Estimating biomass major chemical constituents from ultimate analysis using a random forest model

Intelligent Optimization Design Framework for Alternating Current Pulse Modulation Electrohydrodynamic Printing Process Parameters

To achieve efficient size tuning of printed microstructures on insulating substrates, an integrated process parameter intelligent optimization design framework for alternating current pulse modulation electrohydrodynamic (AC-EHD) printing is proposed for the first time. The framework is comprised of two stages: the construction of a prediction model and the acquisition of process parameters. The first stage employs the elk herd optimizer(EHO)-artificial neural network(ANN) to establish a mapping relationship between printing process parameters and the size of deposited droplets. The analysis of the prediction performance of the EHO-ANN model across various datasets reveals that the model exhibits commendable accuracy and robustness in predicting printed droplet size. In the second stage, the process parameters of AC-EHD printing are intelligently determined by utilizing the error between the model output and the desired droplet size as the fitness value for EHO. By comparing three sets of experimental cases with varying droplet sizes, it is observed that the actual printed droplet sizes closely align with the desired values, thus validating the effectiveness of this framework. The framework proposed in this paper mitigates the time and material wastage caused by adjusting AC-EHD printing process parameters on insulating substrates, thereby significantly enhancing the usability of the technology.

Machine learning constructs the microstructure and mechanical properties that accelerate the development of CFRP pyrolysis for carbon-fiber recycling

The increasing use of carbon-fiber-reinforced plastic (CFRP) has led to its post-end-of-life recycling becoming a research focus. Herein, we studied the macroscopic and microscopic characteristics of recycled carbon fiber (rCF) during CFRP pyrolysis by innovatively combining typical experiments with machine learning. We first comprehensively studied the effects of treatment time and temperature on the mechanical properties, graphitization degree, lattice parameters, and surface O content of rCF following pyrolysis and oxidation. The surface resin residue was found to largely affect the degradation of the mechanical properties of the rCF, whereas oxidation treatment effectively removes this residue and is the critical recycling condition that determines its mechanical properties. In contrast, pyrolysis affected graphitization in a more-pronounced manner. More importantly, a random forest machine-learning model (RF model) that optimizes using a particle swarm algorithm was developed based on 336 data points and used to determine the mechanical properties and microstructural parameters of rCF when treated under various pyrolysis and oxidation conditions. The constructed model was effectively used to forecast the recovery conditions for various rCF target requirements, with the predictions for different recycling conditions found to be in good agreement with the experimental data.

High-throughput prediction of stalk cellulose and hemicellulose content in maize using machine learning and Fourier transform infrared spectroscopy

Cellulose and hemicellulose are key cross-linked carbohydrates affecting bioethanol production in maize stalks. Traditional wet chemical methods for their detection are labor-intensive, highlighting the need for high-throughput techniques. This study used Fourier transform infrared (FTIR) spectroscopy combined with machine learning (ML) algorithms on 200 large-scale maize germplasms to develop robust predictive models for stalk cellulose, hemicellulose and holocellulose content. We identified several peak height features correlated with three contents, used them as input data for model building. Four ML algorithms demonstrated higher predictive accuracy, achieving coefficient of determination (R2) ranging from 0.83 to 0.97. Notably, the Categorical Boosting algorithm yielded optimal models with coefficient of determination (R2) exceeding 0.91 for the training set and over 0.81 for the test set. The approach combined FTIR spectroscopy with ML algorithms offers a precise and high-throughput tool for predicting stalk cellulose, hemicellulose and holocellulose contents, benefiting maize genetic breeding for bioenergy and biofuels.

