Journal of Engineering and Digital Technology (JEDT)

Development of Decision Support System Using ATP-DSS Techniques to Reduce Delays and Efficiency Improvement in the Order Promising Process with Make-to-Order Manufacturing: Case Study of an Electronics Manufacturer

Phimon Yensom — 2026-08-27

The case study company’s current situation, which is an electronic components manufacturer, who is facing a delay of an available-to-promise (ATP) response times from make-to-order (MTO) production, and the impact of customers service efficiency and competitiveness. This research addressed the development of decision support systems using ATP-DSS techniques which is a principle of applying the information system to manage the inventory for prolonged available-to-promise response times, exceeding 24 hours, within an electronics manufacturer's order promising process (OPP), aiming to reduce delay defects and variability. A Real-Time Decision Support System was developed, integrating ATP-DSS techniques, Six Sigma principles, Lean methodologies (ECRS), and Business Process Reengineering (BPR). This integration facilitated enhanced, real-time assessments of inventory levels, production capacity, and supplier part availability. Results demonstrated a significant OPP reduction average 22.23 minutes per order, or 84.62%, with OPP response times within 24 hours improving by an average of 69.71%. Defect rates were eliminated, achieving a 100% reduction or zero mistakes, and process variability stabilized, with the Process Capability Index (Cpk) increasing from 0.95 to 1.37. Moreover, the value-added (VA) activities increased by 7 steps (100%), and necessary but non-value-added (NNVA) activities decreased by 7 steps (87.50%).

Parameter Identification for Interior Permanent Magnet Synchronous Motors in Air Conditioners Emphasis on the Effects of Magnetic Cross-Coupling and Induced Electromotive Force Harmonics

Jadsadakorn Thitipumdacha — 2026-08-27

Interior permanent magnet synchronous motors (IPMSMs) are widely used in air conditioners. The compressor speed is controlled by inverter using sensorless control in which the performance depends mainly on motor parameters. However, the parameters provided by motor manufacturers are preliminary and often insufficiently accurate, especially when the motor structures are redesigned to reduce cost. Such circumstances can easily cause the saturation of iron core and lead to the necessity for accurate identification of motor parameters that should take the effects of magnetic cross-coupling and induced electromotive force harmonics into consideration. This article proposes a systematic procedure for parameter identification for IPMSMs that emphasizes the effects of magnetic cross-coupling and induced electromotive force harmonics. The proposed scheme can optimize the duration of parameters’ fine tuning in the testing process and can be practically applied in product development. Experimental results show that the parameters obtained by the proposed method are more accurate than those provided by motor manufacturers. Consequently, the control performance of the drive system of air conditioners can be improved across the entire operating range of compressors, while the rotor position estimation error can be reduced by up to 52 percent.

Design and Construction of an Automation System for Battery Drop Test Machine

Tharadol Mooyotha — 2026-08-27

This research presents the design and construction of an automated battery drop test machine developed to overcome the high costs of commercial equipment and the physical limitations of existing testers regarding large-scale battery research. The proposed system features an adjustable drop height ranging from 300 mm to 1500 mm and utilizes a stainless-steel hinge clamping mechanism for secure battery positioning. The control architecture is built upon a Siemens LOGO!8 PLC integrated with an HMI touchscreen, employing Finite State Machine (FSM) principles to ensure precise operation in both manual and automatic modes. To prioritize user safety, the machine is housed within a secure test chamber equipped with a CO₂ fire suppression system and water cooling to mitigate thermal runaway risks.

Experimental validation was conducted following the testing protocols outlined in the IEC 62133 and MIL-STD-810H standards, utilizing battery samples weighing up to 80 kg. Results confirmed the machine’s ability to accurately release batteries in various orientations, including 6 surfaces, 12 edges, and 8 corners, with impacts consistently aligning with target positions. The system offers a flexible, cost-effective, and safe alternative for the continuous testing of diverse battery types, particularly those intended for electric vehicles and energy storage systems. Future work will focus on expanding compliance to include UN 38.3 and UL 2054 standards for international product certification.

