AI-driven pulse thermography (PT) has become a crucial tool in non-destructive testing (NDT), enabling automatic detection of hidden anomalies in various industrial components. Current state-of-the-art PT segmentation and depth estimation networks rely on thermal features extracted from Principal Component Thermography (PCT) or Thermographic Signal Reconstruction (TSR) representations. However, relying on PCT and TSR independently constrains the performance of PT inspection models, as these representations possess complementary semantic features. To address this limitation, this work proposes PT-Fusion, a novel feature fusion network for enhanced analysis of subsurface defects in PT setups. PT-Fusion introduces novel fusion modules, Adaptive Weighing Fusion Gate (AWFG) and Gating Enhanced Decoding Block (GEDB), to adaptively fuse thermal features extracted from PCT and TSR representations. A novel data augmentation technique is also proposed based on random data sampling from thermographic sequences to address the scarcity of PT datasets. PT-Fusion is benchmarked against state-of-the-art PT inspection models, including U-Net, attention U-Net, 3D-CNN, TransUNet, and Swin-UNet on the Université Laval IRT-PVC dataset. The results demonstrate that PT-Fusion outperforms U-Net, attention U-Net, and 3D-CNN architectures in defect segmentation and depth estimation accuracies with a margin of 10%. Compared to TransUNet and Swin-UNet, the results show that PT-Fusion's performance is on par with the aforementioned models, with fewer parameters. Video Link .

Automatic anomaly detection in video surveillance is crucial for public and private safety. However, it is challenging because of unclear abnormal events, limited labeled data, and mismatches between different types of data. Traditional video anomaly detection methods mainly focus on spatiotemporal visual features. They often ignore semantic information and interactions between different data types. Additionally, many multimodal approaches use basic fusion methods that do not solve the alignment problems between these types of data. To address these issues, we propose a multimodal framework that includes a Hierarchical Multi-scale Temporal Network (H-MSTN). This network models short-, medium-, and long-term dependencies in visual and textual data. A lightweight cross-modal attention module makes sure the semantics align. Meanwhile, a Multimodal Attention-Based Fusion Transformer (MAFT) refines cross-modal representations in real time. We evaluate this framework using the UCF-Crime and XD-Violence benchmarks. The proposed method achieves 92.42% AUC on UCF-Crime and 88.63% AP on XD-Violence with significantly lower computational cost and faster inference than recent multimodal baselines such as ReFLIP-VAD. These results demonstrate a strong efficiency–accuracy trade-off for real-time deployment while maintaining competitive or improved performance over prior methods such as MVAD and TEVAD.

The increasing digitalization of industrial environments, driven by Industry 4.0, has fostered the integration of Cyber-Physical Systems (CPS), which combine Information Technology (IT) and Operational Technology (OT) to enhance efficiency, automation, and real-time decision-making. However, this convergence has also increased the exposure of Industrial Control Systems (ICS) to cyber threats, highlighting the need for tools and frameworks capable of detecting anomalies and cyberattacks. In this context, the present work proposes a testbed for evaluating Artificial Intelligence techniques aimed at the automatic detection of anomalies in industrial processes. The testbed enables the simulation of both normal and anomalous conditions following the GEMMA methodology. A use case is presented in which a supervised machine learning approach—specifically, the K-Nearest Neighbours (KNN), Random Forest (RF), and Support Vector Machine (SVM) algorithms—is applied to data generated in the Testbed for Anomaly Detection and Protection of Industrial Cyber-Physical Systems (DAyPSCI). The results demonstrate the feasibility of the proposed approach for classifying instances based on their operational behaviour, achieving high accuracy in identifying normal patterns and actuator failures, and acceptable performance in detecting sensor anomalies. These findings validate the testbed as a realistic and effective environment for the development and evaluation of intelligent anomaly-detection systems in industrial CPS. Moreover, the proposed testbed can be considered a solid foundation for future research in predictive maintenance, process optimization, and industrial cybersecurity. Keywords: Testbed, Anomaly detection, Cybersecurity, Digital Twin, Cyber-physical system, Programmable Logic Controller (PLC), Machine Learning, Industrial control.

