Journal of Science Ladkrabang

Development of Experiment Set for Study Newton’s Law of Cooling and Electrical Equivalent of Heat to Determine the Specific Heat Capacity of Vegetable Oil

Ratchaneewan Siri — 2025-06-30

This research aims to develop an experimental setup for studying Newton’s law of cooling and the electrical equivalent of heat in determining the specific heat capacity of vegetable oil. To save time and reduce equipment procurement costs, the researcher used easily accessible and low-cost materials instead of purchasing a preassembled experimental kit. Vegetable oil was selected due to its non-reactivity with equipment, high stability and safety, and suitability for maintaining controlled experimental conditions effectively. The experiment was conducted by studying the cooling of equal volumes of water and vegetable oil in beakers of the same size, using known specific heat capacities of water and the beaker as references for calculation. In addition, the researcher employed the electrical equivalent of heat method by heating the vegetable oil using a resistor connected to a power supply and compared the results obtained from both methods. The developed experimental setup was applied in the Advanced Physics Laboratory course within the Physics program, Division of Physical Science, Faculty of Science. The research findings indicate that the developed setup can accurately determine the specific heat capacity of vegetable oil. The value obtained using Newton’s law of cooling was 3.156 J/g·K, while that obtained using the electrical equivalent of heat method was 3.119 J/g·K. The difference between the two values was only 1.2%, which aligns with the principles of Newton’s law of cooling and the electrical equivalent of heat. Moreover, the results from implementing the experimental setup in the advanced laboratory course showed that the effectiveness of the two methods differed significantly at the 0.05 statistical level.

Aspect-Based Sentiment Analysis for Thai Language from Consumer Reviews Towards Smartphone

Nitthiwat Jensirisak — 2025-06-30

The purpose of this research is to develop a model for aspect-based sentiment analysis for Thai language from consumer reviews towards smartphone which consists of the camera, battery, screen, performance, and price aspects that collected from YouTube with number of 67,907 comments including 6 brands: Apple, Samsung, Xiaomi, Vivo, Oppo, and Huawei. The process consists of: 1) Data collection 2) Data preprocessing 3) Aspect-based sentiment analysis, and 4) Model performance evaluation. This research used machine learning model and deep learning model to compare performance sentiment classification and performance evaluation from precision, recall, F-measure, and accuracy metric. The result suggests WangchanBERTa model is the most reliable method for sentiment classification.

Production of Low-Alcohol Beverages from Pummelo Juice (Citrus maxima (Burm.) Merr.) Cultivated Variety of Khao Taeng Kwa, Fermented with the Mycelia of Pink Oyster Mushroom (Pleurotus flabellatus (Berk & Br.) Sacc.)

Puncharat Pilong — 2025-06-30

Low-alcohol beverage was produced from the juice of pummelo (Citrus maxima (Burm.) Merr.), Khao Taeng Kwa cultivated variety, fermented with pink oyster mushroom (Pleurotus flabellatus) mycelia, as a potential alternative alcohol product. The study aimed to determine the optimal pummelo juice concentration, amount of mycelial starter, and the soluble solid content in the pummelo juice for low-alcohol beverage production. Results demonstrated that P. flabellatus mycelia effectively fermented 100% pummelo juice. The optimal fermentation conditions included 15 mycelial starter pieces, each 0.7 cm in diameter, and a soluble solids content of 22 °Brix. After 12–21 days of fermentation, the final product contained 1.00–1.04% alcohol, highlighting its potential as a promising low-alcohol alternative.

Reducing Effect of Sulfide and Increasing Potential of Biogas Production from High Sulfate Wastewater by Sulfide Oxidation and Denitrification Process

