KMITL

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

Browse

Search Results

Now showing 1 - 10 of 46

Some of the metrics are blocked by your
consent settings
Photothermal solar assisted Madhuca diethyl ether fuel processing for LHR engines with AI-based performance and yield prediction
(2026-12-01)
Dubey, Rakesh
;
Prajapati, Ajeet Kumar
;
Bharadwaj, Shruti
;
Kamchoom, Viroon
;
Onyelowe, Kennedy C.
This study investigates the combustion, performance, and emission characteristics of biodiesel blends derived from Madhuca longifolia oil with diethyl ether (DEE) as an oxygenated additive in a diesel engine. Prior to fuel preparation, Fourier Transform Infrared (FTIR) analysis was conducted to verify the chemical composition of the extracted oil, confirming the presence of triglyceride structures and long-chain fatty acids characteristic of Madhuca longifolia oil. A solar-assisted preheating mechanism was incorporated during oil extraction to reduce energy consumption and improve yield consistency. The system was further integrated with a 250 Wp solar photovoltaic (PV) panel (efficiency ~ 17%, Voc = 37 V, Isc = 8.5 A, MPPT = 30 V/8 A) to power auxiliary loads such as the fuel metering unit, sensors, and control panel. This renewable integration enabled 100% solar contribution for auxiliary components, saving approximately 1.04 kWh/day of grid electricity and achieving an estimated reduction of about 151 kg of CO<inf>2</inf> emissions annually. Four fuel types were evaluated: Diesel, MB100 (pure biodiesel), MB20D80 (20% biodiesel, 80% diesel), and MB5DEE5D90 (5% biodiesel, 5% DEE, 90% diesel). Among these, MB5DEE5D90 demonstrated comparatively improved performance, showing an 8% increase in Brake Thermal Efficiency (BTE) and a 10% reduction in Brake-Specific Fuel Consumption (BSFC) compared with diesel. Emission analysis indicated reductions of approximately 20% in CO and 18% in HC emissions, while life-cycle assessment suggested around 40% lower combustion-phase CO<inf>2</inf> emissions. Heat release rate analysis indicated earlier and more efficient combustion behavior. Additionally, LSTM-based predictive modeling showed lower error margins compared with RNN, demonstrating improved prediction accuracy for engine performance parameters.
Some of the metrics are blocked by your
consent settings
Neural networks for neurocomputing circuits: A computational study of tolerance to noise and activation function non-uniformity when machine learning materials properties
(2025-12-01)
Thant, Ye min
;
Nukunudompanich, Methawee
;
Chueh, Chu Chen
;
Ihara, Manabu
;
Manzhos, Sergei
Dedicated analog neurocomputing circuits are promising for high-throughput, low power consumption applications of machine learning (ML) and for applications where implementing a digital computer is unwieldy (remote locations; small, mobile, and autonomous devices, extreme conditions, etc.). Neural networks (NN) implemented in such circuits, however, must contend with circuit noise and the non-uniform shapes of the neuron activation function (NAF) due to the dispersion of performance characteristics of circuit elements (such as transistors or diodes implementing the neurons). We present a computational study of the impact of circuit noise and NAF inhomogeneity in regression problems as a function of NN architecture and training regimes. We focus on one application that requires high-throughput ML: materials informatics, using as representative problem ML of formation energies vs. lowest-energy isomer of peri-condensed hydrocarbons, formation energies and band gaps of double perovskites, and zero point vibrational energies of molecules from QM9 dataset. We show that in these applications, NNs generally possess low noise tolerance with the model accuracy rapidly degrading with noise level. Single-hidden layer NNs, and NNs with larger-than-optimal sizes are somewhat more noise-tolerant. Models that show less overfitting (not necessarily the lowest test set error) are more noise-tolerant. Importantly, we demonstrate that the effect of activation function inhomogeneity can be palliated by retraining the NN using practically realized shapes of NAFs.
Some of the metrics are blocked by your
consent settings
Low-Cost System for Investigating a Small Motor Fault Classification Based on Current Signal
(2025-01-01)
Taweewat, Pat
;
Suwan-Ngam, Warachart
;
Songsuwankit, Kanoknuch
;
Konghuayrob, Poom
This research presents low-cost system for motor fault classification. This system uses a microcontroller with built-in ADC and communication capability. The two purposes of this article are to investigate the quality of the system for data acquisition and capability of the system for detecting early motor faults both by microcontroller on the system and personal computer. The embedded software on the system is designed to record current signals from a current sensor, compute FFT-based features and classify the fault based on tinyML method. Data communication between the system and the personal computer can be done by both serial port and TCP socket over Wi-Fi. The performances of the system and the personal computer are compared by the experiment as well as the quality of data recorded from the built-in ADC and a digital oscilloscope. The broken rotor bar and bearing fault in a 2.2kW induction motor are investigated. The classifier used in the experiment is a small feed forward neural network which can be implemented on both the proposed low-cost system and the personal computer. Although the recorded electrical current data by built-in ADC is contaminated with noise, the fault classification on the personal computer yield accuracy up to 90%.
