Transfer Modeling for Sparse Industrial Asset Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems face challenges in accurately modeling and predicting the performance of industrial assets with sparse data, such as newly deployed or recently maintained assets, due to insufficient historical data, leading to inefficiencies in failure diagnostics, remaining useful life estimation, and anomaly detection.
Innovation Solution
A system and method that combine relevant data with operational data to create a transfer model, leveraging simulation data or physics models from peer assets, allowing for more accurate analytics tasks and early alerting of potential issues without requiring extensive manual calibration or initial data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional systems use actual historical data for analytics tasks, then model accuracy is improved for assets with sufficient data, but assets with sparse data cannot be accurately modeled
Solution Approach 1:
The patent creates a digital twin (virtual copy) of the physical asset that can be populated with data from multiple sources including historical data, manufacturer data, and operational data. This copying approach allows the model to be built even when direct historical data from the specific asset is sparse, as the digital twin can be initialized with representative data from similar assets or manufacturer specifications.
Solution Approach 2:
The system merges multiple data sources including actual operational data, manufacturer provided data, simulation data, and data from similar assets into a unified digital twin model. This combination allows the model to leverage available sparse data while supplementing it with data from other sources, thereby maintaining model accuracy even when direct historical data is limited.
2Measurement precision
If more historical data is collected for sparse data assets, then model accuracy improves, but time and operational disruption increase
Solution Approach 1:
The system performs preliminary actions by pre-populating the digital twin with manufacturer data, design specifications, and simulation data before the asset begins operation or while having minimal operational data. This preliminary modeling allows analytics tasks to be performed immediately without waiting for extensive historical data accumulation.
Solution Approach 2:
The digital twin acts as an intermediary that bridges the gap between sparse actual operational data and the need for comprehensive modeling data. It intermediates by incorporating data from manufacturer sources, simulation models, and similar assets, allowing accurate analytics to be performed without requiring extensive direct historical data from the specific asset.
3Measurement precision
If manual calibration and initial data collection are performed extensively, then model accuracy improves, but operational costs and complexity increase
Solution Approach 1:
The digital twin system enables self-service by automatically ingesting and integrating data from multiple sources including manufacturer APIs, operational sensors, and simulation models without requiring extensive manual calibration. The system autonomously populates and updates the model, reducing the need for manual intervention and complex calibration procedures.
Solution Approach 2:
The digital twin framework provides a universal platform that can handle multiple data sources and analytics tasks through a single unified model. This multi-functionality eliminates the need for separate calibration processes for different data types or analytics tasks, reducing overall system complexity while maintaining accuracy across various operational conditions.
Data Source
AI summary
According to some embodiments, a system and method are provided to model a sparse data asset. The system comprises a processor and a non-transitory computer-readable medium comprising instructions that when executed by the processor perform a method to model a sparse data asset. Relevant data and operational data associated with the newly operational are received. A transfer model based on the relevant data and the received operational data. An input into the transfer model is received and a predication based on data associated with the received operational data and the relevant data is output.