Reduced-Order Digital Twin Models for Machine Anomaly Root Cause
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine condition monitoring systems struggle to accurately identify the root cause of anomalies and are often dependent on labeled data, which can be difficult to collect, and high-fidelity physics-based digital twin simulations are expensive and time-consuming.
Innovation Solution
The method involves creating a reduced order model (ROM) of a digital twin for a machine type, feeding current data into it, and comparing outputs to detect anomalies. If an anomaly is detected, the data is fed into fault models with specific component failures to identify the root cause by selecting the fault model output that most closely matches the measured data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-fidelity physics-based digital twin simulations are used to identify root cause of anomalies, then measurement precision is improved, but loss of time and use of energy worsen due to expensive and time-consuming computations
Solution Approach 1:
The patent creates reduced-order models (ROMs) that are simplified copies of high-fidelity physics-based digital twin simulations. These ROMs retain the essential predictive capabilities for anomaly detection and root cause identification while requiring significantly less computational resources and time to execute, thus resolving the contradiction between accuracy and computation time
Solution Approach 2:
The patent employs multiple pre-computed ROMs representing different fault conditions that can be quickly evaluated without the computational burden of running full physics-based simulations. These lightweight models enable rapid root cause analysis by comparing actual sensor data against pre-prepared fault signatures, dramatically reducing the time and energy required for anomaly diagnosis
2Loss of information
If multiple fault models are evaluated to identify root cause of anomaly, then loss of information is reduced, but device complexity worsens
Solution Approach 1:
The patent segments the root cause identification process into distinct fault models, each representing a specific component failure mode. By dividing the overall diagnostic task into separate, specialized models (e.g., pump failure model, valve failure model, motor failure model), the system can evaluate multiple potential causes without creating a single overly complex monolithic model, thus managing complexity while comprehensively identifying root causes
Data Source
AI summary
A method for identifying a cause of a machine operating anomaly including creating a reduced order model (ROMs) for a digital twin model of a selected machine type and feeding current data from a deployed machine into the ROM. The method can include comparing a current output from the selected ROM with a measured output from the current data and determining that an operating anomaly exists when the difference between the current output and the measured output exceeds a selected anomaly threshold. The cause of the operating anomaly can be identified by feeding the current data into a plurality of fault models, wherein each fault model includes a particular component failure, comparing a fault model output from each of the plurality of fault models with the measured output from the current data, selecting the fault model with the fault model output most closely matching the measured output, and displaying the identified component failure associated with the selected fault model as the cause of the operating anomaly.