Magnet Temperature Estimation Using Moving Averages and Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing magnet temperature estimation devices in vehicles face significant errors when cumulatively adding temperature rise per unit time, leading to divergent estimated values from actual magnet temperatures.
Innovation Solution
A magnet temperature estimation device utilizing a parameter acquiring unit, a calculating unit for moving averages, and a temperature acquiring unit with a trained 1D convolutional neural network to estimate magnet temperatures based on rotor parameters, reducing estimation errors by processing moving averages of stator coil temperature, motor rotational speed, oil temperature, and oil pump rotational speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If magnet temperature is estimated by cumulatively adding temperature rise per unit time, then temperature estimation can be performed over time, but estimation error builds up and diverges from actual temperature
Solution Approach 1:
The patent applies feedback by using oil temperature measurements as a reference to continuously correct the cumulative temperature rise calculations. The oil temperature serves as a feedback signal that validates and adjusts the estimated magnet temperature, preventing error accumulation and divergence from actual temperature values.
Solution Approach 2:
The patent changes the parameter basis for temperature estimation by shifting from purely cumulative temperature rise calculations to a hybrid approach that incorporates oil temperature measurements. This parameter change allows the system to reset or correct estimation errors by referencing the actual oil temperature, thereby maintaining long-term estimation accuracy.
2Measurement precision
If detailed temperature calculation data is stored for accurate estimation, then estimation accuracy improves, but storage requirements increase
Solution Approach 1:
The patent extracts only the essential parameters needed for temperature estimation (oil temperature, motor operational parameters) while discarding redundant detailed temperature calculation data. By taking out only the critical measurement data and using it with a trained model, the system achieves accurate estimation without storing large volumes of historical temperature data.
Solution Approach 2:
The patent uses a trained model that has been copied from extensive training data to perform real-time estimation. The model captures the complex relationships during the training phase, allowing the system to make accurate predictions using only current input parameters rather than storing and processing all historical data.
3Measurement precision
If comprehensive training data is used to train the estimation model, then model accuracy improves, but training period and computational resources increase
Solution Approach 1:
The patent performs preliminary action by training the estimation model offline using comprehensive training data before actual deployment. This preliminary training phase captures all the complex relationships and patterns, allowing the model to be deployed with high accuracy without requiring extensive real-time computational resources or additional training time during operation.
Data Source
AI summary
Parameters relating to rotation of a motor (2) measured every constant time are acquired and the moving average of each constant interval of the parameters is calculated. The calculated moving averages are input to a training model trained so as to output a temperature of magnets attached to a rotor (7) of the motor (2) when the moving averages of the parameters relating to rotation of the motor (2) are input, and an estimated value of the magnet temperature output from the model is acquired. Next, the acquired estimated value of the magnet temperature is output.