Digital Twin Degradation Modeling for Electric Powertrain Components
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current degradation models for electric powertrain components in heavy-duty vehicles are inadequate, failing to accurately predict quality and quantity of recyclable materials and considering vehicle-to-vehicle operational and environmental variations, leading to conservative lifetime estimates and lack of automated methods for remanufacturing.
Innovation Solution
Implementing AI and ML techniques, specifically using digital twins with Deep Neural Networks (DNNs) to model and detect degradation in components like power inverters, leveraging IoT sensors for real-time data, and creating a cloud-based platform for predictive maintenance and material assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional degradation models are used for electric powertrain components, then the modeling process is simple, but the accuracy of predicting component lifetime and recyclable materials is inadequate
Solution Approach 1:
The patent creates digital twins as virtual copies of physical powertrain components (inverter, motor, generator). These digital twins replicate the physical components' degradation behavior through machine learning models trained on sensor data, enabling accurate prediction of component lifetime and recyclable material quantities without complex physical testing
Solution Approach 2:
The patent replaces traditional mechanical/degradation modeling approaches with AI and machine learning-based digital twin systems. Instead of using conventional physics-based models, the system uses neural networks trained on sensor data to predict component degradation, transforming the modeling approach from mechanical to computational
2Duration of action of moving object
If conservative lifetime estimates are used for powertrain components, then the reliability is maintained, but the vehicle life is unnecessarily limited and productivity is reduced
Solution Approach 1:
The patent implements continuous monitoring of powertrain components using sensor networks that feed real-time data to digital twin models. The system provides feedback on actual component degradation states, allowing dynamic adjustment of maintenance schedules and extending vehicle operational life based on actual condition rather than conservative fixed intervals
Solution Approach 2:
The patent performs preliminary degradation assessment and lifetime prediction through digital twins before actual component failure occurs. By predicting remaining useful life and degradation trends in advance, the system enables proactive maintenance planning that extends vehicle operational life while maintaining reliability
3Productivity
If manual methods are used for assessing degradation and material recovery, then the process is straightforward, but the productivity is low and manufacturing precision is poor
Solution Approach 1:
The patent implements automated degradation assessment systems where digital twins continuously monitor and evaluate powertrain component conditions without manual intervention. The system automatically predicts component lifetime, assesses degradation states, and calculates recyclable material quantities, eliminating the need for manual inspection methods
Solution Approach 2:
The patent creates a universal digital twin platform that can assess multiple powertrain components (inverter, motor, generator) simultaneously using the same AI-based framework. This multi-functional system handles degradation assessment, lifetime prediction, and material recovery estimation across different component types, improving productivity while managing complexity through standardization
4Adaptability or versatility
If vehicle-to-vehicle variations are not considered in degradation models, then the modeling is simpler, but the measurement precision and adaptability are reduced
Solution Approach 1:
The patent creates customized digital twin models for each specific vehicle and component combination, accounting for local variations in operating conditions, component manufacturing batches, and vehicle usage patterns. Each digital twin is trained on vehicle-specific sensor data, providing accurate predictions tailored to individual vehicles rather than using generic models
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
In at least one example embodiment, a computer system (350) includes a memory (352) storing instructions and at least one processor (404) configured to execute the instructions to cause the computer system (350) to obtain sensor data, the sensor data corresponding to measurements of at least one component of an electric powertrain system (305) of at least one vehicle (300) and generate a first digital twin based on the obtained sensor data, the first generated twin associated with a type of the at least one vehicle (300).