KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 10 of 22

Some of the metrics are blocked by your
consent settings
Interpretable ANN-Based Computer Vision System for Mangosteen Ripeness Detection for Export Markets
(2026-01-21)
Lapcharoensuk, Ravipat
;
Tosribunjerd, Naphon
;
Poonpakdee, Pasu
Mangosteen is a high-value tropical fruit widely consumed and exported from Thailand. Mangosteen ripeness classification is crucial for export quality control, but manual grading leads to inconsistency and inefficiency. This study presents a computer vision system using an Artificial neural network to classify mangosteen into ripe, semi-ripe, and unripe stages based on peel color. A dataset of 378 images was collected and processed to extract 40 color-based features across multiple color spaces. Principal Component Analysis demonstrated non-linear separability among the ripeness classes. SMOTE and Gaussian noise augmentation were used to tackle data imbalance and enhance generalizability. The model reached a 95% accuracy rate and displayed flawless precision and recall for the ripe class. Integrated Gradients analysis highlighted the importance of the red-green color component (CIELAB a*) in the classification process. The proposed method demonstrates a low-cost, interpretable, and efficient solution suitable for real-world application in the mangosteen export industry.
Some of the metrics are blocked by your
consent settings
Experimental investigation and ANN prediction of heat transfer enhancement in a heat exchanger tube utilizing twin corrugated twisted tapes
(2025-12-01)
Du, Y.
;
Wongcharee, K.
;
Thianpong, C.
;
Chuwattanakul, V.
;
Chamoli, S.
This report introduces a novel twin-corrugated twisted tape (TC-TT) insert designed to enhance heat transfer in exchanger tubes. The key innovation lies in the twin-corrugated structure, which generates a twin-swirl flow effect. The corrugated surface synergistically increases flow disturbance and expands the effective heat transfer area. The studied parameters were twist ratios (y/w = 3.0, 3.5, and 4.0) and corrugation angles (θ = 45°, 60°, 75°, and 90°) at 6,000 ≤ Re ≤ 20,000. The results show that using twin-corrugated twisted tapes increases the average Nusselt number by roughly 60–135% compared to a plain tube and by 16–35% compared to a conventional single-twisted tape, confirming the effectiveness of this structural modification. This enhancement is primarily due to the combination of double swirling-flows and enhanced effective heat transfer generated by the corrugated surface. Reducing the corrugation angle (θ) and twist ratio (y/w) led to increases in the Nusselt number (Nu), friction factor (f), and thermal performance factor (TPF). Within the studied range, the Nusselt number, friction factor, and thermal performance factor reached maximum values of 5.18, 0.153, and 1.44, respectively, at a twist ratio of 3.0, a corrugation angle of 45°, and Re = 6,000. Regression analysis was utilized to develop correlations for the Nu and f, considering the Re, Pr, y/w, and θ as influencing variables. The proposed correlations for predicting the friction factor and Nusselt number have errors within ±3% and ±2%, respectively. In addition, an artificial neural network (ANN) was developed for predicting the thermal performance values occurring below the experimental study range. The optimal state ANN model shows remarkable prediction accuracy with R2 of 0.965.
Some of the metrics are blocked by your
consent settings
Application of foam-mat drying to produce field crab powder: Foaming process optimization, drying kinetics, and final product characterization
(2025-08-01)
Thuy, Nguyen Minh
;
Nhut Minh, Ngo Ngoc
;
Kha, Nguyen Hoang
;
Bich Thuy, Bui Thi
;
Giau, Tran Ngoc
This study used foam-mat drying to make powder from field crab meat for the first time. In which, the effect of foaming conditions [egg albumin (EA, 7.96–16.44 %) and xanthan gum (XG, 0.04–0.44 %)] and drying temperature (65–80 °C) on powder quality were examined. With appropriate EA and XG levels of 13.16 % and 0.30 %, foam density and foam expansion were 0.395 g/mL and 279.78 %, respectively. The total energy required and specific energy consumption decreased. In contrast, thermal efficiency and energy efficiency rose with drying temperature, reaching 2.494–4.452 %, and 1.419–1.879 %, respectively. Temperature affects effective moisture diffusion coefficient according to the Arrhenius equation with an activation energy of 39.33 kJ/mol. Fitting experimental data to seven thin-layer drying models and an ANN model. The Aghbashlo model scored best, with the highest correlation coefficient. Nevertheless, the ANN model demonstrated slightly superior accuracy compared to the Aghbashlo model, indicating the potential for industrial system control.
