KMITL

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

Browse

Search Results

Now showing 1 - 10 of 11

Some of the metrics are blocked by your
consent settings
Integration of discriminant analysis with Artificial neural networks to decision analytic framework for enhancing automated visual IC inspection accuracy
(2026-05-15)
Kiatcharoenpol, Tossapol
;
Klongboonjit, Sakon
This study aims to enhance the accuracy and reliability of automated visual inspection (AVI) in semiconductor manufacturing by integrating Linear Discriminant Analysis (LDA) and an Optimization Layer by Layer Neural Network (OLLNN). Unlike prior LDA-ANN hybrid approaches that primarily emphasize classification accuracy, this study formalizes a decision-analytic inspection loop that explicitly links discriminant centroids, feasible lighting regions, surrogate nonlinear modeling, and production level validation. A two-stage decision analytic framework is developed. In the first stage, LDA is applied to classify and identify discriminant boundaries and centroids between acceptable and defective image features under three lighting setups: coaxial ring, high ring, and low ring lights. In the second stage, OLLNN is trained using these features to capture nonlinear dependencies between greyscale intensity and lighting parameters, and then a surface response plot is used to ease the optimal parameter selection. The integrating model is validated using experimental IC marking inspection data to evaluate improvements in accuracy, especially false positive rates (Type I error). It was found that for the validation state, the false positive rates are reduced from 5.8% to below 4.6%, and classification accuracy improves significantly across variable illumination conditions. After implementation in mass production, the yield is increased to 99.6% with zero false positive found. This significant development of the integrating model enhances a foundation for adaptive, data-driven control of AVI parameters in smart factory environments that support real-time learning and improvement.
Some of the metrics are blocked by your
consent settings
Evaluating the slope behavior for geophysical flow prediction with advanced machine learning combinations
(2025-12-01)
Onyelowe, Kennedy C.
;
Ebid, Ahmed M.
;
Hanandeh, Shadi
;
Kamchoom, Viroon
Ensuring safety in geotechnical engineering has consistently posed challenges due to the inherent variability of soil. In the case of slope stability problems, performing on-site tests is both costly and time-intensive due to the need for sophisticated equipment (to acquire and move) and logistics. Consequently, the analysis of simulation models based on soft computing proves to be a practical and invaluable alternative. In this research work, learning abilities of the Class Noise Two (CN2), Stochastic Gradient Descent (SGD), Group Method of Data Handling (GMDH) and artificial neural network (ANN) have been investigated in the prediction of the factor of safety (FOS) of slopes. This has been successfully done through literature search, data curation and data sorting. A total of three hundred and forty-nine (349) data entries on the FOS of slopes were collected from literature and sorted to remove odd values and unlogic results, which had been used together in a previous research work. After the sorting process, the remainder of the realistic data entries was 296. The previous work which had included unrealistic data entries had unit weight, γ (kN/m<sup>3</sup>), cohesion, C (kPa),angle of internal friction (Φ°), slope angle (°), slope height H (m), and pore water pressure ratio, r<inf>u</inf> as the studied parameters, which formed the independent variables. After careful checks, the initial results showed poor correlation with the individual factors and the factors were collected into three non-dimensional parameters based on the understanding of the physics of flows, which are: C/γ.h-Cohesion/unit weight x slope height, tan(ϕ)/tan(β)-the tangent of internal friction angle/Tangent of slope angle, and ρ/γ.h-Water pressure/unit weight x slope height, which are deployed as inputs and FOS-the safety factor of the slope as the output. At the end of the exercise, the ANN outclassed the other techniques with SSE of 62%, MAE of 0.27, MSE of 0.21, RMSE of 0.46, average total error of 24%, and R<sup>2</sup> of 0.946 thereby becoming the decisive intelligent model in this exercise. However, there is an advantage the deployment of GMDH, which comes second in order of superiority, has over the ANN. This is the development of a closed-form equation that allows its model to be applied manually in the design of slope stability problems. Overall, the present research models outperformed the eleven (11) models of the previous work due to sorting and elimination of unrealistic data entries deposited in the literature, the application of dimensionless combination of the studied slope stability parameters and the superiority of the selected machine learning techniques.
Some of the metrics are blocked by your
consent settings
Effect of conical air distributors on drying of peppercorns in a fluidized bed dryer: Prediction using an artificial neural network
