating using the measurements, and then the robust optimization is performed with the updated scenarios. Due to the iterative scenario adaptation and robust optimization mechanism, the proposed RB2BO approach is intended to provide robust recipes less conservative compared to the conventional methods, resulting in high objective function values and minimal constraint violations. This is demonstrated through an example of a pectin extraction process by considering feedstock variabilities and terminal constraints for the desired product quality. In add...

Computational fluid dynamics (CFD) is a powerful tool for quantitative prediction of fluid dependent properties in a finite volume. However, the complexity of solving the momentum balances and the continuity equations at each element of the discretized geometry can easily lead to an expensive computational task. Compartment modelling is a potential alternative to speed up the calculation, which is however reached at the expense of the level of accuracy. The most important factor in optimizing a compartment model (CM) concerning the accuracy and the computational time is the quality of the chosen compartments to represent the critical gradients. This work presents a new automated compartmentalization method to characterize an improved network of compartments derived from initial detailed CFD results, with a focus on cylindrical-shaped systems. This method was evaluated with a case study of a 700 L stirred tank bioreactor by estimating the mixing performance and demonstrating its high efficiency.

Dead-end cake filtration as one of the first unit operations deployed in biochemical production plants has been subject of academic investigation for many decades. The recurrent discrete reinitialisation events are challenging from a process control point of view, especially in a continuous downstream line. The complexity that arises when multiple units are operated in parallel seems to have received little attention. This work aims at illustrating this complexity and delineates the arising plantwide control problem. Guidelines for optimised operation are derived from general process understanding and at hand of an industrial case study. The need for a predictive model to solve multiple scheduling problems is identified, and a mathematical model based on conventional filtration theory is derived. Due to raw material variability and operational uncertainties, the predictions are found to be too imprecise for deployment. This is expected to be representative of many bio-based processes, where manual scheduling needs to be integrated effectively into plantwide control structures.

A quantitative description of flow and mixing in bioreactors is necessary to assess the exposure of the microorganisms to suboptimal conditions, which may negatively impact key performance parameters of a process, such as yield and productivity. The traditional approach of quantifying mixing, by means of a terminal mixing time, is inadequate because it rests on the assumption that a homogeneous state is reached, and it provides little information on the mixing process itself. In this study, it has been demonstrated how flow-following sensor devices can be used to obtain detailed knowledge on macroscopic flow in stirred bioreactors under turbulent flow conditions. The flow generated by common impeller types; the Rushton disc turbine and the pitched blade turbine, has been examined in terms of the axial flow features and circulation times under various levels of agitation. It was demonstrated that useful features to characterize the flow in the vessel, such as the axial distribution and axial velocity profile of the sensor devices could be obtained for the examined conditions. Based on these features it was observed that the flow following behavior of the sensor devices were negatively affected at higher impeller speeds. This finding was corroborated by CFD simulations of the vessel. The mean circulation times determined by the sensor devices were found to be highly correlated to the mixing times determined from homogenization of a chemical tracer (tm = 2.2–2.6 t ¯ c ), which demonstrates that flow-following sensor devices can be used to quantify macromixing in bioreactors. Furthermore, the underlying circulation time distributions could be derived, which were found to be well described by the lognormal probability distribution.

The goal of this paper is to review and critically assess different methods to monitor key process variables for ethanol production from lignocellulosic biomass. Because cellulose-based biofuels cannot yet compete with non-cellulosic biofuels, process control and optimization are of importance to lower the production costs. This study reviews different monitoring schemes, to indicate what the added value of real-time monitoring is for process control. Furthermore, a comparison is made on different monitoring techniques to measure the off-gas, the concentrations of dissolved components in the inlet to the process, the concentrations of dissolved components in the reactor, and the biomass concentration. Finally, soft sensor techniques and available models are discussed, to give an overview of modeling techniques that analyze data, with the aim of coupling the soft sensor predictions to the control and optimization of cellulose to ethanol fermentation. The paper ends with a discussion of future needs and developments., Comment: 59 pages, 3 figures, 20 tables

Packed bed chromatography is an important unit operation for purification of product molecules in biopharmaceutical processes. Packed bed chromatographic processes are modeled as advection-diffusion-reaction partial differential equations. The advection term strongly dominates the diffusion term. Therefore, specialized numerical methods must be used for efficient simulation of packed bed chromatographic processes. In this paper, we use a discontinuous-Galerkin method on finite-elements for spatial discretization and low storage explicit Runge-Kutta (LSERK) methods for numerical solution of the resulting system of differential equations. We study the numerical solution of deterministic and stochastic models of packed bed chromatographic processes. The stochastic model and its corresponding numerical solution constitute the first step toward systematic monitoring, fault detection, and optimal predictive control of chromatographic processes. It is also an essential ingredient in uncertainty quantification for efficient and robust design and operation of chromatographic processes. (C) 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.