The Journal of King Mongkut’s University of Technology North Bangkok

บรรณาธิการวารสารวิชาการพระจอมเกล้าพระนครเหนือ

บรรณาธิการ วารสารวิชาการพระจอมเกล้าพระนครเหนือ — 2026-07-21

บรรณาธิการวารสารวิชาการพระจอมเกล้าพระนครเหนือ

A Compartmental Pharmacokinetic Model with Linear and Nonlinear Elimination for Medical Cannabidiol Oil

Thanachok Mahahong — 2026-05-11

This study presents the development of a compartmental model to describe the pharmacokinetics of cannabidiol (CBD) in medical cannabis oil following drug administration. The model incorporates both linear and nonlinear elimination based on Michaelis–Menten kinetics to represent urinary excretion and metabolic transformation of the drug. Ordinary differential equations were formulated based on the law of mass action to simulate the concentration–time relationships in each compartment and were solved numerically using MATLAB with the ode45 function. Pharmacokinetic parameters were estimated through curve fitting to clinical data. Simulation results demonstrate that the model-predicted plasma CBD concentrations are in close agreement with clinical data, with a peak concentration of approximately 1.45 ng/mL occurring at around 120 minutes. In addition, the estimated parameters can be applied to simulate drug concentrations in other compartments. This study is significant because it systematically links mechanistic understanding of CBD distribution and elimination with clinical applications. The developed model can serve as a supportive tool for dose determination and treatment planning, enabling the safe and effective clinical use of medical CBD oil.

Influence of Rice Husk-derived Biochar Rates on Ammonia Emissions and Quality of Compost

Nitiphon Ainthauaong — 2026-06-11

This study aimed to evaluate the effects of adding rice husk biochar to compost at different rates (0, 20, 40, and 60%) on ammonia (NH3) emissions and compost quality. The experiment was conducted in a greenhouse for 54 days using a Completely Randomized Design (CRD) with four replications. Rice husk biochar produced by a traditional pyrolysis method was mixed with rice straw and cattle manure, which served as the main composting materials, at the designated ratios. Ammonia emissions, moisture content, and temperature of the compost piles were measured on days 1, 4, 16, 32, and 54 after composting. At the end of the composting period, compost properties were analyzed, including pH, Electrical Conductivity (EC), organic matter content, total nitrogen content, and the carbon-to-nitrogen (C/N) ratio. The results showed that the addition of rice husk biochar significantly reduced NH3 emissions compared with the treatment without biochar, particularly at the 60% application rate, which achieved the greatest reduction in NH3 emissions (up to 30.22%) throughout the composting process. Moreover, all biochar application rates increased total nitrogen and organic matter contents and decreased the C/N ratio, while pH and EC were not significantly different. However, when compost quality, nitrogen retention, cost, and practical applicability were considered together, the 40% rice husk biochar application rate was identified as the most suitable under the experimental conditions. These findings indicate the potential of rice husk biochar to enhance composting efficiency, reduce nitrogen losses, and support the sustainable management of agricultural residues.

คำแนะนำในการเตรียมต้นฉบับบทความวารสารวิชาการพระจอมเกล้าพระนครเหนือ

คำแนะนำในการเตรียมต้นฉบับบทความ วิชาการพระจอมเกล้าพระนครเหนือ — 2026-07-21

คำแนะนำในการเตรียมต้นฉบับบทความวารสารวิชาการพระจอมเกล้าพระนครเหนือ

Low-Pass Filter Design for Harmonic Detection Based on the Instantaneous Reactive Power Theory in AC Electric Railway Systems

Sooppakit Sirasugol — 2026-07-21

This paper presents the design of a Low-Pass Filter (LPF) for extracting the harmonic component of active power in the harmonic detection process based on instantaneous reactive power theory (PQ theory). The objective is to accurately compute the harmonic current to be used as a reference current for Active Power Filters (APFs), which inject compensating currents to eliminate harmonics in AC electric railway systems. The performance of the proposed LPF is evaluated through MATLAB/Simulink simulations, with test cases classified according to the selected cutoff frequency. The simulation results indicate that the appropriate cutoff frequency should be selected within the mid-range between 0 Hz and the frequency of the first harmonic order of active power. Furthermore, to validate the harmonic detection performance of the PQ theory, this study conducts Hardware-in-the-Loop (HIL) simulations for harmonic mitigation in AC electric railway systems. The results confirm that harmonic detection based on PQ theory, combined with the LPF designed using the proposed method, can accurately compute the reference current. Consequently, the APF can effectively inject compensating currents to mitigate harmonics. The Total Harmonic Distortion (THD) values of the three-phase source currents (iSA, iSB, iSC) are 3.51%, 3.53%, and 3.54%, respectively, which are all below the 5% limit specified in IEEE Std. 519-2022.

