Significant efforts have been made to increase the strength of concrete by using industrial waste such as fly ash and steel slag as partial substitutes for concrete in concrete. However, predicting the concrete’s compressive strength is a challenge as it is influenced by several factors such as the shape and size of the aggregate, the water-ratio balance. This study examines the predictive capability of three deep learning models: Bagging Extreme Gradient Boosted Model (BXGBM), Deep Random Vector Functional Link (DRVFL), and Kernel Extreme Learning Machine (KELM) on the prediction for compressive strength of concrete. The dataset was split into a training and testing set, and the performance measures were analyzed. The statistical metrics include Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and coefficient of determination (R2). According to model BXGBM, the MSE was found to be 3.757, and the R2 of 0.9864 was as in testing, with MSE of 16.63, R2 = 0.941 performed well with good accuracy. The model DRFVL has a training MSE of 29.858, R2 = 0.8922, and has low overall generalizability of 46.030 and 0.8391 on the testing set. KELM also did well with a training MSE of 13.851, R2 of 0.950; testing performance declined with an MSE of 31.05 and R2 of 0.891. The results show that BXGBM is most trustworthy as a model that predicts the compressive strength of concrete, which allows emphasizing its high potential in applying to the practical sphere of concrete technology.

Given the cost, time requirements, and limited accessibility of locked wheel skid trailer (LWST) testing, this research explores alternative approaches to predicting skid number (SN) using more affordable and readily available devices. These alternatives offer quicker results, which is particularly relevant given the standardized nature of SN measurements worldwide. The study focuses on predicting the SN generated by the LWST test through analyzing data from dynamic friction tester (DFT), British pendulum tester (BPT), and circular texture meter (CTM) using statistical and machine learning (ML) techniques. The research includes a thorough assessment of various prediction methods and factors influencing SN accuracy, including strategies for handling missing data. Prediction models employ multiple linear regression (MLR), support vector machine (SVM), artificial neural network (ANN), and ensemble bagged tree (EBT) techniques, utilizing predictors such as dynamic friction number at 20 km/h (DFT20), dynamic friction number at 64 km/h (DFT64), British pendulum number (BPN), and mean profile depth (MPD). The development of forty-eight models, including full, reduced, and individual predictor models, involved employing data imputation methods and stepwise regression. Significant correlations were observed among friction parameters, with DFT20 showing the highest correlation with SN, while MPD exhibited the lowest. Following data imputation, notable enhancements in BPN's correlation were noted. The study emphasizes the superiority of full models in SN prediction accuracy, with EBT models, especially those enhanced by data imputation, demonstrating outstanding performance. Statistical simulations confirmed the reliability of these models, indicating minimal outliers, overfitting, and high accuracy in SN estimation. The prioritization of DFT over BPT and CTM is highlighted, with the potential for further exploration of devices to improve prediction reliability. (c) 2026 Tongji University and Tongji University Press. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

This research introduces a novel approach by combining Long Short-Term Memory (LSTM) networks with Response Surface Methodology (RSM) to optimize and predict the performance of a plate heat exchanger for industrial applications. While both techniques have been individually applied in heat exchanger studies, this work demonstrates their integration to enhance performance prediction and optimization. Experimental data were collected across variable hot fluid flow rates (0.5–1.4 l/min), cold fluid flow rates (1–3 l/min), and feed temperatures (50–60 °C). The optimal heat exchanger effectiveness was found to be 57.41

Photovoltaic (Solar) panels, i.e., PV cells, are the most effective and economical renewable energy source generator. However, their power generation efficiency depends heavily on the cleanliness of their surfaces. The dust, snow, and bird drop cover the panel surface, reducing energy generation capability. Additionally, extreme weather causes electrical and physical damage to panels. This paper presents an approach for identifying these anomalies in aerial images using an advanced convolutional neural network (CNN). The proposed network uses the Swin Transformer (ST) to address the critical challenge of the traditional CNN’s interpretability. Multiple features were extracted using a shifted-patch-based multi-head self-attention at various stages, and these features were classified into six anomaly classes. The results show that the model succeeded with 92.53% accuracy and 92.64% F1-score. Thus, the proposed approach provides a practical solution for analysing renewable energy infrastructure from an aerial perspective.

