Input Variables Research Articles

Production and testing of mahua oil-based biodiesel synthesized through heterogeneous catalyst using experimental and numerical method.

A two-step treatment of mahua oil was conducted to synthesize mahua biodiesel using heterogeneous biomass-based catalyst derived from mahua shell. Mahua oil having higher free fatty acid (FFA) content (about 19%) was esterified to reduce the FFA content up to 1%. The esterification process was carried out using 200mL mahua oil, 5:1 molar ratio (methanol:oil), and 2.25 weight% of H2SO4 at a temperature of 60°C for 3h. Post esterification, a set of 16 experiments were created using a Box-Behnken design (BBD)-based response surface methodology (RSM) approach to conduct the transesterification of the esterified oil. Molar ratio, catalyst loading, reaction temperature, and reaction time were the four input variables chosen for the design of experiments. The optimized conditions for maximum biodiesel yield (87.7%) were found to be 14.88 molar ratio, 3.578% catalyst loading, 69.7°C reaction temperature, and 81.9min reaction time. The Diesel RK engine simulation tool which was experimentally validated for baseline diesel fuel was used for numerical simulation of mahua biodiesel. The performance, combustion, and emission behavior of mahua biodiesel analyzed using numerical simulation presented the sustainability of mahua biodiesel as an alternate fuel.

Evolutionary Algorithms for Strength Prediction of Geopolymer Concrete

Geopolymer concrete (GPC) serves as a sustainable substitute for conventional concrete by employing alternative cementitious materials such as fly ash (FA) instead of ordinary Portland cement (OPC), contributing to environmental and durability benefits. To increase the rate of utilization of FA in the construction industry, distinctive characteristics of two machine learning (ML) methods, namely, gene expression programming (GEP) and multi-expression programming (MEP), were utilized in this study to propose precise prediction models for the compressive strength and split tensile strength of GPC comprising FA as a binder. A comprehensive database was collated, which comprised 301 compressive strength and 96 split tensile strength results. Seven distinct input variables were employed for the modeling purpose, i.e., FA, sodium hydroxide, sodium silicate, water, superplasticizer, and fine and coarse aggregates contents. The performance of the developed models was assessed via numerous statistical metrics and absolute error plots. In addition, a parametric analysis of the finalized models was performed to validate the prediction ability and accuracy of the finalized models. The GEP-based prediction models exhibited better performance, accuracy, and generalization capability compared with the MEP-based models in this study. The GEP-based models demonstrated higher correlation coefficients (R) for predicting the compressive and split tensile strengths, with the values of 0.89 and 0.87, respectively, compared with the MEP-based models, which yielded the R values of 0.76 and 0.73, respectively. The mean absolute errors for the GEP- and MEP-based models for predicting the compressive strength were 5.09 MPa and 6.78 MPa, respectively, while those for the split tensile strengths were 0.42 MPa and 0.51 MPa, respectively. The finalized models offered simple mathematical formulations using the GEP and Python code-based formulations from MEP for predicting the compressive and tensile strengths of GPC. The developed models indicated practical application potential in optimizing geopolymer mix designs. This research work contributes to the ongoing efforts in advancing ML applications in the construction industry, highlighting the importance of sustainable materials for the future.

Multi‐response optimization of characteristics for graphite reinforced biodegradable PVA‐fumaric acid cross‐linked composite: A gray‐fuzzy logic‐based hybrid approach

AbstractIn the current investigation, enhancement in the mechanical properties, thermal stability, and decrement in the moisture absorption capabilities of polyvinyl alcohol was observed by cross‐linking it with different weight percentages (wt. %) of fumaric acid and reinforcing them along with sulfuric acid‐modified graphite particles. The effects of altering the cross‐linking duration and fumaric acid wt. % were investigated in the resulting composite using the Taguchi L16 experimental design method. As in this contemporary era of rapid industrial expansion and fierce competition, the primary goals of manufacturers revolve around the production of high‐quality goods at optimal costs within a specified timeframe. Therefore, to save both cost and time, the authors have utilized a technique known as “gray‐fuzzy optimization” to balance competing goals between mechanical (toughness, elongation and strength) and physical (water intake capacity) attributes. The present approach aimed to reduce the experimentation expenses and increase the effectiveness of composite manufacturing by considerably reducing the intricacy of practical multi‐response optimization issues. Response graphs and analysis of variance tests were also utilized to calculate the percentage contribution of the input variables and provide model significance.Highlights Polyvinyl alcohol is cross‐linked with different weight percentages (wt. %) of fumaric acid, and reinforced them with sulfuric acid‐modified graphite particles. Enhancement in mechanical properties, thermal stability, and decrease in moisture absorption capabilities were perceived for the fabricated composite. Effects of altering the cross‐linking duration and fumaric acid wt. % were investigated in the resulting composite using the Taguchi L16 experimental design. Gray‐fuzzy optimization technique was utilized to balance competing goals between water intake and mechanical attributes like toughness, elongation and strength. Response graphs and analysis of variance (ANOVA) tests were also utilized to evaluate the percentage contribution of input variables and provide model significance.