Forecasting Thai Orchid Exports Using Artificial Neural Networks and Grid Search Integrated with Empirical Mode Decomposition

Nathagorn Tosing — 2026-06-19

Accurate forecasting of orchid exports is essential for production planning and cold-chain logistics in Thailand, yet export time series are often nonlinear and non-stationary. This study develops a hybrid forecasting framework that integrates Empirical Mode Decomposition (EMD) with Artificial Neural Networks (ANN) and tunes the parameters using grid search. The proposed methodology decomposes monthly export signals into the First Intrinsic Mode Function (IMF1) and a residual component, transforming the original series into component-wise learning targets. Separate ANN models are designed for each component to learn distinct data characteristics independently. The framework is validated using 72 months of monthly export data (2019–2024) obtained from the Office of Agricultural Economics. Model performance is evaluated using a rolling-origin expanding-window approach, which refits the model with all information available at each origin and tests forecasts on subsequent observations to reflect practical deployment. The results show that the proposed EMD–ANN model outperforms benchmark models, namely ARIMA and Holt–Winters. For Test Set 1 with 18 observations, the proposed model achieves an accuracy of 88.36%, exceeding the alternative models. When the evaluation is extended by adding 12 additional observations, resulting in 30 test observations in total, the proposed model maintains a comparable accuracy of 88.19%, demonstrating stable performance as the test horizon increases. Therefore, the proposed model provides higher predictive accuracy than traditional approaches while maintaining consistent forecasting performance under expanded testing conditions. It can serve as a robust decision-support tool for exporters and policymakers to improve inventory management, mitigate supply-chain risk, and enhance resource allocation for temperature-controlled logistics.

Lightweight Bricks from Rice Husk Ash Mixed with Cement: A Possibility for Use as Alternative Construction Materials

Chaiyod Na Bangchang — 2026-06-12

Thailand is an agricultural country where rice husks are widely used as biomass fuel, resulting in a large quantity of rice husk ash as a byproduct. Therefore, this research proposes the production of novel lightweight bricks using rice husk ash obtained from brick firing at factories in Phra Nakhon Si Ayutthaya Province, mixed with cement, to evaluate its feasibility as an alternative construction material. For the experiment, the morphology and composition of rice husk ash were analyzed by SEM-EDS. The physical and mechanical properties of the bricks produced from rice husk ash (5–30% by weight) mixed with cement, including density, water absorption, and compressive strength, were then tested according to TIS 2601-2556 (Cellular lightweight concrete blocks using preformed foam). From the results, it was found that the rice husk ash after sieving for particle size reduction has a polygonal shape, an average particle size
of 14.66 μm, and a high silicon dioxide (SiO2) content of 83.45%. The lightweight bricks produced from rice husk ash mixed with cement at 25–30% by weight have a density in the range of 979–1,000 kg/m3, and a compressive strength of 48.5–58.3 kg/cm2. This indicates that the bricks are lightweight and have acceptable mechanical properties for aerated lightweight concrete block materials (TIS 2601-2556, C10), except for water absorption, which slightly exceeds the limit specified in the TIS 2601-2556 standard. These results show that lightweight bricks produced from rice husk ash mixed with cement containing more than 25% by weight of rice husk ash have adequate mechanical properties and the potential to be effectively developed and applied as aerated lightweight concrete blocks (C10 standard), with the addition of a foaming agent to adjust water absorption properties to meet the requirements of the construction industry. In addition, this research promotes the utilization of agricultural waste for value addition and supports the development of sustainable practices in the industry.

Classification of Pomelo Maturity Using Low-Cost Peel Feature Analysis and Machine Learning Algorithms

Ketsarin Chawgien — 2026-06-15

The lack of accurate pre-harvest maturity assessment tools is a critical challenge for pomelo production. Since pomelo is a non-climacteric fruit, its internal quality becomes fixed immediately upon harvest. Consequently, inaccuracies in harvest timing decisions directly result in economic losses. This study presents a cost-effective machine learning framework for maturity classification (early-mature vs. mature) of Khao Nampueng pomelos. A balanced dataset of 1,008 images from 70 pomelos was obtained using a Raspberry Pi camera at two critical times: 180 and 210 Days After Fruit Set (DAFS). To capture the non-linear relationships between peel appearance and maturity, color and texture features were analyzed using the Permutation Feature Importance method integrated with a Support Vector Machine (SVM) utilizing a Radial Basis Function (RBF) kernel. The proposed method yielded an optimized subset of biologically relevant features. The SVM-RBF model trained on this subset achieved robust classification performance, with a test accuracy of 95.6%. Interpretability analyses confirmed that the model’s decision boundary aligned with physiological and biological changes, identifying variation in the chromatic component (b* channel) and micro-texture (LBP) as the most significant predictors. This study demonstrates that cost-effective peel feature analysis driven by machine learning algorithms provides a high-precision tool for optimizing pomelo harvest timing.

