Digital Chemical Engineering最新文献

Modeling and optimization of various processes enable more efficient operations and better planning activities for new process developments. With recent advances in computing power, data driven models, such as Machine Learning (ML), are being extensively applied in many areas of chemical engineering topics. Compared to mechanistic models that often do not reflect the realities of field conditions and the high costs associated with them, these techniques are relatively easier to implement. Data-driven models generated via ML techniques can be regularly updated, thereby giving an accurate picture of the system. Due to these inherent benefits, such tools are increasingly gaining a lot of traction in process systems. Even though data-driven models have the potential to be used as a replacement for traditional optimization tools that can be implemented in various process industries, it was found that applications of such models in process systems were quite limited to reactor modeling, molecular design, as well as safety, and relatability. The challenge still exists for data-driven modeling due to the lack of specialized tools tailored for macro systems and scale up. Most datasets were found to be derived from experimental studies which are limited in nature and only fit into microsystems. Hence, this paper provides a state of the art review on recent applications for data driven modeling research in process systems, and discusses the prominent challenges and future outlooks that were observed.

In industrial processes, control valve stiction is known to be one of the primary causes for poor control loop performance. Stiction introduces oscillatory behaviour in the process, leading to increased energy consumption, variations in product quality, shortened equipment lifespan and a reduction in overall plant profitability. Several detection algorithms using routine operating data have been developed over the last few decades. However, with the exception of a handful of recent publications, few attempts to apply classical supervised learning techniques have been published thus far. In this work, principal component analysis, linear discriminant analysis and a one-class support vector machine are trained to detect stiction using time series features as input. These features are extracted from the data using the tsfresh package for Python. The training data consists of simulated stiction examples generated using the XCH stiction model as well as other sources of oscillation. The classifier is subsequently benchmarked against closed-loop stiction data collected in an industrial setting, with performance exceeding that of existing methods.

Differential evolution (DE) algorithm and genetic algorithm (GA) are used in this study to estimate a set of kinetic parameters for Arabic date juice-based lactic acid production via batch and fed-batch mode of fermentation. Different feeding approaches like feed-forward control, exponential-, and modified exponential- feed are employed to obtain optimum kinetic parameters. The global optimum sets of kinetic parameters for both fermentation methods are found by minimizing the least square error between the experimental data and the simulated model results. In both batch and fed-batch fermentation methods (including different feeding strategies) the DE algorithm resulted in either the least value of the objective function or the least value of the sum of the square of residual errors between the experimental and model-predicted values for biomass growth (X), substrate consumption (S), and product formation (P). Six different strategies of the DE algorithm are used and their performance is compared for exponential feeding fed-batch fermenter. For exponential feeding fed-batch fermenter best suitable DE strategies were found to be best/1/bin and current to best/1/bin based on algorithm control parameters analysis. This manuscript highlights the limitations and improvements in the performance of individual algorithms on the given biochemical fermenters.

The role of carbon capture technology using monoethanolamine (MEA) is critical for achieving the carbon-neutrality goal. However, maintaining the efficient operation of the post-combustion carbon capture is challenging considering the hyperdimensional design space and nonlinear characteristics of the process. In this work, CO₂ capture level from the flue gas in the absorption column is investigated for the post-combustion carbon capture process using MEA. Artificial neural network (ANN) and support vector machine (SVM) models are constructed to model CO₂ capture level under extensive hyperparameters tuning. The comparative performance analysis based on external validation test confirmed the superior modelling and generalization ability of ANN for the carbon capture process. Later, partial derivative-based sensitivity analysis is carried out and it is the found that absorbent-based input variables like lean solvent temperature and lean solvent flow rate are the two most significant input variables on CO₂ capture level in the absorption column. The optimization problem with the ANN model embedded in the nonlinear programming-based optimization environment is solved under different operating scenarios to determine the optimum operating ranges for the input variables corresponding to the maximum CO₂ capture level. This research presents the optimum operating conditions for CO₂ removal from the flue gas for the post-combustion carbon capture process using MEA that contributes to achieving the carbon neutrality goal.

