Electric motors with a permanent magnet (permanent magnet synchronous motor) are used in a lot of applications such as in all kinds of electric vehicles, money changing machines, industrial robots, pumps, compressors, servo drives, HVAC systems, laundry machines, amusement park equipment, refrigerators, microwave ovens, vacuum cleaners, etc.
In high torque and speed operations there is an abundant temperature rise in the magnets which causes a torque decrease, due to flux density decrease, and it may also cause safety issues. For these reasons it is important to know the magnet surface temperature during operations. By knowing the temperature safety issues can be prevented and the torque can be optimally controlled. Measuring this temperature in operational conditions is very difficult and very expensive (rotating parts in a small closed environment) but it was discovered that some other easily and inexpensively measured parameters can be used to train a machine learning model, which then can be used to estimate the surface temperature of the magnet.
In this example we will train a supervised regression model which then can be used to estimate the rotor temperature (equals to the magnet temperature) of a synchronous electric motor with a permanent magnet in operational conditions.
The input data to the machine learning model (around 1 million records) is laboratory measurements data presented below:
|
Measured parameter
|
Range/Notes
|
|
Ambient temperature as measured by a thermal sensor
located closely to the stator.
|
[-8.57, 2.97]
|
|
Coolant temperature. The motor is water cooled.
Measurement is taken at outflow.
|
[-1.43, 2.65]
|
|
Voltage d-component
|
[-1.66, 2.27]
|
|
Voltage q-component
|
[-1.86, 1.79]
|
|
Motor speed
|
[-1.37, 2.02]
|
|
Torque induced by current.
|
[-3.35, 3.02]
|
|
Current d-component
|
[-3.25, 1.06]
|
|
Current q-component
|
[-3.34, 2.91]
|
|
Permanent Magnet surface temperature representing the
rotor temperature. This was measured with an infrared
thermograph unit.
|
[-2.63, 2.92]
Decision variable.
Regression!
|
|
Stator yoke temperature measured with a thermal sensor.
|
[-1.83, 2.45]
|
|
Stator tooth temperature measured with a thermal
sensor.
|
[-2.07, 2.33]
|
|
Stator winding temperature measured with a thermal
sensor.
|
[-2.02, 2.65]
|
|
* data source: [2]
|
The first step is to create an AI-TOOLKIT database with the “Create New AI-TOOLKIT Database” command on the Database tab on the left taskbar. Save the database in a directory of your choice. The second step is to import all data into the database created in the former step with the “Import Data into Database” command. Do not forget to indicate the number of header rows (if any) and the correct zero based index of the decision column (8 in this example)! Next we must create the AI-TOOLKIT project file. Use the “Open AI-TOOLKIT Editor” command and then insert the chosen model template with the “Insert ML Template” button. In this example we will use a supervised SVM model. Use the SVM Parameter Optimizer for finding the optimal values for all parameters automatically. Read the ‘AI in Finance process improvement’ article for more details about these steps.
The final AI-TOOLKIT project file is shown hereunder (notice the EPSILON_SVR parameter value which selects the SVM Regression model):
model:
id: 'ID-uHXdJRNxyH'
type: SVM
path: 'em.sl3'
params:
- svm_type: EPSILON_SVR
- kernel_type: RBF
- gamma: 15.0
- C: 6.31
- p: 0.359
- cache_size: 1000
- max_iterations: 5000
training:
- data_id: 'em'
- dec_id: 'decision'
test:
- data_id: 'em'
- dec_id: 'decision'
input:
- data_id: 'em_input_data'
- dec_id: 'decision'
output:
- data_id: 'em_output_data'
- col_id: 'decision'
After training the machine learning model the final total mean squared error (MSE) on the whole training dataset (nearly 1 million records) with 1000 maximum number of iterations is 0.222, and with 2000 maximum number of iterations is 0.18. There is a significant improvement in the MSE and also in the prediction results if we increase the maximum number of iterations from 1000 to 2000. We could still increase the number of iterations significantly and it is expected to have an even better performance. A neural network model could also be developed for the same purpose to further improve the accuracy of the model.
This article is a slightly modified excerpt from the book “The Application of Artificial Intelligence”. If you are interested in the subject then it is strongly recommended to read the book which contains many more details and real world case studies for several sectors and disciplines! The book explains several examples step-by-step by using the AI-TOOLKIT. The book is going through the publishing process at the time of writing this article. You may use the contact form for info about pre-ordering the book.
References
- [1] The Application of Artificial Intelligence, Zoltan Somogyi.
- [2] Kirchgässner, Wilhelm & Wallscheid, Oliver & Böcker, Joachim. (2019). Empirical Evaluation of Exponentially Weighted Moving Averages for Simple Linear Thermal Modeling of Permanent Magnet Synchronous Machines.
Learn about the application of Artificial Intelligence and Machine Learning from the book "The Application of Artificial Intelligence | Step-by-Step Guide from Beginner to Expert", Springer 2020 (~400 pages) (ISBN 978-3-030-60031-0). Unique, understandable view of machine learning using many practical examples. Introduces AI-TOOLKIT, freely available software that allows the reader to test and study the examples in the book. No programming or scripting skills needed! Suitable for self-study by professionals, also useful as a supplementary resource for advanced undergraduate and graduate courses on AI. More information can be found at the Springer website: Springer book: The Application of Artificial Intelligence.
|
 |