Specific durability problems are often due to inadequate drying, resulting in poor hydration, increased permeability and reduced durability. This study examines the effects of two field-representative curing methods on the mechanical properties and durability of concrete grade G25, G30 and G35 ready mixed. Site Curing 1 (SC1), exposed to ambient conditions and Site Curing 2 (SC2) covered with plastic sheets for the first seven days. Significant durability metrics, including compressive strength, water absorption, volume of permeable voids (VPV), and sorptivity, were evaluated experimentally at 7, 28, and 56 days in different specimen zones. According to the results, the compressive strength of the concrete specimen, SC2 found to be improved values at 56 days. Grade G25 increased from 29.76 MPa (SC1) to 30.37 MPa, G30 from 38.88 MPa to 39.71 MPa, and G35 from 40.26 to 40.74 MPa. Water absorption decreased under SC2, G25 7–6.45

Automobile wash wastewater contains sand and dust, free oil, grease, oil/water emulsion, carbon, asphalt, salts and volatile organic compounds. A bubble column reactor was employed to treat the wastewater generated from automobile wash. The performance of the reactor was evaluated through various parameters, pH, electrical conductivity, turbidity, total suspended solids, chemical oxygen demand, biological oxygen demand and total petroleum hydrocarbonate. Returned activated sludge was used to provide the microbial seeding in the reactor. It was observed a 94

This study investigates the optimization and performance modeling of a spray flash desalination system, integrating experimental design with machine learning techniques, including Random Forest (RF), Gradient Boosting (GB), and Decision Tree (DT) models. The experimental analysis varied four key parameters namely, vacuuming pressure (20–60 kPa), salinity (20,000–40,000 ppm), feeding temperature (25–45 °C), and flow rate (0.3–0.5 L·min− 1) to optimize distillate production. The results showed an optimal distillate output of 7,585 mL·h⁻¹, with a specific energy consumption (SEC) of 125.4 kWh·m⁻³ and a gain output ratio (GOR) of 11.6. Machine learning models demonstrated strong prediction accuracy, with RF achieving a root mean squared error (RMSE) of 545 mL·h⁻¹ and R² of 0.93. The interpretability using SHAP analysis shows that feed temperature has a significant impact on prediction. This research presents a novel hybrid approach combining response surface methodology and machine learning to optimize desalination systems, providing a robust framework for improving energy efficiency and performance in future large-scale desalination applications. Optimized a spray flash desalination system using experimental design and machine learning. Achieved an optimal distillate production of 7,585 ml/h with a specific energy consumption of 125.4 kWh/m³. Developed a hybrid approach combining Response Surface Methodology (RSM) and machine learning techniques (RF, GB, DT). Random Forest model showed strong prediction accuracy with an R² of 0.93 and RMSE of 545 ml/h.

This study investigates the photocatalytic degradation of oil-produced water (OPW) using solar irradiation, the application of response surface methodology, and deep learning models. A total of 70 experimental batch reactor trials were conducted by varying pH from 6 to 9, Titanium dioxide (TiO2) catalyst dosage from 1 to 4 g/L, and reaction time from 180 min, along with chemical oxygen demand (COD) removal efficiency. Using the RSM, the effects of inputs on COD were quantified to determine the optimal conditions. Subsequently, Support Vector Machine (SVM), Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), and Extreme Learning Machine (ELM) were employed to further predict COD based on the RSM study, and the results were evaluated. According to RSM, the optimum value for COD removal efficiency was 95