The true potential of artificial intelligence (AI) is to contribute towards the performance enhancement and informed decision making for the operation of the large industrial complexes like coal power plants. In this paper, AI based modelling and optimization framework is developed and deployed for the smart and efficient operation of a 660 MW supercritical coal power plant. The industrial data under various power generation capacity of the plant is collected, visualized, processed and subsequently, utilized to train artificial neural network (ANN) model for predicting the power generation. The ANN model presents good predictability and generalization performance in external validation test with R² = 0.99 and RMSE =2.69 MW. The partial derivative of the ANN model is taken with respect to the input variable to evaluate the variable’ sensitivity on the power generation. It is found that main steam flow rate is the most significant variable having percentage significance value of 75.3 %. Nonlinear programming (NLP) technique is applied to maximize the power generation. The NLP-simulated optimized values of the input variables are verified on the power generation operation. The plant-level performance indicators are improved under optimum operating mode of power generation: savings in fuel consumption (3 t/h), improvement in thermal efficiency (1.3 %) and reduction in emissions discharge (50.5 kt/y). It is also investigated that maximum power production capacity of the plant is reduced from 660 MW to 635 MW when the emissions discharge limit is changed from 510 t/h to 470 t/h. It is concluded that the improved plant-level performance indicators and informed decision making present the potential of AI based modelling and optimization analysis to reliably contribute to net-zero goal from the coal power plant.

Monoclonal antibodies (mAb) are biopharmaceutical products that improve human immunity. In this work, we propose a multi-actor proximal policy optimization-based reinforcement learning (RL) for the control of mAb production. Here, manipulated variable is flowrate and the control variable is mAb concentration. Based on root mean square error (RMSE) values and convergence performance, it has been observed that multi-actor PPO has performed better as compared to other RL algorithms. It is observed that PPO predicts a 40 % reduction in the number of days to reach the desired concentration. Moreover, the performance of PPO is improved as the number of actors increases. PPO agent shows the best performance with three actors, but on further increasing, its performance deteriorated. These results are verified based on three case studies, namely, (i) for nominal conditions, (ii) in the presence of noise in raw materials and measurements, and (iii) in the presence of stochastic disturbance in temperature and noise in measurements. The results indicate that the proposed approach outperforms the deep deterministic policy gradient (DDPG), twin delayed deep deterministic policy gradient (TD3), and proximal policy optimization (PPO) algorithms for the control of the bioreactor system.

Solvent selection and Controlling of operating parameters play a crucial role in batch cooling crystallization process. Choosing a best solvent for crystallization process involves more experimentation and time. To overcome this problem, an Artificial Neural Network (ANN) model technique is used to predict the carbamazepine form Ⅲ solubility by considering the thermodynamic properties of different solvents i.e. critical temperature, critical pressure, temperature, molecular weight, and acentric factor. The ANN model was trained and evaluated for solubility at various input data sets using experimental solubility data available in the literature. The ANN model with 20 hidden neurons has given the R² value of 0.9943 which shows that the developed ANN model can be used for the selection of best solvent for batch crystallization process. Further, to determine the optimal cooling profile of batch cooling crystallization process, a multi-objective optimization problem is formulated by considering objectives as minimizing the coefficient of variation (CV) and maximizing the Number mean size (NMS) of crystals subjected to population balance equations using “method of moments” technique. Two types of temperature strategies i.e., piece-wise constant and piece-wise linear are developed and solved using NSGA-Ⅱ dynamic optimization procedure. The optimal NMS value attained through piece-wise linear strategy was 197.1 µm. This value has been increased by 28.3 µm from the nominal case (without optimization) and the coefficient of variation has decreased from 0.951 to 0.76. Further, optimal NMS value attained through piece-wise constant strategy was 205 µm. The value has been increased by 36.2 µm and the coefficient of variation has decreased from 0.951 to 0.73. This proves that the crystal attributes can be improved by optimal cooling temperature profile obtained by multi-objective optimization framework. For implementing the optimal cooling profile an advanced model-based control, i.e., Generic Model Control (GMC) was developed. It was observed that the GMC controller has the good tracking profile with no offset with/without disturbances and small value of root mean square error (RMSE) of 0.0016 using piece-wise constant as set point temperature. Using piece-wise linear as set point temperature, the RMSE value was 0.0018. In particular, it is advantageous to operate the batch cooling crystallization process with piece-wise linear strategy for set point trajectory tracking problems.

