• Smart Structures and Systems
• /
• v.33 no.1
• /
• pp.1-16
• /
• 2024

The investigation on vehicle-bridge coupling system (VBCS) is crucial in bridge design, bridge condition evaluation, and vehicle overload control. A real-time hybrid testing (RTHT) method for VBCS (RTHT-VBCS) is proposed in this paper for accurately and economically disclosing the dynamic performance of VBCSs. In the proposed method, one of the carriages is chosen as the experimental substructure loaded by servo-hydraulic actuator loading system in the laboratory, and the remaining carriages as well as the bridge structure are chosen as the numerical substructure numerically simulated in one computer. The numerical substructure and the experimental substructure are synchronized at their coupling points in terms of force equilibrium and deformation compatibility. Compared to the traditional iteration experimental method and the numerical simulation method, the proposed RTHT-VBCS method could not only obtain the dynamic response of VBCS, but also economically analyze various working conditions. Firstly, the theory of RTHT-VBCS is proposed. Secondly, numerical models of VBCS for RTHT method are presented. Finally, the feasibility and accuracy of the RTHT-VBCS are preliminarily validated by real-time hybrid simulations (RTHSs). It is shown that, the proposed RTHT-VBCS is feasible and shows great advantages over the traditional methods, and the proposed models can effectively represent the VBCS for RTHT method in terms of the force equilibrium and deformation compatibility at the coupling point. It is shown that the results of the single-degree-of-freedom model and the train vehicle model are match well with the referenced results. The RTHS results preliminarily prove the effectiveness and accuracy of the proposed RTHT-VBCS.

• Journal of Korea Society of Industrial Information Systems
• /
• v.29 no.3
• /
• pp.51-65
• /
• 2024

Accurate electrical load forecasting is important to the effective operation of power systems in smart grids. With the recent development in machine learning, artificial intelligence-based models for predicting power demand are being actively researched. However, since existing models get input variables as numerical features, the accuracy of the forecasting model may decrease because they do not reflect the semantic relationship between these features. In this paper, we propose a scheme for short-term load forecasting by using features extracted through the large language models for input data. We firstly convert input variables into a sentence-like prompt format. Then, we use the large language model with frozen weights to derive the embedding vectors that represent the features of the prompt. These vectors are used to train the forecasting model. Experimental results show that the proposed scheme outperformed models based on numerical data, and by visualizing the attention weights in the large language models on the prompts, we identified the information that significantly influences predictions.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.1
• /
• pp.83-88
• /
• 2024

In this research, we propose a Semi-Supervised learning based railroad surface defect detection method. The Resnet50 model, pretrained on ImageNet, was employed for the training. Data without labels are randomly selected, and then labeled to train the ResNet50 model. The trained model is used to predict the results of the remaining unlabeled training data. The predicted values exceeding a certain threshold are selected, sorted in descending order, and added to the training data. Pseudo-labeling is performed based on the class with the highest probability during this process. An experiment was conducted to assess the overall class classification performance based on the initial number of labeled data. The results showed an accuracy of 98% at best with less than 10% labeled training data compared to the overall training data.

• Advances in nano research
• /
• v.16 no.6
• /
• pp.623-638
• /
• 2024

This study investigates the thermal post-buckling behavior of concrete eccentric annular sector plates reinforced with graphene oxide powders (GOPs). Employing the minimum total potential energy principle, the plates' stability and response under thermal loads are analyzed. The Haber-Schaim foundation model is utilized to account for the support conditions, while the transform differential quadrature method (TDQM) is applied to solve the governing differential equations efficiently. The integration of GOPs significantly enhances the mechanical properties and stability of the plates, making them suitable for advanced engineering applications. Numerical results demonstrate the critical thermal loads and post-buckling paths, providing valuable insights into the design and optimization of such reinforced structures. This study presents a machine learning algorithm designed to predict complex engineering phenomena using datasets derived from presented mathematical modeling. By leveraging advanced data analytics and machine learning techniques, the algorithm effectively captures and learns intricate patterns from the mathematical models, providing accurate and efficient predictions. The methodology involves generating comprehensive datasets from mathematical simulations, which are then used to train the machine learning model. The trained model is capable of predicting various engineering outcomes, such as stress, strain, and thermal responses, with high precision. This approach significantly reduces the computational time and resources required for traditional simulations, enabling rapid and reliable analysis. This comprehensive approach offers a robust framework for predicting the thermal post-buckling behavior of reinforced concrete plates, contributing to the development of resilient and efficient structural components in civil engineering.

• Journal of Advanced Navigation Technology
• /
• v.28 no.1
• /
• pp.149-154
• /
• 2024

Research on particulate matter is advancing in real-time, and various methods are being studied to improve the accuracy of prediction models. Furthermore, studies that take into account various factors to understand the precise causes and impacts of particulate matter are actively being pursued. This paper trains an LSTM model using seasonal data and another LSTM model using concentration-based data. It compares and analyzes the PM_2.5 prediction performance of the two models. To train the model, weather data and air pollutant data were collected. The collected data was then used to confirm the correlation with PM_2.5. Based on the results of the correlation analysis, the data was structured for training and evaluation. The seasonal prediction model and the concentration-specific prediction model were designed using the LSTM algorithm. The performance of the prediction model was evaluated using accuracy, RMSE, and MAPE. As a result of the performance evaluation, the prediction model learned by concentration had an accuracy of 91.02% in the "bad" range of AQI. And overall, it performed better than the prediction model trained by season.

