Genetic Algorithm Neural Network Research Articles

Application of ionic liquid ultrasound-assisted extraction (IL-UAE) of lycopene from guava (Psidium guajava L.) by response surface methodology and artificial neural network-genetic algorithm

Lycopene-rich guava (Psidium guajava L.) exhibits significant economic potential as a functional food ingredient, making it highly valuable for the pharmaceutical and agro-food industries. However, there is a need to enhance the extraction methods of lycopene to fully exploit its beneficial uses. In this study, we evaluated various ionic liquids to identify the most effective one for extracting lycopene from guava. Among thirteen ionic liquids with varying carbon chains or anions, 1-butyl-3-methylimidazolium chloride demonstrated the highest productivity. Subsequently, a single-factor experiment was employed to test the impact of several parameters on the efficiency of lycopene extraction using this selected ionic liquid. These parameters included extraction time, ultrasonic power, liquid-solid ratio, concentration of the ionic liquid, as well as material particle size. Moreover, models of artificial neural networks using genetic algorithms (ANN-GA) and response surface methodology (RSM) were employed to comprehensively assess the first four key parameters. The optimized conditions for ionic liquid ultrasound-assisted extraction (IL-UAE) were determined as follows: 33 min of extraction time, 225 W of ultrasonic power, 22 mL/g of liquid-solid ratio, 3.0 mol/L of IL concentration, and extraction cycles of three. Under these conditions, lycopene production reached an impressive yield of 9.35 ± 0.36 mg/g while offering advantages such as high efficiency, time savings, preservation benefits, and most importantly environmental friendliness.

Process Optimization for the Extraction of Phenolic Compounds from Pomegranate Peels: Response Surface Methodology-Desirability Function and Artificial Neural Network-Genetic Algorithm

Valorization of agricultural wastes is ongoing topic in industry. Determining the best conditions by artificial neural networks based optimization techniques is the key step to extract valuable compounds efficiently and to obtain high quality extracts. In this study, the response surface methodology (RSM)-desirability function (DF) and artificial neural network (ANN)-genetic algorithm (GA) approaches were compared in modeling and optimization the extraction parameters (temperature, time and ethanol concentration (ratio of ethanol to water, % v/v)) of phenolic compounds in pomegranate peels. The ANN-GA approach providing higher coefficient of determination and lower root mean square deviation showed better predictive capability than the RSM. The optimum time (81.4 min) and ethanol concentration (15.7%) of RSM-DF approach shifted to the lower levels (78.8 min and 15.3%) with the ANN-GA approach while the optimum temperature (54.0°C) shifted to a higher level (59.3°C). The use of these values provided total phenolic content of >1000 mg GAE L-1 and the corresponding antioxidant activity was 11 mmol TE L-1. As a result, increasing temperature up to a critical level decreased the extraction time and ethanol concentration, and it was determined that higher time-temperature combinations must be used for the complete water-based extraction of phenolic compounds from plant wastes in comparison to ethanol-water based extraction.

Advanced machine learning techniques: Forecasting thermal resistance in borehole heat exchanger system through RSM and hybrid DFNN-GA approaches

The thermal efficiency of ground heat exchanger (GHEs) relies on optimizing borehole thermal resistance and total internal thermal resistance for cost-effective and efficient design. This study aims to establish a robust correlation among various parameters, including thermal conductivity of pipe, grout, and ground, borehole depth, borehole and pipe diameter, pipe-pipe distance, fluid velocity, and heat transfer rate of the ground, with the 3D borehole thermal resistance (Rb,3D) and total internal thermal resistance (Ra,3D). This exploration of eight influential factors results in an l-130 design matrix with 130 data points. A 3D computational model, utilizing a four-resistance method, accurately calculates Rb,3D and Ra,3D, while analyzing heat transfer in both solids and fluids. Two prediction models, Response Surface Methodology (RSM) and a hybrid Deep Feedforward Neural Network-Genetic Algorithm (DFNN-GA), establish the correlation between Rb,3D and Ra,3D and the eight variables. The results highlight the efficacy of the prediction models, achieving R2 values of 79.87% and 87.90% for RSM and 94.00% and 99.15% for DFNN-GA, corresponding to the predictive accuracy of Rb,3D and Ra,3D, respectively. Furthermore, lower RSME, MAE, and AARE values are observed. A rigorous comparative analysis assesses the relative effectiveness of these methodologies in achieving the research goals.

