Computers & Chemical Engineering最新文献

Bayesian optimization has proven effective for optimizing expensive-to-evaluate functions in Chemical Engineering. However, valuable physical insights from domain experts are often overlooked. This article introduces a collaborative Bayesian optimization approach that re-integrates human input into the data-driven decision-making process. By combining high-throughput Bayesian optimization with discrete decision theory, experts can influence the selection of experiments via a discrete choice. We propose a multi-objective approach togenerate a set of high-utility and distinct solutions, from which the expert selects the desired solution for evaluation at each iteration. Our methodology maintains the advantages of Bayesian optimization while incorporating expert knowledge and improving accountability. The approach is demonstrated across various case studies, including bioprocess optimization and reactor geometry design, demonstrating that even with an uninformed practitioner, the algorithm recovers the regret of standard Bayesian optimization. By including continuous expert opinion, the proposed method enables faster convergence and improved accountability for Bayesian optimization in engineering systems.

The study and application of contemporary optimization techniques considerably enhance the efficiency of chemical research and manufacturing. With the dynamic progression of modern manufacturing technologies, the emergence of numerous black-box models characterized by inaccessible mathematical formulations and high evaluation costs poses new challenges to traditional optimization methods, leading to difficulty in programming and solving. Hence, based on the trust region filter (TRF) method, we define a new framework to elevate optimization efficiency in this study for chemical systems involving computationally expensive black-box functions. Sampling size per iteration is reduced, and sampling efficiency is improved by incorporating the known data beyond the trust region to assist in constructing reduced models, which is achieved by the application of the Gaussian process. Through comparison and validation of benchmark tests and case studies, this approach demonstrates that using the Gaussian process as the reduced model can lower the number of calls to black-box functions by more than half compared to common linear and quadratic models, and convergence to first-order critical points can still be guaranteed.

Accurate state estimation of wastewater treatment plants is critical for optimizing wastewater treatment processes and reducing operating costs and energy consumption. Due to their large size and numerous state variables, these wastewater treatment plants are considered as high-dimensional systems. The complexity of wastewater treatment plants results in varying and complex process noise statistics, posing challenges for state estimation. This paper proposes a novel state estimation method for wastewater treatment plants subject to inaccurate process noise statistics. The high-dimensional state vector is partitioned into multiple state blocks based on the system architecture, and lost correlations between blocks are compensated by considering time-series correlations. Real-time modification of the process noise covariance matrix is applied to adaptively adjust the inaccurate process noise statistics and compensate for errors from block division. It is verified through simulations that the proposed Bayesian algorithm can achieve satisfactory estimation results while the computational cost is moderate.

Most system identification methods ignore correlations between different identification tasks and do not make full use of historical models when identifying a new process. In this paper, a two-stage transfer learning-based nonparametric system identification method is proposed to improve model accuracy and to reduce identification costs. First, to overcome the transfer capability limitations of the discrete-time model, the target process is modeled as a continuous-time impulse response (IR) function, which consists of a transfer model part and a residual model part, which is regarded as a zero-mean Gaussian process (GP). Then, by utilizing the process geometrical characteristics, the IR functions are classified into three IR domains, and transformation functions within domain or between domains are designed to learn the relationship between the source process model and the target process model. Finally, a two-stage nonparametric identification algorithm based on GP regression is developed: The first stage is performed to select the appropriate type of transformation function through the weighted-derivative dynamical time warping technique, and the second stage is conducted to estimate the transfer model and residual model by using the empirical Bayes approach. Three case studies are conducted to validate the superiority of the proposed identification method.

This work proposes a framework for simulation-based and surrogate-based reduced space Bayesian optimization of process flowsheets. The framework uses global sensitivity analysis for dimensionality reduction via the identification of critical process variables that contribute significantly to the variability of the objective function (e.g. productivity and operating costs). Both simulation- and surrogate-based algorithms are applied to a biopharmaceutical and a chemical process simulator for the production of plasmid DNA and dimethyl ether (DME), respectively. Their capabilities are assessed in terms of the trade-off between computational effectiveness and solution accuracy. Results indicate that simulation-based Bayesian optimization achieves better objective function values, while surrogate-based Bayesian optimization is more computationally effective.