Artificial intelligence modeling and simulation of membrane-based separation of water pollutants via ozone Process: Evaluation of separation

The present study offers a comparative examination of regression models that are utilized for the prediction of concentration (C) in a new hybrid ozone-membrane process for removal of water pollutants. The main focus is on the tracking ozone concentration in the feed side of a membrane contactor system. Computational fluid dynamics (CFD) was carried out to obtain data of ozone concentration (C) for developing some machine learning (ML) models. The models are based on input variables r (m) and z (m). The dataset comprises over 10,000 data, and three different models, namely Convolutional Neural Network (CNN), Support Vector Regression (SVR), and Orthogonal Matching Pursuit (OMP), are evaluated. The hyperparameters of these models are optimized using the Glowworm Swarm Optimization (GSO) technique. Prior to model training, preprocessing steps are applied. The findings suggest that SVR exhibited a noteworthy R2 score of 0.99698, surpassing CNN which obtained a R2 score of 0.98073, and OMP which obtained a R2 score of 0.8748. The aforementioned discoveries offer significant perspectives on the utilization of diverse machine learning models in the prognostication of C, demonstrating their efficacy and proficiency in this particular realm.

Conceptualization of Intelligent Control System for Humanitarian Demining Robotic Complexes Based on Verbal Methods

Introduction. Humanitarian demining is characterized by growing attention to the problems of creating and using robotic systems. The most important aspect of their use is the control method.Problem Statement. At present, the fi rst generation robotic complexes (controlled devices) have been the mostcommon, the second generation complexes (semi-autonomous devices) have been improving. To switch to the third generation complexes (autonomous devices), it is necessary to develop intelligent control systems based on artificial intelligence technologies. The most used for system development are quantitative methods, but such models are “black box models” that do not provide complete clarity about their behavior.Purpose. To develop conceptual model of intelligent control systems for humanitarian demining robotic complexes based on verbal methods.Materials and Methods. Formal logic methods, verbal analysis qualitative methods, and BPMN. Results. Conceptual model of intelligent control systems for humanitarian demining robotic complexes and sym bolic models of representing relevant declarative and procedural knowledge based on verbal methods have been developed.Conclusions. The conceptual model makes it possible to formulate Symbolic Models in the notations of selected verbal methods. At the decision-making level, the ordinary classifi cation method by denotation of the terms used by experts in the selected subject area has been chosen. At the executive level, the BPMN method has been chosen to create process diagrams in graphical notation. The verbal methods make it possible to create “white-box models” that unambiguously interpret the dependence of output and input variables and to explain the model behavior. Symbolic models in the notations of selected verbal methods allow the implementation of an intelligent control systems for a specifi c humanitarian demining robotic complex.

Minimizing Temperature Deviations in Rubber Mixing Process using Artificial Neural Networks

ABSTRACT Rubber mixing is a complex manufacturing process that poses challenges for process control due to the high number of control variables including mixing parameter settings, rheological behaviour, compound viscosity and batch-dependent material variations. Already small deviations from the control variables can influence the compound properties, leading to increased scrap rates. To address these challenges, this paper introduces an Artificial Intelligence (AI)-based approach to enhance process control in rubber mixing by predicting mixing temperatures from input variables. The proposed method utilizes Feedforward Neural Networks (FFN) to enable early identification of batch-specific temperature deviations, enabling systematic improvements with each new application. The FFN was trained on a diverse dataset encompassing various rubber recipes and batches. Post-training, the FFN demonstrated remarkable accuracy, achieving a Mean Absolute Percentage Error (MAPE) of 1.00% on the training dataset and 1.44% on the validation dataset, thereby showcasing its efficacy in predicting temperature fluctuations within the mixing process. Consequently, the FFN can determine the relevant input variables necessary to achieve specific mixing temperatures, providing a foundation for an automated control system in rubber mixing process. This paper outlines the system architecture of the FFN tailored for rubber mixing and provides a comprehensive overview of the experimental results.