Machine learning prediction of stalk lignin content using Fourier transform infrared spectroscopy in large scale maize germplasm

Lignin has been recognized as a major factor contributing to lignocellulosic recalcitrance in biofuel production and attracted attentions as a high-value product in the biorefinery field. As the traditional wet chemical methods for detecting lignin content are labor-intensive, time-consuming and environment-toxic, it is an urgent need to develop high-throughput and environment-friendly techniques for large-scale crop germplasms screening. In this study, we conducted a Fourier transform infrared (FTIR) assay on 150 maize germplasms with a diverse lignin composition to build predictive models for lignin content in maize stalk. Principal component analysis (PCA) was applied to the FTIR spectra for use as model inputs. Classification and advanced gradient boosting machine (GBM) algorithms demonstrated higher predictive accuracy (0.82-0.96) compared to traditional linear and regularization algorithms (0.03-0.04) in the training set. Notably, two optimal models, built using the extreme gradient boosting (XGBoost) and light gradient boosting machine (LightGBM) algorithms, achieved R2 values of over 0.91 in the training set and over 0.82 in the test set. Overall, the combination of FTIR and machine learning (ML) algorithms offers a high-throughput and efficient method for predicting lignin content. This approach holds significant potential for genetic breeding and the effective utilization of maize in industrial production.

Prediction of adsorption performance of ZIF-67 for malachite green based on artificial neural network using L-BFGS algorithm

Given the necessity and urgency in removing organic pollutants such as malachite green (MG) from the environment, it is vital to screen high-capacity adsorbents using artificial neural network (ANN) methods quickly and accurately. In this study, a series of ZIF-67 were synthesized, which adsorption properties for organic pollutants, especially MG, were systematically evaluated and determined as 241.720 mg g-1 (25 ℃, 2 h). The adsorption process was more consistent with pseudo-second-order kinetics and Langmuir adsorption isotherm, which correlation coefficients were 0.995 and 0.997, respectively. The chemisorption mechanism was considered to be π-π stacking interaction between imidazole and aromatic ring. Then, a Python-based neural network model using the Limited-memory BFGS algorithm was constructed by collecting the crucial structural parameters of ZIF-67 and the experimental data of batch adsorption. The model, optimized extensively, outperformed similar Matlab-based ANN with a coefficient of determination of 0.9882 and mean square error of 0.0009 in predicting ZIF-67 adsorption of MG. Furthermore, the model demonstrated a good generalization ability in the predictive training of other organic pollutants. In brief, ANN was successfully separated from the Matlab platform, providing a robust framework for high-precision prediction of organic pollutants and guiding the synthesis of adsorbents.

Databases for biomass and waste biorefinery - a mini-review and SWOT analysis

The world is facing problems of the increasing amount of resources wasted as the world population grows. Biowaste streams form a significant part of the overall waste generation, and a circular economy utilizing this biowaste will significantly reduce waste whilst lowering the anthropogenic carbon footprint. Due to their energy content and high concentration of hydrocarbon molecules, bio-based waste streams have the potential to be transformed into valorized products (energy, fuels, and chemicals) using biorefinery technologies. In this work, a mini-review has been conducted on available, mostly European databases on existing biomass types and biorefinery technologies to provide a framework for a desirable, comprehensive database connecting bio-based waste streams, biorefinery technologies and bioproducts, as well as the geographical distribution of feedstocks and biorefineries. The database assessment utilized the SWOT (strengths, weakness, opportunities, threats) methodology to support benchmark analysis and to identify critical gaps in underlying data structures that could be included in a single database. The results show that current databases are useful but insufficient for waste biorefineries due to limited quality and quantity as well as the usability of data. A comprehensive database or improved database cluster would be necessary, not only for technology development but for better investment and policy decisions. The development of the new database architecture would need to incorporate the aspects: expansion of database scope and content depth, improved usability, accessibility, applicability, update frequency, openness to new contributions, process descriptions and parameters, and technology readiness level.

