This study applies MTAD-GAT to early anomaly detection in unmanned aerial vehicle (UAV) flight telemetry using the ALFA fault dataset. Thirty-one telemetry variables are arranged as lookback windows, and a graph-attention architecture is used to jointly model inter-feature relationships and temporal dependencies. To keep the evaluation compact, we rely on three indicators: PR-AUC for timestep-level anomaly ranking, Flight AUROC for flight-level fault separation, and mean time-to-detection (TTD) for early-warning capability. Under the operational setting, MTAD-GAT achieves a PR-AUC of 0.5282, a Flight AUROC of 0.8889, and a mean TTD of 15.13 timesteps. The same setting also records the highest PR-AUC among the conventional unsupervised baselines, indicating that jointly modeling feature interactions and temporal context is beneficial for UAV fault detection. The contribution of this work is to reframe UAV anomaly detection not only as a timestep-level scoring task but also as a flight-level early-warning problem.

Plastic injection molding is widely used in the automotive, electronics, and other manufacturing industries. Since product quality is significantly affected by process variables such as temperature, pressure, speed, and cycle time, datadriven quality management has become increasingly important. However, conventional quality control methods mainly depend on operator experience and post-process inspection, making it difficult to detect defects and abnormal process conditions in real time. In addition, manufacturing datasets often suffer from severe class imbalance because defective samples are much fewer than normal samples. This study proposes an AI-based defect prediction and anomaly detection system using plastic injection molding process data. For defect prediction, XGBoost was selected as the supervised learning model, and SMOTE was applied to address class imbalance. Recall, F1-score, and PR-AUC were used as the main evaluation metrics instead of accuracy. SHAP analysis was also applied to identify important process variables affecting defect occurrence. For anomaly detection, Isolation Forest and AutoEncoder were used together. Isolation Forest was adopted for fast first-stage detection, while AutoEncoder was used to detect complex nonlinear abnormal patterns. The results indicate that the proposed system can support both defect prediction and process anomaly detection. In the supervised learning results, XGBoost showed the best performance on the CN7 dataset, achieving a recall of 1.000 and an F1-score of 0.800. However, all supervised models showed limited defect detection performance on the RG3 dataset, indicating the effect of severe class imbalance and dataset-specific characteristics. In anomaly detection, AutoEncoder achieved the highest F1-score among the compared models, while Isolation Forest demonstrated practical advantages for real-time application due to its computational efficiency. Therefore, the proposed AI-based system can contribute to smart quality management and data-driven decision-making in plastic injection molding processes.

상수관망의 누수는 수자원 손실 및 시설물 피해의 주요 원인으로, 효과적인 탐지를 위해 다양한 기술이 개발되고 있다. 본 연구는 실제 상수도 누수 사례를 대상으로 청음 데이터를 수집하고, 주파수 및 청각 기반 음향 특징을 추출하여 비지도 학습 기반의 이상 감지 모델을 적용함으로써 누수음을 탐지하는 기법을 제안한다. 청음 신호에 대해 푸리에 변환과 멜 주파수 켑스트럼 계수(MFCC)를 적용하여 총 86개의 음향 특징을 구성하였으며, 랜덤 포레스트를 통해 주요 변수 6개를 선정하였다. 거리 기반 군집 분석을 통해 정상 소음 분포를 구성하고, Isolation Forest 및 Autoencoder 알고리즘을 활용하여 이상 음향을 판별하였다. 비지도 모델에 의해 탐지된 이상 지점이 실제 현장 판단과 부합함을 확인하였다. 정량적 기준 기반의 이상 탐지 결과가 실제 누수 지점과 일관성을 보였다. 본 연구는 실무 적용 가능한 비지도 이상 탐지 접근법을 제시함으로써, 기존 라벨 의존 탐지 방식의 한계를 보완할 수 있음을 시사한다.

