# US 2021/0279387 A1 · Method for obtaining a temporal model of a magnetic bearing, device for modelling the magnetic bearing and associated systems

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0279387 A1
- Document ID: US20210279387A1
- Status: Published application
- Kind: A1
- Application number: 17/174,744
- Filed: 2021-02-12
- Published: 2021-09-09
- Earliest claimed: 2020-03-04
- Assignees: Skf Magnetic Mechatronics
- Applicants: SKF Magnetic Mechatronics
- Inventors: Francois Caire; Lakdar Sadi-Haddad
- Web page: https://patentgrep.com/patent/US20210279387A1

## CPC

- G05B 13/04: involving the use of models or simulators
- G06F 30/27: using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
- F16C 32/0444: {Details of devices to control the actuation of the electromagnets}
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- F16C: SHAFTS; FLEXIBLE SHAFTS; ELEMENTS OR CRANKSHAFT MECHANISMS; ROTARY BODIES OTHER THAN GEARING ELEMENTS; BEARINGS

## Abstract

The method for obtaining a temporal model of a magnetic bearing provides the generation of reference data representative of characteristics of the bearing. The method includes the production of the temporal model of the magnetic bearing from the reference data having temporal data.

## Claims

### Claim 1

A method for obtaining a temporal model of a magnetic bearing comprising:
the generation of reference data representative of characteristics of the bearing, and
the production of the temporal model of the magnetic bearing from the reference data comprising temporal data.

### Claim 2 (depends on 1)

The method according to claim 1, wherein the reference data comprise learning data and validation data, the learning data comprising at least one input reference data set and at least one output reference data set, and the validation data comprising a first validation data set comprising a first input validation data set and a first output validation data set, the generation of learning data comprising:
a) the generation of a first transient temporal data set of wave forms and of random amplitudes,
b) the determination of the input reference data set from the first transient temporal data set of wave forms and of random amplitudes,
c) the injection of the input reference data set into a nonlinear model of the magnetic bearing or in the magnetic bearing,
d) the collection of the output reference data set, the output reference data resulting from the excitation of the nonlinear model of the magnetic bearing or of the magnetic bearing by the input reference data set, and
e) the saving of the input reference data set and of the associated output reference data set, and
the generation of the first validation data set comprising:
f) the generation of a second transient temporal data set of wave forms and of random amplitudes different from the first transient temporal data set,
g) the determination of the first input validation data set from the second transient temporal data set of wave forms and of random amplitudes,
h) the injection of the first input validation data set into the nonlinear model of the magnetic bearing or into the magnetic bearing,
i) the collection of the first output validation data set, the first output validation data set resulting from the excitation of the nonlinear model of the magnetic bearing or of the magnetic bearing by the first input validation data set, and
j) the saving of the first input validation data set and of the associated first output validation data set.

### Claim 3 (depends on 1)

The method according to claim 1, wherein the production of the temporal model comprises a learning step comprising the implementation of an automatic learning algorithm with the learning data.

### Claim 4 (depends on 3)

The method according to claim 3, wherein the production of the model further comprises a first validation step in which the temporal model generates a second output validation data set associated with the first input validation data set.

### Claim 5 (depends on 4)

The method according to claim 4, wherein, if the absolute value of the difference of the contents of the first output validation data set and second output validation data set is greater within a confidence probability, the method comprises the reiteration of the steps a) to e), of the learning step and of the first validation step.

### Claim 6 (depends on 4)

The method according to claim 4, wherein, if the absolute value of the difference of the contents of the first output validation data set and second output validation data set is less than or equal within a confidence probability, the validation data comprise a second validation data set comprising a second input validation data set and a third output validation data set, the generation of the second validation data set comprising:
k) the generation of a third transient temporal data set of wave forms and of random amplitudes, the third temporal data set being different from the first and second temporal data sets,
l) the determination of the second input validation data set from the third transient temporal data set of wave forms and of random amplitudes,
m) the injection of the second input validation data set into the nonlinear model of the magnetic bearing or into the magnetic bearing,
n) the collection of the third output validation data set, the second output validation data set resulting from the excitation of the nonlinear model of the magnetic bearing or of the magnetic bearing by the second input validation data set, and
o) the saving of the second input validation data set and of the associated third output validation data set, and
the production of the model further comprising a second validation step in which the temporal model generates a fourth output validation data set associated with the second input validation data set.

### Claim 7 (depends on 6)

The method according to claim 6, wherein, if the absolute value of the difference of the contents of the third output validation data set and fourth output validation data set is greater within the confidence probability, the method comprises the reiteration of the steps a) to e), of the learning step and of the first and second validation steps.

### Claim 8 (depends on 3)

The method according to claim 3, wherein the automatic learning algorithm comprises a statistical learning algorithm.

### Claim 9 (depends on 2)

The method according to claim 2, wherein each transient temporal data set of wave forms and of random amplitudes, the input reference data set and each input validation data set each comprise at least one value of an excitation current of the windings of the magnetic bearing, a value of radial displacement or axial displacement of a rotor of the magnetic bearing and a value of a bias current of the windings of the magnetic bearing, and the first output reference data set and each output validation data set comprise at least one value of a voltage at the terminals of the windings, of a magnetic flux generated by the windings and a resultant force generated by the windings on the rotor of the magnetic bearing.

### Claim 10 (depends on 2)

The method according to claim 2, wherein, each transient temporal data set of wave forms and of random amplitudes, the input reference data set and each input validation data set each comprise at least one value of an excitation voltage of the windings of the magnetic bearing, a value of radial displacement or axial displacement of a rotor of the magnetic bearing and a value of a bias current of the windings of the magnetic bearing, and the first output reference data set and each output validation data set comprise at least one value of a current in the windings, of a magnetic flux generated by the windings and of a resultant force generated by the windings on the rotor of the magnetic bearing.