Heavy metals removal from wastewater using nanoporous adsorbent: Separation analysis via machine learning model

This paper presents a comprehensive analysis of three predictive models, namely Multi-Layer Perceptron (MLP), LASSO, and Extreme Gradient Boosting (XGB) for estimating concentration (C) of a substance in a given dataset. Adsorption separation was considered for water treatment, and the models were employed for tracking the concentration variations of solute in the process. With x(m) and y(m) as input variables which are the location, and concentration (C) measured in mol/m³ as output, the dataset comprises more than 19,000 data points. The Fireworks Algorithm (FWA) was employed to perform hyper-parameter optimization for the models. Different metrics were utilized to gauge the proficiency of each model, such as the R2 (coefficient of determination) for both the training and testing datasets, RMSE, and MAE. Results indicate that the MLP model obtained the highest score in terms of R2, with values of 0.99751 for the training data and 0.99756 for the testing data, suggesting excellent predictive accuracy. The MLP model also demonstrated the lowest error rates, with an RMSE of 1.3937 E+00 and an MAE of 8.81875E-01. The results revealed that the employed machine learning models are well capable of predicting adsorption process when combined with mass transfer modeling and great accuracy can be obtained.

Microbial Fuel Cells Performance Evaluation through the Application of Fuzzy Logic

Objective: This work aimed to employ fuzzy logic combined with the design of experiments technique to statistically evaluate how the microbial fuel cells operating parameters influence their performance. Theoretical Framework: Microbial fuel cells (MFC) are a technology of interest in the current scenario as they allow simultaneously promoting the biotreatment of waste and the bio generation of electrical energy. Methodology: Through a bibliographical search based on publications on Google Scholar platform over the last 10 years, it was noticed that anode area, external electrical resistance and reactor volume are the most reported input parameters in MFC research and current density and power density are the output parameters most frequently portrayed in these studies, which is why these variables were selected for statistical investigation. Results and Discussion: The results showed that, for both outputs studied, reactor volume and anode area showed a positive effect, while the external electrical resistance showed a negative effect. It was also possible to develop mathematical models that indicated the relationship between the input and output variables studied, with statistical significance for power density model (R2 = 86%). Originality: Applying computational simulations and subsequent design of experiments to obtain results in accordance with those obtained experimentally in the laboratory.

Predicting hourly heating load in residential buildings using a hybrid SSA–CNN–SVM approach

This study proposes a hybrid prediction model using sparrow search algorithm (SSA) to optimize the convolutional neural network (CNN) and support vector machine (SVM), in order to perform accurate prediction of secondary supply temperature (Ts2). The historical operation data of Weifang residential building thermal station was adopted and reasonable data preprocessing was performed to suppress the interference of abnormal data. The input variables of the prediction model were screened using the correlation analysis method, taking the influence of the hysteresis effect into consideration. The SSA–CNN–SVM model was then developed for prediction. The performance of the model was evaluated by the root mean square error, mean absolute error, mean absolute percentage error (MAPE), and absolute value of relative error of each time step. The results obtained demonstrated that the SSA–CNN–SVM model has high prediction accuracy. The MAPE values of the two heat exchange stations were between 2.28 % and 2.4 %. The indoor temperature significantly affected the prediction accuracy of Ts2. After the introduction of the indoor temperature, the MAPE of the predicted values of the hybrid model was reduced by 0.35 %. The maximum reduction in MAPE of the SSA–CNN–SVM model was 1.5 % compared with other prediction models.

Comparative performance of intensive care mortality prediction models based on manually curated versus automatically extracted electronic health record data

IntroductionBenchmarking intensive care units for audit and feedback is frequently based on comparing actual mortality versus predicted mortality. Traditionally, mortality prediction models rely on a limited number of input variables and significant manual data entry and curation. Using automatically extracted electronic health record data may be a promising alternative. However, adequate data on comparative performance between these approaches is currently lacking. MethodsThe AmsterdamUMCdb intensive care database was used to construct a baseline APACHE IV in-hospital mortality model based on data typically available through manual data curation. Subsequently, new in-hospital mortality models were systematically developed and evaluated. New models differed with respect to the extent of automatic variable extraction, classification method, recalibration usage and the size of collection window. ResultsA total of 13 models were developed based on data from 5,077 admissions divided into a train (80%) and test (20%) cohort. Adding variables or extending collection windows only marginally improved discrimination and calibration. An XGBoost model using only automatically extracted variables, and therefore no acute or chronic diagnoses, was the best performing automated model with an AUC of 0.89 and a Brier score of 0.10. DiscussionPerformance of intensive care mortality prediction models based on manually curated versus automatically extracted electronic health record data is similar. Importantly, our results suggest that variables typically requiring manual curation, such as diagnosis at admission and comorbidities, may not be necessary for accurate mortality prediction. These proof-of-concept results require replication using multi-centre data.

