Methods and apparatus to predict machine failures

Patent metadata

Field	Value
Publication number	US-10657454-B2
Application number	US-201916271490-A
Country	US
Kind code	B2
Filing date	Feb 8, 2019
Priority date	Jul 1, 2016
Publication date	May 19, 2020
Grant date	May 19, 2020

Abstract

Methods and apparatus of modeling work vehicle data and predicting machine failures based on the same are disclosed. An example apparatus includes a text miner to text mine first alert data to identify a first machine failure within the first alert data; a failure alert sequence identifier to identify a first alert sequence associated with the first machine failure; a conditional probability determiner to determine a conditional probability of the first alert sequence leading to failure based on i) a number of work machines in which the first alert sequence ended in failure and ii) a number of work machines in which the first alert sequence did not end in failure; and a collator to update a model by correlating the first alert sequence with the first machine failure and first probable parts used to repair the first machine failure.

First claim

What is claimed is:

1. An apparatus, comprising:
a text miner to text mine first alert data to identify a first machine failure within the first alert data;
a failure alert sequence identifier to identify a first alert sequence associated with the first machine failure;
a conditional probability determiner to determine a conditional probability of the first alert sequence leading to failure based on i) a number of work machines in which the first alert sequence ended in failure and ii) a number of work machines in which the first alert sequence did not end in failure; and
a collator to update a model by correlating the first alert sequence with the first machine failure and first probable parts used to repair the first machine failure.

2. The apparatus of claim 1 , wherein the collator is to generate the model by correlating the first machine failure and the first probable parts used to repair the first machine failure based on parts used to repair work machines in which the first alert sequence occurred and resulted in failure.

3. The apparatus of claim 1 , wherein the text miner is to access second alert data from a second work vehicle and parsing the second alert data to identify a second alert sequence from the second work vehicle.

4. The apparatus of claim 3 , further including a comparator to compare the second alert sequence to the first alert sequence; identifying a substantial similarity between the first and second alert sequences; and based on the model and the first alert sequence, identifying the first probable parts used to repair the first machine failure.

5. The apparatus of claim 1 , further including a tallier to tally the number of work machines in which the first alert sequence occurred and resulted in failure.

6. The apparatus of claim 1 , further including a conditional probability significance level determiner to average the conditional probability of the first alert sequence with conditional probabilities of alert sequences having the same length as the first alert sequence to determine a significance level.

7. The apparatus of claim 6 , further including a comparator to compare the significance level to the conditional probability to determine if the conditional probability of the first alert sequence is greater than the significance level.

8. An article of manufacture comprising instructions, which, when executed, cause a machine to at least:
text mine first alert data to identify a first machine failure within the first alert data;
identify a first alert sequence associated with the first machine failure;
determine a conditional probability of the first alert sequence leading to failure based on i) a number of work machines in which the first alert sequence ended in failure and ii) a number of work machines in which the first alert sequence did not end in failure; and
update a model by correlating the first alert sequence with the first machine failure and first probable parts used to repair the first machine failure.

9. The article of manufacture of claim 8 , wherein the instructions, when executed, cause the machine to correlate the first machine failure and the first probable parts used to repair the first machine failure based on parts used to repair work machines in which the first alert sequence occurred and resulted in failure.

10. The article of manufacture of claim 8 , wherein the instructions, when executed, cause the machine to access second alert data from a second work vehicle and parse the second alert data to identify a second alert sequence from the second work vehicle.

11. The article of manufacture of claim 10 , wherein the instructions, when executed, cause the machine to compare the second alert sequence to the first alert sequence; identify a substantial similarity between the first and second alert sequences; and based on the model and the first alert sequence, identify the first probable parts used to repair the first machine failure.

12. The article of manufacture of claim 8 , wherein the instructions, when executed, cause the machine to tally the number of work machines in which the first alert sequence occurred and resulted in failure.

13. The article of manufacture of claim 8 , wherein the instructions, when executed, cause the machine to average the conditional probability of the first alert sequence with conditional probabilities of alert sequences having the same length as the first alert sequence to determine a significance level.

14. The article of manufacture of claim 13 , wherein the instructions, when executed, cause the machine to compare the significance level to the conditional probability to determine if the conditional probability of the first alert sequence is greater than the significance level.

15. A method, comprising:
text mining first alert data to identify a first machine failure within the first alert data;
identifying a first alert sequence associated with the first machine failure;
determining a conditional probability of the first alert sequence leading to failure based on i) a number of work machines in which the first alert sequence ended in failure and ii) a number of work machines in which the first alert sequence did not end in failure; and
updating a model by correlating the first alert sequence with the first machine failure and first probable parts used to repair the first machine failure.

16. The method of claim 15 , further including generating the model by correlating the first machine failure and the first probable parts used to repair the first machine failure based on parts used to repair work machines in which the first alert sequence occurred and resulted in failure.

17. The method of claim 15 , further including accessing second alert data from a second work vehicle and parsing the second alert data to identify a second alert sequence from the second work vehicle.

18. The method of claim 17 , further including comparing the second alert sequence to the first alert sequence; identifying a substantial similarity between the first and second alert sequences; and based on the model and the first alert sequence, identifying the first probable parts used to repair the first machine failure.

19. The method of claim 15 , wherein determining the conditional probability includes tallying the number of work machines in which the first alert sequence occurred and resulted in failure.

20. The method of claim 15 , further including:
averaging the conditional probability of the first alert sequence with conditional probabilities of alert sequences having the same length as the first alert sequence to determine a significance level;
comparing the significance level to the conditional probability to determine if the conditional probability of the first alert sequence is greater than the significance level; and
updating the model when the conditional probability is a greater value than the significance level.

Assignees

• Deere & Co

Inventors

• SINGH Mahesh Pratap
• Pophale Gauri Narayan
• Kudale Mayuri
• Das Hridoy

Classifications

• G05B23/0283 Primary

  Predictive maintenance, e.g. involving the monitoring of a system and, based on the monitoring results, taking decisions on the maintenance schedule of the monitored system; Estimating remaining useful life [RUL] (preventive maintenance, i.e. planning maintenance according to the available resources without monitoring the system G06Q10/06) · CPC title

• G06N7/005 Primary

  Physics · mapped topic

• G06N7/01 Primary

  Probabilistic graphical models, e.g. probabilistic networks · CPC title