Engineering Access

Integrated Smart Geotechnical Monitoring Using Ensemble Transfer Learning and Multi-Agent Systems

Prashant Pande, Jayant Raut, Rajesh Bhagat, Boskey Bohoriya, Amol Tatode — Wed, 01 Jul 2026 00:00:00 +0700

The growing complexities in modern infrastructure projects necessitate advanced geotechnical monitoring systems to ensure sustainability and safety. Traditional techniques often fall short due to their inability to provide real-time analysis, adapt to new data patterns, and prevent network failures, leading to delayed detection of soil-related issues. This research proposes an integrated smart monitoring system using IoT and AI technologies, featuring three key methodologies: Ensemble Transfer Learning (ETL), Autonomous Multi-Agent Systems (AMAS), and Long Short-Term Memory Networks with Attention Mechanisms (LSTM-AM). ETL improves prediction accuracy by 20% and reduces false positives by 15%. AMAS minimizes network downtime by 40% and data loss by 30%. LSTM-AM increases failure prediction accuracy by 25% and reduces unexpected failures by 30%. Overall, the proposed system enhances monitoring accuracy by 25%, resilience by 35%, and prevention rate by 20%, offering a robust solution for geotechnical challenges in urban infrastructure.

Experimental Analysis of Self Compacting Concrete using Bagasse Ash, Metakaolin and Glass Fiber

Monali Wagh, U. P. Waghe, Anshul Nikhade, Harshal Nikhade — Wed, 01 Jul 2026 00:00:00 +0700

Self-compacting concrete (SCC) is a concrete that has the capacity to flow under its own weight without segregation. SCC has the ability to pass through the closed reinforcement, fill all the spaces of the formwork without segregation. In India, industrial and agricultural waste are produced in large quantities. Use of these wastes as a replacement for cement or sand gives a solution to the problem of disposal of these wastes. It decreases the cost of the structure as well as CO₂ emissions. This project is aimed at identifying the optimum amount of agricultural waste bagasse ash and metakaolin as replacements for cement with and without 0.1% GF and compared the fresh and strength properties of self-compacting concrete with the control mix. The result shows that the SCCB10M5 mix has better flowability as compared to others, and the strength properties were highest for the SCC1B10M5 mix.

Optimization of Hybrid 1+ Proportional-Integral and Proportional-Integral-Derivative Controller for Load Frequency Regulation in Nonlinear Two-Area Power Systems Using Catch Fish Techniques

Wed, 01 Jul 2026 00:00:00 +0700

This investigation investigates the evolution of a hybrid frequency controller that is specifically engineered for nonlinear two-area power systems. The controller ensures reliable frequency regulation by incorporating the most efficient components of both 1+PI and PID control methodologies, particularly when faced with complex issues such as GRC and GDB. In order to achieve the optimal calibration of the control parameters, the CFOA is implemented. This algorithm effectively traverses the search space and avoids suboptimal local optima by utilizing established fishing techniques. The performance criterion for the proposed method is ITAE, and simulations are conducted on a two-area power system model with 0.2 per unit load variations. The results suggest that the hybrid controller significantly reduces overshoot, undershoot, and settling time, while simultaneously guaranteeing a continuous tie-line power flow. The controller’s efficacy is extraordinary, with an ITAE of approximately 0.0336. These results emphasize its suitability for complex multi-area power networks and suggest potential applications in future systems, such as renewable energy or fluctuating demands.

Optimizing Firewall Log File Classification with Multilayer Perceptrons

Wed, 01 Jul 2026 00:00:00 +0700

Cyber threats remain a major concern for network security and data integrity, often causing serious consequences through unauthorized access, data breaches, and malware attacks. Intrusion detection systems are essential for countering these risks by continuously monitoring and classifying network activities. This study investigates the use of multilayer perceptrons (MLPs) to optimize firewall log file classification and improve the identification of network events. Four activation functions (Linear, Sigmoid, Tanh, and ReLU) were examined to determine the optimal configuration. The MLP model consistently achieved high performance, with accuracy exceeding 99%, and the ReLU activation function demonstrated superior effectiveness. The synthetic minority oversampling technique was applied to handle class imbalance in the firewall log dataset, improving the detection accuracy of the minority reset-both class. Moreover, feature selection using XGBoost reduced the input set from eleven to five key attributes, achieving 99.84% accuracy and improving computational efficiency. Experimental evaluations confirmed that the proposed model effectively recognizes complex nonlinear relationships in network data. These results demonstrate the potential of MLPs to enhance the accuracy, efficiency, and robustness of firewall log file classification for modern intrusion detection systems