Hybrid Prediction-Driven High-Throughput Sustainability Screening for Advancing Waste-to-Dimethyl Ether Valorization

Assessing the prospective climate preservation potential of novel, innovative, but immature chemical production techniques is limited by the high number of process synthesis options and the lack of reliable, high-throughput quantitative sustainability pre-screening methods. This study presents the sequential use of data-driven hybrid prediction (ANN-RSM-DOM) to streamline waste-to-dimethyl ether (DME) upcycling using a set of sustainability criteria. Artificial neural networks (ANNs) are developed to generate in silico waste valorization experimental results and ex-ante model the operating space of biorefineries applying the organic fraction of municipal solid waste (OFMSW) and sewage sludge (SS). Aspen Plus process flowsheeting and ANN simulations are postprocessed using the response surface methodology (RSM) and desirability optimization method (DOM) to improve the in-depth mechanistic understanding of environmental systems and identify the most benign configurations. The hybrid prediction highlights the importance of targeted waste selection based on elemental composition and the need to design waste-specific DME synthesis to improve techno-economic and environmental performances. The developed framework reveals plant configurations with concurrent climate benefits (-1.241 and -2.128 kg CO2-eq (kg DME)-1) and low DME production costs (0.382 and 0.492 € (kg DME)-1) using OFMSW and SS feedstocks. Overall, the multi-scale explorative hybrid prediction facilitates early stage process synthesis, assists in the design of block units with nonlinear characteristics, resolves the simultaneous analysis of qualitative and quantitative variables, and enables the high-throughput sustainability screening of low technological readiness level processes.

Multifunctional catalyst-assisted sustainable reformation of lignocellulosic biomass into environmentally friendly biofuel and value-added chemicals

Rapid urbanization is increasing the world's energy demand, making it necessary to develop alternative energy sources. These growing energy needs can be met by the efficient energy conversion of biomass, which can be done by various means. The use of effective catalysts to transform different types of biomasses will be a paradigm change on the road to the worldwide goal of economic sustainability and environmental protection. The development of alternative energy from biomass is not easy, due to the uneven and complex components present in lignocellulose; accordingly, the majority of biomass is currently processed as waste. The problems may be overcome by the design of multifunctional catalysts, offering adequate control over product selectivity and substrate activation. Hence, this review describes recent developments involving various catalysts such as metallic oxides, supported metal or composite metal oxides, char-based and carbon-based substances, metal carbides and zeolites, with reference to the catalytic conversion of biomass including cellulose, hemicellulose, biomass tar, lignin and their derivative compounds into useful products, including bio-oil, gases, hydrocarbons, and fuels. The main aim is to provide an overview of the latest work on the use of catalysts for successful conversion of biomass. The review ends with conclusions and suggestions for future research, which will assist researchers in utilizing these catalysts for the safe conversion of biomass into valuable chemicals and other products.

Application of the biogas residue of anaerobic co-digestion of gentamicin mycelial residues and wheat straw as soil amendment: Focus on nutrients supply, soil enzyme activities and antibiotic resistance genes

Land utilization of the biogas residue (BR) produced by anaerobic co-digestion of gentamicin mycelial residues (GMRs) and wheat straw is a promising method to achieve the deep recycling of GMRs. This study evaluated the feasibility and efficacy of application of using BR as a soil amendment by using a pot experiment. Results indicated that BR could improve the soil fertility better than commercial chicken manure fertilizer (CMF) in terms of the soil enzyme activities and nutrients supply. Random Forest (RF) model was applied to predict soil enzyme activities and identify key influencing factors. Combining the Random Forest (RF) model with the Three-dimensional Excitation-emission Matrix and Parallel Factor (3D-EEM-PARAFAC) analysis, revealing that humic-like substances provided by BR protected soil enzymes, thus improving soil fertility. Furthermore, gentamicin and antibiotic resistance genes (ARGs)/mobile genetic elements (MEGs) introduced by BR decreased greatly after cultivation, implying a low risk of antimicrobial resistance. This study suggested that reasonable application of BR could improve soil nutrients supply, soil enzyme activity and control antimicrobial resistance risk.

