Grease consistency is a critical quality factor in industrial lubrication processes, as it significantly affects mechanical performance, operational stability, and product durability. In grease manufacturing, fluctuations in process variables such as feed temperature, evaporation time, flow rate, and environmental conditions can cause inconsistencies in quality, which may lead to operational defects or increased maintenance costs. To address this challenge, this study proposes a predictive modeling approach for forecasting grease consistency with the aim of enhancing process quality. Real manufacturing process data were collected from a grease production facility, and irrelevant or highly correlated variables were eliminated through multicollinearity analysis and dimensionality reduction. Multiple machine learning regression techniques were applied and evaluated to identify the most effective model for predicting grease consistency. Through systematic comparison, the final predictive model was developed to provide accurate consistency estimation based on selected process variables. The proposed model enables proactive quality control by allowing consistency deviations to be detected early, thereby supporting process optimization and decision-making in manufacturing environments. This research demonstrates the applicability of data-driven predictive modeling in the grease industry and contributes to the development of intelligent quality management strategies in modern manufacturing. The findings suggest that machine learning-based consistency prediction can play a key role in improving production efficiency and ensuring stable product performance.

Anomaly detection is a key technique for ensuring the reliability and stability of systems across various industrial domains. Autoencoder-based reconstruction models are particularly effective in learning normal patterns and detecting deviations. However, conventional loss functions such as Mean Squared Error (MSE) and Mean Absolute Error (MAE) are limited in capturing anomalies that follow heavy-tailed or asymmetric distributions, which are commonly observed in real-world industrial settings. To address this limitation, we propose a Mixture Negative Log-Likelihood (Mixture NLL) loss function based on a combination of Gaussian, Laplace, and Student-t distributions. The loss is constructed using the probability density functions of each distribution, with key parameters such as standard deviation, scale, and degrees of freedom learned during training. The mixture weights representing the contribution of each distribution are also jointly optimized. Experimental results on real-world time-series anomaly detection datasets demonstrate that the proposed MixtureLoss consistently outperforms conventional loss-based Autoencoder models, particularly in detecting tail-end anomalies.

This study proposes a weighted ensemble deep learning framework for accurately predicting the State of Health (SOH) of lithium-ion batteries. Three distinct model architectures—CNN-LSTM, Transformer-LSTM, and CEEMDAN-BiGRU—are combined using a normalized inverse RMSE-based weighting scheme to enhance predictive performance. Unlike conventional approaches using fixed hyperparameter settings, this study employs Bayesian Optimization via Optuna to automatically tune key hyperparameters such as time steps (range: 10-35) and hidden units (range: 32-128). To ensure robustness and reproducibility, ten independent runs were conducted with different random seeds. Experimental evaluations were performed using the NASA Ames B0047 cell discharge dataset. The ensemble model achieved an average RMSE of 0.01381 with a standard deviation of ±0.00190, outperforming the best single model (CEEMDAN-BiGRU, average RMSE: 0.01487) in both accuracy and stability. Additionally, the ensemble's average inference time of 3.83 seconds demonstrates its practical feasibility for real-time Battery Management System (BMS) integration. The proposed framework effectively leverages complementary model characteristics and automated optimization strategies to provide accurate and stable SOH predictions for lithium-ion batteries.

본 연구의 목적은 남북한 간 교류협력 재개에 대비하여 향후 민간 차원의 남북교역 추진 시 활용 가능한 표준계약서 조항을 도출하는 것 이다. 이를 위해 기존의 남북교역 거래 관행과 계약체결 실태를 조사하 고, 실제 남북교역 현장에서 사용되었던 분야별·업체별 계약서를 분석 하였다. 그동안 남북한 당사자 간 상거래는 남북관계의 특수성과 북한 리스크 요인 등으로 사업추진 과정에서 예측 불가능한 위험과 불확실 성이 지속되어 왔다. 이러한 상황에서 대외무역과 국제상거래 관행에 부합하는 남북한 간 표준계약 양식의 도입과 적용 필요성이 꾸준히 제 기되어 왔다. 본 연구는 남북교역 표준계약서를 제시하기 위해 기존 남 북교역 계약서상의 관행을 참고하되, 국제적으로 통용되는 무역계약 조 항의 관점에서 구성하였다. 즉, 중장기적으로 남북관계의 특수성보다는 국제규범 및 관례 등에 기초한 보편성을 중심으로 한 계약서 작성이 필요하다는 것에 주안점을 두었다. 이를 통해 남북교역 시 공정거래 관 행이 확립되고 신뢰관계 형성을 위한 초석이 될 수 있기를 기대한다.

This study proposes a weight optimization technique based on Mixture Design of Experiments (MD) to overcome the limitations of traditional ensemble learning and achieve optimal predictive performance with minimal experimentation. Traditional ensemble learning combines the predictions of multiple base models through a meta-model to generate a final prediction but has limitations in systematically optimizing the combination of base model performances. In this research, MD is applied to efficiently adjust the weights of each base model, constructing an optimized ensemble model tailored to the characteristics of the data. An evaluation of this technique across various industrial datasets confirms that the optimized ensemble model proposed in this study achieves higher predictive performance than traditional models in terms of F1-Score and accuracy. This method provides a foundation for enhancing real-time analysis and prediction reliability in data-driven decision-making systems across diverse fields such as manufacturing, fraud detection, and medical diagnostics.

Defective product data is often very few because it is difficult to obtain defective product data while good product data is rich in manufacturing system. One of the frequently used methods to resolve the problems caused by data imbalance is data augmentation. Data augmentation is a method of increasing data from a minor class with a small number of data to be similar to the number of data from a major class with a large number of data. BAGAN-GP uses an autoencoder in the early stage of learning to infer the distribution of the major class and minor class and initialize the weights of the GAN. To resolve the weight clipping problem where the weights are concentrated on the boundary, the gradient penalty method is applied to appropriately distribute the weights within the range. Data augmentation techniques such as SMOTE, ADASYN, and Borderline-SMOTE are linearity-based techniques that connect observations with a line segment and generate data by selecting a random point on the line segment. On the other hand, BAGAN-GP does not exhibit linearity because it generates data based on the distribution of classes. Considering the generation of data with various characteristics and rare defective data, MO1 and MO2 techniques are proposed. The data is augmented with the proposed augmentation techniques, and the performance is compared with the cases augmented with existing techniques by classifying them with MLP, SVM, and random forest. The results of MO1 is good in most cases, which is believed to be because the data was augmented more diversely by using the existing oversampling technique based on linearity and the BAGAN-GP technique based on the distribution of class data, respectively.