Enhancing Sustainable Agriculture: Detection of Plant Leaf Diseases Using YOLO-Based Object Detection

Wed, 01 Jul 2026 00:00:00 +0700

This research explores the utilization of YOLO-based object detection for the early detection of plant leaf diseases, with the objective of promoting sustainable agricultural practices. This paper presents a comprehensive analysis of the YOLOv10s architecture, highlighting its advanced features designed to improve detection accuracy and computational efficiency. Our approach includes thorough data preparation, which entails splitting datasets into training, testing, and validation subsets, along with the implementation of hyperparameter optimization and pretraining-fine-tuning strategies. The findings indicate that YOLOv10s significantly surpasses earlier versions, achieving high detection accuracy while remaining viable for resource-constrained environments. Comparative analysis indicated that YOLOv9 outperformed in detecting healthy leaves (precision: 0.668, recall: 0.611) and leaf spot disease (precision: 0.632, recall: 0.626), whereas YOLOv10s exhibited a balanced performance in leaf blight detection. While YOLOv11 demonstrated incremental advancements, these were insufficient to justify its added complexity. This study highlights the transformative potential of deep learning technologies in agriculture, facilitating rapid disease detection and reducing reliance on chemical pesticides, thus supporting sustainable agricultural objectives.

Hybrid ANN-GA and RSM-GA Models for Optimizing GMAW Parameters to Enhance Impact Strength of Dissimilar Aluminum Alloys

Yodprem Pookamnerd, Thanatep Phatungthane, Chuthong Summatta — Wed, 01 Jul 2026 00:00:00 +0700

Optimizing welding parameters is essential for achieving superior mechanical performance, particularly in joining dissimilar aluminum alloys such as AA6061 and AA7075 using Gas Metal Arc Welding (GMAW). Unlike previous studies that primarily rely on single-model frameworks or conventional optimization techniques, this research uniquely integrates both ANN-GA and RSM-GA models to comprehensively evaluate and compare their performance on dissimilar aluminum alloy welds (AA6061–AA7075), offering a dual-model perspective rarely explored in the literature. This study employs a comprehensive approach integrating Artificial Neural Networks (ANN) and Genetic Algorithms (GA) to optimize parameters for enhanced impact strength. The Response Surface Methodology (RSM) with GA is also applied to analyze parameter interactions. Materials, including AA6061, AA7075, and ER5356 filler wire, were selected for their chemical compositions, balancing strength, corrosion resistance, and weldability. Using robotic GMAW ensured precision and repeatability, while Charpy impact testing evaluated joint toughness. The ANN-GA model identified optimal parameters—welding current of 143.08 A, speed of 2.61 mm/s, and wire feed rate of 3.86 m/min—achieving an impact energy of 20.00 J. Similarly, the RSM-GA framework yielded comparable results with parameters of 148.55 A, 2.52 mm/s, and 3.68 m/min. The comparative analysis highlighted the ANN-GA model's superior predictive accuracy (RMSE = 0.79815, MAPE = 3.7979), while RSM-GA provided insights into parameter trends. The outcomes provide a robust foundation for improving weld performance and optimizing parameters in industrial applications, demonstrating the practical value and adaptability of AI-driven welding models. The findings advocate integrating ANN and RSM models to enhance parameter optimization, offering practical applications in aerospace, automotive, and structural engineering industries.

A New Predictive Algorithm for Enhancing Regression Performance Using Insights from Random Vine Anchorage Mechanisms

