Детальная информация

This thesis tackles the challenge of semantic interoperability in real-time industrial automation. It introduces the Real-Time Semantic Broker, a modular framework that ingests SCADA, OPCUA, and Modbus streams and converts raw tag-value data into semantically enriched RDF triples within sub-second intervals. The methodology combines protocol adapters, an in-memory tag-to-ontology mapping engine, and a non-blocking streaming RDF generator. Online learning streaming for anomaly detection and adaptive forecasting is embedded directly into the pipeline, enabling continuous model updates without batch retraining. Initial proof-of-concept demonstrates that the broker accelerates decision-making, boosts equipment reliability through early anomaly warnings, and optimizes process throughput with context-aware predictions. Secure bidirectional OPC UA↔Modbus bridging preserves on-site data sovereignty while enabling seamless cross-system integration. Applicable to manufacturing, energy, and utilities, this approach empowers smarter, more responsive operations. Future work will explore edge-native deployments, standardized semantic profiles, and closed-loop control integration to pave the way for fully autonomous industrial ecosystems.

Данная диссертация посвящена проблеме семантической интероперабельности в области промышленной автоматизации в режиме реального времени. В работе представлен семантический брокер реального времени-модульный фреймворк, предназначенный для приема потоков данных от систем SCADA, OPCUA и Modbus с последующим преобразованием исходных значений тегов в семантически обогащенные RDF-триплеты за интервал времени менее одной секунды. Предложенная методология объединяет протокольные адаптеры, движок отображения тегов на онтологии, функционирующий в оперативной памяти, и неблокирующий потоковый генератор RDF. В конвейер обработки данных непосредственно встроены онлайн-алгоритмы машинного обучения для выявления аномалий и адаптивного прогнозирования, обеспечивающие непрерывное обновление моделей без необходимости пакетного переобучения. Проведённые первоначальные эксперименты, демонстрирующие работоспособность концепции, показали, что брокер ускоряет процесс принятия решений, повышает надежность оборудования за счет раннего предупреждения об аномалиях и оптимизирует производительность процессов посредством контекстно-ориентированных прогнозов. Обеспечиваемый брокером безопасный двунаправленный мост OPC UA↔Modbus сохраняет суверенитет данных на месте их сбора, одновременно обеспечивая беспрепятственную интеграцию между различными системами. Предлагаемый подход может применяться в производственной сфере, энергетике и коммунальном хозяйстве, обеспечивая более интеллектуальное и оперативное управление операциями. Дальнейшие исследования будут направлены на изучение возможности развертывания на периферийных устройствах (edge-native), стандартизацию семантических профилей и интеграцию с замкнутыми контурами управления, что откроет путь к созданию полностью автономных промышленных экосистем.

