Apparatus and method for determining friction coefficient of brake friction material

1. An apparatus for determining a friction coefficient of a brake friction material, the apparatus comprising:

a model generation device configured to generate a friction coefficient meta model to determine the friction coefficient based on information of an operation state of a brake using raw data acquired through a preceding test evaluation process,

wherein the model generation device comprises:

a preprocessor configured to process the raw data to acquire data required for machine learning;

a processor configured to train a model through machine learning using data acquired by the preprocessor as training data; and

a postprocessor configured to further process the model completely trained by the machine learning unit to extract a final friction coefficient meta model to determine the friction coefficient corresponding to the operation state of the brake from an input parameter indicating the operation state of the brake; and

wherein the raw data comprises data on a rotation speed and a temperature of a brake disc or a rotor and a brake hydraulic pressure, indicating the operation state of the brake, and data on the friction coefficient corresponding to the operation state of the brake.

2. The apparatus of claim 1 , wherein the information of the operation state of the brake comprises a rotation speed and a temperature of a brake disc or a rotor and a brake hydraulic pressure.

3. The apparatus of claim 1 , wherein the raw data is acquired depending on a state of the brake friction material,

wherein the preprocessor comprises a data group classifier configured to group the raw data input thereto based on the state of the brake friction material and to classify data groups so that the data groups are identified, and

wherein the machine learning unit trains the model through machine learning using the data groups identified depending on the state of the brake friction material.

4. The apparatus of claim 1 , wherein the preprocessor comprises at least one of:

a missing data processor configured to acquire data missing from the raw data input thereto through linear interpolation;

a negative value processor configured to substitute a negative value among the raw data input thereto with a positive value obtained through linear interpolation;

an outlier processor configured to remove data related to the friction coefficient greater than a predetermined reference value from the raw data input thereto;

an invalid data processor configured to remove data in an invalid period from the raw data input thereto; or

a stabilization data period extractor configured to extract a period that meets a movement standard deviation criterion of the friction coefficient from the raw data input thereto, to secure data in an extracted valid period as data for model training, and to remove data in remaining periods.

5. The apparatus of claim 4 , wherein in the outlier processor, the reference value is a value set depending on characteristics of the brake friction material.

6. The apparatus of claim 5 , wherein the reference value is a value set depending on a material of the brake friction material.

7. The apparatus of claim 4 , wherein in the invalid data processor, the invalid period is a period in which a vehicle speed is less than a reference vehicle speed set as a vehicle speed at which evaluation in the preceding test evaluation process is completed.

8. The apparatus of claim 4 , wherein the stabilization data period extractor accepts a period in which a standard deviation of the friction coefficient from an arbitrary starting time point to an n th time point set at predetermined regular time intervals is maintained at a predetermined reference value or less from the arbitrary starting time point to an m th time point (where m<n) as a valid period, determines data in the accepted valid period as data for model training, and removes data in remaining periods other than the accepted valid period.

9. A method of determining a friction coefficient of a brake friction material, the method comprising:

acquiring raw data required to generate a friction coefficient meta model to determine the friction coefficient based on information of an operation state of a brake through a preceding test evaluation process;

preprocessing, by a preprocessor, the acquired raw data according to a predetermined algorithm to acquire data required for machine learning;

training, by a processor, a model through machine learning using data acquired by the preprocessor as training data; and

further processing, by a postprocessor, the model completely trained by the machine learning unit to extract a final friction coefficient meta model to determine the friction coefficient corresponding to an operation state of a brake from an input parameter indicating the operation state of the brake;

wherein the raw data comprises data on a rotation speed and a temperature of a brake disc or a rotor and a brake hydraulic pressure, indicating the operation state of the brake, and data on the friction coefficient corresponding to the operation state of the brake.

10. The method of claim 9 , further comprising:

determining, by a controller in which the final friction coefficient meta model is input and stored, the friction coefficient using the stored friction coefficient meta model based on real-time information of the operation state of the brake collected while a vehicle is traveling.

11. The method of claim 9 , wherein the information of the operation state of the brake comprises a rotation speed and a temperature of a brake disc or a rotor and a brake hydraulic pressure.

12. The method of claim 9 , wherein the raw data is acquired depending on a state of the brake friction material,

wherein the preprocessor comprises a data group classifier configured to group the raw data input thereto based on the state of the brake friction material and to classify data groups so that the data groups are identified, and

wherein the machine learning unit trains the model through machine learning using the data groups identified depending on the state of the brake friction material.

13. The method of claim 12 , wherein the state of the brake friction material comprises a new brake friction material, a brake friction material that has been burnished, a brake friction material that is fading, and a brake friction material that has faded, and

wherein the raw data is data acquired depending on the state of the brake friction material through the preceding test evaluation process.

14. The method of claim 9 , wherein the preprocessor comprises at least one of:

a missing data processor configured to acquire data missing from the raw data input thereto through linear interpolation;

a negative value processor configured to substitute a negative value among the raw data input thereto with a positive value obtained through linear interpolation;

an outlier processor configured to remove data related to the friction coefficient greater than a predetermined reference value from the raw data input thereto;

an invalid data processor configured to remove data in an invalid period from the raw data input thereto; or

a stabilization data period extractor configured to extract a period that meets a movement standard deviation criterion of the friction coefficient from the raw data input thereto, to secure data in an extracted valid period as data for model training, and to remove data in remaining periods.

15. The method of claim 14 , wherein in the outlier processor, the reference value is a value set depending on characteristics of the brake friction material.

16. The method of claim 15 , wherein the reference value is a value set depending on a material of the brake friction material.

17. The method of claim 14 , wherein, in the invalid data processor, the invalid period is a period in which a vehicle speed is less than a reference vehicle speed set as a vehicle speed at which evaluation in the preceding test evaluation process is completed.

18. The method of claim 14 , wherein the stabilization data period extractor accepts a period in which a standard deviation of the friction coefficient from an arbitrary starting time point to an n th time point set at predetermined regular time intervals is maintained at a predetermined reference value or less from the arbitrary starting time point to an m th time point (where m<n) as a valid period, determines data in the accepted valid period as data for model training, and removes data in remaining periods other than the accepted valid period.