Weerachai Jonburom, Charinee Chaichana — Wed, 01 Jul 2026 00:00:00 +0700

This study proposes a novel regression algorithm termed the Random Vine Anchorage Mechanism (RVAM), inspired by adaptive anchoring behaviors observed in climbing vines. Unlike conventional regression approaches that rely on fixed feature weighting or global optimization objectives, RVAM introduces an anchorage-based learning strategy that dynamically reinforces informative features while attenuating redundant or weakly relevant ones through a vine-structured dependency model.The primary objective of this research is to enhance regression performance under challenging data conditions, including high dimensionality, feature correlation, and noise perturbation. To achieve this, RVAM integrates three key mechanisms: adaptive feature anchorage initialization, random vine-based dependency modeling, and normalized anchorage weight updating. These components jointly enable the proposed algorithm to capture complex feature interactions while maintaining robustness and numerical stability.The effectiveness of RVAM is evaluated using multiple benchmark datasets specifically designed for regression analysis. Experimental results demonstrate that the proposed method consistently outperforms conventional regression models, including linear regression, LASSO, Elastic Net, Random Forest, and Gradient Boosting, across multiple evaluation metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), coefficient of determination (R²), and Mean Absolute Percentage Error (MAPE). On average, RVAM achieves performance improvements ranging from 8% to 15% compared with baseline methods, particularly in datasets characterized by multicollinearity and complex feature dependencies. These findings highlight the potential of biologically inspired anchorage mechanisms as an effective and generalizable strategy for improving regression performance in complex engineering and data-driven applications.

A Measurement Model of Labor Productivity in High-Rise Building Construction Projects

Tewakun Chankampom, Korb Srinavin — Wed, 01 Jul 2026 00:00:00 +0700

Construction labor productivity is a critical determinant of project performance in residential high-rise construction, particularly within dense urban environments. This study empirically examines the hierarchical structure of factors influencing labor productivity in Thai residential high-rise construction projects located in Bangkok and its metropolitan areas. A quantitative approach was employed using questionnaire data collected from 336 construction personnel, including engineers, site supervisors, and skilled workers. Confirmatory factor analysis was applied, with a second-order measurement model used to capture the multidimensional nature of labor productivity. The results validate a five-factor first-order structure comprising Equipment/Tools, Materials, Safety, Motivation, and Weather, which collectively constitute an overarching productivity construct. Among these factors, Equipment/Tools shows the highest standardized loading, highlighting its prominent role in urban high-rise construction. The second-order CFA provides a quantified hierarchy of productivity determinants, offering insights beyond conventional single-level analyses. The findings are context-specific and should be interpreted within the scope of Thai residential high-rise construction projects. Practically, the validated hierarchical structure supports more informed prioritization of productivity-related factors under typical resource constraints in high-rise projects. This study contributes to construction management research by operationalizing labor productivity as an empirically grounded hierarchical construct and provides a foundation for future contextual extensions.

Optimization of Post-Weld Heat Treatment Parameters for Improving Tensile Strength of Dissimilar AA6061–AA7075 Friction Stir Welded Joints Using the Taguchi Method

Somchat Sonasang, Panuwat Thosa, Thanatep Phatungthane — Wed, 01 Jul 2026 00:00:00 +0700

This research presents a novel approach to optimizing post-weld heat treatment (PWHT) parameters for dissimilar friction stir welded joints between AA6061 and AA7075 aluminum alloys. Unlike prior studies that often overlook multi-variable interactions, this study employs a Taguchi L9 design of experiments to systematically investigate the combined effects of solution treatment and aging conditions. The optimal PWHT configuration — solutionizing at 480 °C followed by aging at 180 °C for 6 hours — achieved a peak tensile strength of 307.38 MPa. Elongation performance was strongly influenced by aging duration, with extended times promoting superior ductility. Statistical analysis using ANOVA confirmed aging time as the most significant factor. The novelty of this work lies in the integrated experimental–statistical methodology and its application to dissimilar aluminum joints, providing a robust strategy for tailoring mechanical properties through customized thermal cycles. These findings offer practical insights for engineering applications requiring high-performance welded structures.

Structural Equation Modeling for Successful Water Management in the Northeastern Mekong River Basin under the Royal Irrigation Department

Parichat Panyaprachum, Pongsatorn Tantrabundit — Wed, 01 Jul 2026 00:00:00 +0700

This study aims to: (1) examine the influence of causal factors that affect the success of water management, and (2) verify the consistency of the developed model with empirical data. This is a quantitative research study that employed questionnaires distributed to 375 staff members responsible for water management under the Royal Irrigation Department in the Northeastern Mekong River Basin. The participants were from 15 agencies and had at least one year of work experience. The data were analyzed using Structural Equation Modeling (SEM) techniques. The findings from the quantitative research were also analyzed qualitatively using the documentary analysis method. The results revealed that: (1) Continuous quality improvement, becoming a learning organization, leadership, farmer participation, and environmental conditions all had a statistically significant influence (at the 0.05 level) on the success of water management by the Royal Irrigation Department, with influence values of 0.93, 0.90, 0.80, and 0.71, respectively. Therefore, a key strategy for the Royal Irrigation Department to achieve successful water management involves the proper and timely management of a learning organization in response to changing circumstances. Furthermore, there is an urgent need to enhance the efficiency and effectiveness of leadership, farmer participation, and environmental management to ensure sustainable water management development in the future.