Anomaly detection is crucial for ensuring the reliability and safety of mechanical systems across industries such as power generation, manufacturing, and transportation. In these mechanical systems, data is usually collected in time-series form using sensors such like vibration, current or sound for anomaly detection. Time-series anomaly detection methods often face limitations due to insufficient training data and poor generalization across complex operating conditions and varying loads. To address these challenges, this study proposes a transfer learning-based anomaly detection model, leveraging pre-trained knowledge to deliver robust performance and adaptability in data-scarce scenarios and diverse industrial environments. To this end, time-series signals are transformed into spectrograms through Short-Time Fourier Transform(STFT), followed by feature extraction through a Convolutional Autoencoder to obtain low-dimensional latent features. These features are used to detect anomalies using classification such as Random Forest and eXtreme Gradient Boosting. Building on this approach, this research validates the model's performance through migration tasks using the Case Western Reserve University(CWRU) Bearings dataset. Furthermore, to show cross-condition generalization, the proposed model was validated on the Hanoi University of Science and Technology(HUST), Sumair–Umar Bearing Fault(SUBF) dataset v2.0, and a dataset collected using microphone sensor in motor dynamo tests. Consequently, unlike other studies limited by specific operating conditions, the proposed model exhibits strong generalization performance across benchmark datasets. Experimental results highlight the effectiveness of combining STFT, CAE, and tree-based classifiers in addressing data scarcity and enhancing generalization, making it highly suitable for real-world industrial applications. Future work will focus on noise-robust techniques and broader fault types to further improve performance.

The purpose of this study is to evaluate the applicability of an unsupervised outlier-detection method as a surrogate safety measure (SSM) to estimate the effect of AI-based Bike-Safe monitoring system. An SSM that utilizes near-miss data immediately before an accident occurs must be developed to compensate for inadequate bicycle accident data and missing reports. In particular, the omission level of accident reports related to bicycle users is higher on bicycle paths, which implies that the importance of an SSM in safety management is much greater than in the general road environment. Therefore, the unsupervised outlier-detection method was set as the SSM because it can be learned without a label, is suitable for streaming data, and is generalizable under limited data. Additionally, the DeepAnT(deep learningbased anomaly detection) model was selected as the most appropriate time-series outlier-detection method. Using the time-series prediction module of the learned DeepAnT model, we analyzed the frequency of outliers or avoidance behaviors based on a linear relationship between estimated and observed values. The history data of the acceleration change rate of each bicycle were applied to the DeepAnT model to evaluate the possibility of using alternative safety indicators. Thus, those data are expected to be applicable as an alternative safety indicator for bicycle paths.

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.

Anomaly detection technique for the Unmanned Aerial Vehicles (UAVs) is one of the important techniques for ensuring airframe stability. There have been many researches on anomaly detection techniques using deep learning. However, most of research on the anomaly detection techniques are not consider the limited computational processing power and available energy of UAVs. Deep learning model convert to the model compression has significant advantages in terms of computational and energy efficiency for machine learning and deep learning. Therefore, this paper suggests a real-time anomaly detection model for the UAVs, achieved through model compression. The suggested anomaly detection model has three main layers which are a convolutional neural network (CNN) layer, a long short-term memory model (LSTM) layer, and an autoencoder (AE) layer. The suggested anomaly detection model undergoes model compression to increase computational efficiency. The model compression has same level of accuracy to that of the original model while reducing computational processing time of the UAVs. The proposed model can increase the stability of UAVs from a software perspective and is expected to contribute to improving UAVs efficiency through increased available computational capacity from a hardware perspective.

안정적이고 효율적인 수자원 공급을 보장하는 것은 가정, 산업, 공공 보건 분야 복지에 필수적이다. 상수도 시스템에서 이상을 감지하기 위해 데이터 모델, 수리 모델 기반 시뮬레이션 등 다양한 접근 방식을 통해 이상감지 역량이 꾸준히 향상되어 왔으나, 현장 적용 및 검증에 한계가 있어 실질적인 활용은 폭 넓게 이루어지지 못하고 있다. 실제 적용 가능한 이상감지 시스템 측면에서, 본 연구에서는 유량 및 압력 계측 데이터를 활용하여 시스템 내 이상 발생을 신속하게 감지하고 개략적인 위치를 파악하기 위한 실시간 이상감지 및 탐색 모델을 제안하였다. 제안된 모델은 유량수지 분석, 유량-수두손실 관계, EPANET 기반 수리 해석 방법을 통합하여 이상 감지 및 위치 파악의 정확성을 개선시키고자 하였다. 현장 실험 결과, 제안된 모델은 높은 신뢰도에서 시스템 내 이상유량의 발생을 효과적으로 감지하고, 발생 구간을 파악할 수 있는 것으로 나타났다. 본 연구 성과는 시스템의 실시간 이상 감지 및 운영관리를 위한 실효성 있는 접근 방식을 제공함으로써 상수도 시스템의 지속 가능하고 회복력 있는 운영관리에 기여할 것으로 기대된다.

