Tire wear estimation using a hybrid machine learning system and method
A tire tread wear system may include one or more vehicle sensors and a processor. The processor may include a control module, a geometrical model, a machine learning model, and a switch. The geometrical model may be configured to collect data from the vehicle sensors to determine a dynamic rolling radius of a tire. The geometrical model may be configured to output a tread wear estimation based on the dynamic rolling radius of the tire. The machine learning model may be configured to collect data from the vehicle sensors. The machine learning model may be configured to output a tread wear estimation based on a correlation of the tread wear estimation output from the geometrical model and one or more data instances with a tire tread state. The switch may be configured to activate the geometrical model, the machine learning model, or a combination thereof.
1. A tire tread wear system comprising:
a plurality of vehicle sensors; and
a processor comprising:
a control module configured to determine a tire event;
a geometrical model configured to collect data from the plurality of vehicle sensors to determine a dynamic rolling radius of a tire and output a tread wear estimation based on the dynamic rolling radius of the tire;
a machine learning model configured to collect data from the plurality of vehicle sensors and output a tread wear estimation based on a correlation of the tread wear estimation output from the geometrical model and a plurality of data instances with a tire tread state, wherein the geometric model is used to train the machine learning model; and
a logic switch configured to activate the geometrical model and the machine learning model.
2. The tire tread wear system of claim 1 , wherein the control module is configured to activate the geometrical model based on a determination that the tire event is an initial tire setup.
3. The tire tread wear system of claim 2 , wherein the logic switch is configured to activate the machine learning model on a condition that the control module determines that the tire event is a tire change.
4. The tire tread wear system of claim 3 , wherein the logic switch is configured to deactivate the geometrical model.
5. The tire tread wear system of claim 1 , wherein the geometrical model is configured to collect data on a condition that the vehicle is at a constant speed.
6. The tire tread wear system of claim 1 , wherein the machine learning model is configured to collect data continuously and independently from a speed of the vehicle.
7. The tire tread wear system of claim 1 , wherein the data collected by the geometrical model includes global positioning system (GPS) data and wheel velocity data.
8. The tire tread wear system of claim 1 , wherein the data collected by the machine learning model includes at least one of wheel velocity data, steering wheel angle data, brake pressure data, longitudinal acceleration data, or total distance data.
9. The tire tread wear system of claim 1 , wherein the machine learning model utilizes a multi-class support vector machine (SVM) with a radian kernel to predict the tire tread state and classify the tire tread state into a plurality of groups.
10. The tire tread wear system of claim 9 , wherein a first group of the plurality of groups is 100% tread remaining, a second group of the plurality of groups is 75% tread remaining, a third group of the plurality of groups is 50% tread remaining, and a fourth group of the plurality of groups is 25% tread remaining.
11. A tire tread wear system comprising:
a plurality of vehicle sensors;
a processor comprising:
a control module configured to determine a tire state of a vehicle;
a geometrical model configured to collect data from the plurality of vehicle sensors to determine a dynamic rolling radius of a tire and output a tread wear estimation based on the dynamic rolling radius of the tire;
a machine learning model that is trained using the geometric model, the machine learning model configured to collect data from the plurality of vehicle sensors and output a tread wear estimation based on a correlation of the tread wear estimation output from the geometrical model and a plurality of data instances with a tire tread state;
a logic switch configured to activate the geometrical model and the machine learning model; and
a display configured to:
receive the geometrical model tread wear estimation and the machine learning model tread wear estimation;
generate a visualization for the tire based on the geometrical model tread wear estimation and the machine learning model tread wear estimation; and
display the generated visualization for the tire.
12. The tire tread wear system of claim 11 , wherein the control module is configured to activate the geometrical model based on a determination that the tire event is an initial tire setup.
13. The tire tread wear system of claim 12 , wherein the logic switch is configured to activate the machine learning model on a condition that the control module determines that the tire event is a tire change.
14. The tire tread wear system of claim 13 , wherein the logic switch is configured to deactivate the geometrical model.
15. The tire tread wear system of claim 11 , wherein the data collected by the geometrical model includes global positioning system (GPS) data and wheel velocity data.
16. The tire tread wear system of claim 11 , wherein the data collected by the machine learning model includes at least one of wheel velocity data, steering wheel angle data, brake pressure data, longitudinal acceleration data, or total distance data.
17. The tire tread wear system of claim 11 , wherein the machine learning model utilizes a multi-class support vector machine (SVM) with a radian kernel to predict the tire tread state and classify the tire tread state into a plurality of groups.
18. The tire tread wear system of claim 17 , wherein a first group of the plurality of groups is 100% tread remaining, a second group of the plurality of groups is 75% tread remaining, a third group of the plurality of groups is 50% tread remaining, and a fourth group of the plurality of groups is 25% tread remaining.
19. A tire tread wear system comprising:
a plurality of vehicle sensors;
a wireless modem configured to transmit data from the plurality of sensors;
a server configured to receive the data from the wireless modem, the server comprising:
a control module configured to determine a tire state of a vehicle;
a geometrical model configured to collect data from the plurality of vehicle sensors to determine a dynamic rolling radius of a tire and output a tread wear estimation based on the dynamic rolling radius of the tire;
a machine learning model that is trained using the geometric model, the machine learning model configured to collect data from the plurality of vehicle sensors and output a tread wear estimation based on a correlation of the tread wear estimation output from the geometrical model and a plurality of data instances with a tire tread state; and
a logic switch configured to activate the geometrical model and the machine learning model.