• Journal of the Korean Geotechnical Society
• /
• v.37 no.5
• /
• pp.47-63
• /
• 2021

Recently, the demand for three-dimensional (3D) underground maps from the perspective of digital twins and the demand for linkage utilization are increasing. However, the vastness of national geotechnical survey data and the uncertainty in applying geostatistical techniques pose challenges in modeling underground regional geotechnical characteristics. In this study, an optimal learning model based on multi-layer perceptron (MLP) was constructed for 3D subsurface lithological and geotechnical classification in Seoul, South Korea. First, the geotechnical layer and 3D spatial coordinates of each borehole dataset in the Seoul area were constructed as a geotechnical database according to a standardized format, and data pre-processing such as correction and normalization of missing values for machine learning was performed. An optimal fitting model was designed through hyperparameter optimization of the MLP model and model performance evaluation, such as precision and accuracy tests. Then, a 3D grid network locally assigning geotechnical layer classification was constructed by applying an MLP-based bet-fitting model for each unit lattice. The constructed 3D geotechnical layer map was evaluated by comparing the results of a geostatistical interpolation technique and the topsoil properties of the geological map.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.30 no.1
• /
• pp.108-117
• /
• 2024

In this study, simulations based on computational fluid dynamics were performed for self-propulsion performance prediction of a catamaran that has asymmetrical inside and outside hull form and numerous knuckle lines. In the simulations, the Moving Reference Frame (MRF) or Sliding Mesh (SDM) techniques were used, and the rotation angle of the propeller per time step was different to identify the difference using the analysis technique and condition. The propeller rotation angle used in the MRF technique was 1˚ and those used in the SDM technique were 1˚, 5˚, or 10˚. The torque of the propeller was similar in both the techniques; however, the thrust and resistance of the hull were computed lower when the SDM technique was applied than when the MRF technique was applied, and higher as the rotation angle of the propeller per time step in the SDM technique was smaller in the simulations for several revolutions of the propeller to estimate the self-propulsion condition. The revolutions, thrust, and torque of the propeller in the self-propulsion condition obtained using linear interpolation and the delivered power, wake fraction, thrust deduction factor, and revolutions of the propeller obtained using the full-scale prediction method showed the same trend for both the techniques; however, most of the self-propulsion efficiency showed the opposite trend for these techniques. The accuracy of the propeller wake was low in the simulations when the MRF technique was applied, and slight difference existed in the expression of the wake according to the rotation angle of the propeller per time step when the SDM technique was applied.

• Radiation Oncology Journal
• /
• v.15 no.1
• /
• pp.71-78
• /
• 1997

Purpose : To collect beam data for dynamic wedge fields using conventional measurement tools without the multi-detector system, such as the linear diode detectors or ionization chambers. Materials and Methods : The accelerator CL 2100 C/D has two photon energies of 6MV and 15MV with dynamic wedge an91es of 15o, 30o, 45o and 60o. Wedge transmission factors, percentage depth doses(PDD's) and dose Profiles were measured. The measurements for wedge transmission factors are performed for field sizes ranging from $4\times4cm^2\;to\;20\times20cm^2$ in 1-2cm steps. Various rectangular field sizes are also measured for each photon energy of 6MV and 15MV, with the combination of each dynamic wedge angle of 15o 30o. 45o and 60o. These factors are compared to the calculated wedge factors using STT(Segmented Treatment Table) value. PDD's are measured with the film and the chamber in water Phantom for fixed square field. Converting parameters for film data to chamber data could be obtained from this procedure. The PDD's for dynamic wedged fields could be obtained from film dosimetry by using the converting parameters without using ionization chamber. Dose profiles are obtained from interpolation and STT weighted superposition of data through selected asymmetric static field measurement using ionization chamber. Results : The measured values of wedge transmission factors show good agreement to the calculated values The wedge factors of rectangular fields for constant V-field were equal to those of square fields The differences between open fields' PDDs and those from dynamic fields are insignificant. Dose profiles from superposition method showed acceptable range of accuracy(maximum 2% error) when we compare to those from film dosimetry. Conclusion : The results from this superposition method showed that commissionning of dynamic wedge could be done with conventional dosimetric tools such as Point detector system and film dosimetry winthin maximum 2% error range of accuracy.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.

• Korean Journal of Remote Sensing
• /
• v.38 no.5_1
• /
• pp.627-646
• /
• 2022

Recently, the seriousness of climate change-related problems caused by global warming is growing, and the average temperature is also rising. As a result, it is affecting the environment in which various temperature-sensitive creatures and creatures live, and changes in the ecosystem are also being detected. Seasons are one of the important factors influencing the types, distribution, and growth characteristics of creatures living in the area. Among the most popular and easily recognized plant seasonal phenomena among the indicators of the climate change impact evaluation, the blooming day of flower and the peak day of autumn leaves were modeled. The types of plants used in the modeling were forsythia and cherry trees, which can be seen as representative plants of spring, and maple and ginkgo, which can be seen as representative plants of autumn. Weather data used to perform modeling were temperature, precipitation, and solar radiation observed through the ASOS Observatory of the Korea Meteorological Administration. As satellite data, MODIS NDVI was used for modeling, and it has a correlation coefficient of about -0.2 for the flowering date and 0.3 for the autumn leaves peak date. As the model used, the model was established using multiple regression models, which are linear models, and Random Forest, which are nonlinear models. In addition, the predicted values estimated by each model were expressed as isopleth maps using spatial interpolation techniques to express the trend of plant seasonal changes from 2003 to 2020. It is believed that using NDVI with high spatio-temporal resolution in the future will increase the accuracy of plant phenology modeling.