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 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).

본 논문에서는 다이나믹크리깅 대리모델 기반 자동차 브레이크 패드 마모량 측정센서 브라켓의 설계최적화를 소개한다. 브레이크 작동시 마찰재 바닥의 온도가 600°C 이상으로 상승하고, 이 열이 전달되어 센서의 기능을 상실시킨다. 따라서 열전달을 최소화하는 브라켓 형상의 설계최적화는 필수적이다. 최적화에 소요되는 계산비용을 절감하기 위해 다이나믹크리깅 대리모델로 열전달 시뮬레 이션을 대체하였다. 다이나믹크리깅은 최적의 상관함수와 기저함수를 선정하였으며, 정확한 대리모델을 도출하였다. 최적화 결과 센 서위치의 온도가 초기모델에 비해 7.57% 감소하였으며, 이를 열전달 시뮬레이션으로 다시 한번 확인하여 대리모델 기반 최적설계가 유의미함을 검증하였다.

Effects-Based Operations (EBO) refers to a process for achieving strategic goals by focusing on effects rather than attrition-based destruction. For a successful implementation of EBO, identifying key nodes in an adversary network is crucial in the process of EBO. In this study, we suggest a network-based approach that combines network centrality and optimization to select the most influential nodes. First, we analyze the adversary’s network structure to identify the node influence using degree and betweenness centrality. Degree centrality refers to the extent of direct links of a node to other nodes, and betweenness centrality refers to the extent to which a node lies between the paths connecting other nodes of a network together. Based on the centrality results, we then suggest an optimization model in which we minimize the sum of the main effects of the adversary by identifying the most influential nodes under the dynamic nature of the adversary network structure. Our results show that key node identification based on our optimization model outperforms simple centrality-based node identification in terms of decreasing the entire network value. We expect that these results can provide insight not only to military field for selecting key targets, but also to other multidisciplinary areas in identifying key nodes when they are interacting to each other in a network.

In the automobile manufacturing industry, lightweight design is one of the essential challenges to be solved fundamentally. The vehicle wheels are classified as safety related components as the main substructure of the vehicle. In this study, we illustrate a technique for selecting the appropriate number of spokes. Based on the basic model of the selected number of spokes, we propose a method to maintain stiffness and design lightweight using topology optimization software. Based on the basic model of the selected number of spokes, it was redesigned to be lightweight while maintaining stiffness by utilizing topology optimization software. By comparing and reviewing the structural analysis results of the basic model and the redesigned model, a design technique that can maintain structural safety and reduce wheel mass was proposed.

The multi-layered heat source model is a model that can cover most of existing studies and can be defined with a simple formula. Based on the methodology performed in previous studies, the welding heat source was found through experiments and FEM under the welding power conditions of three cases and the parameters of the welding heat source were analyzed according to the welding power. In this study, parameters of fiber laser welding heat source according to welding power were searched through optimization algorithm and finite element analysis, and the correlation was analyzed. It was confirmed that the concentration of the welding heat source in the 1st layer was high regardless of the welding power, and it was confirmed that the concentration of the welding heat source in the 5th layer (last layer) increased as the welding power increased. This reflects the shape of the weld bead that appears during actual fiber laser welding, and it was confirmed that this study represents the actual phenomenon.

This study explored the usefulness and implications of the Bayesian hyperparameter optimization in developing species distribution models (SDMs). A variety of machine learning (ML) algorithms, namely, support vector machine (SVM), random forest (RF), boosted regression tree (BRT), XGBoost (XGB), and Multilayer perceptron (MLP) were used for predicting the occurrence of four benthic macroinvertebrate species. The Bayesian optimization method successfully tuned model hyperparameters, with all ML models resulting an area under the curve (AUC) > 0.7. Also, hyperparameter search ranges that generally clustered around the optimal values suggest the efficiency of the Bayesian optimization in finding optimal sets of hyperparameters. Tree based ensemble algorithms (BRT, RF, and XGB) tended to show higher performances than SVM and MLP. Important hyperparameters and optimal values differed by species and ML model, indicating the necessity of hyperparameter tuning for improving individual model performances. The optimization results demonstrate that for all macroinvertebrate species SVM and RF required fewer numbers of trials until obtaining optimal hyperparameter sets, leading to reduced computational cost compared to other ML algorithms. The results of this study suggest that the Bayesian optimization is an efficient method for hyperparameter optimization of machine learning algorithms.

A60 급 갑판 관통 관은 선박과 해양플랜트에서 화재사고가 발생할 경우 화염의 확산을 방지하고 인명을 보호하기 위해 수평구조에 설치되는 방화장치이다. 본 연구에서는 다양한 대리모델과 다중 섬유전자 알고리즘을 이용하여 A60 급 갑판 관통 관의 방화설계에 대한 이산변수 근사최적화를 수행하였다. A60 급 갑판 관통 관의 방화설계는 과도 열전달해석을 통해 평가하였다. 근사최적화에서 관통 관의 길이, 지름, 재질, 그리고 단열재의 밀도는 이산설계변수로 적용하였고, 제한조건은 온도, 생산성 및 가격을 고려하였다. 대리모델 기반의 근사최적설계 문제는 제한조건을 만족하면서 A60 급 갑판 관통 관의 중량을 최소화할 수 있는 이산설계변수를 결정하도록 정식화 하였다. 반응표면모델, 크리깅, 그리고 방사기저함수 신경망과 같은 다양한 대리모델이 근사최적화에 사용되었다. 근사최적화의 정확도를 검토하기 위해 최적해의 결과는 실제 계산 결과와 비교하였다. 근사최적화에 사용된 대리모델 중 방사기저함수 신경망 모델이 A60 급 갑판 관통 관의 방화설계에 대해 가장 정확한 최적설계 결과를 나타내었다.