Efficient global sensitivity analysis method for dynamic models in high dimensions

AbstractDynamic models generating time‐dependent model predictions are typically associated with high‐dimensional input spaces and high‐dimensional output spaces, in particular if time is discretized. It is computationally prohibitive to apply traditional global sensitivity analysis (SA) separately on each time output, as is common in the literature on multivariate SA. As an alternative, we propose a novel method for efficient global SA of dynamic models with high‐dimensional inputs by combining a new polynomial chaos expansion (PCE)‐driven partial least squares (PLS) algorithm with the analysis of variance. PLS is used to simultaneously reduce the dimensionality of the input and output variables spaces, by identifying the input and output latent variables that account for most of their joint variability. PCE is incorporated into the PLS algorithm to capture the non‐linear behavior of the physical system. We derive the sensitivity indices associated with each output latent variable, based on which we propose generalized sensitivity indices that synthesize the influence of each input on the variance of entire output time series. All sensitivities can be computed analytically by post‐processing the coefficients of the PLS‐PCE representation. Hence, the computational cost of global SA for dynamic models essentially reduces to the cost for estimating these coefficients. We numerically compare the proposed method with existing methods by several dynamic models with high‐dimensional inputs. The results show that the PLS‐PCE method can obtain accurate sensitivity indices at low computational cost, even for models with strong interaction among the inputs.

Factors influencing the development of artificial intelligence in orthodontics.

Since developing AI procedures demands significant computing resources and time, the implementation of a careful experimental design is essential. The purpose of this study was to investigate factors influencing the development of AI in orthodontics. A total of 162 AI models were developed, with various combinations of sample sizes (170, 340, 679), input variables (40, 80, 160), output variables (38, 76, 154), training sessions (100, 500, 1000), and computer specifications (new vs. old). The TabNet deep-learning algorithm was used to develop these AI models, and leave-one-out cross-validation was applied in training. The goodness-of-fit of the regression models was compared using the adjusted coefficient of determination values, and the best-fit model was selected accordingly. Multiple linear regression analyses were employed to investigate the relationship between the influencing factors. Increasing the number of training sessions enhanced the effectiveness of the AI models. The best-fit regression model for predicting the computational time of AI, which included logarithmic transformation of time, sample size, and training session variables, demonstrated an adjusted coefficient of determination of 0.99. The study results show that estimating the time required for AI development may be possible using logarithmic transformations of time, sample size, and training session variables, followed by applying coefficients estimated through several pilot studies with reduced sample sizes and reduced training sessions.

Climate classification for Northeast Brazil using reanalysis data and the Absolute Aridity Index

The climate of the Northeast Region of Brazil (NEB) has been intensively studied and analyzed for climate classification. The aridity index of the United Nations Environment Programme (UNEP) (AIUNEP) has been used for this purpose, but without fully satisfactory results. The input variables needed for its calculation are precipitation and reference potential evapotranspiration (ET0). However, although rainfall stations recording routine measurements of precipitation are well distributed in the NEB, they do not provide the necessary variables for estimating ET0. Thus, interpolation is used to calculate ET0, but this can generate errors. Another objective climate classification approach is the Thornthwaite method, based on the determination of the moisture index (Im), whose calculation also requires weather station data. Thus, seeking to circumvent the problem of paucity of stations and improve the spatial distribution of information on meteorological variables in the NEB, the present work had as one of its objectives to validate reanalysis data from ERA5 of the European Center for Medium-range Weather Forecast (ECMWF) and the unified gauge-based analysis of global daily precipitation project of the Climate Prediction Center/National Oceanic and Atmospheric Administration (CPC/NOAA). After validation, climate classifications were developed for the NEB using the AIUNEP and Im. It was observed that the Thornthwaite climate classification overestimated the aridity in the NEB, while the IAUNEP tended to underestimate it. For this reason, a new climate classification index, called the absolute aridity index (Iab), was suggested, yielding satisfactory results.