Due to industrialization, urbanization, and the need for sustainable water resources, the treatment of organic water has become a major environmental concern. Semiconductor photocatalysts degrade organic contaminants through a clean, energy-efficient process called photocatalysis. In this study, batch reactor experiments were performed under natural solar irradiation to examine the influence of zinc oxide (ZnO) dosage, pH, and reaction time (inputs) on photocatalytic performance, with total organic carbon (TOC) as the output. Three machine learning algorithms, namely Artificial Rabbit Optimization–Support Vector Regression (ARO–SVR), Support Vector Regression (SVR), and AdaBoost, were proposed and compared. Model performance was assessed by the coefficient of determination (R2), root mean squared error (RMSE), mean absolute error (MAE), and mean squared error (MSE). The proposed ARO–SVR model outperformed the other models, with training and testing R2 values of 0.9782 and 0.9731, RMSE values of 0.040 and 0.0453, MAE values of 0.0402 and 0.046, and MSE values of 0.0016 and 0.0020. By contrast, conventional SVR and AdaBoost showed lower prediction accuracy and larger testing errors, indicating a lower generalization ability. The results show that combining photocatalytic technology in seawater treatment with an optimized machine learning framework can reliably and efficiently predict seawater treatment performance and minimize the need for extensive lab experiments.

The effective design and operation of submerged multi-jet electrolyte flow systems is necessary for a variety of industrial applications, including electrochemical machining and electroplating. The optimization and prediction of pressure drop in such systems utilizing Response Surface Methodology (RSM) in conjunction along with artificial neural network is thoroughly examined in this study. A prototype submerged multi-jet electrolyte flow system is used to gather experimental data under various operating conditions. Then, using critical input factors, ANN were employed to build prediction models for pressure drops. ANN was used to forecast the pressure drop at several heights between 0.01 and 0.25 m, the optimum pressure drops, 1476.2 N/m2, was recorded at 0.25 m. Additionally, the Response Surface Methodology is used to create empirical models that shows the process parameters and pressure drop are related. The recommended methodology offers a systematic and efficient approach to predict and optimize pressure drop in submerged multi-jet electrolyte flow systems. In industrial applications, this enables enhanced process performance and efficiency.

The availability of dependable electricity in the world market is an issue of concern, especially in the rural areas where much of the rural population does not have access to basic power. There are some renewable energy technologies that provide alternative power to the traditional fossil-based system, and ocean wave energy is an opportunity that is not fully exploited yet. This paper presents and experimentally tests two simple wave energy conversion systems, which can be used to generate small scale electricity. The previous system involves a micro-hydro generator system powered by pressurized water flowing into it whereas the latter system involves a raft-powered wave power generator that transforms the movement of the waves into electricity. The two prototypes were planned, made and experimented under controlled laboratory conditions to test their electrical performance. According to the experimental findings, the micro-hydro generator was able to generate the highest voltage of up to 11.9 V at a pressure situation of 12 bar, and with the approximation current output of 20 mA, which equates to a power output of 0.1–0.2 W. The suggested devices depict a streamlined structure whereby the fabrication cost of the product is about 33.81 and 14.35 respectively. The findings emphasize the potential of creating affordable wave energy harvesting instruments that can be used in political small-scale applications and educative displays. The originality of this research work is the creation and comparative analysis of two low-tech and economically feasible wave energy prototype designs which ultimately strive to simplify structural designs and still retain functional energy conversion potential.

In this comprehensive review, the emerging field of polymer photocatalysts is examined, including the problems and prospects that may arise in the development and use of polymer photocatalysts to provide solutions for environmental remediation and green energy. The review explains the special characteristics of the polymeric materials their tunability, easy modulation, and their synergy with inorganic semiconductors. The most important issues include stability, photocatalytic efficiency, and mechanisms underlying the reaction pathways are reviewed critically. Moreover, the study will draw attention to the recent developments in the fabrication method and the application of nanotechnology in the development of photocatalytic performance. The vision of future directions and the development of biodegradable and bio-based photocatalytic materials in this review will be beneficial in helping researchers and practitioners in addressing the existing shortcomings and use different strategies to exploit the potential of polymeric photocatalysis to a more sustainable future.