Wikanda Thongnueakhaeng — 2025-06-30

This research focuses on mitigating the impact of sulfide on methane-producing bacteria in biogas production system to enhance the efficiency of biogas production from high-sulfate wastewater by controlling the appropriate conditions within the system to facilitate the removal of organic substances, sulfate, and sulfide under anaerobic conditions. Experimental runs were conducted both in laboratory and rubber sheet establishment. At the laboratory scale, nitrate is added to the wastewater before it enters the system, with the Sulfate/Nitrate (S/N) ratio controlled at 2.0. The study examined Hydraulic Retention Times (HRT) of 10, 20, and 30 days. In the establishment, nitrate was sourced from nitrification system of wastewater from establishment without chemical addition. Results showed that operation at HRT of 10, 20, and 30 days provided organic removal efficiencies in form of COD (Chemical Oxygen Demand) at 61.18±1.52, 82.99±2.24, and 84.68±2.32%, respectively. At HRT of 30 days, the maximum biogas production rate was found to be 0.35±0.4% L-CH₄/gCOD_removed, with methane content comprising 71.2±0.4% of the gas. The experiment in establishment was found to work similarly to laboratory experiments, with average COD removal efficiency of 80.56±1.05% and average biogas production rate of 0.30±0.09 L-CH₄/gCOD_removed. The system in this study was able to reduce the effects of sulfides occurring in the system on methane-producing microorganisms. This has led to an increase in the activity of methane-producing microorganisms from the existing system of the establishment from average of 0.035±0.001 gCH₄/gVSS/d increased to 0.060±0.001 gCH₄/gVSS/d.

Corn Price Prediction Model Using Deep Learning

Sirinthra Samchairat — 2025-06-30

Animal feed corn is a highly demanded agricultural commodity across various industries. However, its prices fluctuate unpredictably each year, causing difficulties for farmers in planning their crops. Therefore, accurate price forecasting is crucial for helping farmers plan effectively. This research presents the development of a corn price prediction model using both historical price data and other important features. The study compares the performance of four models: ARIMA, ARIMAX, LSTM, and GRU, using data from 2015 to 2021 to predict corn prices for 2020 and 2021. The comparison between models that use only historical price data and those that incorporate additional features shows that the GRU model, utilizing both historical price data and features such as total corn exports, export prices, Chicago Board of Trade futures prices, and total export value, performs the best. The GRU model achieves an RMSE of 0.0780 and an MAE of 0.0662, demonstrating the highest accuracy in the test datasets. Accurate price forecasting is crucial for farmers and stakeholders in the corn industry, as it enables more efficient planning and resource management.

Feasibility Study on Preparation of Fiber Sheet from Banana Stem Fiber and Bagasse Fiber Combined with Binder from Cassava Flour for Application in Forming Food Packaging

Jitsopa Chaliewsak — 2025-06-30

The aims of this research were to prepare fiber sheets with various amounts of banana stem fiber, bagasse fiber, and binder, to analyze the effect of fibers and binder on the physical properties of fiber sheets, and to cast food packaging samples from the fiber sheets. Banana stem fiber and bagasse fiber were extracted with a 5% w/w sodium hydroxide solution at 90°C for 120 min. Then, fiber sheets were formed with various amounts of fiber and binder from cassava flour. The fiber sheets were analyzed for physical appearance, chemical structure, water absorption, oil absorption, and mechanical properties. Then, food packaging samples were cast from the fiber sheets by the compression molding process. The results showed that the physical appearance of the fiber sheets depended on the type and amount of fiber and the amount of binder. The fiber sheets with binder presented a smooth surface, no fluff of fiber, but unparalleled fiber formation because the binder seized fiber into the sheets, and the color of the fiber sheet depended on type and amount of fiber. All of the fiber sheets presented very high water absorption (3.4–5.7 times) and high oil absorption (1.6–3.7 times). All of the fiber sheets could be cast into food packaging samples, but the food packaging samples are not suitable for soups and fried foods. The fiber sheet BN30BG30 was suitable for further study and developing food packaging.

Efficiency Comparison of Classification Models for Pali and Sanskrit in Thai Languages Using Machine Learning Techniques

Natratanon Kanraweekultana — 2025-06-30

This research aims to test and compare the performance of classification models for Pali and Sanskrit in Thai language using Machine learning techniques. The study focuses on improving the accuracy in distinguishing between words from these two languages, which often exhibit similarities in pronunciation and spelling. Five models were tested: Random Forest, Decision Tree, K-Nearest Neighbors (K-NN), Naive Bayes, and Support Vector Machine (SVM). The evaluation process employed 10-fold cross-validation to assess model performance. The results indicate that the SVM model is the most efficient, with an accuracy of 95.75% and a precision of 90.90%. K-NN follows closely with an accuracy of 92.86%, while Naive Bayes achieves 81.29%. The Random Forest model, however, shows the lowest performance with an accuracy of only 55.27%. These findings highlight the SVM model's effectiveness in accurately classifying Pali and Sanskrit words in Thai language. The results can be applied to further developments in language classification, translation, and educational technology tools for language learning.