In this study, a welding heat source model was presented and verified during fiber laser welding. The multi-layered heat source model is a model that can cover most of existing studies and can be defined with a simple formula. It consists of a total of 12 parameters, and an optimization algorithm was used to find them. As optimization algorithms, adaptive simulated annealing, multi island genetic algorithm, and Hooke-Jeeves technique were applied for comparative analysis. The parameters were found by comparing the temperature distribution when the STS304L was bead on plate welding and the temperature distribution derived through finite element analysis, and all three models were able to derive a model with similar trends. However, there was a deviation between parameters, which was attributed to the many variables. It is expected that a more clear welding heat source model can be derived in subsequent studies by giving a guide to the relational expression and range between variables and increasing the temperature measurement point, which is the target value.

Welding is the most widely used technology for manufacturing in the automobile, and shipbuilding industries. Fiber laser welding is rapidly introduced into the field to minimize welding distortion and fast welding speed. Although it is advantageous to use finite element analysis to predict welding distortion and find optimized welding conditions, there are various heat source model for fiber laser welding. In this study, a welding heat source was proposed using a multi-layered heat source model that encompasses most of the existing various welding heat source models: conical shape, curved model, exponential model, conical-cylindrical model, and conical-conical model. A case study was performed through finite element analysis using the radius of each layer and the ratio of heat energy of the layer as variables, and the variables were found by comparing them with the actual experimental results. For case study, by applying Adaptive simulated annealing, one of the global optimization algorithms, we were able to find the heat source model more efficiently.

In this study, a model to optimize residual chlorine concentrations in a water supply system was developed using a multi-objective genetic algorithm. Moreover, to quantify the effects of optimized residual chlorine concentration management and to consider customer service requirements, this study developed indices to quantify the spatial and temporal distributions of residual chlorine concentration. Based on the results, the most economical operational method to manage booster chlorination was derived, which would supply water that satisfies the service level required by consumers, as well as the cost-effectiveness and operation requirements relevant to the service providers. A simulation model was then created based on an actual water supply system (i.e., the Multi-regional Water Supply W in Korea). Simulated optimizations were successful, evidencing that it is possible to meet the residual chlorine concentration demanded by consumers at a low cost.

The estimation of heat source model is very important for heat transfer analysis with finite element method. Part I of this study used adaptive simulated annealing which is one of the global optimization algorithm for anticipating the parameters of the Goldak model. Although the analysis with 3D model which depicted the real situation produced the correct answer, that took too much time with moving heat source model based on Fortran and Abaqus. This research suggests the procedure which can reduce time with maintaining quality of analysis. The lead time with 2D model is reduced by 90% comparing that of 3D model, the temperature distribution is similar to each other. That is based on the saturation of heat transfer among the direction of heat source movement. Adaptive simulated annealing with 2D model can be used to estimate more proper heat source model and which could enhance to reduce the resources and time for experiments.

Anticipation of welding deformation with finite element method is a very interested topic in the industries, adequate heat source model is essential for concluding reasonable results. This study is related to estimate the parameters of Goldak heat source model, and global optimization algorithm is applied to this research. The heat affected zone (HAZ) boundary line of bead on plate (BOP) welding is used as the target, parameters of heat sources are used as the variables. Adaptive simulated annealing is applied and the optimal result is obtained out of 1,000 candidates. The convergence of finite element method and the global optimization is meaningful for estimation of welding deformation, which could enhance to reduce the resources and time for experiments.

In this study, optimizations were carried out for a heater core and an evaporator installed in a passenger car. The main geometric parameters of each component were selected as design variables, and the main performances examined were the outlet temperature and the pressure drop. In addition, the sensitivity analysis was performed to grasp the dominant design variable. The thermal flow analysis for each component was performed using the commercial program STAR CCM+. On the other hand, EasyDesign, a commercial program based on DOE and metamodel-based optimization, was used as an optimization tool. The optimized performances of the heater core were compared with performance target, and finally, the improved designs of the heater and the evaporator were presented.

본 연구는 3개의 운전변수(압력, 공기량, 운전시간)를 실험 설계하고 마이크로 버블의 종말부상 속도(Terminal rise velocity)를 반응 값으로 하여 예측식 모델과 최적 조건을 수립하였다. 다항식 회귀분석을 통해 펌프의 압력(X1) 4.5bar, 공기량(X2) 3.3L/min 그리고 운전시간(X3)이 2.2min에서 종말상승속도(Terminal rise velocity)에 대한 최적값인 5.14 cm/min (85.7㎛/sec)을 얻었다. 또한, 레이저 입자계수 측정장치를 이용하여 2~5㎛ 및 25~50㎛ 영역에서의 가장 높은 마이크로버블 직경크기 분포를 확인하였다.