A Machine Learning Approach for Predicting Thermal Comfort in a Bangkok Academic Library: Integrating Logit and Random Forest Models

Wipawadee Wongsuwan — 2026-08-27

A study of thermal comfort was conducted in a 4-story academic library located in Bangkok, Thailand. Since limited thermal comfort studies have been conducted for academic libraries located in a tropical climate. The field measurements by a set of sensors and questionnaire surveys were conducted among 234 library occupants. The measurements led to the average air temperature, relative humidity, and air velocity of about 24.39°C, 55.39%, and 0.073 m/s, respectively. Hence, the thermal comfort was evaluated using the Predicted Mean Vote (PMV) and the Predicted Percentage of Dissatisfied (PPD) based on the ASHRAE 55 standard, and the Actual Mean Vote (AMV) from the occupants’ perception. PMV and AMV were found to be -0.74 and -0.12, implying that the library environment was slightly cooled. Although the conventional standard PMV was widely used, it might not accurately represent real-time thermal preferences and cannot be practically integrated into the Heating, Ventilation, and Air-Conditioning (HVAC) control. In addition to the conventional method, two Machine Learning (ML) models, Random Forest (RF) and Support Vector Machine (SVM), were investigated for their predictive performance. For the ML model development, five environmental variables (T_a, T_mrt, V_a, RH, and CLO) were used. It was found that the RF model performed well with an R-squared of about 0.9876, while SVM achieved only 0.8067. Therefore, the key findings support the application of the ML approach, especially RF, to enhance HVAC control for energy management strategies while maintaining the thermal comfort of library occupants in academic buildings located in tropical climates.

Application of Microbubble Technology for Improved Efficiency and Reduction in Treatment Time of 1-Methylcyclopropene for Postharvest Ripening Delay of Banana

Hafnee Lateh — 2026-08-27

This research is aimed at evaluating feasibility of using 1-methylcyclopropene microbubbles (1-MCP-MBs) as an alternative postharvest treatment for delaying the ripening of bananas with performance comparison of treating bananas with gaseous fumigation using conventional treatments and without treatment. The effects of immersion time (5, 10, and 15 min), 1-MCP concentration (500, 1000, and 1500 ppb), and microbubble generation pressure (4, 5, and 6 kg/cm²) were tested systematically for 10 days storage period. The results showed the high efficacy of 1-methylcyclopropene microbubbles (1-MCP-MBs) in control of fruit ripening and preventing weight loss when compared with the conventional fumigation with gaseous 1-MCP. The optimal conditions for treating were 1-MCP at 500 ppb, an immersion duration of 5 min, and a generation pressure of 5 kg/cm². These situations were able to create stable microbubble formation with a high specific surface area resulting in increased adhesion of 1-MCP to the fruit peel. The optimized storage condition resulted in weight loss of 1.68 ± 0.07% during storage for 10 days in comparison to the control weight loss of 3.14 ± 0.56%. Moreover, the treatment using the 1-MCP resulted in a good preservation of the chloroplast integrity and chlorophyll stability, and thus fruit freshness was maintained for over 10 days. Contrastingly, fruits in the control group experienced severe senescing and ripening in 6–8 days. The overall results indicated high potential for 1-MCP-MB technology to be applied as a postharvest technology that can increase competitiveness of banana exports in terms of cost, time, and reduced chemical usage by improved operational efficiency.

Enhancing Accessibility for Visually Impaired Users through a Hybrid Real-Time Multi-Object Recognition System