Automatic anomaly detection is vital in domains such as healthcare, finance, and cybersecurity, where subtle deviations may signal fraud, failures, or impending risks. This paper proposes an unsupervised anomaly-detection method called Anomaly Detection Based on Markovian Geometric Diffusion (AD-MGD). The technique is applicable to uni- and multidimensional datasets, employing Markovian Geometric Diffusion to uncover nonlinear structures in the relationships among instances. For multidimensional data, the scale parameter, which is crucial to the performance of the method, is tuned using Shannon entropy. The approach includes a global search followed by local refinement of the scale parameter, promoting adaptability to the data context. Experimental evaluations on synthetic and real datasets show that AD-MGD consistently outperforms classical methods such as KNN, LOF, and IForest in terms of area under the ROC curve (AUC), particularly in heterogeneous data scenarios. The results highlight the potential of AD-MGD in critical anomaly-detection applications, advancing the use of diffusion techniques in data mining.

Monitoring the technical condition of bridge structures is a crucial aspect of their operation, as it enables early detection of damage and optimization of maintenance costs. Thanks to the implementation of monitoring systems, it is possible not only to extend the service life of bridges but also to increase user safety and reduce repair expenses. Recent publications indicate that the use of machine learning (ML) methods in bridge monitoring can significantly enhance the effectiveness and precision of conducted analyses. ML techniques are widely applied in the assessment of technical condition – they allow for automatic anomaly detection, damage classification, and forecasting of the durability of individual structural elements.

Anomaly detection is crucial to financial institutions in detecting fraud and financial crime. Financial institutions are becoming more profitable because of the increasing use of cyber technology, but this has also made them more vulnerable to financial crimes and fraud. Artificial intelligence's quick development, especially in the field of machine learning and deep learning, has offered various automatic anomaly detection models to detect anomalies better and faster than the traditional methods in fraud and financial crime; however, given the large volume of financial transactions and the challenges of identifying such transactions as fraudulent or not. Unsupervised machine learning-based anomaly detection is more effective in detecting fraud and financial crimes in the financial industry. To tackle these problems, we provide in this study an improved long short term memory (LSTM) based time-series anomaly detection method. The suggested scheme's main components include an enhanced LSTM model that might produce more accurate time-series predictions for financial crimes and fraud in financial institutions, as well as a technique for figuring out the right error threshold for anomaly detection based on prediction mistakes. To improve anomaly detection performance even more, we also suggest a pruning strategy to lower the quantity of false anomalies. The difficulty of the highly unequal distribution of financial transaction data is overcome by our approach, which dynamically establishes a threshold of prediction errors to identify anomalies instead of depending on scarce anomaly labels. We evaluated performance through a series of intensive trials in a financial transaction activity. In comparison to current anomaly detection techniques, the experimental findings show that the suggested approach performs better and is effective.

Accurate and efficient digital archive management is a crucial component of Electronic-Based Government Systems (SPBE) in Indonesia. The Integrated Dynamic Archival Information System (SRIKANDI), widely used by government agencies, continues to face various challenges such as incomplete metadata, inconsistent classification, and difficulties in archive retrieval and retention scheduling. This study aims to optimize the SRIKANDI system by implementing machine learning algorithms XGBoost for document classification and One-Class SVM (OCSVM) for automatic anomaly detection in metadata. The methodology involves data preprocessing, feature selection, label generation, and the application of classification and anomaly detection models on archival data from the Meteorological, Climatological, and Geophysical Agency (BMKG), Central Java. The XGBoost model achieved a classification accuracy of 77%, showing strong performance in identifying "Destructible" archives but limited ability in detecting the "Permanent" category due to data imbalance. Meanwhile, the OCSVM model successfully identified 16 anomalous entries (9.14%) out of 175 archives, with key indicators including extreme item counts and illogical retention periods. The results demonstrate that integrating machine learning into digital archival systems significantly improves classification accuracy, operational efficiency, and metadata integrity. Furthermore, this approach supports proactive auditing and validation of archival metadata. The findings offer valuable insights for developing AI-powered archival classification and anomaly detection systems to enhance accountability, transparency, and data governance in the public sector.

