Environmental pollution has led to global warming, which threatens human life. In response, hydrogen is gaining attention as a next-generation energy source that does not emit carbon. Due to its explosive nature, special care must be taken in the safe storage and transportation of hydrogen. Among various storage methods, liquefied storage, which can reduce its volume to 1/800, is considered efficient. However, since its boiling point reaches -253°C, the design of an insulation system is essential. For the design of insulation systems applied to large containers, a membrane-type design is required, which necessitates the use of cryogenic adhesives. To evaluate whether the cryogenic adhesive is properly implemented, assessments such as tensile and shear tests are necessary. This study presents a methodology for shear evaluation. Conventional methods for shear evaluation of adhesives result in slippage, preventing proper assessment. Therefore, a method involving drilling holes in the gripper and pulling from the holes must be applied. Optimal design concerning the size and location of the holes is required, and this study derives optimal values based on finite element analysis. By conducting experiments based on the results of this study, it is expected that the risk of gripper damage will be minimized, allowing for accurate evaluation of the adhesive’s performance.
Abstract Handling imbalanced datasets in binary classification, especially in employment big data, is challenging. Traditional methods like oversampling and undersampling have limitations. This paper integrates TabNet and Generative Adversarial Networks (GANs) to address class imbalance. The generator creates synthetic samples for the minority class, and the discriminator, using TabNet, ensures authenticity. Evaluations on benchmark datasets show significant improvements in accuracy, precision, recall, and F1-score for the minority class, outperforming traditional methods. This integration offers a robust solution for imbalanced datasets in employment big data, leading to fairer and more effective predictive models.
본 논문에서는 다목적 구조물인 다중연결 해양부유체를 대상으로 변형 기반 모드 차수축소법을 적용하고 차수축소모델의 구조응 답 예측 성능을 향상시키기 위해 유전 알고리즘 기반의 센서 배치 최적화를 수행하였다. 다중연결 해양부유체의 차수축소모델 생성 에 필요한 변형 기반 모드 데이터를 얻기 위해 다양한 규칙파랑하중조건에 대한 유체-구조 연성 수치해석을 수행하고 변형 기반 모드 의 직교성, 자기상관계수를 이용하여 주요 변형 기반 모드를 선정하였다. 다중연결 해양부유체의 경우 차수축소모델의 구조응답 예 측 성능이 계측 및 예측 구조응답 위치에 따라 민감하기 때문에 유전 알고리즘 기반의 최적화를 수행하여 최적의 센서 배치를 도출하 였다. 최적화 결과, 모든 센서 배치 조합에 대한 차수축소모델 생성 및 예측 성능 평가 대비 약 8배의 계산 비용을 절감하였으며, 예측 성능 평가 지표인 평균 제곱근 오차가 초기 센서 배치보다 84% 감소하였다. 또한, 다중연결 해양부유체 모형시험 결과를 이용하여 불 규칙파랑하중에 대한 최적화된 센서 배치의 차수축소모델의 구조응답 예측 성능을 평가 및 검증하였다.
This study introduces and experimentally validates a novel approach that combines Instruction fine-tuning and Low-Rank Adaptation (LoRA) fine-tuning to optimize the performance of Large Language Models (LLMs). These models have become revolutionary tools in natural language processing, showing remarkable performance across diverse application areas. However, optimizing their performance for specific domains necessitates fine-tuning of the base models (FMs), which is often limited by challenges such as data complexity and resource costs. The proposed approach aims to overcome these limitations by enhancing the performance of LLMs, particularly in the analysis precision and efficiency of national Research and Development (R&D) data. The study provides theoretical foundations and technical implementations of Instruction fine-tuning and LoRA fine-tuning. Through rigorous experimental validation, it is demonstrated that the proposed method significantly improves the precision and efficiency of data analysis, outperforming traditional fine-tuning methods. This enhancement is not only beneficial for national R&D data but also suggests potential applicability in various other data-centric domains, such as medical data analysis, financial forecasting, and educational assessments. The findings highlight the method's broad utility and significant contribution to advancing data analysis techniques in specialized knowledge domains, offering new possibilities for leveraging LLMs in complex and resource- intensive tasks. This research underscores the transformative potential of combining Instruction fine-tuning with LoRA fine-tuning to achieve superior performance in diverse applications, paving the way for more efficient and effective utilization of LLMs in both academic and industrial settings.
This paper presents a path planning optimization model for the engineering units to install obstacles in the shortest time during wartime. In a rapidly changing battlefield environment, engineering units operate various engineering obstacles to fix, bypass, and delay enemy maneuvers, and the success of the operation lies in efficiently planning the obstacle installation path in the shortest time. Existing studies have not reflected the existence of obstacle material storage that should be visited precedence before installing obstacles, and there is a problem that does not fit the reality of the operation in which the installation is continuously carried out on a regional basis. By presenting a Mixed Integrer Programming optimization model reflecting various constraints suitable for the battlefield environment, this study attempted to promote the efficient mission performance of the engineering unit during wartime.
This research presented the procedural framework of developing and optimizing an artificial intelligence model for predicting the change of bread texture by different baking enhancers. Emphasis was placed on the impact of various baking enhancers on the Mixolab thermo-mechanical properties of wheat flour and consequent alterations in bread texture. The application of baking enhancers positively contributed to dough formation and stability, producing bread with a soft texture. However, a relatively low Pearson correlation coefficient was observed between a single Mixolab parameter and bread texture (r<0.59). To more accurately predict the texture of bread from the thermo-mechanical features of wheat flour with baking enhancers, five AI models (multiple linear regression, decision tree, stochastic gradient descent, random forest, and multilayer perceptron neural network) were applied, and their prediction performance was compared. The multilayer perceptron neural network model was further utilized to enhance the prediction of bread texture by mitigating overfitting risks. Finally, the hyperparameter tuning (activation function [Leaky ReLU], regularization [0.0001], and dropout [0.1]) led to enhanced model performance (R2 = 0.8109 and RMSE = 0.1096).