Thermal Digital Twin Training for Machine Temperature Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital twin systems for estimating machine temperatures are complex, costly, and require real-time knowledge of power losses, necessitating additional hardware and manual intervention for updates, which is time-consuming and resource-intensive.
Innovation Solution
A method using a thermal digital twin replicated by a machine learning model that records temperature values and operating point parameters under different conditions, trained with self-tuning analytical models, enabling automated adaptation without manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a thermal digital twin with high accuracy is used to predict machine temperatures, then temperature prediction accuracy is improved, but the system complexity and hardware resource requirements increase due to the need for real-time power loss knowledge and additional digital twins
Solution Approach 1:
The patent extracts and eliminates the requirement for real-time power loss knowledge from the thermal digital twin system. By reformulating the thermal model to use only measurable operating point parameters (voltage, current, speed, torque), the complex electromagnetic-digital twin coupling is removed, significantly reducing system complexity while maintaining temperature prediction accuracy
Solution Approach 2:
The patent creates a universal thermal digital twin that can predict temperatures across multiple operating conditions using a single unified model. The model uses general operating point parameters that are universally measurable, eliminating the need for separate digital twins or complex electromagnetic models for different scenarios
2Measurement precision
If real-time knowledge of power losses is obtained by employing another digital twin coupled with electromagnetic models, then power loss estimation accuracy is improved, but hardware cost and system complexity increase
Solution Approach 1:
The patent replaces the complex electromagnetic model and additional digital twin system with a simplified thermal model that directly uses measurable operating point parameters. This substitution eliminates the need for expensive electromagnetic modeling hardware and software while maintaining the ability to estimate power losses through the thermal model's inherent relationships
Solution Approach 2:
The patent creates a simplified copy of the thermal system that directly relates operating point parameters to temperature without requiring a separate electromagnetic model copy. This single thermal digital twin copy performs both temperature prediction and implicit power loss estimation, eliminating the need for multiple digital twin copies
3Measurement precision
If manual operations and testing are performed to develop and maintain digital twins, then model accuracy is maintained, but time consumption and human resource requirements increase
Solution Approach 1:
The patent enables the thermal digital twin to self-adapt to changing operating conditions by using only measurable operating point parameters as inputs. The model automatically adjusts to new conditions without requiring manual retraining or human intervention, as it relies on directly measurable parameters that inherently reflect the current system state
Solution Approach 2:
The patent creates a dynamic thermal model that continuously adapts to changing operating conditions by using real-time operating point parameters. The model structure allows automatic adjustment to new operating regimes without manual intervention, making the system flexible and responsive to changes in machine behavior
4Measurement precision
If additional hardware resources are allocated to support multiple digital twins and electromagnetic models, then prediction accuracy is improved, but system cost increases
Solution Approach 1:
The patent creates a universal thermal digital twin that performs multiple functions: temperature prediction, implicit power loss estimation, and adaptation to various operating conditions. This single multi-functional model eliminates the need for separate hardware resources for electromagnetic models and multiple specialized digital twins
Data Source
Figure 1
Figure 2
Figure 3~5
AI summary
A method is disclosed for estimating internal temperatures of a machine using a thermal digital twin that replicates a temperature sensor of the machine. The method comprises recording temperature values received from one or more temperature sensors of the machine while operating the machine under at least two different operating conditions; capturing one or more operating point parameters for each of the different operating conditions; and obtaining a base thermal model for the thermal digital twin by training the thermal machine learning model using the recorded temperatures and the captured one or more operating point parameters, the training further comprising using input from a self-tuning analytical model. A device, use of the device, a computer program and computer program product are also disclosed.