Controlling recast layer and material removal in reverse µEDM for sustainable fabrication of microelectrode arrays

Microelectrode arrays (MEAs) are essential for researching neurons and heart tissue because they record and stimulate neuronal activity. For manufacturing MEAs on conductive materials, reverse microelectrical discharge machining, or reverse µEDM, presents a viable method. However, achieving precise MEAs requires control over recast layers generated during micromachining. This study focuses on recast layer thickness (RLT) and material removal rate (MRR) as key performance indicators for sustainable reverse µEDM fabrication of MEAs. RLT affects surface stability, while MRR influences energy consumption and cost. Optimising these responses depends on three input factors: feed rate, capacitance, and voltage. ANOVA analysis reveals that capacitance is crucial in both MRR and recast layer thickness (RLT) in reverse µEDM, contributing to 88.71% and 94.04% simultaneously. A hybrid artificial neural network-genetic algorithm for global optimisation was employed. The optimised parameters 46.5nF capacitance, 110 V voltage, and 5 µm/sec feed yielded a minimum RLT of 1.58 µm and a maximum MRR of 0.3 mm³/min. These advancements signify significant improvements in the precision and efficiency of MEAs fabrication through reverse µEDM, contributing to its sustainable implementation.

A Comparative Study on the Estimation of Wind Speed and Wind Power Density Using Statistical Distribution Approaches and Artificial Neural Network-Based Hybrid Techniques in Çanakkale, Türkiye

In recent years, wind energy has become remarkably popular among renewable energy sources due to its low installation costs and easy maintenance. Having high energy potential is of great importance in the selection of regions where wind energy investments will be made. In this study, the wind power potential in Çanakkale Province, located in the northwest of Türkiye, is examined, and the wind speed is estimated using hourly and daily data over a one-year period. The data, including 12 different meteorological parameters, were taken from the Turkish State Meteorological Service. The two-parameter Weibull and Rayleigh distributions, which are the most widely preferred models in wind energy studies, are employed to estimate the wind power potential using hourly wind speed data. The graphical method is implemented to calculate the shape (k) and scale (c) parameters of the Weibull distribution function. Daily average wind speed estimation is performed with artificial neural network–genetic algorithm (ANN-GA) and ANN–particle swarm optimization (ANN-PSO) hybrid approaches. The proposed hybrid ANN-GA and ANN-PSO algorithms provide correlation coefficient values of 0.94839 and 0.94042, respectively, indicating that the predicted and measured wind speed values are notably close. Statistical error indices reveal that the ANN-GA model outperforms the ANN-PSO model.

Towards accuracy improvement in solution of inverse kinematic problem in redundant robot: A comparative analysis

AbstractThe redundant manipulators have more DOFs (degree of freedoms) than it requires to perform specified task. The inverse kinematic (IK) of such robots are complex and high nonlinear with multiple solutions and singularities. As such, modern Artificial Intelligence (AI) techniques have been used to address these problems. This study proposed two AI techniques based on Neural Network Genetic Algorithm (NNGA) and Particle Swarm Optimization (PSO) algorithm to solve the inverse kinematics (IK) problem of 3DOF redundant robot arm. Firstly, the forward kinematics for 3 DOF redundant manipulator has been established. Secondly, the proposed schemes based on NNGA and PSO algorithm have been presented and discussed for solving the inverse kinematics of the suggested robot. Thirdly, numerical simulations have been implemented to verify the effectiveness of the proposed methods. Three scenarios based on triangle, circular, and sine-wave trajectories have been used to evaluate the performances of the proposed techniques in terms of accuracy measure. A comparison study in performance has been conducted and the simulated results showed that the PSO algorithm gives 7% improvement compared to NNGA technique for triangle trajectory, while 2% improvement has been achieved by the PSO algorithm for circular and sine-wave trajectories.