• The KIPS Transactions:PartB
• /
• v.10B no.6
• /
• pp.657-664
• /
• 2003

Reliable tracking of moving humans is essential to motion estimation, video surveillance and human-computer interface. This paper presents a new approach to human motion tracking that combines appearance-based and model-based techniques. Monocular color video is processed at both pixel level and object level. At the pixel level, a Gaussian mixture model is used to train and classily individual pixel colors. At the object level, a 3D human body model projected on a 2D image plane is used to fit the image data. Our method does not use inverse kinematics due to the singularity problem. While many others use stochastic sampling for model-based motion tracking, our method is purely dependent on nonlinear programming. We convert the human motion tracking problem into a nonlinear programming problem. A cost function for parameter optimization is used to estimate the degree of the overlapping between the foreground input image silhouette and a projected 3D model body silhouette. The overlapping is computed using computational geometry by converting a set of pixels from the image domain to a polygon in the real projection plane domain. Our method is used to recognize various human motions. Motion tracking results from video sequences are very encouraging.

• Journal of the Korea Society for Simulation
• /
• v.31 no.1
• /
• pp.11-18
• /
• 2022

Recently, research to classify human activity using imagery has been actively conducted for the purpose of crime prevention and facility safety in public places and facilities. In order to improve the performance of human activity classification, most studies have applied deep learning based-transfer learning. However, despite the increase in the number of backbone network models that are the basis of deep learning as well as the diversification of architectures, research on finding a backbone network model suitable for the purpose of operation is insufficient due to the atmosphere of using a certain model. Thus, this study applies the transfer learning into recently developed deep learning backborn network models to build an intelligent system that classifies human activity using imagery. For this, 12 types of active and high-contact human activities based on sports, not basic human behaviors, were determined and 7,200 images were collected. After 20 epochs of transfer learning were equally applied to five backbone network models, we quantitatively analyzed them to find the best backbone network model for human activity classification in terms of learning process and resultant performance. As a result, XceptionNet model demonstrated 0.99 and 0.91 in training and validation accuracy, 0.96 and 0.91 in Top 2 accuracy and average precision, 1,566 sec in train process time and 260.4MB in model memory size. It was confirmed that the performance of XceptionNet was higher than that of other models.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.40 no.5
• /
• pp.449-456
• /
• 2022

This study proposes a deep learning algorithm to predict crop yield using GIS (Geographic Information System) to extract soil properties from Soilgrids and soil suitability class maps. The proposed model modified the structure of a published CNN-RNN (Convolutional Neural Network-Recurrent Neural Network) based crop yield prediction model suitable for the domestic crop environment. The existing model has two characteristics. The first is that it replaces the original yield with the average yield of the year, and the second is that it trains the data of the predicted year. The new model uses the original field value to ensure accuracy, and the network structure has been improved so that it can train only with data prior to the year to be predicted. The proposed model predicted the yield per unit area of autumn cabbage for kimchi by region based on weather, soil, soil suitability classes, and yield data from 1980 to 2020. As a result of computing and predicting data for each of the four years from 2018 to 2021, the error amount for the test data set was about 10%, enabling accurate yield prediction, especially in regions with a large proportion of total yield. In addition, both the proposed model and the existing model show that the error gradually decreases as the number of years of training data increases, resulting in improved general-purpose performance as the number of training data increases.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.8
• /
• pp.341-354
• /
• 2023

The title of a document is the brief summarization of the document. Readers can easily understand a document if we provide them with its title in their preferred styles and the languages. In this research, we propose a cross-lingual and style-based title generation model using multiple adapters. To train the model, we need a parallel corpus in several languages with different styles. It is quite difficult to construct this kind of parallel corpus; however, a monolingual title generation corpus of the same style can be built easily. Therefore, we apply a zero-shot strategy to generate a title in a different language and with a different style for an input document. A baseline model is Transformer consisting of an encoder and a decoder, pre-trained by several languages. The model is then equipped with multiple adapters for translation, languages, and styles. After the model learns a translation task from parallel corpus, it learns a title generation task from monolingual title generation corpus. When training the model with a task, we only activate an adapter that corresponds to the task. When generating a cross-lingual and style-based title, we only activate adapters that correspond to a target language and a target style. An experimental result shows that our proposed model is only as good as a pipeline model that first translates into a target language and then generates a title. There have been significant changes in natural language generation due to the emergence of large-scale language models. However, research to improve the performance of natural language generation using limited resources and limited data needs to continue. In this regard, this study seeks to explore the significance of such research.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.5
• /
• pp.331-337
• /
• 2023

In this paper, we propose a deep learning model to detect lumbar 3 (L3) CT images to determine the occurrence and degree of sarcopenia. In addition, we would like to propose an optimization technique that uses oversampling ratio and class weight as design parameters to address the problem of performance degradation due to data imbalance between L3 level and non-L3 level portions of CT data. In order to train and test the model, a total of 150 whole-body CT images of 104 prostate cancer patients and 46 bladder cancer patients who visited Gangneung Asan Medical Center were used. The deep learning model used ResNet50, and the design parameters of the optimization technique were selected as six types of model hyperparameters, data augmentation ratio, and class weight. It was confirmed that the proposed optimization-based L3 level extraction model reduced the median L3 error by about 1.0 slices compared to the control model (a model that optimized only 5 types of hyperparameters). Through the results of this study, accurate L3 slice detection was possible, and additionally, we were able to present the possibility of effectively solving the data imbalance problem through oversampling through data augmentation and class weight adjustment.