Collaborative optimization for multirobot manufacturing system reliability through integration of SysML simulation and maintenance knowledge graph

Introduction. Hearing loss affects 1 to 2 out of every 1000 newborns. Detecting anomalies in the inner ear is a challenging task even for experienced specialists.Aim. To develop a fully automated sequence of commands with a pipeline data transfer for the classification of inner ear defects and processing of CT images of inner ear anomalies in patients using this program data.Materials and мethods. This study presents the first automated method for classifying congenital inner ear anomalies. In the experimental part, a 3D cochlear structure network was developed for 346 standard and 121 atypical structures using a common segmentation scheme trained exclusively on normal anatomy. From 2018 to 2024, 98 patients were examined at the Federal State Budgetary Institution of Science, Otolaryngology, Federal Medical and Biological Agency of Russia, including 54 (55.5%) boys and 44 (44.5%) girls aged from 8 months to 6 years (average age 2.5 years) with inner ear developmental anomalies and severe hearing impairments, who subsequently underwent cochlear implantation.Results. We achieved a generalized average accuracy of 77% across 7 different pathological subgroups of disorders compared to the professional diagnosis of an otolaryngologist specializing in congenital inner ear defects.Discussion. Although automatic detection of various types of inner ear anomalies is essentially a classification task, the lack of representative and heterogeneous datasets that accurately represent the diversity of these congenital developmental defects necessitates the use of a parametric approach. This method is employed with standard data to extract implicit information that could potentially detect anomalies in a non-controlled manner.Conclusions. We proposed the first method for the automatic detection of congenital anomalies of the inner ear and demonstrated that the use of 3D information about the shape of the cochlea, extracted using a model trained exclusively on standard structures, is sufficient for classifying developmental defects.

The rapid advancement of Artificial Intelligence (AI) is transforming the construction sector, particularly in site monitoring and safety management. Real-time monitoring enables the automatic detection of work progress issues, anomalies, and hazardous situations. However, no existing Deep Learning (DL)-based system is specifically designed to utilize Unmanned Aerial Vehicles (UAVs) for excavation area monitoring. This study presents an automated workflow that integrates UAV imagery with DL architectures, featuring a 1D Convolutional Neural Network (1D-CNN) for classifying excavation work phases and a VGG16 network for detecting safety fences. These technologies are incorporated into a Decision Support System (DSS), which automates report generation and enhances decision-making by providing structured, data-driven insights. The system was validated in a real-world case study involving an oil and gas construction company, demonstrating its ability to streamline site management tasks and improve safety oversight. Compared to traditional monitoring methods, our approach leverages UAV technology and DL methodologies to provide higher accuracy, efficiency, and scalability in excavation site monitoring. This contribution supports the digital transformation of construction management, offering a practical and innovative solution for real-time progress tracking and compliance verification.