Some of the metrics are blocked by your
consent settings
Comparison of Reduced-Length FFT-Based Feature for Induction Motor Fault Classification
(2025-01-01)
Taweewat, Pat
;
Suwan-Ngam, Warachart
;
Songsuwankit, Kanoknuch
;
Konghuayrob, Poom
This research presents a comparison of FFT-based features which can be used for classifying induction motor faults via neural network. In this paper, the misalignment and rotor bar damage faults are investigated by using stator current as input data only. As the length of the full FFT can include both informative data corresponding to the faults and uninformative data such as noise from environment or electrical supply, only relevant magnitude from FFT bins should be selected and used instead. This paper proposed to use threshold level determined from the magnitude of FFT bins in dataset as a criterion for the selection. From experimental results, an input feature vector created by proposed method can create short input feature vector length to be used by neural network efficiently. The trained neural network performs classification task at 99.98% in accuracy. Comparing to using dimension reduction by PCA, thresholding method needs basic computation, and yields result close to PCA method.
Some of the metrics are blocked by your
consent settings
Machine learning approach to predict the strength of concrete confined with sustainable natural FRP composites
(2024-07-01)
Ali Talpur, Shabbir
;
Thansirichaisree, Phromphat
;
Poovarodom, Nakhorn
;
Mohamad, Hisham
;
Zhou, Mingliang
Recent earthquakes have highlighted the need to strengthen existing structures with substandard designs. NFRPs provide a sustainable, cost-effective alternative for strengthening, but accurately predicting their performance remains a challenge. This study investigates the use of machine learning algorithms for predicting the compressive strength concrete specimens confined with various NFRPs. Four algorithms were employed: decision tree, random forest, neural network, and gradient boosting regressor. A diverse dataset encompassing various geometries, material properties, and confinement configurations was used to train and evaluate the models. Gradient boosting regressor (GBR) achieved the highest performance, with an average R-squared value of 0.94 and low mean absolute error (MAE) and root mean squared error (RMSE) during training and k-fold cross-validation. Neural network and random forest also demonstrated satisfactory performance, with average R-squared values of 0.88 and 0.86, respectively, during cross-validation. These results suggest that machine learning holds promise for predicting the compressive strength of concrete confined with NFRPs. GBR offers the most accurate predictions, making it a valuable tool for engineers seeking to optimize the design and performance of strengthened structures using sustainable materials.
Some of the metrics are blocked by your
consent settings
Atmospheric precursors from multiple satellites associated with the 2020 Mw 6.5 Idaho (USA) earthquake
(2024-01-01)
Qasim, Muhammad
;
Shah, Munawar
;
Shahzad, Rasim
;
Jamjareegulgarn, Punyawi
Remote sensing has became a powerful tool for identifying lithosphere and atmosphere anomalies associated with the impending Earthquakes (EQs) in the vicinity of seismic breeding zone. In this paper, Land Surface Temperature (LST) of both the daytime and nighttime from the Moderate Resolution Imaging Spectroradiometer (MODIS) along with Air Temperature (AT), Relative Humidity (RH), Air Pressure (AP) and Outgoing Longwave Radiations (OLR) are studied for the 2020 Idaho (USA) EQ of Mw 6.5. We found the EQ induced surface and atmospheric parameters anomalies just one day prior to EQ main shock using statistical analysis. Moreover, we observed a sharp increment in LST and AT followed by the drop in both AP and RH, which are responsible for cooling the hot gases emitted from epicenter during preparation period. Also, we observed a large increase in OLR on the same day confirming these anomalous variations to be related with the main shock. Furthermore, these abrupt variations are also confirmed using neural networks (nonlinear autoregressive network with exogenous inputs (NARX), multilayer perceptron (MLP) and continuous wavelet transformation (CWT). These multi-parameter and multi-technique analyses can contribute to assist in the main shock forecasting in future with an enhanced cluster of satellite observations.
Some of the metrics are blocked by your
consent settings
INTELLIGENT RECOGNITION OF PHYSICAL EDUCATION CURRICULUM RESOURCES BASED ON DEEP NEURAL NETWORK AND THE GAME MODEL STUDY
(2023-01-01)
Yang, Xuelin
;
Sumettikoon, Piyapong
;
Wu, Xiang