Optimization-based scheduling in the chemical industry is highly beneficial but also highly difficult due to its combinatorial complexity. Different modeling and optimization techniques exist, each with individual strengths. We propose Benders decomposition to integrate mixed-integer linear programming (MILP) and discrete-event simulation (DES) to solve flow shop scheduling problems with makespan minimization objective. The basic idea is to generate valid Benders cuts based on sensitivity information of the DES sub problem, which can be found in the critical paths of DES solutions. For scaled literature flow shops, our approach requires at least an order of magnitude fewer iterations than a genetic algorithm and provides optimality gap information. For a real-world case study, our approach finds good solutions very quickly, making it a powerful alternative to established methods. We conclude that the Benders-DES algorithm is a promising approach to combine rigorous MILP optimization capabilities with high-fidelity DES modeling capabilities.

In this paper, we propose a NMPC scheme based on NARX-Laguerre model in its discrete-time structure in order to control the pH neutralization process. This model is the result of the development of discrete-time NARX model parameters on five independent Laguerre bases introducing the benefit of an essentially diminished number of parameters compared to the classical NARX model. The decrease in parametric complexity actually depends on choosing the ideal poles that characterize the Laguerre bases. In this paper, we develop an analytical technique to optimize the NARX-Laguerre poles. The parameters of the NARX-Laguerre model are reached using a recursive technique. The proposed model is utilized to integrate a nonlinear model predictive control, where we create a $j$-step ahead predictor on the prediction horizon $[k+1,k+N_p]$ and formulate the optimization problem using a performance criterion taking into consideration the restrictions imposed on the input and output of the process. The proposed nonlinear model predictive control strategy is validated on pH neutralization process.

The shortage of trained plant operators who can control complex systems in the process and energy industry is leading to an increasing need for more autonomy of such plants. In future, such systems will be controlled by autonomous agents executing intelligent control algorithms. Due to the critical nature of such plants and processes, new control strategies should not be introduced untested. In order to test suitable algorithms, robust and comprehensive simulation environments are required. In this paper, a framework, called Autonomous Systems Training Environment, is proposed for the evaluation of control algorithms in autonomous plant operations. Furthermore, an exemplary use case for a process engineering system is implemented in Matlab/Simulink, taking into account different levels of control (e.g., regulatory control, operator control, safety control). Faults, which represent a major challenge in the autonomous control, are also considered.

Simulated moving bed chromatographic (SMB) chromatographic processes are widely used for separations in pharmaceutical and biotechnological industries. In the present work, the control of such processes is proposed based on a semi-centralized control scheme that utilizes a combination of model predictive control (MPC) and classical PID control. The necessary information from the process, i.e. the internal concentration profiles, for the MPC is obtained by a moving horizon estimator (MHE) in real-time from the available limited measurements (the cycle-averaged concentrations of the extract and raffinate product streams and the dimensionless retention times of the concentration waves in the regeneration zones). As a test case study, the separation of a hypothetical system governed by the Langmuir isotherms in the nonlinear concentration range of the isotherms is considered. First, the stand-alone MHE is validated in open loop mode with no plant-model mismatch under deterministic and stochastic conditions. In the latter case, the true measurements are subject to random normally distributed noise. To evaluate the performance of the proposed control strategy, a reference tracking (change of the requirements for both the purities and the retention times) scenario is simulated. The investigated scenarios are: (i) no plant-model mismatch, (ii) MHE with ideal noiseless measurements and finally (iii) MHE under the influence of measurement noise. Results show that the controller is able to follow the change of the references from reduced purity to complete separation and vice versa closely.

This work presents a methodology for capacity planning under uncertainty in general multi-stage manufacturing networks. Multiple manufacturing lines are available and the production time on each line must be divided between a set of materials. The methodology generates mixed-integer linear programming models that represent capacity expansions with economy of scale costs functions and production planning details. A deterministic model is solved to create a baseline and the impact of uncertainty is investigated by sensitivity analysis and stochastic programming. The applicability of the methodology is exemplified through two case studies derived from industrial pharmaceutical manufacturing. The methodology identifies bottlenecks that limit supply and, where required, activates, and assigns capacity expansion projects for satisfying demand subject to uncertainty. The methodology determines the best use of existing resources and the location and size of capacity expansions thereby generating a portfolio of recommendations for decision making on integrated planning of capacity and production.