Comparison of an Ultrasound-Assisted Aqueous Two-Phase System Extraction of Anthocyanins from Pomegranate Pomaces by Utilizing the Artificial Neural Network-Genetic Algorithm and Response Surface Methodology Models.

As a by-product of pomegranate processing, the recycling and reuse of pomegranate pomaces (PPs) were crucial to environmentally sustainable development. Ultrasound-assisted aqueous two-phase extraction (UA-ATPE) was applied to extract the anthocyanins (ACNs) from PPs in this study, and the central composite design response surface methodology (CCD-RSM) and artificial neural network-genetic algorithm (ANN-GA) models were utilized to optimize the extraction parameters and achieve the best yield. The results indicated that the ANN-GA model built for the ACN yield had a greater degree of fit and accuracy than the RSM model. The ideal model process parameters were optimized to have a liquid-solid ratio of 49.0 mL/g, an ethanol concentration of 28 g/100 g, an ultrasonic time of 27 min, and an ultrasonic power of 330 W, with a maximum value of 86.98% for the anticipated ACN yield. The experimental maximum value was 87.82%, which was within the 95% confidence interval. A total of six ACNs from PPs were identified by utilizing UHPLC-ESI-HRMS/MS, with the maximum content of cyanidin-3-O-glucoside being 57.01 ± 1.36 mg/g DW. Therefore, this study has positive significance for exploring the potential value of more by-products and obtaining good ecological and economic benefits in the future.

Hybrid genetic algorithms in agent-based artificial market model for simulating fan tokens trading

In recent years cryptographic tokens have gained popularity as they can be used as a form of emerging alternative financing and as a means of building platforms. The token markets innovate quickly through technology and decentralization, and they are constantly changing, and they have a high risk. Negotiation strategies must therefore be suited to these new circumstances. The genetic algorithm offers a very appropriate approach to resolving these complex issues. However, very little is known about genetic algorithm methods in cryptographic tokens. Accordingly, this paper presents a case study of the simulation of Fan Tokens trading by implementing selected best trading rule sets by a genetic algorithm that simulates a negotiation system through the Monte Carlo method. We have applied Adaptive Boosting and Genetic Algorithms, Deep Learning Neural Network-Genetic Algorithms, Adaptive Genetic Algorithms with Fuzzy Logic, and Quantum Genetic Algorithm techniques. The period selected is from December 1, 2021 to August 25, 2022, and we have used data from the Fan Tokens of Paris Saint-Germain, Manchester City, and Barcelona, leaders in the market. Our results conclude that the Hybrid and Quantum Genetic algorithm display a good execution during the training and testing period. Our study has a major impact on the current decentralized markets and future business opportunities.

Retracted: Study on College English Online Teaching Model in Mixed Context Based on Genetic Algorithm and Neural Network Algorithm

Retracted: Study on College English Online Teaching Model in Mixed Context Based on Genetic Algorithm and Neural Network Algorithm

Analyzing Temperature Distribution, Mass Transport, and Cell Performance in PEM Fuel Cells with Emphasis on GDL Face Permeability and Thermal Contact Resistance Parameters.