New motor development requires high-speed load testing using dynamo equipment to calculate the efficiency of the motor. Abnormal noise and vibration may occur in the test equipment rotating at high speed due to misalignment of the connecting shaft or looseness of the fixation, which may lead to safety accidents. In this study, three single-axis vibration sensors for X, Y, and Z axes were attached on the surface of the test motor to measure the vibration value of vibration. Analog data collected from these sensors was used in classification models for anomaly detection. Since the classification accuracy was around only 93%, commonly used hyperparameter optimization techniques such as Grid search, Random search, and Bayesian Optimization were applied to increase accuracy. In addition, Response Surface Method based on Design of Experiment was also used for hyperparameter optimization. However, it was found that there were limits to improving accuracy with these methods. The reason is that the sampling data from an analog signal does not reflect the patterns hidden in the signal. Therefore, in order to find pattern information of the sampling data, we obtained descriptive statistics such as mean, variance, skewness, kurtosis, and percentiles of the analog data, and applied them to the classification models. Classification models using descriptive statistics showed excellent performance improvement. The developed model can be used as a monitoring system that detects abnormal conditions of the motor test.

In this study, we present a sewer pipe inspection technique through a combination of active sonar technology and deep learning algorithms. It is difficult to inspect pipes containing water using conventional CCTV inspection methods, and there are various limitations, so a new approach is needed. In this paper, we introduce a inspection method using active sonar, and apply an auto encoder deep learning model to process sonar data to distinguish between normal and abnormal pipelines. This model underwent training on sonar data from a controlled environment under the assumption of normal pipeline conditions and utilized anomaly detection techniques to identify deviations from established standards. This approach presents a new perspective in pipeline inspection, promising to reduce the time and resources required for sewer system management and to enhance the reliability of pipeline inspections.

Deep learning-based computer vision anomaly detection algorithms are widely utilized in various fields. Especially in the manufacturing industry, the difficulty in collecting abnormal data compared to normal data, and the challenge of defining all potential abnormalities in advance, have led to an increasing demand for unsupervised learning methods that rely on normal data. In this study, we conducted a comparative analysis of deep learning-based unsupervised learning algorithms that define and detect abnormalities that can occur when transparent contact lenses are immersed in liquid solution. We validated and applied the unsupervised learning algorithms used in this study to the existing anomaly detection benchmark dataset, MvTecAD. The existing anomaly detection benchmark dataset primarily consists of solid objects, whereas in our study, we compared unsupervised learning-based algorithms in experiments judging the shape and presence of lenses submerged in liquid. Among the algorithms analyzed, EfficientAD showed an AUROC and F1-score of 0.97 in image-level tests. However, the F1-score decreased to 0.18 in pixel-level tests, making it challenging to determine the locations where abnormalities occurred. Despite this, EfficientAD demonstrated excellent performance in image-level tests classifying normal and abnormal instances, suggesting that with the collection and training of large-scale data in real industrial settings, it is expected to exhibit even better performance.

Smart factory companies are installing various sensors in production facilities and collecting field data. However, there are relatively few companies that actively utilize collected data, academic research using field data is actively underway. This study seeks to develop a model that detects anomalies in the process by analyzing spindle power data from a company that processes shafts used in automobile throttle valves. Since the data collected during machining processing is time series data, the model was developed through unsupervised learning by applying the Holt Winters technique and various deep learning algorithms such as RNN, LSTM, GRU, BiRNN, BiLSTM, and BiGRU. To evaluate each model, the difference between predicted and actual values was compared using MSE and RMSE. The BiLSTM model showed the optimal results based on RMSE. In order to diagnose abnormalities in the developed model, the critical point was set using statistical techniques in consultation with experts in the field and verified. By collecting and preprocessing real-world data and developing a model, this study serves as a case study of utilizing time-series data in small and medium-sized enterprises.