Characteristics prediction of hydrothermal biochar using data enhanced interpretable machine learning

Hydrothermal biochar is a promising sustainable soil remediation agent for plant growth. Demands for biochar properties differ due to the diversity of soil environment. In order to achieve accurate biochar properties prediction and overcome the interpretability bottleneck of machine learning models, this study established a series of data-enhanced machine learning models and conducted relevant sensitivity analysis. Compared with traditional support vector machine, artificial neural network, and random forest models, the accuracy after data enhancement increased in average from 5.8% to 15.8%, where the optimal random forest model showed the average of accuracy was 94.89%. According to sensitivity analysis results, the essential factors influencing the predicting results of the models were reaction temperature, reaction pressure, and specific element of biomass feedstock. As a result, data-enhanced interpretable machine learning proved promising for the characteristics prediction of hydrothermal biochar.

Opportunities and challenges of machine learning in bioprocesses: Categorization from different perspectives and future direction

Recent advances in machine learning (ML) have revolutionized an extensive range of research and industry fields by successfully addressing intricate problems that cannot be resolved with conventional approaches. However, low interpretability and incompatibility make it challenging to apply ML to complicated bioprocesses, which rely on the delicate metabolic interplay among living cells. This overview attempts to delineate ML applications to bioprocess from different perspectives, and their inherent limitations (i.e., uncertainties in prediction) were then discussed with unique attempts to supplement the ML models. A clear classification can be made depending on the purpose of the ML (supervised vs unsupervised) per application, as well as on their system boundaries (engineered vs natural). Although a limited number of hybrid approaches with meaningful outcomes (e.g., improved accuracy) are available, there is still a need to further enhance the interpretability, compatibility, and user-friendliness of ML models.

Machine learning technology in biohydrogen production from agriculture waste: Recent advances and future perspectives

Agricultural waste biomass has shown great potential to deliver green energy produced by biochemical and thermochemical conversion processes to mitigate future energy crises. Biohydrogen has become more interested in carbon-free and high-energy dense fuels among different biofuels. However, it is challenging to develop models based on experience or theory for precise predictions due to the complexity of biohydrogen production systems and the limitations of human perception. Recent advancements in machine learning (ML) may open up new possibilities. For this reason, this critical study offers a thorough understanding of ML's use in biohydrogen production. The most recent developments in ML-assisted biohydrogen technologies, including biochemical and thermochemical processes, are examined in depth. This review paper also discusses the prediction of biohydrogen production from agricultural waste. Finally, the techno-economic and scientific obstacles to ML application in agriculture waste biomass-based biohydrogen production are summarized.

Prediction of carbon dioxide production from green waste composting and identification of critical factors using machine learning algorithms

Controlling carbon dioxide produced from green waste composting is a vital issue in response to carbon neutralization. However, there are few computational methods for accurately predicting carbon dioxide production from green waste composting. Based on the data collected, this study developed novel machine learning methods to predict carbon dioxide production from green waste composting and made a comparison among six methods. After eliminating the extreme outliers from the dataset, the Random Forest algorithm achieved the highest prediction accuracy of 88% in the classification task and showed the top performance in the regression task (root mean square error = 23.3). As the most critical factor, total organic carbon, with the Gini index accounting for about 59%, can provide guidance for reducing carbon emissions from green waste composting. These results show that there is great potential for using machine learning algorithms to predict carbon dioxide output from green waste composting.

RETRACTED: Chemistry-Informed Neural Networks modelling of lignocellulosic biomass pyrolysis

This article has been retracted: please see Elsevier Policy on Article Withdrawal (http://www.elsevier.com/locate/withdrawalpolicy). This article has been retracted at the request of the authors and the Editor-in-Chief. The article has reused text from the papers published by other authors in Combustion and Flame 240 (2022) 111992 https://doi.org/10.1016/j.combustflame.2022.111992 and the Journal of Physical Chemistry A 125 (2021) 1082–1092 https://doi.org/10.1021/acs.jpca.0c09316 without proper citation and discussion of the two articles. One of the conditions of submission of a paper for publication is that authors declare explicitly that their work is original and has not appeared in a publication elsewhere. As such this article represents a misuse of the scientific publishing system. The scientific community takes a strong view on this matter and apologies are offered to readers of the journal that this was not detected during the submission process.