Diabetic Retinopathy Stage Classification Using Deep Learning Techniques

Tassanee Hattiya — 2026-06-16

This research presents the development of a classifier for detecting the stages of Diabetic Retinopathy (DR) using Convolutional Neural Networks (CNNs), a deep learning–based classification technique. Five architectures, including DenseNet121, EfficientNetB0, EfficientNetB1, MobileNetV2, and NASNetMobile, were explored. A dataset of 5,000 retinal fundus images obtained from Kaggle was categorized into five stages of Diabetic Retinopathy. The research methodology consisted of two main steps: 1) data preparation and 2) model generation. Each CNN architecture was trained for 100 epochs. The experimental results indicated that EfficientNetB0 achieved the highest performance, reaching an accuracy of 99.25% on the training dataset. When evaluated on the test dataset, the model achieved an accuracy of 81.30% and a macro-average F1-score of 81.15%. Furthermore, the model demonstrated exceptional effectiveness in identifying high-risk stages, achieving a precision of 92.82% for the most severe stage of Diabetic Retinopathy and a recall of 89.00% for the severe stage. These findings suggest that the developed classifier has significant potential to support medical screening processes and effectively reduce the risk of vision loss in diabetic patients.

Two-Stage Unsupervised Hybrid Model for Silent Risk Detection in Domestic Violence

Wutthiphong Khuandin — 2026-06-19

Detecting latent risk groups for domestic violence is a challenging task because the available data are typically unlabeled. This study presents a two-stage hybrid unsupervised model designed to detect and identify latent high-risk groups by integrating the strengths of the Gaussian Mixture Model (GMM), Autoencoder (AE), and Local Outlier Factor (LOF). In the first stage, the model isolates clearly high-risk cases (59 out of 1,161 records) from the entire dataset. The second stage utilizes the remaining 1,102 records to identify latent high-risk groups through logical fusion strategies, including AND, OR, and weighted fusion. The clustering quality and structural validity are evaluated using the Silhouette Index, Calinski–Harabasz Index (CH), and Davies–Bouldin Index (DBI), along with statistical validation using the Kruskal–Wallis test and the chi-square test for categorical variables. The results indicate that the proposed GAAL model (GMM + AE + LOF with AND logic) achieves the highest Silhouette score and the lowest DBI value (2.8478), indicating superior cluster clarity and structural stability. The t-SNE visualization further confirms the distinct separation between the identified groups, consistent with the quantitative metrics. In addition, the variable Age is found to differ significantly among clusters based on the Kruskal–Wallis test. These findings demonstrate the structural quality and detection capability of the proposed hybrid approach in uncovering hidden risk patterns from unlabeled social data.

Enhancing Insider Threat Detection Using STRIDE Threat Modeling with Evolutionary Algorithms

Songpon Nakharacruangsak — 2026-06-12

Insider threats are among the most critical risks affecting the security of information systems, as insiders often possess legitimate access that makes detection by traditional methods challenging. This study aims to develop an effective detection approach by integrating the STRIDE threat modeling framework with evolutionary optimization techniques, namely the Genetic Algorithm (GA) and Differential Evolution (DE), to refine detection rules based on real behavioral data. Using the LANL dataset comprising 50,000 records with 15% insider threat instances, the proposed models were evaluated against widely used machine learning classifiers, including the Support Vector Machine (SVM), Random Forest (RF), and Multi-Layer Perceptron (MLP). Experimental results show that the baseline STRIDE model achieved an accuracy of 0.823 and an AUC of 0.832, while STRIDE combined with GA and DE significantly improved performance, reaching an accuracy of 0.878 with an AUC of 0.894, and an accuracy of 0.889 with an AUC of 0.902, respectively. These results outperform SVM (AUC = 0.875) and RF (AUC = 0.891), and are comparable to MLP (AUC = 0.895), while maintaining superior interpretability through rule-based modeling. The findings highlight the potential of integrating STRIDE with evolutionary optimization techniques to achieve accurate, flexible, and interpretable insider threat detection that is practical for real-world organizational environments.

The Green Food Industry as a Carbon Sink: Integrating Regenerative Innovations and Regulations for Net-Negative Emissions

Patchanee Yasurin — 2026-07-21

The roadmap for achieving net-zero emissions in global food systems by 2050 indicates that net-zero emissions can be achieved by integrating cost-effective technologies to reduce emissions from land-use change, improving rice and livestock production, and accelerating renewable energy use in food processing by 2040–2050 [1]. To meet the 1.5°C warming limit established by the Paris Agreement, the international community must commit to swift and far-reaching cuts in global greenhouse gas emissions. Achieving this benchmark requires a radical shift away from historical emission patterns, necessitating immediate, large-scale decarbonization efforts across all sectors of the global economy. By prioritizing aggressive mitigation strategies now, nations can work toward stabilizing

ปกวารสารวิชาการพระจอมเกล้าพระนครเหนือ

ปกวารสาร มจพ. วารสารวิชาการพระจอมเกล้าพระนครเหนือ — 2026-07-21