Refinery industrial processes are very complex with nonlinear dynamics resulting from varying feedstock characteristics and also from changes in product prioritization. Along these processes, there are key properties of intermediate compounds that must be monitored and controlled since they directly affect the quality of the end products commercialized by these manufacturers. However, most of these properties can only be measured through time-consuming and expensive laboratory analysis, which is impossible to obtain in high frequencies, as required to properly monitor them. In this sense, developing soft sensors is the most common way to obtain high-frequency estimations for these measurements, helping advanced control systems to establish the correct setpoints for temperatures, pressures, and other sensors along the refining process, controlling the quality of end products. Since the amount of labeled data is scarce, most academic research has focused on employing semi- supervised learning strategies to develop machine learning (ML) models as soft sensors. Our research, on the other hand, goes in another direction. We aim to elaborate a framework that leverages the knowledge of domain experts and employs data augmentation techniques to build an enhanced fully labeled dataset that could be fed to any supervised ML algorithm to generate a quality soft sensor. We applied our framework together with Automated ML to train a model capable of predicting a specific key property associated with the production of Naphtha compounds in a refinery: the ASTM 95% distillation temperature of the Heavy Naphtha. Although our framework is model agnostic, we opted by using Automated ML for the optimization strategy, since it applies a diverse set of models to the dataset, reducing the bias of utilizing a single optimization algorithm. We evaluated the proposed framework on a case study carried out in an industrial refinery in Brazil, where the previous model in production for estimating the ASTM 95% distillation temperature of the Heavy Naphtha was based entirely on the physicochemical knowledge of the process. By adopting our framework with Automated ML, we were capable of improving the R² score by 120%. The resulting ML model is currently operating in real-time inside the refinery, leading to significant economic gains.

The thermochemical conversion of biomass is a promising technology due to its cost-effectiveness and feedstock flexibility, with pyrolysis being a particularly noteworthy method for its diverse product range. Despite the potential of pyrolysis, commercialization remains elusive, and there is a growing need to fully understand its dynamics to facilitate process scaling up. However, waste biomass pyrolysis is complex, time-consuming, and capital-intensive. Machine Learning (ML) has emerged as a possible means of supporting and accelerating pyrolysis research despite these challenges. This study provides a comprehensive overview of the use of ML in pyrolysis, from biorefinery to end-of-life product management. In addition, the success of ML in process optimization and control, predicting product yield, real-time monitoring, life-cycle assessment (LCA), and techno-economic analysis (TEA) during biomass pyrolysis is highlighted. Several ML methods have been utilized in a bid to study pyrolysis; the potentiality of artificial neural networks (ANNs) to learn extremely non-linear input-output correlations has led to the widespread adoption of these networks. Furthermore, the current knowledge gaps in ML research in pyrolysis and future recommendations for its application are identified. Finally, this study demonstrates the potential of ML in accelerating research and development as well as the scalability of pyrolysis of biomass.

Biofilms are formed by multicellular colonies of microorganisms that are protected by hard extracellular matrices. Eradication of biofilms is a challenging task due to their recalcitrant nature and thus biofilm formation poses a global threat to public health. In this regard, antibiofilm peptides are a promising class of therapeutics that are active against biofilms. However, large-scale experimental screening and testing of peptides for antibiofilm activity is a resource-intensive task. In this study, a machine learning-aided design framework is proposed to aid in screening of antibiofilm peptides. An SVM-based binary classification model is developed using amino acid compositions, sequence, and physicochemical properties of peptides as independent features. The physicochemical property-based model developed in this study achieved the highest accuracy of 97.9%, which is found to be substantially higher than the other feature representation techniques. The explainability of this model is performed using SHAP analysis. Results obtained show that amphiphilicity, aliphaticity and cationicity have positive correlation whereas steric parameter, length, and volume have negative correlation with antibiofilm activity of peptides. The developed model can be accessed freely via web tool: AntiBFP.