Some of the metrics are blocked by your
consent settings
Developing a novel artificial model to predict the foaming properties and β-carotene content of lucuma (Pouteria lucuma) during foam-mat drying and process optimization
(2024-12-01)
Thuy, Nguyen Minh
;
Duong, Le Thi Thuy
;
Giau, Tran Ngoc
;
Hao, Hong Van
;
Minh, Vo Quang
Drying fruit puree by the foam drying method has become popular due to its simplicity, low cost, short drying time, and low thermal degradation. The objective of the study was to investigate the effect of foaming conditions on foam properties (foam expansion, foam density) and content of β-carotene in lucuma powder using Box-Behnken design (BBD) with 3 factors and 3 levels, including water:lucuma ratio (1:1–3:1), egg albumin concentration (EA, 5–15 %), and xanthan gum (XG, 0.1–0.3 %). Response surface methodology (RSM) and artificial neural network (ANN) were used for model establishment. The results showed that as the EA increased, the foam volume increased significantly, while the foam density decreased. The ANN-coupled BBD model structure of 3–10-3 demonstrated a high level of accuracy in predicting the impact of foaming formulation on responses, with a coefficient of determination exceeding 0.99. The optimal conditions by stimulation multiple-objective RSM for lucuma foam-mat drying were achieved with a water:lucuma ratio, EA, and XG of 2.53:1, 10.8 %, and 0.22 %, respectively. Based on these ideal conditions, the foam density, foam expansion, and β-carotene content of the dried powder were found to be 0.23 ± 0.04 g/mL, 228 ± 2 %, and 237.1 ± 0.1 μg/g, respectively. The obtained experimental values were very close to the model-predicted results, with very low differences identified when the validation was performed. These findings provide information for controlling the drying process using an artificial model and further applying lucuma powder in various fields in the food industry.
Some of the metrics are blocked by your
consent settings
Artificial intelligence optimization for producing high quality foam-mat dried tomato powder and its application in nutritional soup
(2024-12-01)
Thuy, Nguyen Minh
;
Giau, Tran Ngoc
;
Hao, Hong Van
;
Minh, Vo Quang
;
Tai, Ngo Van
The present work aims to investigate the effect of foam-mat drying on drying rate and lycopene content of tomato powder using a three-level Box-Behnken experimental design of Response Surface Methodology (RSM). Three process parameters included egg albumin (EA) ranging from 3 to 9 % as a foaming agent, carboxymethyl cellulose (CMC) from 0.2 to 0.6 % as a foam stabilizer and drying temperatures (60, 70, and 80oC). The responses measured drying rate (DR) and lycopene content, which are the indication of drying process and product quality. Optimization of drying process using RSM and artificial neural network coupled genetic algorithm (ANN-GA) models has been also investigated. Foam mat dried tomato powder produced under optimal conditions was then used to prepare a nutritious soup powder with 4 designed recipes with other nutritious ingredients. The results showed that the ANN-GA model (network structure of 3-10-2) could predict and optimize better than the RSM model. The optimal conditions for foam-mat drying process were EA of 6.67 %, CMC of 0.381 %, and drying temperature of 70.6oC. These gave the DR and lycopene to be 3.004 g water/g dry matter/min and 392.8 μg/g, respectively. Validation optimal condition was performed and showed that the experimental values obtained were greatly close to the predicted values. From the 4 designed soup formulas, the macronutrient content in formula F2 met the range for AMDR with the percentage of calories from protein, lipid, and carbohydrate being 21.28 %, 20.28 %, and 58.44 %, respectively. It was proven that ANN-GA is a more reliable and robust predictive modelling tool for foam-mat tomato powder production optimization compared to RSM model. Also, the promising application of tomato powder in nutritious soup production also was shown in this study, which could further research in larger scale.