Machine learning-based model for predicting arrival time of container ships

Forecasting container ship arrival times is challenging, requiring a thorough analysis for accuracy. This study investigates the effectiveness of machine learning (ML) techniques in maritime transportation. Using a dataset of 581 samples with 8 input variables and 1 output variable (arrival time), ML models are constructed. The Pearson correlation matrix reduces input variables to 7 key factors: freight forwarder, dispatch location, loading and discharge ports, post-discharge location, dispatch day of the week, and dispatch week. The ranking of ML performance for predicting the arrival time of container ships can be arranged in descending order as GB-PSO > XGB > RF > RF-PSO > GB > KNN > SVR. The best ML model, GB-PSO, demonstrates high accuracy in predicting the arrival time of container ships, with R2 = 0.7054, RMSE = 7.4081 days, MAE = 5.1891 days, and MAPE = 0.0993% for the testing dataset. This is a promising research outcome as it seems to be the first time that an approach involving the use of minimal and easily collectible input factors (such as freight forwarder, dispatch time and place, port of loading, post port of discharge, port of discharge) and the combination of a machine learning model has been introduced for predicting the arrival time of container ships.

Designing Pathway-Controlled Multicomponent Ultrashort Peptide Hydrogels with Diverse Functionalities at the Nanoscale for Directing Cellular Behavior.

Tuning self-assembling pathways by implementing different external stimuli has been extensively studied, owing to their effective control over structural and mechanical properties. Consequently, multicomponent peptide hydrogels with high structural tunability and stimuli responsiveness are crucial in dictating cellular behavior. Herein, we have implemented both coassembly approach and pathway-dependent self-assembly to design nonequilibrium nanostructures to understand the thermodynamic and kinetic aspects of peptide self-assembly toward controlling cellular response. Our system involved an ultrashort peptide gelator and a hydrophilic surfactant which coassembled through different pathways, i.e., heat-cool and sonication methods with variable energy input. Interestingly, it was possible to access diverse structural and mechanical properties at the nanoscale in a single coassembled system. Further, the hydrophilic surfactant provided additional surface functionalities, thus creating an efficient hydrophilic matrix for cellular interaction. Such diverse functionalities in a single coassembled system could lead to the development of advanced scaffolds, with applications in various biomedical fields.

Reliable smart models for estimating frictional pressure drop in two-phase condensation through smooth channels of varying sizes

Reliable and comprehensive predictive tools for the frictional pressure drop (FPD) are of particular importance for systems involving two-phase flow condensation. However, the available models are only applicable to specific operating conditions and channel sizes. Thus, this study aims at developing universal models to estimate the FPD during condensation inside smooth mini/micro and conventional (macro) channels. An extensive databank, comprising 8037 experimental samples and 23 working fluids from 50 reliable sources, was prepared to achieve this target. A comprehensive investigation on the literature models reflected the fact that all of them are associated with high deviations, and their average absolute relative errors (AAREs) exceed 26%. Hence, after identifying the most effective input variables through the Spearman's correlation analysis, three soft-computing paradigms, i.e., multilayer perceptron (MLP), gaussian process regression (GPR) and radial basis function (RBF) were employed to establish intelligent and dimensionless predictive tools for the FPD based on the separated model suggested by Lockhart and Martinelli. Among them, the most accurate results were presented by the GPR approach with AARE and R2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^{2}$$\\end{document} values of 4.10%, 99.23% respectively, in the testing step. The truthfulness and applicability of the models were explored through an array of statistical and visual analyses, and the results affirmed the obvious superiority of the newly proposed approaches over the literature correlations. Furthermore, the novel predictive tools excellently described the physical variations of the condensation FPD versus the operating parameters. Ultimately, the order of importance of factors in controlling the condensation FPD was clarified by a sensitivity analysis.

Deriving Input Variables through Applied Machine Learning for Short-Term Electric Load Forecasting in Eskilstuna, Sweden

As the demand for electricity, electrification, and renewable energy rises, accurate forecasting and flexible energy management become imperative. Distribution network operators face capacity limits set by regional grids, risking economic penalties if exceeded. This study examined data-driven approaches of load forecasting to address these challenges on a city scale through a use case study of Eskilstuna, Sweden. Multiple Linear Regression was used to model electric load data, identifying key calendar and meteorological variables through a rolling origin validation process, using three years of historical data. Despite its low cost, Multiple Linear Regression outperforms the more expensive non-linear Light Gradient Boosting Machine, and both outperform the “weekly Naïve” benchmark with a relative Root Mean Square Errors of 32–34% and 39–40%, respectively. Best-practice hyperparameter settings were derived, and they emphasize frequent re-training, maximizing the training data size, and setting a lag size larger than or equal to the forecast horizon for improved accuracy. Combining both models into an ensemble could the enhance accuracy. This paper demonstrates that robust load forecasts can be achieved by leveraging domain knowledge and statistical analysis, utilizing readily available machine learning libraries. The methodology for achieving this is presented within the paper. These models have the potential for economic optimization and load-shifting strategies, offering valuable insights into sustainable energy management.