Assessment of Productivity in the Development and Production Processes of Cricket Products and Their Impact on Thailand’s Manufacturing Sector

Nathaporn Phong-a-ran, Pongsatorn Tantrabundit — Wed, 01 Jul 2026 00:00:00 +0700

This research aims to evaluate the productivity in the development and production processes of cricket products and their impact on Thailand’s production sector. The advanced statistical tools were employed to conduct an in-depth assessment of these impacts using the Computable General Equilibrium Model (CGE). This model requires a database derived from Thailand’s Social Accounting Matrix (SAM). The study highlights the differences in various types of environmental costs, including energy costs, fertilizer and pesticide costs, and sanitary and similar services costs. The analysis results indicate that the development and production of cricket products affect Thailand’s top 10 production sectors. The sector most impacted was Sector 105: Iron and Steel, with a Multiplier value of 3.230, a Forward Linkage value of 0.970, and a real benefit of 0.148. The second most impacted was Sector 043: Canning and Preserving of Meat, with a Multiplier value of 2.614, Forward Linkage value of 0.785, and a real benefit of 0.700. The least affected sector among the top 10 was Sector 059: Coffee and Tea, with a Multiplier value of 2.272, Forward Linkage value of 0.683, and a real benefit of 0.706. The findings suggest that strategies to mitigate the negative impacts of energy production and consumption across all activities should follow the approach outlined in this research. Since no previous studies in Thailand have investigated this topic in such a manner, this research provides a critical foundation for future strategic planning

Development of Generative Chatbot towards Thai Rice Knowledge Using Retrieval-Augmented Generation

Wed, 01 Jul 2026 00:00:00 +0700

This paper presents the development of a generative chatbot to support Thai rice farmers by extending rice-relevant knowledge from reliable external sources. The developed chatbot system is based on a lightweight pretrained large language model and is improved through fine-tuning using collected documents related to Thai rice farming. In addition, approved Thai rice information is utilized as an external knowledge base within a retrieval-augmented generation framework to enable responses to domain-specific and technical questions related to Thai rice cultivation. Experimental evaluation compares chatbot systems with and without the application of retrieval-augmented generation. The results indicate that the generative chatbot incorporating retrieval-augmented generation performs better than the baseline system in terms of response relevance and factual reliability. User-driven evaluation shows that the RAG-based chatbot produces 91.43% acceptable responses and 8.57% partially acceptable responses, with no unacceptable responses observed. To further assess system reliability, a grounding-based evaluation was conducted using a controlled query set with document-level reference attribution. The proposed system achieves document-level grounding recall of 0.896 (micro-average) and 0.983 (macro-average), indicating effective utilization of the external knowledge base. These findings suggest that retrieval-augmented generation is an effective approach for developing generative chatbots in domain-specific and localized agricultural advisory contexts, where accuracy, grounding, and trustworthiness of information are essential.

SEM-SAM Model for Strategic Management Toward Efficient and Sustainable Emerging Industries for Sustainable Future

Rachada Fongtanakit — Wed, 01 Jul 2026 00:00:00 +0700

This research aims to develop a model to define the causal relationships of factors influencing future energy consumption in Thailand’s industrial sectors, aligned with the country’s sustainable development goals. The study proposes a Structural Equation Modeling based on the Social Accounting Matrix (SEM-SAM model) as a key tool for effective national management in line with the carbon neutrality target by 2050. The findings reveal that from 1992 to 2025, there has been consistent growth in both the economic and social sectors. However, this growth has simultaneously led to significant environmental degradation. The study indicates that CO2 emissions resulting from industrial energy consumption have surpassed the carrying capacity threshold of 65.05 Mt CO2 Eq. (2025-2034), with a growth rate (2034/2025) of 30.45%, reaching 75.01 Mt CO2 Eq. (2025-2034) during this period. As a solution, the research introduces a new scenario policy that incorporates increased use of biodiesel fuel and clean technologies, which helps reduce CO2 emissions to only 45.05 Mt CO2 Eq. (2025-2034). These findings highlight the potential of the proposed model as a decision-making tool for steering national policy toward a sustainable future.