Some of the metrics are blocked by your
consent settings
Prediction of the germination rate and antioxidant properties of VD20 Rice by utilizing Artificial neural network-coupled response surface methodology and product characterization
(2024-10-01)
Loan, Le Thi Kim
;
Tat, Truong Quoc
;
Minh, Pham Do Trang
;
Thao, Vo Thi Thu
;
Hoang, Pham Thi Minh
The current research aims to predict and optimize process conditions to produce germinated VD20 with a high rate of germination and antioxidant properties. Box-Behnken design (BBD) was used to build models with three factors [soaking time (ST: 4–6 h), germination time (GT: 18–22 h), and germination temperature (33–37 °C)] and three replications. The data set from the BBD experiment was used to create an artificial neural network (ANN) model that estimated the change in responses by variable factors. The ANN model was extremely accurate, with an overall correlation coefficient (R) of 0.9997 and showed the best fit with actual and predicted data. The germination conditions were further optimized using multi-objective RSM, which revealed that the optimal conditions were ST of 5.34 h, GT of 20.78 h, and germination temperature of 35.6 °C. The experimental validation revealed a high level of agreement between the results of the BBD models forecasted and the actual experimental values. The ANN-coupled BBD methodology is a promising hybrid method for modeling, forecasting, and optimizing the impact of process conditions on the quality of germinated grain. In addition, the raw and germinated VD20 rice were analyzed for their techno-functional properties, estimated glycemic index (eGI), and FTIR. Lower peak viscosity, values of breakdown and setback, and phytic acid were found after rice was germinated. The result revealed that high antioxidant content and activity, which were confirmed by the FTIR pattern, and low digestion behaviors (eGI = 64.23) were the attributes of the germinated product. Furthermore, the results of pasting, thermal, swelling power, and solubility showed the wide range of further application of this material, which should receive more consideration in future research.
Some of the metrics are blocked by your
consent settings
Ability of near infrared spectroscopy to detect anthracnose disease early in mango after harvest
(2024-08-01)
Seehanam, Pimjai
;
Sonthiya, Katthareeya
;
Maniwara, Phonkrit
;
Theanjumpol, Parichat
;
Ruangwong, Onuma
Determining anthracnose-infested mango can involve laborious and time-consuming assays, resulting in delayed postharvest management and decreased fruit marketability. Near infrared spectroscopy (NIRS) is proposed to detect the fungus in fully matured ‘Namdokmai Sithong’ mango. Inoculation of Colletotrichum gloeosporioides (1 × 106 conidia/mL) was artificially made onto one side of the fruit’s peel at the center of mango fruit while the other side was left intact. Interactance measurements were conducted at both inoculated and intact locations for 104 mango samples every 24 h until anthracnose symptoms visibly appeared. The classification approaches included a partial least squares discriminant analysis (PLS-DA) and a conventional artificial neural network (ANN). Results of our study revealed increased absorbance values corresponding with days after inoculation. Relatively high classification accuracies were obtained from all chemometrics approaches (˃ 89%). In the early hours after inoculation (24 h), the best classification result was obtained from the ANN model (98.1%), confirming that early detection was possible. Applications of PLS-DA and ANN are discussed.
