KMITL

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

Browse

Search Results

Now showing 1 - 10 of 13

Some of the metrics are blocked by your
consent settings
An operational machine learning framework for calibrating COSMIC radio occultation TEC to ground-based GNSS-derived TEC
(2026-03-15)
Okoh, Daniel
;
Habarulema, John Bosco
;
Nava, Bruno
;
Cesaroni, Claudio
;
Baki, Paul
The Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) provides global Radio Occultation (RO) measurements of ionospheric total electron content (TEC), but these values are systematically underestimated relative to ground-based Global Navigation Satellite System (GNSS)-derived TEC due to the exclusion of the plasmaspheric contribution. This study presents a machine learning calibration framework that transforms COSMIC TEC into GNSS-equivalent values. Using co-located COSMIC and GNSS observations from 2006 to 2025, we developed neural network models (ROTEC-A and ROTEC-B) trained on (19 and 22) input features respectively, including COSMIC profile parameters, spatiotemporal descriptors, and optionally, solar and geomagnetic activity indices. Results show that the calibration effectively mitigates systematic underestimation, reducing mean bias from 6.97 TECU (uncalibrated COSMIC) to near zero (0.02–0.03 TECU). The calibrated products also substantially reduce skewness in residuals, yielding nearly symmetric error distributions suitable for data assimilation. Across various latitudinal, local time, and seasonal sectors, mean absolute errors were reduced by 50–75%, with the best performance at mid-latitudes and slightly elevated errors in high-latitude and equatorial regions. Although, the inclusion of solar and geomagnetic indices yielded marginal improvements, statistical tests confirmed no significant advantage over the baseline model. The operationally oriented framework outputs calibrated GNSS-equivalent TEC in near real-time, providing enhanced ionospheric monitoring capability, especially over GNSS-sparse regions such as oceans and deserts. These results demonstrate the potential of COSMIC RO data, once calibrated, to serve as a reliable complement to GNSS observations for ionospheric research, space weather monitoring, and operational applications.
Some of the metrics are blocked by your
consent settings
Monitoring Pesticide Residue on Bok Choi using Convolution Neural Network with NIR spectral Data
(2024-12-29)
Lapcharoensuk, Ravipat
;
Fhaykamta, Chawisa
;
Anurak, Watcharaporn
;
Chadwut, Wasita
Deep learning (DL) has been applied in agriculture, especially quality control in agricultural processing. One key area of interest is the detection and monitoring of pesticide residues in crops. The most popular measurement tool for nondestructive monitoring of pesticide residues is near-infrared spectroscopy (NIRS). A combination of CNN model with NIR spectral data was developed for monitoring pesticide residue on bok choi. The NIR spectral of bok choi with and without pesticide residue (chlorpyrifos) was collected in wavelength range between 908 and 1676 nm. A simple structure of CNN was modified for a one-dimensional task and this deep learning architecture was trained for classification of the bok choi samples. The results showed prefect prediction with 100% accuracy, precision, recall and specificity. This study also found that deep learning for NIR spectroscopy data requires less processing than traditional machine learning while still achieving great results.
Some of the metrics are blocked by your
consent settings
ACCURATE STATE OF CHARGE ESTIMATION OF LITHIUM-ION BATTERY USING RECURRENT AND NON-RECURRENT NEURAL NETWORKS FOR WLTP DRIVING PROFILES
(2024-12-01)
Praisan, Akkarawat
;
Polmai, Sompob
;
Kittiratsatcha, Supat
Estimating the state of charge (SoC) of a battery is essential to maximize its performance and ensure reliable operation and battery life. Nowadays, many countries are increasingly adopting electric vehicles (EVs) with lithium-ion batteries due to their high specific energy and long service life. This paper presents a method for estimating the state of charge of lithium-ion batteries using artificial neural networks, specifically the Feedforward Neural Network (FNN) and Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM), through a data-driven approach. The training and testing of the networks are conducted using recorded datasets of the battery, based on the WLTP driving profiles class 2 and class 3. These driving profiles are specifically designed for testing electric vehicles, thereby enhancing the realism of the state of charge estimation by the network. In terms of the analytical aspect, the FNN was able to train the network faster due to its simpler structure, requiring less computation. On the other hand, the LSTM demonstrated more accurate SoC estimation with fewer response oscillations, thanks to its ability to learn and adapt network parameters internally.
Some of the metrics are blocked by your
consent settings
DEVELOPMENT OF AN EMBEDDED EXERCISE POSTURE PREDICTION SYSTEM FOR OFFICE SYNDROME USING MACHINE LEARNING
(2024-12-01)
Sisuk, Noppadon
;
Onkhong, Tada
;
Kaewpoonsuk, Anucha
;
Wardkein, Paramote
;
Prompak, Kriangsak
Office syndrome is a condition with a high prevalence among people of working age and an upward tendency. In this study, a wearable device implant based on machine learning will be created for exercise of office syndrome. Developing a machine learning model that can distinguish 3 exercise postures for implantable devices was the objective of the study. The Edge Impulse online software was used for data collecting, preprocessing, training, and testing. After that, the model was successfully installed and tested on an Arduino Nano 33 BLE microcontroller. In addition to designing embedded systems, this work also developed applications through the App Inventor program for exercising properly. From testing the system, the model showed an average classification accuracy of 97.6% of exercise postures. The results of testing the embedded system in conjunction with the presented application on 10 subjects showed that the system was able to provide predictive accuracy for wrist exercises, shoulder exercises and stretch exercises with 95%, 97%, and 93%, respectively.
Some of the metrics are blocked by your
consent settings
Decentralized control of cooperative robotics system using neural networks for the purpose of an object balancing flat plate tasks
(2020-02-01)
Jaisumroum, Nattapon
;
Chotiprayanakul, Pholchai
;
Sriseubsai, Wipoo
