Engineering and Technology Horizons

Enhancing Planning and Control for Sustainable Custom and Project-Based Furniture Manufacturing

2025-02-20T13:41:43+07:00

This study investigates production planning and control (PPC) challenges in the Thai furniture industry, particularly in custom and project-based manufacturing. A mixed-methods approach was adopted, combining structured interviews, on-site observations, and the analysis of historical production data to identify inefficiencies in scheduling, task coordination, and resource utilization. The factory implemented integrated PPC strategies such as standardized task planning, workforce optimization, and refined scheduling. As a result, it achieved a 23.58% reduction in delivery delays, a 14% improvement in labor utilization, and a 77.78% increase in the production rate. These outcomes highlight the effectiveness of agile coordination and structured planning in enhancing operational efficiency and supporting more sustainable manufacturing operations in high-variability, make-to-order environments.

Min-Max Policy Implementation for Inventory Management in Steel Supply Sector

2025-03-04T08:30:47+07:00

This study explores the application of the Min-Max inventory management policy to improve stock control efficiency at a steel supply company in Chiang Mai Province, Thailand. The company operates two branches: a central warehouse and a retail store located 40 km apart. Due to the absence of a formal inventory management policy, the retail branch frequently experiences stock shortages, leading to daily replenishment trips from the main warehouse, causing operational inefficiencies and excessive transportation costs. The research focuses on four top-selling products from the company’s highest sales-value category. A Min-Max policy was proposed where the minimum stock level was set based on the average daily demand multiplied by the minimum lead time, and the maximum level based on average demand multiplied by the most likely lead time. Historical demand data from May 2022 to May 2023 were analyzed to determine appropriate Min-Max thresholds. Demand analysis revealed low average daily sales coupled with high variability, indicating the need for a structured inventory approach. Trace-driven simulations were conducted using historical demand data to assess the impact of the Min-Max policy. The simulation results showed a significant reduction in the number of transportation trips and cost compared to the company’s current practice of daily restocking. The study concludes that implementing the Min-Max policy can reduce operational inefficiencies and transportation costs for the company. However, the results are limited to the selected products and may not reflect the entire inventory's performance. The simplicity of the Min-Max policy makes it practical for SMEs, though further refinements such as demand forecasting techniques could optimize its performance in environments with high demand variability.

Comparative Study of Modern VPN Solutions: Impact of Cloudflare, ZeroTier, and WireGuard on Network and Server Performance

2025-03-11T11:13:00+07:00

This research investigates of the performance of three popular VPN solutions namely Cloudflare, ZeroTier and WireGuard, by measuring their effect on network performance and server resource usage across multiple metrics such as file upload/download speeds, round-trip time (RTT), web latency, and server CPU usage. The aim is to find the best solution for certain workloads by benchmarking these solutions in a controlled manner. The results of these experiments showed large performance differences. The results were consistent for all tests: WireGuard provided the fastest upload and download speed (19 seconds and 52 seconds, for 1000 MB files, respectively), the lowest web latency (50 milliseconds for 1000 connections), and the most efficient CPU utilization (24% at 1000 connections). For small size of packets (less than 700 bytes), Cloudflare provided competitive RTTs around 10 milliseconds and balanced performance for light workloads. However, it was not scalable indicated by web latency about 200 milliseconds and CPU utilization higher than 32% in high-concurrency scenarios. Conversely with lower workloads, ZeroTier struggled with download of heavy file sizes and lots of connections such as downloading with 1000 MB in size took 84 seconds and up to 62% of CPU utilization. WireGuard emerges as the best-suited high-performance solution for scalable applications. Cloudflare and ZeroTier offer trade-offs helpful to particular use cases, providing perspective on which VPN solution to choose depending on workload requirements and resource constraints.

Gain Enhancement of a Dual-Band S-Patch Antenna Array for 5G Application

2025-03-17T09:01:52+07:00

This paper proposes the dual-band S-patch antenna with gain enhancement by using the planar array configuration and dielectric superstrate. The design of the proposed antenna is focused on the base station antenna in the 5G frequency bands n41 (2.6 GHz) and n78 (3.5 GHz). The dual-band S-patch antenna is arranged in the
2 × 6 elements planar array antenna and the FR4 dielectric superstrate is on the top of the array. The simulated results indicated that the operating frequency of 2.55–2.65 GHz and 3.46–3.61 GHz with uni-directional radiation pattern. The antenna gain can be improved with 18.70 dBi at 2.6 GHz and 19.30 dBi at 3.5 GHz. The prototype antenna is fabricated and the measured results are in good agreement. With the simple design of the proposed antenna, it would be useful for the installation of the base station antenna.

Energy Consumption Prediction and Anomaly Detection for Boiler Feed Pump in Power Plant Using Machine Learning and Deep Learning