Temperature distribution, mass transport, and current density are crucial parameters to characterize the durability and output performance of proton exchange membrane fuel cell (PEMFC), which are affected by thermal contact resistance (TCR) and gas diffusion layer (GDL) face permeability within both cathode and anode GDL porous jumps. This study examined the effects of TCR and GDL face permeability on a single PEM fuel cell's temperature profiles, mass transport, and cell performance using a three-dimensional, nonisothermal computational model with an isotropic gas diffusion layer (GDL). This model calculates the ideal thermal contact resistance by comparing the expected plate-cathode electrode temperature difference to the numerical and experimental literature. The combined artificial neural network-genetic algorithm (ANN-GA) method is also applied to identify the optimum powers and their operating conditions in six cases. Theoretical findings demonstrate that TCR and suitable GDL face permeability must be considered to optimize the temperature distribution and cell efficiency. TCR and GDL face permeability lead to a 1.5 °C rise in maximum cell temperature at 0.4 V, with a "Λ" shape in temperature profiles. The TCR and GDL face permeability also significantly impacts electrode heat and mass transfer. Case 6 had 1.91, 6.58, and 8.72% higher velocity magnitudes, oxygen mass fractions, and cell performances than case 1, respectively. Besides, the combined ANN-GA method is suitable for predicting fuel cell performance and identifying operation parameters for optimum powers. Therefore, the findings can improve PEM fuel cell performance and give a reference for LT-PEMFC design.

Impact of extrusion processing on bioactive compound enriched plant-based extrudates: A comprehensive study and optimization using RSM and ANN-GA

The present study aimed to investigate the influence of extrusion process parameters, specifically screw speed (300 to 500 rpm) and temperature (130 to 150 °C) on the stability of phytosterols and the physical, functional, and structural characteristics of extrudates made from composite blends containing pea protein isolate (PPI). Response Surface Methodology (RSM) and Artificial Neural Network-Genetic Algorithm (ANN-GA) were employed to optimize the process parameters.Extrudates produced at lower temperatures exhibited higher phytosterol retention compared to those produced at higher temperatures. X-ray diffraction analysis showed that incorporating PPI significantly increased crystallinity from 22% to 33%. Fourier Transform Infrared analysis revealed a reduction in the Amide II stretching of the secondary structure in pea protein isolate following extrusion. Scanning electron microscopy images demonstrated that the addition of PPI and functional oil altered the structure of the extrudates.Statistical comparison of the process parameters using higher R2 values, lower root mean square as well as mean absolute percentage error indicated that the ANN model outperformed the RSM model in predicting all responses. Overall, incorporating PPI and phytosterols into corn-based extrudates improved their nutritional quality. These findings contribute to meeting the demand for fortified foods.

Artificial intelligence model for prediction of cardiovascular disease: An empirical study

Cardiovascular disease (CVD) is a disease related to the heart and blood vessels. Prediction of CVD is essential for early detection and diagnosis, which is however compounded by the complex interplay between medical history, physical examination outcomes, and imaging results. While the existing automated systems are fraught with the usage of irrelevant and redundant attributes, artificial intelligence (AI) helps in the identification of potential CVD populations by prediction models. This work aims at developing an AI model for predicting CVD using different classifications of machine learning techniques. The CVD dataset was obtained from the UCI repository containing about 76 cardiac attributes for training in various machine learning models, which include a hybrid of artificial neural network-genetic algorithm (ANN-GA), artificial neural network, support vector machine (SVM), K-means, K-nearest neighbor (KNN), and decision tree (DT). The performance of the models was measured in terms of accuracy, means square error, sensitivity, specificity, and precision. The results showed that the hybrid model of ANN-GA performs better with an accuracy of 86.4%, compared to the SVM, K-means, KNN, and DT measured at 84.0%, 59.6%, 79.0%, and 77.8%, respectively. It was observed that the system performs better as the number of datasets increases in the database, with a fewer selection of attributes using genetic algorithm for selection. Thus, the ANN-GA model is recommended for CVD prediction and diagnosis.

