Machine operation classifier

• US 9,691,025 B2
• Filed: 09/16/2014
• Issued: 06/27/2017
• Est. Priority Date: 09/16/2014
• Status: Active Grant

• Current Assignee: Caterpillar Incorporated
• Original Assignee: Caterpillar Incorporated
• Inventors: Kirsch, Karl
• Primary Examiner(s): Gonzales, Vincent
• Application Number:
  US14/487,857
• Publication Number:
  US 20160078340A1
• Time in Patent Office: 1,015 Days
• CPC Class Codes: E02F 9/264 Sensors and their calibration for indicating the position of the work tool
• CPC Class Codes: E02F 9/267 Diagnosing or detecting failure of vehicles
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: G06N 99/00 Subject matter not provided for in other groups of this subclass
• CPC Class Codes: G07C 5/08 Registering or indicating performance data other than driving, working, idle, or waiting time, with or without registering driving, working, idle or waiting time

Abstract

A process for developing machine classification systems includes using human experts to associate expected operations with various machine states including drawbar pull, tool position, tool commands, gear, and ground speed, among others, to create a classification system that can be used in a particular machine. The classification system operates in real time to infer operations such as dig, dump, travel, and push from machine state inputs and logs the operations for use in operational analysis and maintenance of the machine.

Claims

1. A method of developing a machine operation classifier, the method comprising:

i) identifying, via a user interface of a computer, an operation of a machine;

ii) compiling, at the computer, a list of conditions that are associated with the operation of the machine, the step of compiling including compiling a list of true/false statements corresponding to one of tool events, direction events, gear events, and power events; and

wherein the step of compiling the list of true/false statements for the tool events includes documenting a tool position value and tool position tolerance, the tool position value being a position of a tool and the tool position tolerance being a range of tool positions over which the tool is considered to be in a predetermined position;

repeating, at the computer, steps i and ii for one or more operations that the machine is expected to perform; and

generating, via the computer, a classifier algorithm, wherein the classifier algorithm outputs the operation of the machine selected from the identified operations of the machine in response to identification of conditions in the associated list of conditions when the classifier algorithm is executed on a processor of the machine, wherein the operation of the machine includes the operation of one of a construction machine, a mining machine, or an earthmoving machine.

2. The method of claim 1, wherein the operation of the machine is one of an idle operation, a travel operation, a reverse operation, a dig/load operation, a carry/haul operation, a dump/spread operation, a compact operation, a grade operation, and a ditch operation.

3. The method of claim 1, wherein the list of conditions include a raw condition and a derived condition.

4. The method of claim 1, wherein compiling the list of conditions includes documenting a tool command corresponding to a signal indicating an operator-initiated action for a tool.

5. The method of claim 4, wherein the tool is one of a blade, a bucket, a scraper, a ripper, and a payload bed.

6. The method of claim 1, wherein compiling the list of true/false statements for the tool events includes determining a tool position from a cylinder position.

7. A method of creating an operation classification algorithm for a machine, the method comprising:

developing, at a computer, a catalog of operations performed by the machine;

cataloging events associated with each of the operations, wherein the events include tool events, and each tool event is one of a tool active event, a tool inactive event, a tool engaged event, and a tool disengaged event, wherein the tool active event is derived from a tool command value and the tool engaged event is a function of a measured tool position and a derived tool position;

for each event, documenting, via a user interface of a computer, one or more machine conditions associated with the event;

for each machine condition, developing, at the computer, a calculation used to determine the one or more conditions from one of a current machine state or a combination of current and previous machine states; and

generating a classification algorithm that monitors, in the machine, the one or more machine conditions and outputs a current operation of the machine using the one or more conditions to identity events associated with the operation.

8. The method of claim 7, wherein the operations performed by the machine include an idle operation, a travel operation, a reverse operation, a dig/load operation, a carry/haul operation, a dump/spread operation, a compact operation, a grade operation, and a ditch operation.

9. The method of claim 7, wherein the calculation is an integration of a tool position command with a high and low limit set at opposite range limits of a tool.

10. The method of claim 7, wherein generating the classification algorithm comprises developing a selection module that:

monitors and stores the current machine state;

develops one or more machine conditions from one or more machine states;

identifies events from the current machine state or the combination of current and previous machine states; and

selects an operation from the catalog of operations based on the events.