Near Dry Powder Mixed Electric Discharge Machining of AA7050 Hybrid Composites Utilizing Composite Tool Materials

Electric Discharge Machining (EDM) is a prominent technique for producing components with complex geometry and intricate shapes. However, productivity depends on the selection of optimal input variables. This research investigates EDM and Near-Dry EDM (NDEDM) on AA7050 hybrid composites, exploring the impact of tool materials, Al2O3 powder concentration, and dielectric fluids on Material Removal Rate (MRR) and Surface Roughness (Ra). NDEDM offers a higher MRR due to localized heating and energy concentration, while hydrocarbon oil EDM provides lower Ra due to cooling and lubrication. Tool material properties, thermal conductivity and bonding strength, affect MRR and TWR. Aluminum composite tools via stir casting display a strong bonding interface, leading to efficient energy transfer and higher MRR, while copper composite tools via friction stir processing exhibit weaker bonding, potentially reducing MRR. Hydrocarbon oil significantly lowers TWR due to effective lubrication and cooling, while near dry EDM results in slightly higher TWR. Al2O3 powder enhances MRR, with an optimal concentration of 1 g/l, while higher concentrations hinder MRR. The presence of Al2O3 powder enhanced the surface quality by reducing the size of the recast layers and producing a smoother surface. Globules, craters, microcracks, pock marks and re-casted layers are the noticeable features spotted on the surface. The results were optimized using the TOPSIS optimization technique and process parameters 32 A current, 30 µs Ton, 3 mm gap distance machined under 3 g/l Al2O3 incorporated hydrocarbon oil dielectric medium using copper composites tool proffers optimal machining performance.

Development of an early warning system for higher education institutions by predicting first‐year student academic performance

AbstractIn this study, an early warning system predicting first‐year undergraduate student academic performance is developed for higher education institutions. The significant factors that affect first‐year student success are derived and discussed such that they can be used for policy developments by related bodies. The dataset used in experimental analyses includes 11,698 freshman students' data. The problem is constructed as classification models predicting whether a student will be successful or unsuccessful at the end of the first year. A total of 69 input variables are utilized in the models. Naive Bayes, decision tree and random forest algorithms are compared over model prediction performances. Random forest models outperformed others and reached 90.2% accuracy. Findings show that the models including the fall semester CGPA variable performed dramatically better. Moreover, the student's programme name and university placement exam score are identified as the other most significant variables. A critical discussion based on the findings is provided. The developed model may be used as an early warning system, such that necessary actions can be taken after the second week of the spring semester for students predicted to be unsuccessful to increase their success and prevent attrition.

Developing interpretable machine learning model for evaluating young modulus of cemented paste backfill

Cemented paste backfill (CPB), a mixture of wet tailings, binding agent, and water, proves cost-effective and environmentally beneficial. Determining the Young modulus during CPB mix design is crucial. Utilizing machine learning (ML) tools for Young modulus evaluation and prediction streamlines the CPB mix design process. This study employed six ML models, including three shallow models Extreme Gradient Boosting (XGB), Gradient Boosting (GB), Random Forest (RF) and three hybrids Extreme Gradient Boosting-Particle Swarm Optimization (XGB-PSO), Gradient Boosting-Particle Swarm Optimization (GB-PSO), Random Forest-Particle Swarm Optimization (RF-PSO). The XGB-PSO hybrid model exhibited superior performance (coefficient of determination R2 = 0.906, root mean square error RMSE = 19.535 MPa, mean absolute error MAE = 13.741 MPa) on the testing dataset. Shapley Additive Explanation (SHAP) values and Partial Dependence Plots (PDP) provided insights into component influences. Cement/Tailings ratio emerged as the most crucial factor for enhancing Young modulus in CPB. Global interpretation using SHAP values identified six essential input variables: Cement/Tailings, Curing age, Cc, solid content, Fe2O3 content, and SiO2 content.