(2022-08-01)
Chuwattanakul, V.
;
Wongcharee, K.
;
Pimsarn, M.
;
Chokphoemphun, S.
;
Chamoli, S.
The effect of conical air distributors on the drying of peppercorns in a fluidized bed dryer was experimentally studied. A flat perforated sheet was installed in the column at the base of the bed. Conical air distributors consisted of two parts. The first was a solid cone located below an air duct, while the second part was a perforated metal cone placed on the flat perforated sheet. Experiments were carried out using perforated metal cones with three different height to base diameter ratios, (h/H) values of 0.5, 1.0, and 1.5 and three different air velocities, 1.2Umf, 1.6Umf, and 2.0Umf. An air distributor, consisting of a solid cone and a perforated metal cone with h/H = 1.0 and an air velocity of 2.0Umf, showed the best drying performance. It was also discovered that increasing the air velocity accelerated the drying process. A neural network was created to predict the moisture content of peppercorns during the drying process. The split, sample type, spilt ratio, momentum, and learning rate, as well as the numbers of hidden layers, hidden nodes, and training cycles all had an impact. A maximum coefficient of determination of 0.996 was found for the best model.
Some of the metrics are blocked by your
consent settings
FORECASTING MODELS FOR FIRST YEAR PREMIUM OF LIFE INSURANCE
(2022-01-01)
Banditvilai, Somsri
;
Kuharattanachai, Choojai
The objective of this research is to study forecasting models for the first year premium of life insurance. The premium data are gathered from the Office of Insurance Commission (OIC) during January 2003 to November, 2021. The data are divided into 2 sets. The first set from January, 2003 to December 2020 is used for constructing and selection the forecasting models. The second one from January 2021 to November 2021 is used for computing the accuracy of the forecasting model. The forecasting models are chosen by considering the minimum Root Mean Square Error (RMSE). The Mean Absolute Percentage Error (MAPE) is used to measure the accuracy of the model. The results showed that the multiplicative model with initial values from 18 years Decomposition method give the appropriate model for the first year premium of life insurance and yields the MAPE = 17.29%
Some of the metrics are blocked by your
consent settings
Using machine learning-based algorithms to analyze erosion rates of a watershed in Northern Taiwan
(2020-03-01)
Nguyen, Kieu Anh
;
Chen, Walter
;
Lin, Bor Shiun
;
Seeboonruang, Uma
This study continues a previous study with further analysis of watershed-scale erosion pin measurements. Three machine learning (ML) algorithms-Support Vector Machine (SVM), Adaptive Neuro-Fuzzy Inference System (ANFIS), and Artificial Neural Network (ANN)-were used to analyze depth of erosion of a watershed (Shihmen reservoir) in northern Taiwan. In addition to three previously used statistical indexes (Mean Absolute Error, Root Mean Square of Error, and R-squared), Nash-Sutcliffe Effciency (NSE) was calculated to compare the predictive performances of the three models. To see if there was a statistical difference between the three models, theWilcoxon signed-rank test was used. The research utilized 14 environmental attributes as the input predictors of the ML algorithms. They are distance to river, distance to road, type of slope, sub-watershed, slope direction, elevation, slope class, rainfall, epoch, lithology, and the amount of organic content, clay, sand, and silt in the soil. Additionally, measurements of a total of 550 erosion pins installed on 55 slopes were used as the target variable of the model prediction. The dataset was divided into a training set (70%) and a testing set (30%) using the stratified random sampling with sub-watershed as the stratification variable. The results showed that the ANFIS model outperforms the other two algorithms in predicting the erosion rates of the study area. The average RMSE of the test data is 2.05 mm/yr for ANFIS, compared to 2.36 mm/yr and 2.61 mm/yr for ANN and SVM, respectively. Finally, the results of this study (ANN, ANFIS, and SVM) were compared with the previous study (Random Forest, Decision Tree, and multiple regression). It was found that Random Forest remains the best predictive model, and ANFIS is the second-best among the six ML algorithms.
Some of the metrics are blocked by your
consent settings
A rule-based training for artificial neural network packet filtering firewall
(2019-11-01)
Khunkitti, Akharin
;
Chongsujjatham, Ponsuda
The Artificial Neural Network has been used in many network applications, including firewalls. Training process of neural network is very important to define the intelligence of the systems. Many artificial neural network firewalls used direct network packets for training process, which may be difficult to get training samples and may not follow their firewall's policies. This research work proposes a rule-based training for artificial neural network packet filtering firewall. The developed neural network model is trained by generating samples from legacy firewall ruleset. Each rule has been converted to random training samples. All firewall's rules are used to generate the training sample data, rule by rule. The accuracy results show high accuracy with some behavior studies. The number of samples per rule, number of rules and rule style, including default rule and rule-scope effects, have been studied for the best accuracy results. This study also concludes the styles of firewall ruleset for the best accuracy of the proposed system.