Pitiphat Joembunthanaphong — 2026-08-27

Visual impairment creates substantial challenges in performing object recognition tasks essential for independent living. Existing assistive vision systems typically rely on either classical computer vision techniques or deep learning models, limiting their robustness and practical applicability in real-world environments. To address these limitations, this study proposes a hybrid intelligent vision framework that integrates YOLOv8-based object detection for robust object localization, barcode recognition for reliable product verification, and ORB feature matching for fine-grained identification when barcode information is unavailable, thereby combining the complementary strengths of deep learning and classical computer vision. The dataset was designed to represent four common information-access tasks encountered by visually impaired individuals, including product labels, Thai banknotes, short shelf-life food products, and pharmaceutical items. An expanded dataset comprising 1,105 object classes and 7,927 images was constructed under diverse environmental conditions. The complete recognition pipeline was evaluated as an integrated mobile assistive system through both controlled and uncontrolled experiments to assess recognition accuracy, robustness, and the feasibility of practical deployment. Experimental results show that the proposed framework achieved 97.6% recognition accuracy, outperforming both the ORB–Barcode baseline (96.2%) and the standalone YOLOv8 model (97.0%), while maintaining near-real-time performance with an average processing time of 2.30 s per inference. User evaluation further demonstrated statistically significant improvements in task success rates among trained participants. The proposed framework improves recognition robustness while preserving computational efficiency on resource-constrained mobile devices, providing a practical, scalable, and user-validated assistive solution that enhances accessibility and supports independent living for visually impaired individuals.

Thai Antipyretic Medicinal Plant Recognition Using CNNs Integrated into a LINE Chatbot

Walairach Nunsong — 2026-08-27

This study developed a leaf-image recognition system for Thai antipyretic medicinal plants using deep convolutional neural networks (CNNs). Recognizing medicinal plants can be difficult because several species have similar leaf shapes and textures. An automatic recognition tool may help non-expert users identify medicinal plants more consistently. The main contributions are the collection of a 20-species Thai antipyretic herb leaf dataset, the evaluation of eight CNN architectures, and the integration of the selected model into a LINE Chatbot prototype.
A new dataset of leaves from 20 medicinal plants for antipyretics found in southern Thailand was created.
Data augmentation was applied to increase the number and variability of training images. Eight CNN architectures, namely AlexNet, ResNet, VGG, DenseNet, GoogLeNet, EfficientNet, MobileNet, and SqueezeNet, were evaluated for recognizing 20 Thai antipyretic herb species. Lastly, a CNN architecture, which is the best performing, was integrated with a LINE Chatbot for practical application. The experimental results showed that SqueezeNet achieved the highest accuracy of 92.8% with a processing time of 0.004 seconds per image. The SqueezeNet model was therefore selected for integration into the LINE Chatbot prototype.

Development of an n8n-Based Automated Framework for Penetration Testing Using Large Language Models and Structured Evidence

Annop Monsakul — 2026-08-27

This research aims to develop and evaluate an evidence-driven automated framework for penetration testing by integrating large language models with hallucination-aware verification. The proposed framework addresses key limitations of traditional penetration testing, including heavy reliance on expert analysts, fragmented use of multiple security tools, heterogeneous tool outputs, and inconsistent results across repeated testing cycles. The framework employs n8n as the central workflow orchestration platform within a Docker container stack. It classifies assessment targets into four categories: network service systems, modern web applications, API security systems, and systems involving secret leakage or insecure configuration. Based on the target category, the framework selects an appropriate testing strategy, collects outputs from relevant security assessment modules, transforms them into structured evidence, and stores them in a PostgreSQL database. GPT-4o is then used through prompt engineering and structured output, without model fine-tuning, to support vulnerability interpretation, evidence correlation, and risk prioritization. To reduce the risk of unsupported or hallucinated findings, the framework incorporates a hallucination-aware verification mechanism requiring every reported finding to be supported by tool-generated evidence or execution logs before inclusion in the final report. The experiment was conducted using four deliberately vulnerable systems: Metasploitable3, OWASP Juice Shop, OWASP crAPI, and OWASP WrongSecrets. Each target was tested six times, resulting in 24 campaign runs. The experimental results indicate that the proposed framework consistently collected structured evidence, achieving structured-evidence consistency above 97% across all targets. The verification pass rate of LLM-generated findings ranged from 83.02% to 90.32%, while the average traceability score ranged from 89.83% to 96.83%. These results suggest that combining workflow automation, structured evidence management, and evidence-based verification of large language model outputs can improve the reliability, consistency, and traceability of automated penetration testing.