Search

Ensuring the interpretability of machine learning models in chemical engineering remains challenging due to inherent limitations and data quality issues, hindering their reliable application. In this study, a qualitatively implicit knowledge-guided machine learning framework is proposed to improve plasma gasification modelling. Starting with a pre-trained machine learning model, parameters are further optimized by integrating the heuristic algorithm to minimize the data fitting errors and resolving implicit monotonic inconsistencies. The latter is comprehensively quantified through Monte Carlo simulations. This framework is adaptive to different machine learning techniques, exemplified by artificial neural network (ANN) and support vector machine (SVM) in this study. Validated by a case study on plasma gasification, the results reveal that the improved models achieve better generalizability and scientific interpretability in predicting syngas quality. Specifically, for ANN, the root mean square error (RMSE) and knowledge-based error (KE) reduce by 36.44% and 83.22%, respectively, while SVM displays a decrease of 2.58% in RMSE and a remarkable 100% in KE. Importantly, the improved models successfully capture all desired implicit monotonicity relationships between syngas quality and feedstock characteristics/operating parameters, addressing a limitation that traditional machine learning struggles with., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier Ltd. All rights reserved.)

Simple Summary: Our objective is to establish an in vitro rumen fermentation model that can dynamically simulate the fermentation process of various total mixed ration (TMR) diets in the rumen of dairy cows, enabling a quantitative investigation of rumen methane and rumen acetic acid concentrations. The models were assessed for their prediction accuracy and precision, while the independent verification experiments confirmed the models' generalization ability across different total mixed ration (TMR) ratios (C:F). These results show that in vitro rumen models constructed by machine learning methods can be used as a tool to quantify rumen fermentation parameters (methane and acetic acid) and guide the dietary structure optimization of dairy cows. The aim of this study is to identify an alternative approach for simulating the in vitro fermentation and quantifying the production of rumen methane and rumen acetic acid during the rumen fermentation process with different total mixed rations. In this experiment, dietary nutrient compositions (neutral detergent fiber (NDF), acid detergent fiber (ADF), crude protein (CP), and dry matter (DM)) were selected as input parameters to establish three prediction models for rumen fermentation parameters (methane and acetic acid): an artificial neural network model, a genetic algorithm-bp model, and a support vector machine model. The research findings show that the three models had similar simulation results that aligned with the measured data trends (R2 ≥ 0.83). Additionally, the root mean square errors (RMSEs) were ≤1.85 mL/g in the rumen methane model and ≤2.248 mmol/L in the rumen acetic acid model. Finally, this study also demonstrates the models' capacity for generalization through an independent verification experiment, as they effectively predicted outcomes even when significant trial factors were manipulated. These results suggest that machine learning-based in vitro rumen models can serve as a valuable tool for quantifying rumen fermentation parameters, guiding the optimization of dietary structures for dairy cows, rapidly screening methane-reducing feed options, and enhancing feeding efficiency. [ABSTRACT FROM AUTHOR]

The biomass poultry litter and sewage sludge co-gasification is a good thermochemical method to produce hydrogen energy meanwhile to mitigate the consumption of fossil fuels. However, the operation optimization in the complex process system is important while computationally difficult because of high nonlinearity and many first-principle constraints. To address the optimization of this biohydrogen production process, a methodology framework for achieving the optimal operations is thus presented. The complete process system is first simulated and decomposed into upstream gasification process and downstream hydrogen purification for reducing computational complexity through the thermodynamic equilibrium-based simulation. Totally 1400 data points by pairing total 15 operating conditions with a composite sustainability index objective are generated through the random sampling and data classification strategies, which are then used for the construction of the artificial neural network (ANN)-based prediction models. ANN models of two subprocesses demonstrate the satisfactory prediction accuracy with R2 value of 0.98 and 0.99, respectively, which are then integrated with mixed-integer linear programming (MILP) for the optimization of upstream process and downstream process step-by-step. The MILP problems based on the ANN models are solved with lower optimization time of 45∼100 s compared to the heuristic algorithm optimization (3–6 h) based on the thermodynamic equilibrium-based simulation. The optimal sustainability index values of two processes are 0.80 and 0.91 which are both improved compared to the existing optimization results (0.79 and 0.84). This study emphasizes the optimization potential of the integration approach of machine learning-based modelling and mathematical programming for developing the optimal waste-to-energy processes. [Display omitted] • A feasible data screening process is conducted based on ANN classification models. • The ANN regression model achieves good results with over 96 % prediction accuracy. • Only 98.54 and 46.36 s are needed to achieve the sub-processes optimization. • The process optimization based on data-driven models shows a much higher efficiency. [ABSTRACT FROM AUTHOR]

Operations optimization in an organic Rankine cycle (ORC) based combined system is important while computationally difficult by using mechanistic models due to complex nonlinearities and constraints. In this study, a hybrid framework integrating machine learning and mathematical programming has been proposed to optimize the operations of the system for the best exergy performance. The combined system is first decomposed into two single ORCs for reducing computational complexity. Classification models and regression models based on artificial neural network (ANN) and linear regression are developed using simulation data, where classifications can be employed for high-throughput screening feasible inputs which meet the mechanistic constraints in ORC. The results demonstrate high performances of machine learning with at least 99% accuracies for classifications and with mean relative errors of less than 1% for regressions. These data-driven models and the relation of two ORCs were then embedded with mathematical programming for optimization and maximum net exergy of 28.66 MW is obtained. By linear expansion of ReLU operators in ANN, mixed-integer linear programming (MILP) based on machine learning models achieve high efficiency with ∼0.1 s required for optimization compared to mixed-integer nonlinear programming (MINLP) (>1000 s) and heuristic optimization based on mechanistic models (>10 h). • Machine learning and math programming are integrated for ORC optimization. • Classification is used for high-throughput identifying the mechanistic constraints. • Machine learning achieve >99% accuracy and < 1% mean relative error. • ReLU activation function is expressed by linear equations. • The method's efficiency is validated by optimization of ORC-based combined system. [ABSTRACT FROM AUTHOR]

• Fuel composition forecasting model is proposed for waste tires pyrolysis process. • The proposed model is developed based on the GBRT algorithm. • The proposed model exhibits high accuracy and strong generalization ability. • Over 60% of the forecasting results have REP values lower than 30%. Waste tire pyrolysis oil holds significant potential as an energy source and chemical feedstock, playing a crucial role in enhancing resource efficiency and promoting environmental sustainability. This study aims to develop a reliable forecasting model for determining the composition proportions of waste tire pyrolysis oil and assessing its performance. This study employed the gradient boosting decision tree (GBDT) method to develop the model that accurately forecasts the composition proportions of aromatic, aliphatic, and other compounds present in waste tire pyrolysis oil. Importance in XGboost and Cramer's V analysis in correlation analysis were used to select the input variables for the forecasting mode. Additionally, this study normalized the percentages of the three components to ensure consistency. To evaluate the performance of the proposed models, we compared them with 13 models from the Decision Tree series, BPNN series, SVM series, and HDMR series, thereby verifying the accuracy of our models. Furthermore, this study conducted random sampling to examine the stability and generalization ability of our proposed models. The experimental results demonstrate that the forecasting model exhibits excellent performance in estimating the proportions of aromatic compounds, aliphatic compounds, and other compounds. Over 60% of the samples yielded relative error percentage (REP) values below 30%. This indicates that this proposed model is capable of accurately forecasting the composition proportions of waste tire pyrolysis oil. The proposed model holds the potential to facilitate dynamic optimization of the waste tire resource reuse process and contribute to the sustainable development of waste tires. [ABSTRACT FROM AUTHOR]