Computers & Chemical Engineering最新文献

Liquid phase oxidation of furfural using hydrogen peroxide offers a promising route for bio-based C₄ furanones and diacids; however, only dilute water-based process designs have been previously suggested that have limited techno-economic potential. In this study, a conceptual process design is presented, where aqueous furfural is extracted using an organic solvent, coupled with peroxide oxidation and product recovery in the presence of the solvent. To address the problem of solvent selection, the COSMO-RS-based solvent screening framework is applied, where quantum mechanics-based thermodynamics are utilized in pinch-based process models. About 2500 solvent candidates were identified as feasible. Focusing on a set of 400 solvent candidates revealed energy consumption values (Q_reb,tot/ṁ_{prod recov}) between approximately 2 MWh/tonne and 33 MWh/tonne, signifying the potential of the solvent-based process in outperforming the reference aqueous process (49.4 MWh/tonne). The study provides potential solvent candidates and future directions to consider in more costly computational and experimental efforts.

Sustainable chemical engineering through demand side management (DSM) and renewable feedstock integration e.g. in biorefineries are key to optimizing the use of fluctuating energy resources and minimizing environmental impact while conserving resources. This contribution presents the results of the economic evaluation of integrating DSM into biofuel biorefineries through a dynamic simulation approach. A previously developed decision support tool for DSM implementation was extended to describe the size of intermediate buffer tanks as a function of oversizing up- and downstream processes. Design optimization of the process cascade determined the oversizing that allows the optimal balance of operational cost reduction through flexibility and capital cost increase through oversizing. Scheduling optimization validated the results of the steady-state optimization and show that, by considering interactions between processes, buffer tank capacity can be reduced, while increasing DSM potential.

In this paper, a controller is proposed to automate the injection of propofol and remifentanil during general anesthesia using bispectral index (BIS) measurement. To handle the parameter uncertainties due to inter- and intra-patient variability, an extended estimator is used coupled with a Model Predictive Controller (MPC). Two methods are considered for the estimator: the first one is a multiple extended Kalman filter (MEKF), and the second is a moving horizon estimator (MHE). The state and parameter estimations are then used in the MPC to compute the next drug rates. The methods are compared with a PID from the literature. The robustness of the controller is evaluated using Monte-Carlo simulations on a wide population, introducing uncertainties in all parts of the model. Results both on the induction and maintenance phases of anesthesia show the potential interest in using this adaptive method to handle parameter uncertainties.

This work investigates the integrated production–inventory-routing problem (PIRP) for liquid oxygen supply chains consisting of multiple plants, customers, and heterogeneous vehicles. Solving such a problem is challenging, especially when dealing with large-scale industrial scales. A two-level procedure is adopted that determines decisions regarding production, inventory, and product allocation by simplifying the routing component in the first level. Then, the routing decisions are considered in the lower level for which 3-index mathematical programming formulations are presented. To address the combinatorial complexity of the lower-level decisions, we propose alternative multi-trip heterogeneous vehicle routing problem (MTHVRP) formulations together with a column generation-based solution strategy. A set of test instances and a real-world case study demonstrate the applicability and performance of the formulations and solution methods.

This article addresses the plant-model mismatch detection problem for linear multiple-input and multiple-output systems operating under the constrained dynamic matrix control (DMC) with the assumption of unknown noise models. An autocovariance-based mismatch detection method that uses sum-of-norms regularization is proposed, aiming to detect parameter jumps and estimate the noise model separately. The intention of introducing regularization is not only to be able to segment the mismatch so that the mismatch is piece-wise constant in time, but also to make the method robust to colored noise. Moreover, a method to alleviate mis-detection caused by unknown operating conditions is proposed. We show that the method can detect significant jumps in parameters and thus provide a priori knowledge for system re-identification and timing of updating the model. Finally, the feasibility of the proposed method under closed-loop conditions is analyzed from a stochastic perspective and demonstrated with illustrative examples.

Recent advances have been made in self-stabilizing Economic Nonlinear Model Predictive Control (eNMPC) formulation without pre-calculated setpoints, which leverages norm-based steady-state optimality conditions to enhance system robustness. To enable practical implementation, a generalized time-domain formulation is proposed, accommodating the discrete-time nature of control instrumentation and the continuous-time nature of first-principles models. A case study involving a modular membrane reactor illustrates the applicability of self-stabilizing eNMPC in real-world industrial scenarios.

Gasification of lignocellulosic biomass can be used to produce syngas used as a biorefinery feedstock. To facilitate the commercialisation of the gasification process, models are used to predict the outputs, simulate the impacts of irregular circumstances, and analyse process feasibility. This paper presents a hybrid model combining Aspen Plus and machine learning (ML) algorithms to enhance the prediction of gasification outputs. A base case gasification process flowsheet simulation was implemented in Aspen Plus based on assumed thermodynamic equilibrium conditions which can lead to inaccurate results. To address this, six ML algorithms were applied to collected experimental data and analysed for accuracy and efficiency. The feature importance, accuracy improvement, and the effect of implementing the ML predictions in the gasification block on the rest of the flowsheet were investigated. This paper emphasises the need of higher accuracy models and the great potential of ML approaches to offer high accurate predictions.

This study introduces a multi-period centralized storage optimization model aimed at designing an efficient hydrogen supply chain system, considering cost and emissions as dual objectives. It integrates multiple energy sources, production and storage methods, transport combinations, demand scenarios, and carbon capture systems, offering a comprehensive decision-making approach for hydrogen network design. Employing the mixed-integer linear programming methodology, the proposed model resolves these complexities. The research applies this model to a case study in France, generating six unique scenarios for 10 and 15 cities, and compares them against two distinct decentralized models. The findings consistently highlight the centralized storage model’s cost benefits across various demand scenarios, including cases of unrestricted emissions as well as cases with limited emission targets. The cost-effectiveness of this proposed model enhances its feasibility within the current context of decarbonization.

This work seeks to allow for the online operation and training of model-free reinforcement learning (RL) agents but limit the risk to system equipment and personnel. The parallel implementation of RL alongside more conventional process control (CPC) allows for the RL algorithm to learn from CPC. The past performance of both methods are assessed on a continuous basis allowing for a transition from CPC to RL and, if needed, transitioning back to CPC from RL. This allows for the RL algorithm to slowly and safely assume control of the process without significant degradation in control performance. It is shown that the RL can derive a near optimal policy even when coupled with a suboptimal CPC. It is also demonstrated that the coupled RL-CPC algorithm learns at a faster rate than traditional RL methods of exploration while the algorithm’s performance does not deteriorate below CPC, even when exposed to an unknown operating condition.

(Economic) nonlinear model predictive control ((e)NMPC) requires dynamic models that are sufficiently accurate and computationally tractable. Data-driven surrogate models for mechanistic models can reduce the computational burden of (e)NMPC; however, such models are typically trained by system identification for maximum prediction accuracy on simulation samples and perform suboptimally in (e)NMPC. We present a method for end-to-end reinforcement learning of Koopman surrogate models for optimal performance as part of (e)NMPC. We apply our method to two applications derived from an established nonlinear continuous stirred-tank reactor model. The controller performance is compared to that of (e)NMPCs utilizing models trained using system identification, and model-free neural network controllers trained using reinforcement learning. We show that the end-to-end trained models outperform those trained using system identification in (e)NMPC, and that, in contrast to the neural network controllers, the (e)NMPC controllers can react to changes in the control setting without retraining.