Artificial Neural Network and Genetic Algorithm-based prediction of photovoltaic panel performance with porous foam gradient and nano-enhanced phase change material

Photovoltaic panels (PV) become less efficient as the temperature rises. The use of porous foam gradient, phase change materials (PCMs), and nanoparticle-PCM (nePCM) can effectively lower the temperature of PV panels. This study employed numerical analysis using the finite volume method to investigate the effects of different types of PCMs (n-octadecane, decanoic acid natural, lauric acid, paraffin wax, and RT-42), mono and hybrid nanoparticles/PCM, and the position of PCM in the design system. The results indicate that the highest electrical efficiency of 9 % is achieved when the PCM is placed under the PV panel and the thickness of the PCM layer is 0.6 times the height of the microchannel. Furthermore, the use of a porous medium with a variable porosity coefficient that increases along the y-direction can significantly improve thermal and electrical efficiency (approximately 5–42 % and 5–48 % respectively). Additionally, incorporating PCM and hybrid nanoparticles-PCM enhances PV performance compared to the absence of nanoparticles. For TiO2, Ag, Al2O3, GO, and MWCNT nanoparticles, the thermal efficiency, electrical efficiency, and exergetic efficiency increase by approximately 24.18 %, 26.15 %, 26.28 %, 26.30 %, and 26.33 % respectively. Similarly, for hybrid nanoparticles such as GO-Al2O3, GO-Al2O3-MWCNT, MWCNT-GO, and MWCNT-Al2O3, these values increase by approximately 26.32 %, 26.3419 %, 26.3432 %, and 26.335 % respectively for thermal efficiency, and 5.16 %, 5.1702 %, 5.1704 %, and 5.1689 % respectively for electrical efficiency. Moreover, n-octadecane exhibits the highest melting fraction, while paraffin wax has the lowest melting fraction. Lastly, the genetic algorithm-neural network method was employed to estimate the PV electrical efficiency.

Response surface methodology and artificial neural network-genetic algorithm for modeling and optimization of bioenergy production from biochar-improved anaerobic digestion

Biochar can be used to improve the anaerobic digestion (AD) of agricultural wastes for higher methane production. However, the interaction of biochar addition with other factors of the anaerobic co-digestion (Co-AD) process has rarely been investigated. In this study, process models based on response surface methodology (RSM) and artificial neural network (ANN) were compared in modeling the methane yield (MY, mL CH4/g VS added) from the Co-AD of poultry litter and wheat straw with biochar addition. Box-Behnken design was applied, with the controlling parameters being carbon to nitrogen ratio (C/N), total solids (TS, %), and biochar addition (Biochar, % TS). Numerical optimization and genetic algorithm (GA) were used as optimization tools for RSM and ANN, respectively. A significant second-order quadratic model was built by RSM (R2 = 0.9981 and RMSE = 0.91), which demonstrated significant interactions between C/N and TS (p < 0.0001), and between C/N and Biochar (p < 0.05). The trained ANN (3−3−1) was less accurate (R2 = 0.9926, RMSE = 1.80) compared to RSM. The optimization results by RSM and ANN coupled with GA (ANN-GA) were both validated with prediction errors <0.5%. The optimization results by ANN-GA should be used since it generated a higher maximum MY of 290.7 ± 0.2 mL CH4/g VS added, under the optimal conditions of C/N ratio 24.46, TS 5.03%, and Biochar 8.73% TS, showing an improvement of 20.6% (compared to the control) through process optimization. The methods can also be applied in other scenarios for process modeling and optimization. The optimized results could support real applications of using additives including biochar, active carbon, nanoparticles, etc., to promote the bioenergy production from AD of agricultural wastes.