Nowadays, withmore and more physical education curriculum resources, schools or teachers have more and more choices for physical education curriculum resources. However, because some teachers need a deep understanding of curriculum training programs and standards, the selected curriculum resources cannot promote their curriculum development. This paper puts forward the researchon the intelligent recognition and gamemodelof physical education curriculum resources based on neural networks. The specific research conclusions are as follows: The intelligent consciousness and movement model of physical education curriculum resources based entirely on the technical knowledge of the BP neural community and deepneuralcommunityare proposed. WiththehelpofMATLAB7.1 neuralnetwork toolbox to implement the specificrecommendation system, a three-layerBPnetwork is established, and the NEWFF function is used to create the neural network. Useful resources in each direction generate a directionrecognition vector according to the route guidancestandard, calculate the course recommendation degree according to selection statistics and scoring, and input the courseresourcerecognitionvector and recommendationdegreeinto theneuralnetwork. When the number of hidden layer nodes is 10, and the learning training algorithm selects the L-M optimization algorithm, the error between the actual output and the expected output of the network meets the requirements. It shows that the accuracy of the recommendation model meets the requirements; that is, the relationship between the recognition vector of physical education course resources and the recommendation degree of course resources reflected by the neural network basically reflects the functional relationship between them and the model can be used to make corresponding recommendations.
Some of the metrics are blocked by your
consent settings
An Advanced Mixed Methodology Model for Teaching of Physical Education in the Post Covid-19 Era: A Case Study on Junior Middle School Basketball Class
(2022-01-01)
Yang, Xuelin
;
Sumettikoon, Piyapong
Objective: The study's primary purpose is to examine the effectiveness of physical education instruction post-covid-19, focusing on the efficacy of basketball instruction in enhancing teacher-student interaction. Methodology: The research utilized an advanced mixture methodology. Compared to traditional blended instruction, our method considers the students' demand for basketball courses and the post-COVID-19 learning environment. Results: In the post-COVID-19 era, our hybrid online and offline teaching method may be effective in middle school physical education, as demonstrated by our analysis and test case. Students are more satisfied with this method of instruction, and their enthusiasm for acquiring physical education knowledge increases, resulting in enhanced physical education in junior high schools. We examine basketball training for junior high school students and propose a hybrid online/offline training approach to improve effectiveness and satisfaction. Our hybrid teaching model, ADDIE, is designed to accomplish a combination of online and offline instruction and utilizes a neural network to detect errors in students' feedback videos and increase the efficiency of identifying wrong actions. In the final stage of case verification, experimental results indicate that the students' satisfaction with our method is 8.9, showing its potential to meet the requirements of junior middle school physical education instruction after COVID-19. Implications: The study will aid policymakers and educators in formulating post-covid-19 policies regarding the efficacy of physical education instruction. Novelty: The study is among the first to examine the efficacy of physical education instruction in the post-covid era.
Some of the metrics are blocked by your
consent settings
Automatic para-rubber trees classification in Thailand from LANDSAT-8 imagery using DCED neural network
(2021-06-02)
Supunyachotsakul, C.
;
Suksangpanya, N.
Image classification has been one of the main processes performed on satellite imagery where there have been many studies attempt to developing an automatic approach since manual classification is known to be a time-consuming process. Automatic classifying specific types of vegetation in satellite imagery has been a challenging field of study. In this work, a neural network method, specifically deep convolutional encoder-decoder or known as DCED, is applied to the automatic classifying coverage area of Para-rubber trees in Thailand from LANDSAT-8 satellite imagery. The main procedure in this work comprises ground truth preparation, suitable model development with training and validation, and, finally, classification. As a result, the model developed in this work gives average accuracy of 86.9% in training datasets and 70.9% in validation datasets.
Some of the metrics are blocked by your
consent settings
Pig Carcass Assessment on Image Segmentation
(2021-01-01)
Tanthong, Jitpanu
;
Moodleah, Samart
;
Vittayakorn, Sirion
Pork is the most commonly consumed meat across the world: about one-third of all meat consumed is pork, ahead of beef and chicken. Every day a massive number of pig carcasses enter the production pipeline. When entering the pipeline, the carcasses are graded by the slaughterhouses to determine the market price of the meat. The grading criteria could depend on a variety of factors such as the tenderness, color, pH value, water holding capacity as well as the proportion of red meat inside the carcasses. Since the grade can be used to determine the market price and the commercial usage of the meat, this process is crucial. Unfortunately, the grading process is not only time consuming but also requires expertise. To mitigate this problem, in this work we propose: 1) a pig carcass image dataset segmented by experts, 2) an LSQ index image dataset and 3) an algorithm for carcass quality analysis based on the ratio of red meat inside the carcasses, or the Lenden-Speck-Quotient (LSQ). Our experimental results demonstrate that the performance of the proposed LSQ index algorithm is reliable and agrees with experts' annotation with MAPE 5.55%.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results