Method for developing machine operation classifier using machine learning

Patent metadata

Field	Value
Publication number	US-10032117-B2
Application number	US-201414489191-A
Country	US
Kind code	B2
Filing date	Sep 17, 2014
Priority date	Sep 17, 2014
Publication date	Jul 24, 2018
Grant date	Jul 24, 2018

Abstract

A method for developing machine operation classifiers for a machine is disclosed. The method includes receiving training data associated with the machine from one or more on-board engineering channels associated with the machine and determining one or more training features based on the training data values. The method also includes determining one or more training labels associated with the one or more training features and building a predictive model for determining machine operation classifiers using a computer. Building the predictive model may include feeding the one or more training features and the one or more training labels associated with the one or more training features to a machine learning algorithm and determining a predictive model from the machine learning algorithm. The predictive model may be used for receiving new data associated with the machine and determining a predicted label based on the new data.

First claim

What is claimed is:

1. A method for developing machine operation classifiers for a machine, the method comprising:
receiving training data associated with the machine from one or more on-board engineering channels associated with the machine;
determining one or more training features based on the training data values;
determining one or more training labels associated with the one or more training features upon a time period of input of the training data, wherein each training data label will accompany a certain time period of training data values and correspond to one of a plurality of predetermined machine operations;
building a predictive model, using a computer, for determining machine operation classifiers, building the predictive model includes;
feeding the one or more training features and the one or more training labels associated with the one or more training features to a machine learning algorithm; and
determining a predictive model from the machine learning algorithm, the predictive model for receiving new data associated with the machine and determining a predicted label based on the new data.

2. The method of claim 1 , wherein the one or more on board engineering channels includes a plurality of on-board engineering channels associated with the machine and the method further comprises determining one or more elected channels from the plurality of on-board engineering channels.

3. The method of claim 2 , wherein determining the one or more elected channels includes recursive feature elimination on the plurality of on-board engineering channels.

4. The method of claim 2 , wherein determining the predictive model further includes optimizing the predictive model based on input from the one or more elected channels.

5. The method of claim 1 , wherein the machine learning algorithm utilizes, at least, a trained neural network.

6. The method of claim 1 , wherein the machine learning algorithm utilizes, at least a decision tree.

7. The method of claim 1 , wherein the machine learning algorithm utilizes, at least, support vector machine weights.

8. The method of claim 1 , wherein determining one or more training labels associated with the one or more training features is performed using video, the video being synchronized with the training data.

9. A method for determining a predicted machine operation for a machine using a machine operation classifier, the method including:
receiving first data values associated with the machine from one or more on-board engineering channels associated with the machine;
determining one or more first features from the first data values;
determining a first label for the first data values upon a time period of input of the first data by using a predictive model, wherein the first label will accompany a certain time period of first data values and correspond to one of a plurality of predetermined machine operations, the predictive model being built by:
feeding one or more training features and one or more training labels associated with the one or more training features to a machine learning algorithm; and
determining the predictive model from the machine learning algorithm.

10. The method of claim 9 , wherein the machine learning algorithm includes at least one of a trained neural network, a decision tree, and support vector machine weights.

11. A system for developing machine operation classifiers for a machine, the system comprising:
one or more on-board engineering channels for providing training data associated with the machine;
an input module for receiving the training data from the one or more on-board engineering channels, determining one or more training features based on the training data values, and determining one or more training labels associated with the one or more training features upon a time period of input of the training data, wherein each training data label will accompany a certain time period of training data values and correspond to one of a plurality of predetermined machine operations;
a machine learning module for building a predictive model for determining machine operator classifiers, building the predictive model by the machine learning module includes:
feeding the one or more training features and the one or more training labels associated with the one or more training features to a machine learning algorithm; and
determining a predictive model from the machine learning algorithm.

12. The system of claim 11 , further comprising a predictive modelling module, the predictive modelling module receiving new data associated with the machine from the one or more on-board engineering channels and determining a predicted label based on the new data by using the predictive model.

13. The system of claim 11 , further comprising one or more machine sensors associated with the machine and the one or more on-board engineering channels, the on-board engineering channels determining the training data from the machine data provided by the machine sensors.

14. The system of claim 13 , wherein the one or more system sensors include at least one of a ground speed sensor, a track speed sensor, a slope sensor, a gear sensor, and a hydraulic sensor.

15. The system of claim 11 , wherein the one or more on board engineering channels includes a plurality of on-board engineering channels associated with the machine and the machine learning module determines one or more elected channels from the plurality of on-board engineering channels.

16. The system of claim 15 , wherein determining the one or more elected channels include performing recursive features elimination of the plurality of on-board engineering channels.

17. The system of claim 15 , wherein determining the predictive model further includes optimizing the predictive model based on input from the one more elected channels.

18. The system of claim 11 , wherein the machine is an excavator.

19. The system of claim 11 , wherein the machine is a grader.

20. The system of claim 11 , wherein the machine is a wheel loader.

Assignees

• Caterpillar Inc

Inventors

• Hodel Benjamin
• Kim Sangkyum
• Lee Paul

Classifications

• G05B13/0265

  the criterion being a learning criterion · CPC title

• G06N99/005 Primary

  Physics · mapped topic

• G06N20/00 Primary

  Machine learning · CPC title