Comparative transient assessment and optimization of battery and hydrogen energy storage systems for near-zero energy buildings

This research study aims to address the integration of solar energy systems that can be incorporated into buildings to provide zero-energy buildings. In this study, a typical building is considered located in Bandar Abbas city, Iran. Since renewable energy is not always available, energy storage becomes crucial to supplying the energy needed. The primary objective of this case study is to compare two different methods of energy storage: battery and hydrogen storage. TRNSYS, a powerful program for transient simulation is used to model the energy system. TRNSYS's energy system simulation provides consumers with several notable benefits, but it lacks optimization methods. A neural network-genetic algorithm optimization is suggested as a solution to solve this issue. The results indicate that in systems with battery and H2 storage, 39 % and 37 % of the buildings' required electricity can be generated by PV panels, respectively. The findings show that integrating a storage system into the primary system enhances the system's energy security and reduces reliance on grid power. Following that, the optimal value related to the power of the electrolyzer, and fuel cell is computed using a neural network-genetic algorithm optimization approach. The optimal configuration has the least loss of power supply probability (LPSP), CO2 emission, and installation costs. MATLAB software is used to perform the optimization process. According to evaluated data, the hydrogen storage system results in total CO2 generation, LPSP, and total cost of 342 (ton/year), 0.841, and 0.637 (EUR.hr−1), respectively. Also, a parametric study on the battery-based system showed that with the best capacity (7 kW), the values of LPSP, CO2, and cost are 0.5412, 249 (ton/year), and 0.2751 (EUR.hr−1), respectively. Consequently, the battery is an appropriate storage solution for this case study.

Process optimization of microwave-assisted aqueous extraction of tannins and saponins from Malabar Spinach (Basella alba) leaves using ANN-GA and RSM methodology

The effect of microwave-assisted aqueous extraction (MAE) conditions on the recovery and quantification of tannins and saponins from Basella alba leaves has been explored. The experiment was carried out according to the Box-Behnken experimental design of response surface methodology (RSM) with 17 runs. The independent variables of the study were microwave power (100 to 300 W), treatment time (5 to 15 min), and feed/solvent ratio (0.025 to 0.05). The effect of extraction conditions on the response variables, total condensed tannin, total hydrolyzed tannin, and total saponin content, were predicted using RSM and artificial neural network-genetic algorithm (ANN-GA) methodology. The extracted values of total condensed tannin, total hydrolyzed tannin, and total saponin content, using RSM and ANN-GA were compared and evaluated for goodness of fit. The predicted values using RSM and ANN-GA models for total condensed tannin were found to be 0.18 and 0.26 mg TAE g−1, total hydrolyzed tannin were 0.14 and 0.15 mg TAE g−1, and total saponin content were 1.66 and 1.49 mg DE g−1, respectively. When compared to RSM, ANN-GA demonstrated higher value of R2 and lower values of other statistical parameters. Additionally, the predicted value of ANN-GA was more closely aligned with the experimental value.

Tertiary treatment using ultrafiltration in an existing sewage treatment plant for industrial reuse: a modelling approach using artificial neural network with uncertainty estimation

Abstract Navi Mumbai Municipal Corporation of Maharashtra state, India, unified a tertiary treatment plant (TTP) of 20 × 103 m3/day capacity with ultrafiltration technology in an existing Koparkhairane sewage treatment plant (STP) for producing effluent quality usable for industrial purposes. As prior art, an artificial neural network-genetic algorithm (ANN-GA) along with uncertainty estimation using prediction interval is employed to model secondary treated effluent (STE) flow rate (QT) and other quality parameters such as biochemical oxygen demand, chemical oxygen demand, and total suspended solids (TSS) to conclude the reliability of the range in which the input available to TTP. ANN-GA model provides a coefficient of determination above 0.90 for all STE parameters modelled other than TSS. Inferring that a good quantity and quality of 20 × 103 m3/day STP treated water is currently available, where a decreasing trend of QT is also noticed and highlighted. Further, the Wilcoxon signed-rank test on the quality parameter of effluent TTP for industrial reuse standard infers TSS shows infringement during the initial period but started adhering to standards over time. The research delineates at the outset of exploring water reuse policy in India, emphasizing Maharashtra state, modelling STE using ANN-GA and performance evaluation of TTP.

