Hybrid Prognostics for Component Health Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for determining the health state of machine components face challenges due to non-observable damage and sparse data availability, making it difficult to accurately predict component health and prevent failures.
Innovation Solution
The proposed system uses a hybrid prognostics approach that combines usage models, damage models, and prediction models, including physics-based and machine learning methods, to estimate component health states by obtaining usage data and condition indicators, and calculates damage estimates and health state probabilities using Bayesian models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional predictive models are used to estimate component health state, then the system can operate with limited data sources, but the measurement precision and reliability of health state estimation deteriorates due to non-observable damage and sparse data
Solution Approach 1:
The patent introduces virtual sensors as intermediary elements that bridge the gap between observable and unobservable component states. These virtual sensors are created through data fusion techniques that combine multiple data sources (sensor readings, operational data, maintenance records) to generate estimates of unobservable damage states. The virtual sensors act as mediators that translate available information into meaningful health state indicators without requiring direct physical sensors for every parameter.
Solution Approach 2:
The patent replaces physical sensing mechanisms with computational models and algorithms. Instead of installing physical sensors to directly measure internal component damage, the system uses machine learning models, physics-based models, and data fusion algorithms to computationally infer health states. This substitution allows the system to estimate unobservable parameters through information processing rather than direct physical measurement.
2Reliability
If conventional predictive models are used with sparse data, then the device complexity remains low, but the reliability of failure prediction deteriorates
Solution Approach 1:
The patent segments the prognostics system into distinct functional modules: data acquisition modules, data fusion modules, virtual sensor creation modules, health state estimation modules, and remaining useful life prediction modules. Each module performs a specific function and processes specific types of data. This segmentation allows the complex system to be managed through modular components that can be independently developed, tested, and maintained, reducing the practical complexity despite the advanced functionality.
Solution Approach 2:
The patent creates a multi-functional platform that handles diverse data types (sensor data, operational data, maintenance records), multiple analysis methods (machine learning, physics-based models, data fusion), and various output metrics (health state estimates, remaining useful life predictions, failure probabilities). This universal system architecture allows the same infrastructure to serve multiple prognostics needs across different component types and failure modes, amortizing the complexity across broad applicability.
3Measurement precision
If more data sources are integrated to improve health state estimation, then the measurement precision improves, but the device complexity and data processing requirements increase
Solution Approach 1:
The patent introduces data fusion algorithms as intermediary processing layers that harmonize data from multiple heterogeneous sources. These fusion algorithms act as mediators that standardize different data formats, reconcile conflicting information, and integrate diverse data types into a unified health state representation. This intermediary processing layer manages the complexity of data integration by providing a systematic approach to combining multiple data sources without requiring custom integration logic for each combination.
Data Source
AI summary
A system for determining a health state of a component includes a processor and a non-transitory computer-readable storage medium that stores a plurality of computer-executable instructions thereon. The processor, in response to executing the plurality of computer-readable instructions, may be configured to perform one or more steps. The steps may include obtaining, from a first data source, usage data for the component; obtaining, from a second data source, a condition indicator for the component; running a usage model to produce usage parameters for the component based on the usage data; running a damage model to produce a damage estimate for the component based on the usage parameters; and running a prediction model to produce a health state estimate of the component based on the damage estimate and the condition indicator.