본 논문에서는 대규모 실시간 매칭의 생존 게임에서 플레이를 위한 유저들의 소셜 관계에 대해 연구한다. 특 히 “사전 팀 구성”을 통한 자의적인 팀 구성이 어떤 방식으로 유저들을 연결하는 지 연구하고자 한다. 다수 의 사람 간 집단 역학에서 나타나는 특성이나 패턴에 대한 조사를 중심으로 하였으며, 개인의 특성은 보조적 인 수단으로만 사용된다. 이번 연구에서는 게임을 플레이하는 유저들의 익명화 된 대규모 데이터를 활용하며 이에 대한 간소화된 집계 방법을 제안한다. 데이터 세트에는 사전 팀 구성에 관한 11,259만 줄의 속성이 포 함되어 있으며, 데이터에서 우리는 250만개의 노드와 1,182만개의 무방향 에지가 있는 협업 네트워크를 구성 하여 대규모 게임 내 협동 네트워크를 만듭니다. 연결 정도, 경로 길이, 클러스터링 및 소속 하위 컴포넌트의 크기 등 네트워크에 관한 수치를 통해 게임내 소셜 활동에 대한 이해를 높이고자 한다. 본 논문에서는 다음 의 두가지 특성을 중심으로 결론을 제시한다. 첫째, 네트워크 내에는 대규모로 연결된 2개(전체의 44% 및 2%)와 나머지의 파편화된 하위 컴포넌트로 구성 되어있다. 이 대규모 컴포넌트 중 작은 쪽은 한국 유저로만 구성되어 있다. 둘째, 컴포넌트 크기 별 평균 연결 거리와 군집화 계수, k-core를 확인함으로써 기타 다른 네 트워크 대비 이웃 간 연결이 강하면서 전체적으로는 비교적 멀리 떨어져 있음을 확인한다.

본 연구는 네트워크 이상 감지 및 예측을 위해 벡터 자기회귀(VAR) 모델의 사용을 비교 분석한다. VAR 모 델에 대한 간략한 개요를 제공하고 네트워크 이상 체크로 사용 가능한 두 가지 버전을 검토하며 두 종류의 VAR 모델을 통한 경험적인 평가를 제시한다. VAR-Filtered moving-common-AR 모델이 단일 노드 이상 감지 성능에서 우수하며, VAR-Adaptive Learning 버전은 몇 개의 노드 간 이상을 효과적으로 식별하는 데 특히 효 과적이며 두 가지 주요VAR 모델의 전반적인 성능 차이에 대한 근본적인 이유도 분석한다. 각 기술의 장단점 을 개요로 제공하고 성능 향상을 위한 제안도 제시하고자 한다.

해상교통관제센터(VTS)의 관제사는 구역 내 교통 상황을 빠르고 정확하게 파악하여 관제가 필요한 선박에게 정보를 제공하는 역할을 수행한다. 그러나 교통량이 급격히 증가하는 경우 관제사의 업무 부하로 인해 관제 공백이 발생하기도 한다. 이러한 이유에서 관 제사의 업무 부하를 줄이고, 일관성 있는 관제 정보를 제공할 수 있는 관제 지원 기술의 개발이 필요한 실정이며, 본 논문에서는 구역 내 이상 운항 선박을 자동으로 식별하는 모델을 제안하였다. 제안하는 이상 운항 식별 모델은 규칙 기반 모델, 위치 기반 모델, 맥락 기반 모 델로 구성되며, 대상 해역의 교통 특성에 최적화된 교통 네트워크 모델을 사용하는 특징이 있다. 구현된 모델은 시범센터(대산항 VTS)에 서 수집되는 실해역 데이터를 적용하여 실험을 수행하였다. 실험을 통해 실해역의 다양한 이상 운항 상황이 자동으로 식별됨을 확인하였 고, 전문가 평가를 통해 식별 결과를 검증하였다.

Recently, many studies have been conducted to improve quality by applying machine learning models to semiconductor manufacturing process data. However, in the semiconductor manufacturing process, the ratio of good products is much higher than that of defective products, so the problem of data imbalance is serious in terms of machine learning. In addition, since the number of features of data used in machine learning is very large, it is very important to perform machine learning by extracting only important features from among them to increase accuracy and utilization. This study proposes an anomaly detection methodology that can learn excellently despite data imbalance and high-dimensional characteristics of semiconductor process data. The anomaly detection methodology applies the LIME algorithm after applying the SMOTE method and the RFECV method. The proposed methodology analyzes the classification result of the anomaly classification model, detects the cause of the anomaly, and derives a semiconductor process requiring action. The proposed methodology confirmed applicability and feasibility through application of cases.

When developing a new motor, a high-speed load test is performed using dynamo equipment to calculate the efficiency of the developed motor using the collected dynamo data. When connecting the test motor and the dynamo used as a load, abnormal noise and vibration may occur in the test equipment rotating at high speed due to misalignment of the connecting shaft and looseness of the connection, which may lead to a safety accident. In this study, three vibration sensors are attached to the surface of bearing parts of the test motor to measure the vibration value, and statistics such as kurtosis, skewness, and percentiles are obtained in order to clearly express the pattern of the measurement data. With these statistics, machine learning models are developed. The developed model in this way can be used as a diagnostic system that can detect abnormal conditions of the motor test equipment through monitoring the motor vibration data during the motor test.