Auto-Coaxial Switch Controller for Redundancy in FM Broadcasting Transmitter

Wed, 01 Jul 2026 00:00:00 +0700

This paper presents the design and implementation of a Programmable Logic Controller (PLC)-based automatic coaxial switch controller for redundant Frequency Modulation (FM) broadcasting transmitters. The proposed system facilitates autonomous switching between the main transmitter (Tx1) and the backup transmitter (Tx2), ensuring seamless connection to the antenna and uninterrupted broadcast service. Experimental results indicate that the output power (Pout) during normal operation (signal from Tx1) and after switching (signal from Tx2) are comparable, measured at 939.61 W and 944.06 W, respectively. Furthermore, the switching process requires no more than 30 seconds per event, guaranteeing continuous FM broadcasting throughout the year, with the total switching time representing less than 1% of the annual transmission duration. Consequently, the proposed automated controller operates independently without the need for a computer, master clock, or human intervention.

Post-Quantum CRYSTALS-Kyber with Thai Text Seeds: Unicode Normalization and Performance Evaluation

Nattapon Junlachaiworakun, Kritsanapong Somsuk — Wed, 01 Jul 2026 00:00:00 +0700

This research evaluates Post-Quantum CRYSTALS-Kyber through the generation of seeds from Thai text, highlighting Unicode normalization and performance analysis. This approach relies on user-specified seeds derived from Thai text rather than on random seeds. The proposed method includes Thai text normalization, UTF-8 encoding, and seed generation by using the SHAKE256 hash function for Deterministic Random Bit creation (DRBG) to generate key pairs, followed by system verification through encapsulation and decapsulation operations. Kyber512, Kyber768, and Kyber1024 have been chosen for three investigations. The experimental results demonstrate that Thai seeds consistently maintained 100% accuracy throughout the evaluation. Although the seed generation process requires additional durations of 0.0109ms, 0.0095ms, and 0.0108ms, the effect on total processing time remains small at ±0.6%. Furthermore, the examination of four Unicode normalization forms (NFC, NFD, NFKC, NFKD) verifies that visually indistinguishable Thai texts containing different byte sequences could be normalized to provide identical seed results when SHAKE256 is applied with complete consistency. In addition, these results indicate that Thai seeds not only improve user convenience and linguistic significance but also maintain the security and efficiency of Kyber without reducing performance relative to random seeds. Therefore, this study provides the essential foundation for the establishment of hybrid cryptosystems that accommodate the Thai language. Furthermore, it might be adapted for effective post-quantum security systems in the future. This work bridges linguistic usability and quantum-resistant cryptography, paving the way for culturally inclusive post-quantum systems.

Optimal Period Partitioning and Time-of-Use Electricity Pricing for Residential Customers Using Moving Boundary Technique and Multi-Objective Differential Evolution

Nobsouphon Khoupaseuth, Nattaya Rajitroj, Rongrit Chatthaworn — Wed, 01 Jul 2026 00:00:00 +0700

The widespread adoption of Electric Vehicles (EVs) has significantly altered residential electricity consumption patterns, with most EV charging concentrated during evening peak hours, thereby intensifying stress on power grids and operational costs. While Time-of-Use (TOU) pricing is widely used for demand response, many prior studies rely on single-objective or weighted-sum methods, and multi-objective approaches provide limited quantitative evaluation of Pareto front quality. This study addresses these challenges by employing Multi-Objective Differential Evolution (MODE) to optimize TOU electricity pricing for simultaneously minimizing peak demand and total customer energy costs, incorporating price elasticity of demand to reflect realistic consumer response behavior. The quality and diversity of Pareto-optimal solutions are evaluated using the Hypervolume Indicator (HV), ensuring a well-distributed and high-performing set of pricing strategies across the entire trade-off surface, and using the Moving Boundary Technique (MBT) to determine optimal time boundaries for off-peak, mid-peak, and peak periods based on actual hourly load characteristics. The load data utilizes a real residential dataset from Thailand’s Provincial Electricity Authority (PEA), demonstrates the superiority of the optimized TOU three rates compared with the conventional TOU two rates, the proposed approach yields three distinct solution categories: a minimum-peak solution achieving 11.4% peak demand reduction, a minimum-cost solution delivering 34.7% total customer cost savings, and a balanced solution offering 9.5% peak reduction alongside 29% cost savings. Additionally, the minimum-peak solution reduces CO2 emissions by 745 tons daily through strategic load shifting to solar-rich hours.