Multivariable Fault Detection Using Dynamic ML Models and Data Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fault detection systems for multi-variable systems, such as HVAC, are limited by their rule-based and model-based approaches, which are inflexible, complex, and prone to false alarms, making it difficult to adapt to changing conditions and apply to different systems effectively.
Innovation Solution
A method using dynamic machine learning fault detection models, including Dynamic Bayesian Networks and Hidden Markov Models, processes operational data to detect normal or faulty operation, with data fusion and confidence level calculations to provide accurate fault diagnosis, and employs Dempster-Shafer theory for decision-making.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If rule-based fault detection systems are used, then fault detection can be achieved with simple implementation, but the system lacks adaptability to changing conditions and different system types
Solution Approach 1:
The patent replaces rule-based fault detection mechanisms with machine learning models that automatically learn fault patterns from data. The system uses supervised learning algorithms to train classifiers on historical fault data, enabling adaptive fault detection without manual rule configuration. This substitution transforms the rigid rule-based approach into a flexible data-driven system that can adapt to different HVAC system types and changing operational conditions.
Solution Approach 2:
The system dynamically adjusts detection parameters by training machine learning models on varying operational data. The models learn to adapt to different system configurations, environmental conditions, and fault types by modifying their internal parameters during the training phase. This enables the same fault detection system to be applied across different HVAC system types without manual reconfiguration.
2Measurement precision
If qualitative model-based fault detection systems are used, then analytical fault identification can be achieved, but the systems become complex and computationally intensive
Solution Approach 1:
The patent extracts the essential fault detection functionality from complex qualitative models and implements it through streamlined machine learning classifiers. By using supervised learning on labeled fault data, the system achieves accurate fault identification without requiring complex analytical models. The machine learning approach extracts only the necessary patterns from historical data, eliminating unnecessary model complexity while maintaining detection accuracy.
Solution Approach 2:
The system creates simplified representations of fault patterns through machine learning models that are trained on historical data. Instead of using complex qualitative models that require detailed system knowledge, the patent uses data-driven copies of fault patterns that can be automatically learned and applied. This copying approach maintains detection accuracy while significantly reducing model complexity and computational requirements.
3Reliability
If custom rules or analytical models are developed for each specific system, then accurate fault detection for that system can be achieved, but the development process becomes time-consuming and difficult to maintain
Solution Approach 1:
The patent develops a universal fault detection system using machine learning that can be applied across multiple HVAC system types and configurations. The supervised learning models are trained on diverse historical data from various system types, enabling them to generalize and detect faults in different systems without custom development. This universal approach maintains high detection accuracy while eliminating the need for time-consuming custom rule or model development for each specific system.
Solution Approach 2:
The system enables self-service fault detection by automatically learning from historical data without requiring expert intervention for rule creation or model development. The machine learning models autonomously identify fault patterns and update their parameters based on new data, reducing the need for manual maintenance and reconfiguration. This self-learning capability significantly reduces the time and expertise required compared to traditional custom rule-based approaches.
Data Source
AI summary
A method for detecting faulty operation of a multi-variable system is described. The method includes receiving operational data from a plurality of components of the multi-variable system and processing the operational data in accordance with a plurality of dynamic machine learning fault detection models to generate a plurality of fault detection results. Each fault detection model uses a plurality of variables to model one or more components of the multi-variable system and is adapted to detect normal or faulty operation of an associated component or set of components of the multi-variable system. The plurality of fault detection results are output.