(1) Background and objective: Mobility is crucial for healthy aging, and its loss significantly impacts the quality of life, healthcare costs, and mortality among older adults. Clinical mobility assessment methods, though precise, are resource-intensive and economically impractical, and most of the existing solutions for automatic detection of mobility anomalies are either obtrusive or improper for long time monitoring. This study explores the feasibility of using non-intrusive, low-cost binary sensors for continuous, remote detection of mobility anomalies in older adults, aiming to identify both sudden mobility events and gradual mobility loss. (2) Method: The study utilized publicly available datasets (CASAS Aruba and HH120) containing annotated activity data recorded from binary sensors installed in residential environments. After data preprocessing-including filtering irrelevant sensor events and aggregation into behaviorally meaningful places (BMPs)-a time series forecasting model (Prophet) was used to predict normal mobility patterns. A fuzzy inference module analyzed deviations between observed and predicted sensor data to determine the probability of mobility anomalies. (3) Results: The system effectively identified periods of prolonged inactivity indicative of potential falls or other mobility disruptions. Preliminary evaluation indicated a detection rate of approximately 77-81% for point mobility anomalies, with a false positive rate ranging from 12 to 16%. Additionally, the approach successfully detected simulated gradual declines in mobility (1% per day reduction), evidenced by statistically significant regression trends in activity levels over time. (4) Conclusions: The study argues that non-intrusive binary sensors, combined with lightweight forecasting models and fuzzy inference, may provide a practical and scalable solution for detecting mobility anomalies in older adults. Although performance can be further enhanced through improved data preprocessing, predictive modeling, and anomaly threshold tuning, the proposed system effectively addresses key limitations of existing mobility assessment approaches.

Towards automatic anomaly detection and diagnosis in positional arc-directed energy deposition based on deep learning

The implementation of innovative technologies is radically transforming approaches to crime investigation and prevention, opening new avenues to enhance their effectiveness. Technologies such as artificial intelligence, biometric systems, cutting-edge data collection and preservation methods, and video surveillance with automatic anomaly detection, among others, allow for faster and more accurate identification of crimes, their patterns, and the ability to predict the further actions of offenders. This not only enables the investigation of crimes that have already occurred but also helps to prevent them at various stages. At the same time, technologies significantly reduce the level of human error in crime investigation processes, improving the accuracy and reliability of the results obtained. The automation of processes not only enhances the precision of investigations but also allows law enforcement agencies to reduce time spent on routine tasks, enabling them to focus resources on more complex aspects of the investigation. In the context of globalization, the integration of technologies into international cooperation between law enforcement agencies from different countries is also crucial. This enables the rapid exchange of information and coordination of actions in real time. All of this requires continuous study and adaptation of new technological capabilities to the specifics of forensic practice. This topic is of particular relevance, especially in light of the global changes occurring in the modern world. On one hand, technological progress is accelerating, while on the other, the nature of crime itself is evolving, becoming increasingly complex, dynamic, and globalized. Traditional methods of combating crime, which relied on human factors and conventional tools, are not always capable of effectively addressing new challenges. Contemporary crimes are increasingly digital and cybernetic in nature, demanding the use of specific technological tools for investigation. Therefore, the study of the role of innovative technologies in crime investigation and prevention is not only scientifically significant but also practically essential for adapting the justice system to modern challenges in a globalized and technologically advanced world.

In information systems, data is collected, stored and processed, as well as information, knowledge and digital products are protected. The best solution for these requirements is an ISP network monitoring system. ISP networks are distributed and heterogeneous, which makes monitoring crucial for their health. For continuous monitoring, some monitoring systems use open-source software tools. It contributes to the monitoring of systems, applications, services, networks, and infrastructure. We can use these tools to detect network or server problems, share network services or devices among specific groups, generate alerts by simplifying the monitoring process, and save logs for creating reports. Our paper discusses how an ISP can monitor its network using a system such as Nagios. This open-source system provides real-time monitoring and alerts network administrators to anomalies as soon as they occur. We will also present the advantages that Nagios offers to an Internet Service Provider (ISP) for monitoring network equipment and services. We focus on using Nagios as a tool for automatic detection of anomalies occurring in the network topology as well as notification to the Network Monitoring Center (NMC) via email or web interface access.

In order to improve the service quality of the power grid and ensure the safe and stable operation of the power grid, an in-depth study is carried out on the automatic detection and positioning of abnormal power metering. First, the data clustering algorithm is used to cluster the electric energy metering data. Next, the graph convolutional neural network algorithm is used to automatically detect and locate the abnormal data of electric energy metering. A graph is constructed, and the subgraphs are divided. These subgraphs are then sent to the graph convolutional neural network for processing. Feature extraction is carried out on the points and edges of the graph. This process completes the automatic detection and location of the electric energy metering anomaly based on the graph convolutional neural network. The experimental results show that the proposed algorithm can effectively improve the efficiency and accuracy of automatic detection and positioning of electric energy metering anomalies, and can better meet the needs of actual power work.