• INTRODUCTION
• 1. Background and Preliminaries
  • 1.1 Industry 4.0: Concepts and Goals
  • 1.2 Industrial Control Applications and SCADA
    • 1.2.1 Functionality and Architecture
    • 1.2.2 Data Characteristics: Real-Time Streams, Tags, Values
    • 1.2.3 Heterogeneity in SCADA Environments
  • 1.3 The Imperative of Interoperability
    • 1.3.1 Defining Interoperability Levels
    • 1.3.2 Semantic Interoperability Conflicts in Industrial Control Environments
  • 1.4 Semantic Technologies: Foundations for Meaning
    • 1.4.1 Resource Description Framework (RDF)
    • 1.4.2 Ontologies: RDFS and OWL
    • 1.4.3 SPARQL Protocol and RDF Query Language
    • 1.4.4 Knowledge Graphs
  • 1.5 Machine Learning Fundamentals for Real-Time Data
    • 1.5.1 Online Learning vs. Batch Learning
    • 1.5.2 Key Concepts in Online Learning
    • 1.5.3 Use Cases in Industrial Control Environments: Anomaly Detection and Forecasting
• 2. Related Work
  • 2.1 Additional Close-Related Work
    • 2.1.1 SCADA System for Process Data Exchange in Master-Slave RF and IoT Network [1]
    • 2.1.2 The Importance of Semantic Interoperability in the Internet of Things [2]
    • 2.1.3 An Approach to Semantic Interoperability for Product Development Through Automatic Requirement Extraction and Semantic Reconciliation [3]
    • 2.1.4 Semantic Interoperability: Challenges in the Digital Transformation Age [4]
    • 2.1.5 Semantic Interoperability: Information Integration by Using Ontology Mapping in Industrial Environment [5]
    • 2.1.6 SCADA & Simulation Integration Framework [6]
  • 2.2 Semantic Data Integration Approaches
    • 2.2.1 Ontology-Based Data Integration (OBDI)
    • 2.2.2 Data Linking and Fusion Frameworks
    • 2.2.3 Limitations in Real-Time Industrial Contexts
  • 2.3 Knowledge Graphs in Industrial Applications
    • 2.3.1 Enterprise Knowledge Graphs
    • 2.3.2 Semantic Modeling of Standards
    • 2.3.3 Static vs. Real-Time KGs
  • 2.4 Real-Time Data Processing and Streaming Analytics
    • 2.4.1 Stream Processing Engines
    • 2.4.2 Semantic Stream Processing Challenges
  • 2.5 Machine Learning Applications in SCADA/Industrial IoT
    • 2.5.1 Existing Approaches
    • 2.5.2 Gaps: Semantic Context and Online Adaptation
  • 2.6 Positioning the Thesis Contribution
• 3. A Semantic Interoperability for Real-Time Industrial Applications
  • 3.1 Rationale for Real-Time Semantic Interoperability
    • 3.1.1 Methodology Overview
  • 3.2 Proposed Architecture: Real-Time Semantic Broker
    • 3.2.1 Data Acquisition Layer
    • 3.2.2 Semantic Transformation Engine
    • 3.2.3 Ontology Management Component
    • 3.2.4 Real-Time Semantic Data Bus
    • 3.2.5 Consumer Layer
  • 3.3 Real-Time Semantic Modeling Process
    • 3.3.1 Input Collection
    • 3.3.2 Tag-to-Ontology Mapping
    • 3.3.3 Graph Construction
    • 3.3.4 Stream Publication
    • 3.3.5 Formal Algorithm Specification
    • 3.3.6 Formal Semantic Model Examples
  • 3.4 Ontology Design Principles
  • 3.5 Resolving Interoperability Conflicts
  • 3.6 Performance Considerations
  • 3.7 Applications and Use Cases
    • 3.7.1 Link Sync: Unified Real-Time Visualization
    • 3.7.2 Real-Time Semantic Interoperability (Bridging)
    • 3.7.3 Semantic Querying (SPARQL)
    • 3.7.4 Machine Learning Insights (Anomaly Detection and Forecasting)
  • 3.8 Chapter Summary
• 4. Online Machine Learning Integration for Real-Time Industrial Applications
  • 4.1 Rationale: Context-Aware Intelligence
  • 4.2 Architectural Integration of the ML Engine
    • 4.2.1 Subscribing to Relevant Data Streams
    • 4.2.2 Interaction with Semantic Layer/Ontology
  • 4.3 Online Learning Algorithms for Industrial Data Streams
    • 4.3.1 Addressing the "No History" Challenge
    • 4.3.2 Overview of Suitable Algorithms
    • 4.3.3 Mathematical Formulation
  • 4.4 Leveraging Semantic Context for Enhanced ML
    • 4.4.1 Semantic Feature Grouping for Multivariate Models
    • 4.4.2 Contextual Model Selection and Configuration
    • 4.4.3 Interpreting ML Outputs (Alerts, Forecasts)
  • 4.5 Use Case Implementation: Anomaly Detection
    • 4.5.1 Data Flow and Processing Steps
    • 4.5.2 Model Learning and Scoring in Real-Time
    • 4.5.3 Generating Contextual Alerts
  • 4.6 Use Case Implementation: Short-Term Forecasting
    • 4.6.1 Selecting Tags and Models Based on Semantics
    • 4.6.2 Real-Time Prediction and Model Updates
    • 4.6.3 Presenting Forecasts with Context
  • 4.7 Challenges and Considerations for Online ML Integration
  • 4.8 Chapter Summary
• 5. Application and Insights
  • 5.1 Strategic Impact and Value Proposition
  • 5.2 Deployment Considerations and Roadmap
    • Phase 1: Assessment and Governance
    • Phase 2: Pilot Integration
    • Phase 3: Scale-Out and Optimization
  • 5.3 Lessons Learned and Best Practices
  • 5.4 Future Outlook: Extending Industrial Intelligence
  • 5.5 Chapter Summary
• 6. Future Direct
  • 6.1 Summary of Contributions and Findings
  • 6.2 Revisiting the Research Questions
  • 6.3 Limitations of the Current Work
  • 6.4 Future Research Directions
    • 6.4.1 Scalability and Performance Optimization
    • 6.4.2 Advanced Online ML Algorithms and XAI
    • 6.4.3 Automated Ontology Learning and Mapping Refinement
    • 6.4.4 Federated Learning and Edge Processing
    • 6.4.5 Towards Closed-Loop Control and Autonomous Systems
• CONCLUSION
• REFERENCE LIST
• APPENDIX A
• (mandatory information)
• Tag-to-Ontology Mapping Algorithm
• APPENDIX B
• (mandatory information)
• Streaming RDF Generation Algorithm
• APPENDIX C
• (mandatory information)
• Continuous Enrichment Loop
• APPENDIX D
• (mandatory information)
• Namespaces Configuration
• APPENDIX E
• (mandatory information)
• Ontologies Mapping
• APPENDIX F
• (mandatory information)
• Abbreviations
• APPENDIX G
• (mandatory information)
• Software and Libraries
• APPENDIX H
• (mandatory information)
• Turtle Representation
• APPENDIX I
• (mandatory information)
• OWL Class Definitions
• APPENDIX J
• (mandatory information)
• SPARQL Query Example
• APPENDIX K
• (mandatory information)
• Python Graph Generation Code