Some of the metrics are blocked by your
consent settings
Influence of physical and biochemical composition of three cellulose fibers on cracking of soil
(2019-01-01)
Boddu, Rishita
;
Hong, Min
;
Yongkang, Deng
;
Fengjiao, Chen
;
Garg, Ankit
Different soil improvement techniques have been used to intensify the engineering properties of soil. Three different lignocellulose fiber-reinforced (jute, coir and water hyacinth (WH)) have been explored on the desiccation potential of compacted clayey silt coil. The experimental methodology involved the mixing of fibers with soil at requisite amount and subjecting them to natural environment with controlled irrigating. The controlled irrigation comprised of 15 wetting/drying cycles for 105 days. Parameters like matric suction and water content were focused upon and recorded along with the surface crack formation. The data obtained from the field experiments were analyzed using the Artificial Neural Network (ANN) approach, which is developed in house using C++ language. From the analysis, it can be comprehended that coir is more effective as a reinforcement due to its multifilament nature and higher lignin content which is suitable in resisting crack formation. Further, optimization analysis and sensitivity analysis suggested mechanism of cracking for each fiber.
Some of the metrics are blocked by your
consent settings
Visualization and Estimation of Temperature from Glowing Hot Object by Artificial Neural Network and Image Analysis Technique
(2018-01-01)
Jirasuwankul, N.
This paper proposes alternative technique to estimate temperature of glowing hot objects with application of artificial neural network (ANN) and image analysis techniques. Regardless of using the cutting edge technology or sophisticated sensor such as 2D thermo-imaging equipment, an approximated thermo-imagery of the glowing hot object can be reconstructed by a well-trained ANN model together with image analysis in RGB color space. By training the model with data along the Blackbody locus from the CIE-1931 chromaticity chart and using three normalized individual monochromatic R, G and B images as inputs, the processed image having correlated color temperature (CCT) is finally obtained. Experimental results show that averaging error of the estimated temperature can be achieved with 10% for the reddish-yellowish hot objects and less than 10% for the bright-yellow one respectively.
Some of the metrics are blocked by your
consent settings
Thai sign language recognition by using geometric invariant feature and ANN classification
(2017-02-21)
Adhan, Suchin
;
Pintavirooj, Chuchart
Hand sign language is the primary communication tool for people with hearing-impaired or deaf. People can use it to communicate effectively but the challenge is to communicate with the computer. Human computer interaction (HCI) will have a positive impact on their use. Thus, this is to bring the hand gestures in HCI as an important research area. This research focuses on 2D image recognition utilizing an evolved geometric invariant feature and also have developed a two-layer feedforward neural network to identify and translate hand gesture pose of the 42 letters in the Thai Sign Language (TSL) alphabet to Thai alphabets. We designed glove with six different colored markers for using in the experiment. The result shows that this system is able to recognize 42 TSL alphabets with an average accuracy of 96.19 %.
Some of the metrics are blocked by your
consent settings
Comparison study of muscular-contraction classification between independent component analysis and artificial neural network
(2008-12-01)
Sueaseenak, Direk
;
Wibirama, Sunu
;
Chanwimalueang, Theerasak
;
Pintavirooj, Chuchart
;
Sangworasil, Manus
We developed a multi-channel electromyogram acquisition system using PSOC microcontroller to acquire multichannel EMG signals. An array of 4 x 4 surface electrodes was used to record the EMG signal. The obtained signals were classified by a back-propagation-type artificial neural network. B-spline interpolation technique has been utilized to map the EMG signal on the muscle surface. The topological mapping of the EMG is then analyzed to classify the pattern of muscle contraction using independent component analysis. The proposed system was successfully demonstrated to record EMG data and its surface mapping. The comparison study of muscular contraction classification using independent component analysis and artificial neural network demonstrates shows that performance of ANN classification is as comparable as that of the ICA. The computational time of ANN is also less than that of the ICA. © 2008 IEEE.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results