2025-04-02T07:00:18+07:00

Enhancing energy efficiency and operational reliability is crucial in power plant management, particularly for high-energy-consuming machines such as boiler feed water pumps (BFPs). These pumps play a vital role in the continuous generation of steam and electricity and must operate 24/7 to maintain power production stability. This study proposes the development of predictive models based on machine learning and deep learning techniques to accurately predict energy consumption and applies best models to detect anomalous behaviors in BFPs, enabling timely and preventive interventions. A dataset comprising 43,082 hourly records over five years, with 18 critical operational features, was analyzed using preprocessing and feature engineering techniques. Various predictive models were trained and evaluated, including Multiple Linear Regression, Regularized Regressions (Ridge, Lasso, ElasticNet), Support Vector Regression (SVR), Decision Tree, Ensemble Methods (Random Forest, XGBoost, CatBoost, LightGBM), and Deep Learning Architectures (DNN, RNN, GRU, LSTM). Among these models, SVR demonstrated the highest accuracy (MSE: 13.5573, R²: 0.9838), followed closely by LightGBM. Feature importance analysis revealed that boiler feed pump discharge pressure and bearing housing vibration levels were the most influential variables in energy consumption prediction. Anomaly detection using the Interquartile Range (IQR) method classified deviations into two warning levels, enabling proactive maintenance strategies. Additionally, a Graphical User Interface (GUI) web application was developed for real-time monitoring, integrating predictive models, anomaly detection, and an automated email alert system to assist operators in responding to abnormal energy consumption events promptly. These results highlight the potential of predictive analytics and real-time monitoring in optimizing power plant operations, providing a foundation for extending predictive capabilities to other critical energy-intensive systems.

Improving Heavy Maintenance Management Efficiency under Limited Depot Resources: A Case Study of MRT Pink Line

2025-03-27T08:41:34+07:00

This research focuses on optimizing the heavy maintenance scheduling of the Pink Line MRT using two models: a non-flexible model (fixed at 120,000 km) and a flexible model, which allows a ±10% adjustment in the accumulated mileage (108,000 - 132,000 km). The study employs Mixed-Integer Linear Programming (MILP) and a two-year simulation to analyze the effects of key constraints, including depot capacity, repair duration, and flexibility levels. The results indicate that a 10% flexibility reduces unused accumulated mileage by 55.97% and increases utilized mileage by 10%, without requiring additional resources. However, increasing the flexibility to 15% yields diminishing returns, leading to higher operational costs and potential safety risks. Conversely, reducing flexibility to 5% helps control costs but increases maintenance frequency, affecting operational stability. Additionally, sensitivity analysis reveals that a depot capacity of C = 2 (2 maintenance tracks per day) with a 5-day repair duration is optimal, balancing efficiency and resource allocation. C = 1 leads to maintenance congestion and reduced operational efficiency, whereas C = 2 effectively distributes maintenance workload without delays. Although C = 3 shortens repair time, it offers only marginal benefits compared to the increased costs. The findings highlight the importance of strategic flexibility management and optimized depot capacity to reduce maintenance frequency, enhance resource utilization, and improve overall train operation management. This research provides valuable insights for railway maintenance planning, contributing to cost reduction and long-term operational efficiency.

Analysis of Rear Differential Component Clustering in Transmission Systems Using Hierarchical Cluster Analysis with and without Procurement Strategy Matrix Variables

2025-04-23T08:02:10+07:00

The automotive industry has faced significant challenges due to the large number of Tier 2 suppliers for Rear Differential components, with 17 suppliers providing 32 different parts. This situation has resulted in increased production costs and more complex supply chain management. This study aimed to analyze the clustering of Rear Differential components in transmission systems using Hierarchical Cluster Analysis, with the goal of supporting cost reduction in the automotive industry. Two clustering models were compared: Model 1, which excluded procurement strategy matrix variables (Special Requirements, Raw Material Grade, Raw Material Type, Manufacturing Process, Tier 2 Supplier Information, and Company Location), and Model 2, which incorporated an additional variable related to the Procurement Strategy Matrix. The decision criteria for determining the optimal number of clusters were based on four key factors: 1) Product design, 2) Characteristics, 3) Materials, and 4) Manufacturing. The clustering results for both models revealed the same optimal number of 13 clusters; however, the similarity matrix between the clusters differed. Furthermore, the number of members within each cluster varied. Based on the criteria for determining the optimal number of clusters, Model 2, which included the Procurement Strategy Matrix variable, demonstrated superior clustering efficiency compared to Model 1. Ultimately, this research identified 13 optimal clusters, reducing the number of Tier 2 suppliers from 17 to 13, representing a 23.53% reduction.

Development of an Image Processing System for Defect Detection in Nam Dok Mai Golden Mangoes

2025-03-21T12:07:30+07:00

This study proposes an image processing-based approach for detecting surface defects in Nam Dok Mai mangoes. Each fruit was photographed from two sides to capture comprehensive defect characteristics. The images were subsequently converted into the HSV color space to highlight darker defect regions, such as brown or black, followed by morphological dilation to refine defect boundaries and facilitate accurate area measurement. Detected defects were quantified in square centimeters and categorized into four quality classes: Extra Class, Class I, and Class II, according to the Thai Agricultural Standard TAS 5-2567. Additionally, a fourth class, Bad Quality, was introduced to represent defects exceeding the Class II size threshold. The annotated dataset was prepared using Roboflow, where labeling and data augmentation were conducted to enhance sample diversity. The dataset was partitioned into a training set (80%) and a testing set (20%). While image processing techniques were employed for initial dataset preparation, the primary objective was to develop a Mask R-CNN model capable of autonomously detecting defects directly from raw images, thereby eliminating the reliance on manual preprocessing. Following the training phase, the Mask R-CNN model was evaluated for its ability to detect and classify mango defects. Experimental results demonstrated high Precision and F1-Score values, particularly in the Extra Class and Bad Quality groups. The model achieved an overall accuracy of 70.71%, reflecting its strong potential for real-world application. It is anticipated that this system could significantly improve the accuracy and efficiency of mango sorting processes in the agricultural sector, contributing to standardized and reliable quality control.