Machine learning prediction of lignin content in poplar with Raman spectroscopy

Based on features extracted from Raman spectra, regularization algorithms, SVR, DT, RF, LightGBM, CatBoost, and XGBoost were used to develop prediction models for lignin content in poplar. Firstly, Raman features extracted from FT-Raman spectra after data processing were used as input of models and determined lignin contents were output. Secondly, grid-search combined with cross-validation was used to adjust the hyper-parameters of models. Finally, the predictive models were built by aforementioned algorithms. The results indicated regularization algorithms, SVR, DT held test R² were >0.80 which means the predictive values from model still deviate from measured ones. Meanwhile, RF, LightGBM, CatBoost, and XGBoost were better than above algorithms, and their test R² were >0.91 which suggesting the predictive values was nearly close to measured ones. Therefore, fast and accurate methods for predicting lignin content were obtained and will be useful for screening suitable lignocellulosic resource with expected lignin content.

Machine learning based analysis of reaction phenomena in catalytic lignin depolymerization

Heterogeneously catalyzed lignin solvolysis opens the possibility of transforming low value biomass into high value, useful aromatic chemicals, however, its reaction behavior is poorly understood due to the many possible interactions between reaction parameters. In this study, a novel predictive model for bio-oil yield, char yield and reaction time is developed using Random Forest (RF) regression method using data available from the literature to study the impact of surface properties of the catalyst and the weight averaged molecular weight of the lignin (M_w) used in the reaction. The models achieved a coefficient of determination (R2) score of 0.9062, 0.9428 and 0.8327, respectively, and feature importance for each case was explained and tied to studies that provide a mechanistic explanation for the performance of the model. Surface properties and lignin M_w showed no importance to the prediction of bio-oil yield and average pore diameter contributed 3% of feature importance to reaction time.

Recent advances of thermochemical conversion processes for biorefinery

Lignocellulosic biomass is one of the most promising renewable resources and can replace fossil fuels via various biorefinery processes. Through this study, we addressed and analyzed recent advances in the thermochemical conversion of various lignocellulosic biomasses. We summarized the operation conditions and results related to each thermochemical conversion processes such as pyrolysis (torrefaction), hydrothermal treatment, gasification and combustion. This review indicates that using thermochemical conversion processes in biorefineries is techno-economically feasible, easy, and effective compared with biological processes. The challenges experienced in thermochemical conversion processes are also presented in this study for better understanding the future of thermochemical conversion processes for biorefinery. With the aid of artificial intelligence and machine learning, we can reduce time-consumption and experimental work for bio-oil production and syngas production processes.

Machine learning modeling and analysis of biohydrogen production from wastewater by dark fermentation process

Dark fermentation process for simultaneous wastewater treatment and H₂ production is gaining attention. This study aimed to use machine learning (ML) procedures to model and analyze H₂ production from wastewater during dark fermentation. Different ML procedures were assessed based on the mean squared error (MSE) and determination coefficient (R²) to select the most robust models for modeling the process. The research showed that gradient boosting machine (GBM), support vector machine (SVM), random forest (RF) and AdaBoost were the most appropriate models, which were optimized by grid search and deeply analyzed by permutation variable importance (PVI) to identify the relative importance of process variables. All four models demonstrated promising performances in predicting H₂ production with high R² values (0.893, 0.885, 0.902 and 0.889) and small MSE values (0.015, 0.015, 0.016 and 0.015). Moreover, RF-PVI demonstrated that acetate, butyrate, acetate/butyrate, ethanol, Fe and Ni were of high importance in decreasing order.

The role of machine learning to boost the bioenergy and biofuels conversion

The development and application of bioenergy and biofuels conversion technology can play a significant role for the production of renewable and sustainable energy sources in the future. However, the complexity of bioenergy systems and the limitations of human understanding make it difficult to build models based on experience or theory for accurate predictions. Recent developments in data science and machine learning (ML), can provide new opportunities. Accordingly, this critical review provides a deep insight into the application of ML in the bioenergy context. The latest advances in ML assisted bioenergy technology, including energy utilization of lignocellulosic biomass, microalgae cultivation, biofuels conversion and application, are reviewed in detail. The strengths and limitations of ML in bioenergy systems are comprehensively analysed. Moreover, we highlight the capabilities and potential of advanced ML methods when encountering multifarious tasks in the future prospects to advance a new generation of bioenergy and biofuels conversion technologies.