ABSTRACT Water contamination caused by organic pollutants leads to severe deterioration of water quality and major environmental concern worldwide. The implementation of effective monitoring and remediation measures is necessary for sustainable water treatment. A batch reactor using titanium oxide (TiO2) as a catalyst to remove total organic carbon (TOC) and chemical oxygen demand (COD) in seawater by solar photocatalysis was used. Response surface methodology was carried out with input parameters such as TiO2 dosage 1–4 g/L, pH value 6–9 and reaction time 60–300 min and percentage reduction of TOC and COD as output variables. Adaptive Boosting (AdaBoost), Gradient Boosting (GB), Bayesian Ridge Regression (BRR) and Long Short-Term Memory networks (LSTM) were used for prediction. The accuracy of models is determined by metric analysis. GB model recorded the highest R2=0.993 (COD) and 0.980 (TOC) with minimal error MSE of 3.43 COD and 0.009 TOC. Whereas RSM showed R2=0.953, 0.950 for COD and TOC. AdaBoost demonstrated R2 of 0.973 of COD, in addition BRR and LSTM showed lower prediction accuracy. The findings shed light on the possibility of machine learning methods in improving prediction of water quality parameters and optimizing of water treatment processes.

Radiators play an important role in automobiles as they directly affect the energy utilization and operational dependability. This paper presents an integrated experimental, response surface, and Machine Learning (ML) based techniques to investigate and predict the performance of a typical automobile radiator. Experimental data is collected by varying hot water flow rate (0.5–2.5 l/min), cold air velocity (1–5 m/s), and feeding temperature (50–60 °C). The operating conditions were optimized using Response Surface Methodology (RSM), and the optimal radiator effectiveness was determined to be 73.2

In this research, the Long Short-Term Memory (LSTM) technique coupled with response surface methodology (RSM) is used to optimize and predict the performance of a standard centrifugal pump for industrial applications. Model development is done through experimental data collection by varying aspiration pressure (9-95 kPa), discharge pressure (2-57 kPa), motor speed (2200-2600 rpm), and torque (0.31-0.84 Nm). The optimal pump efficiency was found to be 62.57 %. It is observed that the RSM model is capable of predicting the pump efficiency with an R2 of 0.999 in comparison to the LSTM techniques, which exhibited an R2 of 0.995. The Mean Absolute Error (MAE) is 5.179, the Mean Squared Error (MSE) is 64.38, while the Root Mean Squared Error (RMSE) is found to be 8.02 for the LSTM technique. Likewise, the MAE, MSE, and RMSE were found to be 0.23, 0.08, and 0.29, respectively, for the RSM model. The results of this study highlight that the pump efficiency might be effectively optimized and predicted with high accuracy by utilizing RSM along with the LSTM approach. This study presents a crucial data-driven strategy for optimization and performance prediction of a typical centrifugal pump. The proposed framework not only provides accurate performance predictions but also offers practical guidance for improving pump efficiency, reducing energy consumption, and supporting scalable implementation in industrial pumping systems.

In recent years, the sand crisis has garnered increasing attention, and researchers have investigated a variety of substitutes to mitigate the consumption of river sand. Common alternatives are industrial waste items that would otherwise need to be disposed of on land. For concrete, one such by-product that can be utilized in place of alluvial sand is steel slag (SS). The purpose of this study is to ascertain the ideal replacement percentages for the required mechanical and structural qualities by investigating the viability of using steel slag as a fine aggregate in concrete. Experimental studies were carried out using different ratios of steel slag in place of traditional fine aggregate. Compressive strength, workability, and other characteristics of the concrete mixes were evaluated. The experimental findings show that using steel slag as a partial substitute for fine aggregate has encouraging potential. The steel slag with 15

Water bodies are getting highly polluted by heavy metals, particularly copper, which is a severe issue of concern for the environment that requires an effective and sustainable method of remediation. Though traditional machine learning models performed well at predicting the percentage removal of contaminants, their main weakness is not accounting for the temporal lag inherent in chemical adsorption and filtration. A hybrid framework of Gated Recurrent Unit and Convolutional Neural Network to predict the percentage removal of copper were used. CNN offers automated feature extraction, and GRU processes them as a sequence using its gating mechanism, retaining previous time-step information in the current prediction. The model is trained with 80