A numerical-experimental approach to simulate the sintering behavior of “cer-cer” fuels by NITE process

The Nano-powder Infiltration and Transient Eutectic (NITE) process is a promising method for fabricating Fully ceramic microencapsulated (FCM) fuels at relatively low temperatures. However, when fabricating large-scale samples with high volume fractions of fuel particles, low-density or defective fuels are sometimes obtained. The same issue arises when it comes to ceramic particle-ceramic matrix fuel. To address this issue, it is crucial to properly control the relative density and internal stress of these fuels during the sintering process. This study presents a model that uses a unified approach to describe compaction and sintering simultaneously, and the model is successfully adjusted by experimental data using neural network genetic algorithms. The findings provide valuable insights into the fabrication of FCM fuels and the ceramic particle-ceramic matrix fuels. The proposed method can guide future research and development efforts in this area.

Rapid screening, isolation, and activity evaluation of potential xanthine oxidase inhibitors in Polyporus umbellatus and mechanism of action in the treatment of gout.

Studies show that Polyporus umbellatus has some pharmacological effects in enhancing immunity and against gout. We aimed to establish new techniques for extraction, biological activity screening, and preparation of xanthine oxidase inhibitors (XODIs) from P.umbellatus. First, the extraction of P.umbellatus was investigated using the back propagation (BP) neural network genetic algorithm mathematical regression model, and the extraction variables were optimised to maximise P.umbellatus yield. Second, XODIs were rapidly screened using ultrafiltration, and the change of XOD activity was tested by enzymatic reaction kinetics experiment to reflect the inhibitory effect of active compounds on XOD. Meanwhile, the potential anti-gout effects of the obtained active substances were verified using molecular docking, molecular dynamics simulations, and network pharmacology analysis. Finally, with activity screening as guide, a high-speed countercurrent chromatography (HSCCC) method combined with consecutive injection and two-phase solvent system preparation using the UNIFAC mathematical model was successfully developed for separation and purification of XODIs, and the XODIs were identified using MS and NMR. The results verified that polyporusterone A, polyporusterone B, ergosta-4,6,8(14),22-tetraen-3-one, and ergosta-7,22-dien-3-one of P.umbellatus exhibited high biological affinity towards XOD. Their structures have been further identified by NMR, indicating that the method is effective and applicable for rapid screening and identification of XODIs. This study provides new ideas for the search for natural XODIs active ingredients, and the study provide valuable support for the further development of functional foods with potential therapeutic benefits.

Prediction of compressive strength and feature importance analysis of solid waste alkali-activated cementitious materials based on machine learning

This study examines the analysis and treatment of diverse types of solid waste alkaline-activated cementitious materials utilizing machine learning techniques. To account for the variability of solid waste alkaline-activated cementitious materials, three crucial chemical components (aluminum trioxide, calcium oxide, and silicon dioxide) were chosen as representative features. A total of 274 pertinent datasets were collected and split into training and testing sets with an 80% to 20% ratio for analysis. Six machine learning models, including multilayer feedforward neural network, genetic algorithm neural network, support vector machine, random forest, radial basis function neural network, and long short-term memory network, were employed to construct predictive models. The optimal hyperparameters were identified via grid search. The performance of the models was assessed using the training and testing sets, revealing that all models exhibited strong predictive and generalization capabilities on both sets. Among the models, the support vector machine model achieved the highest performance, yielding an R2 value of 0.9054, MAE of 4.1460, and NRMSE of 0.0997. Through the utilization of SHAP (Shapley Additive Explanations) analysis, the interplay between factors and the significance of features were examined. Calcium oxide, water-to-binder ratio, silicon dioxide, modulus of water glass, and aluminum trioxide were recognized as the primary factors that influence the compressive strength. The research findings offer guidance for optimizing the performance of alkaline-activated cementitious materials made from solid waste.