Improved Estimation of Bio-Oil Yield Based on Pyrolysis Conditions and Biomass Compositions Using GA- and PSO-ANFIS Models

This paper incorporates the adaptive neurofuzzy inference system (ANFIS) technique to model the yield of bio-oil. The estimation of this parameter was performed according to pyrolysis conditions and biomass compositions of feedstock. For this purpose, this paper innovates two optimization methods including a genetic algorithm (GA) and particle swarm optimization (PSO). Primary data were gathered from previous studies and included 244 data of biodiesel oils. The findings showed a coefficient determination (R²) of 0.937 and RMSE of 2.1053 for the GA-ANFIS model, and a coefficient determination (R²) of 0.968 and RMSE of 1.4443 for PSO-ANFIS. This study indicates the capability of the PSO-ANFIS algorithm in the estimation of the bio-oil yield. According to the performed analysis, this model shows a higher ability than the previously presented models in predicting the target values and can be a suitable alternative to time-consuming and difficult experimental tests.

A comparative study of machine learning methods for bio-oil yield prediction - A genetic algorithm-based features selection

A novel genetic algorithm-based feature selection approach is incorporated and based on these features, four different ML methods were investigated. According to the findings, ML models could reliably predict bio-oil yield. The results showed that Random forest (RF) is preferred for bio-oil yield prediction (R2 ~ 0.98) and highly recommended when dealing with the complex correlation between variables and target. Multi-Linear regression model showed relatively poor generalization performance (R2 ~ 0.75). The partial dependence analysis was done for ML models to show the influence of each input variable on the target variable. Lastly, an easy-to-use software package was developed based on the RF model for the prediction of bio-oil yield. The current study offered new insights into the pyrolysis process of biomass and to improve bio-oil yield. It is an attempt to reduce the time-consuming and expensive experimental work for estimating the bio-oil yield of biomass during pyrolysis.

Keystone taxa shared between earthworm gut and soil indigenous microbial communities collaboratively resist chlordane stress

Chlordane is an organochlorine pesticide that is applied extensively. Residual concentrations that remain in soils after application are highly toxic to soil organisms, particularly affecting the earthworm gut and indigenous soil microorganisms. However, response mechanisms of the earthworm gut and indigenous soil microorganism communities to chlordane exposure are not well known. In this study, earthworms (Metaphire guillelmi) were exposed to chlordane-contaminated soils to investigate their response mechanisms over a gradient of chlordane toxicity. Results from high-throughput sequencing and network analysis showed that the bacterial composition in the earthworm gut varied more significantly than that in indigenous soil microbial communities under different concentrations of chlordane stress (2.3-60.8 mg kg^-1; p < 0.05). However, keystone species of Flavobacterium, Candidatus Nitrososphaera, and Acinetobacter remained stable in both the earthworm gut and bacterial communities despite varying degrees of chlordane exposure, and their relative abundance was slightly higher in the low-concentration treatment group (T1, T2) than in the high-concentration treatment group (T3, T4). Additionally, network analysis demonstrated that the average value of the mean degree of centrality, closeness centrality, and eigenvector centrality of all keystone species screened by four methods (MetagenomeSeq, LEfSe, OPLS-DA, Random Forest) were 161.3, 0.5, and 0.63, respectively, and that these were significantly higher (p < 0.05) than values for non-keystone species (84.9, 0.4, and 0.2, respectively). Keystone species had greater network connectivity and a stronger capacity to degrade pesticides and transform carbon and nitrogen than non-keystone species. The keystone species, which were closely related to the microbial community in soil indigenous flora and earthworm intestinal flora, could resist chlordane stress and undertake pesticide degradation. These results have increased understanding of the role of the earthworm gut and indigenous soil bacteria in resisting chlordane stress and sustaining microbial equilibrium in soil.