The Internet of Things (IoT) is being prominently used in smart cities and a wide range of applications in society. The benefits of IoT are evident, but cyber terrorism and security concerns inhibit many organizations and users from deploying it. Cyber-physical systems that are IoT-enabled might be difficult to secure since security solutions designed for general information/operational technology systems may not work as well in an environment. Thus, deep learning (DL) can assist as a powerful tool for building IoT-enabled cyber-physical systems with automatic anomaly detection. In this paper, two distinct DL models have been employed i.e., Deep Belief Network (DBN) and Convolutional Neural Network (CNN), considered hybrid classifiers, to create a framework for detecting attacks in IoT-enabled cyber-physical systems. However, DL models need to be trained in such a way that will increase their classification accuracy. Therefore, this paper also aims to present a new hybrid optimization algorithm called “Seagull Adapted Elephant Herding Optimization” (SAEHO) to tune the weights of the hybrid classifier. The “Hybrid Classifier + SAEHO” framework takes the feature extracted dataset as an input and classifies the network as either attack or benign. Using sensitivity, precision, accuracy, and specificity, two datasets were compared. In every performance metric, the proposed framework outperforms conventional methods.

Abstract The complexity and fusion dynamism of the modern industrial and chemical sectors have been increasing with the rapid progress of IR 4.0–5.0. The transformative characteristics of Industry 4.0–5.0 have not been fully explored in terms of the fundamental importance of explainability. Traditional monitoring techniques for automatic anomaly detection, identifying the potential variables, and root cause analysis for fault information are not intelligent enough to tackle the intricate problems of real-time practices in the industrial and chemical sectors. This study presents a novel dynamic machine learning based explainable fusion approach to address the issues of process monitoring in industrial and chemical process systems. The methodology aims to detect faults, identify their key causes and feature variables, and analyze the root path of fault propagation with the time and magnitude of one cause variable to another impact. This study proposed using a time domain multivariate granger-entropy-aided dynamic independent component analysis (DICA)—distributed canonical correlation analysis approach, incorporating the dynamics time wrapping supported time delay-signed directed graph. The proposed methodology utilized the application to industrial and chemical processes and verified using the continuous stirred tank reactor and Tennessee Eastman process as practical application benchmarks. The framework’s validations and efficiency are evaluated using established techniques such as classic computed ICA and DICA as standard model scenarios. The outcomes and results showed that the newly developed strategy is preferable to previous approaches regarding explainability and robust detection and identification of the actual root causes with high FDRs and low FARs.

Deep Learning Convolution Auto Encoder Framework for Automatic Anomaly Detection in Quadratic Frequency Modulated Thermography

A convolutional autoencoder framework for ECG signal analysis.

The challenges of designing intelligent systems that allow automated classification of problems and decision-making in complex environments are diverse. Particularly difficult are dispersed environments, where the created system must take into account the local nature of the data. Such data may differ in the number of attributes, but also in other aspects, which often requires specific normalization or some other form of information unification. This article contains conclusions from the work on the design of an intelligent system for early detection of problems in truck combustion engines. The environment in this situation is scattered and the data is often heterogeneous. The process included the selection of relevant attributes occurring in different vehicle manufacturers and attempts to automatically detect anomalies. The next phase was to build a classifier based on artificial neural networks. In the end, the system took the form of a rule-based expert system. The work makes one see that less complex expert systems can, in many cases, significantly outperform more advanced techniques that require labeled data sets of very good quality. The advantage of expert systems for local datasets is that they can be customized on the basis of available knowledge, be it expert or catalog data, without affecting other distributed models. However, their greatest advantage is that they can be practically applied even when the initial amount of information and training data is relatively small.