Planning a Timetable for Deciding to Registration in Course

Supapich Kruesri — 2025-06-30

The manual scheduling of university courses is a complex and challenging process due to the diverse registration requirements across academic programs. This research investigates the University Course Timetabling Problem (UCTP) by applying Integer Programming Problem (IP) models to construct mathematical formulations for solving optimization problems and managing intricate constraints. The primary objective is to propose a systematic course registration methodology and develop timetables that enable students to complete their academic programs in the shortest possible duration. The findings reveal that the proposed timetable planning approach effectively reduces scheduling complexity and facilitates students' timely completion of their courses. This study provides a guideline for course registration. However, class scheduling should also consider additional factors to ensure appropriate and individualized decision-making.

Effect of Using Rice Flour Substituted Wheat Flour on Quality of Noodle Mixed Jellyfish Powder Stored in Different Conditions

Benjawan Thumthanaruk — 2025-06-30

Noodles are a staple food widely consumed across the globe. However, the development of noodles with enhanced nutritional diversity remains limited. This study aimed to 1) produce jellyfish-enriched noodles by partially substituting wheat flour with rice flour at 0%, 10%, and 15%, combined with 5% jellyfish powder (based on total flour weight), to enhance collagen protein content. The resulting jellyfish noodles were subjected to sensory evaluation using a 5-point hedonic scale. Saltiness and fishy odor intensity were assessed using a 5-point just-about-right (JAR) scale. Physical properties such as color, tensile strength, and cooking quality were also analyzed; 2) to investigate the effect of storage conditions on noodle quality, including appearance, physical properties, and sensory acceptance by 50 panelists. Results indicated no statistically significant differences (p>0.05) in color, appearance, odor, taste, and overall acceptability between the jellyfish noodles and the control sample (no jellyfish powder). Saltiness and fishy odor intensities were evaluated using a scale of 1 (very weak) to 5 (very strong), with classifications as follows: 1 -very weak, 2-slightly weak, 3-just right, 4-slightly strong, and 5-very strong. The mean scores for saltiness ranged from 2.16 to 2.38, and for fishy odor from 2.50 to 2.84, indicating slight intensities and suggesting that jellyfish powder had no significant impact on flavor quality. Regarding cooking quality, all jellyfish noodle samples exhibited higher total soluble solids than the control, resulting in lower cooked weight. The 10% rice flour substitution sample showed higher tensile strength than the 15% substitution group. Moreover, adding jellyfish powder increased chewiness and boosted protein content to 8.85%, with collagen content reaching 0.425 grams per 100 grams of noodles. Shelf-life evaluation revealed that the jellyfish noodles had a storage life of 2 days at room temperature and 11 days under refrigerated conditions, as determined by visible spoilage. The noodles exhibited a light yellow color and only slight changes in texture over time. However, stored samples showed increased cooking loss and reduced post-cooking weight, indicating changes in cooking performance over time.

Detecting Falls Inside the Building by Deep Learning

Phetcharat Suk-ubon — 2025-06-30

This article presents the detection of falls inside the building using deep learning. The method is to apply Convolutional Neural Network (CNN) for regularization by Dropout technique to create classification models. The data are collected in the form of video files from lnVia laboratory room, University de Franche-Comté, and separated into 60% of Training set, 20% of Test set, and 20% of Validation set. The three experiments are designed for different patterns according to the different inputs: 1) Grayscale images 2) Motion History Images (MHI), and 3) MHI combined with Grayscale image. Regarding the efficiency evaluation of the three best models, the classification model for Grayscale images has 0.9204 accuracy. The model for classifying MHI achieves 0.9193 accuracy. On the other hand, the model for classifying MHI combined with Grayscale images results in 0.9575 accuracy, which is the best model of all 3 experiments.