Some of the metrics are blocked by your
consent settings
Equatorial spread-F forecasting model with local factors using the long short-term memory network
(2023-12-01)
Thammavongsy, Phimmasone
;
Supnithi, Pornchai
;
Myint, Lin Min Min
;
Hozumi, Kornyanat
;
Lakanchanh, Donekeo
The predictability of the nighttime equatorial spread-F (ESF) occurrences is essential to the ionospheric disturbance warning system. In this work, we propose ESF forecasting models using two deep learning techniques: artificial neural network (ANN) and long short-term memory (LSTM). The ANN and LSTM models are trained with the ionogram data from equinoctial months in 2008 to 2018 at Chumphon station (CPN), Thailand near the magnetic equator, where the ESF onset typically occurs, and they are tested with the ionogram data from 2019. These models are trained especially with new local input parameters such as vertical drift velocity of the F-layer height (Vd) and atmospheric gravity waves (AGW) collected at CPN station together with global parameters of solar and geomagnetic activity. We analyze the ESF forecasting models in terms of monthly probability, daily probability and occurrence, and diurnal predictions. The proposed LSTM model can achieve the 85.4% accuracy when the local parameters: Vd and AGW are utilized. The LSTM model outperforms the ANN, particularly in February, March, April, and October. The results show that the AGW parameter plays a significant role in improvements of the LSTM model during post-midnight. When compared to the IRI-2016 model, the proposed LSTM model can provide lower discrepancies from observational data. Graphical Abstract: [Figure not available: see fulltext.].
Some of the metrics are blocked by your
consent settings
Artificial Intelligence based Faults Identification, Classification, and Localization Techniques in Transmission Lines-A Review
(2023-12-01)
Kanwal, Shazia
;
Jiriwibhakorn, Somchat
An overview of the many methods used for fault detection, classification and location in the power system, particularly in transmission lines, is provided in this review, it also includes an experimental result of adaptive neuro-fuzzy inference system -based fault detection , fault classification and fault location. Being in operation outdoor environment, transmission lines are more vulnerable to various faults which may lead to system collapse in severe cases. Therefore, to ensure the reliable and safe operation of power system it is imperative to critically monitor the faults in transmission lines. In this regard, researchers around the globe have developed several techniques and constantly putting efforts to further improve the protection efficacy. The brief yet thorough analysis and comparison of the artificial intelligence-based techniques, hybrid methodologies and most recent approaches in the context of power system faults have been discussed and presented. In addition, the research work and the experimental results of an adaptive neuro-fuzzy inference system-based techniques have also been discussed for IEEE-9 bus system. The mean square error for testing data of ANFIS-based fault detection, classification, is zero and for fault location Mean square error is 5.32km. This piece of work could be helpful in the development of a comprehensive understanding of various artificial intelligence-based techniques within the realm of fault detection, classification and localization in transmission lines.
Some of the metrics are blocked by your
consent settings
Fast and Effective Technique in Evaluation of Lightning Impulse Voltage Parameters
(2021-01-01)
Yutthagowith, Peerawut
;
Kitwattana, Krit
;
Kunakorn, Anantawat
This paper presents an approach for the waveform parameter evaluation of lightning impulse voltage in high voltage tests according to the IEC standards. Such waveform parameters are composed of peak voltage (U<inf>p</inf>), front time (T<inf>1</inf>), time to half (T<inf>2</inf>), and the overshoot rate (B<inf>e</inf>). An artificial neural network with a back-propagation learning algorithm was applied to determine a base curve and its parameters from 14 points along the recorded waveform between 20% of the peak voltage on the wave front part to 40% of the peak voltage on the wave tail part. The 29 waveforms recommended by the standard were used in the training process of the development of the network model, and some experimental cases were also utilized for verification of the proposed method. It is found that the waveform parameters evaluated by the proposed approach are in the tolerances of the standard requirements. Maximum absolute deviations of U<inf>p</inf>, T<inf>1</inf>, T<inf>2</inf>, and B<inf>e</inf> are 0.06%, 2.00%, 0.12%, and 0.79%, respectively. Due to that no iteration process in the proposed approach is required, the efficiency in calculation process is significantly faster than the standard recommended approach.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results