In this article, the manipulation to handle the object on a plate using neural networks, design of 1-DOF robot arm under cooperative control, will be explained. The robot's system specifies the object position and velocity, an assignment oriented components for cooperative control. The novelty of this experiment is that under the decentralized control, the robot estimates the position and speed of the object for control end-effector of robot arm using the camera to track position and speed of the object according to training and assigned collaborative tasks which differ from other experiments that use sensors. The experiment includes three robot manipulators which were capable balancing the objects on flat plate with dataset to training and control servo motor assigned to the corresponding position and the end-effector in decentralized to control robotics. Overall, neural network method can be a training scheme using a cooperative robotics in a decentralized control.
Some of the metrics are blocked by your
consent settings
Investigation of deep learning optimizer for water pipe leaking detection
(2019-07-01)
Arunsuriyasak, Peerachai
;
Boonme, Phattraporn
;
Phasukkit, Pattarapong
Nowadays, Deep learning plays an important role in complex problems. Thus, one of important algorithm part is an optimizer. This paper aims to improve algorithm using optimizers. Adam optimizer, a powerful and effective optimizer, was used to adjust parameters in Deep Neural Networks model. Which, object datasets consist leaking water pipe, non-leaking water pipe are used to classify 2 object labels. Nevertheless, RMSprop and Adadelta are alternative optimizers that can be used in Deep Neural Network. Other than that, this experiment has been shown Adam gave an accuracy at 98.973% for leaking water pipe and 97.466% for non-leaking water pipe. While, Adadelta gave 76.755% and 70.448%. And RMSprop gave 98.973% and 97.466%.
Some of the metrics are blocked by your
consent settings
License plate identification from Myanmar vehicle images under different environmental conditions
(2018-01-01)
Khin, Ohnmar
;
Phothisonothai, Montri
;
Choomchuay, Somsak
We have developed a license plate identification method for Myanmar vehicles that are captured under dissimilar conditions, e.g., angle of image capturing, different types of license plates, and real environmental conditions. In this study, car license plate recognition (CLPR), bounding box, horizontal and vertical dilations, skew angle detection, and plate detection were proposed to identify license numbers from different vehicle images. To recognize the characters, a new algorithm based on deep learning, a subset of artificial intelligence (AI), is proposed. The neural nets are progressing rapidly in many fields. The applied model of neural network is used for classification. The recognition part is a very challenging task. Compared with the traditional method, the neural network has obvious advantages. The benefit of this research is to eliminate the need of license plate recognition (LPR) under different conditions. In mobile phones, there are many sensors used to detect the presence of nearby objects. Accelerometers in mobile phones are used. Developed for the Samsung mobile phone, sensors can yield sensor readings but it not much else. Each car was viewed from four different angles under different conditions. In our experiment, the results showed an average accuracy of 97%, which was substantially applied to license plate identification under different environmental conditions. To extend the experiment, the vehicle images were also collected under different conditions, such as dark and cloudy weather and various sizes and positions of plates.
Some of the metrics are blocked by your
consent settings
Analysis of electrical losses in transformers using artificial neural networks
(2014-01-01)
Suttisinthong, N.
;
Pothisarn, C.
This paper proposes a technique to analysis electrical losses in distribution transformers 1-phase 30 kVA using of back-propagation neural networks (BPNN). Experimental data at various temperature of transformers obtained from manufacturer, are employed as an input pattern for BPNN while output pattern which corresponding to total losses in transformers. The total number of test set are 150 sets in order to verify the validity of the proposes technique. The results show that average accuracy obtained from the proposes technique gives satisfactory accuracy.
Some of the metrics are blocked by your
consent settings
Selection of proper activation functions in back-propagation neural network algorithm for single-circuit transmission line
(2014-01-01)
Suttisinthong, N.
;
Seewirote, B.
;
Ngaopitakkul, A.
;
Pothisarn, C.
This paper proposes an appropriate activation function for the fault classification decision algorithm. The decision algorithm based on the hybrid of discrete wavelet transform (DWT) and back-propagation neural network (BPNN) has been proposed to classify the fault type. The DWT is employed to decompose high frequency component of current signals. The maximum coefficient from the first scale at 1/4 cycle of phase A, B, and C of post-fault current signals and zero sequence current obtained by the DWT have been used as an input variable in a decision algorithm. The activation functions in each hidden layer and output layer have been varied, and the results obtained from the decision algorithm have been investigated with the variation of fault inception angles, fault types, and fault locations. The results have illustrated that the use of Hyperbolic tangent sigmoid function in the first and the second layers with Linear function in the output layer is the most appropriate scheme for the transmission system.
Some of the metrics are blocked by your
consent settings
Improvement of single wavelength-based Thai jasmine rice identification with elliptic Fourier descriptor and neural network analysis
(2012-12-01)
Suwansukho, Kajpanya
;
Sumriddetchkajorn, Sarun
;
Buranasiri, Prathan
Instead of considering only the amount of fluorescent signal spatially distributed on the image of milled rice grains this paper shows how our single-wavelength spectral-imaging-based Thai jasmine (KDML105) rice identification system can be improved by analyzing the shape and size of the image of each milled rice variety especially during the image threshold operation. The image of each milled rice variety is expressed as chain codes and elliptic Fourier coefficients. After that, a feed-forward back-propagation neural network model is applied, resulting in an improved average FAR of 11.0% and FRR of 19.0% in identifying KDML105 milled rice from the unwanted four milled rice varieties. © Copyright SPIE.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results