Maintenance recommendations using lifecycle clustering

Patent metadata

Field	Value
Publication number	US-11780609-B2
Application number	US-201916439062-A
Country	US
Kind code	B2
Filing date	Jun 12, 2019
Priority date	Jun 12, 2019
Publication date	Oct 10, 2023
Grant date	Oct 10, 2023

Abstract

Methods and systems are provided for predictive maintenance of a vehicle component. One method involves mapping a current instance of a component of a vehicle to one of plurality of degradation groups of prior lifecycles for other instances of the component based on a relationship between performance measurement data for the current instance and historical performance measurement data associated with that respective degradation group, obtaining contextual data associated with operation of the vehicle, and determining a maintenance recommendation for the current instance of the component based on the contextual data using a predictive maintenance model associated with the mapped degradation group.

First claim

What is claimed is:

1. A method of monitoring a component of a vehicle, the method comprising:
obtaining measurement data characterizing operation of the component for an individual lifecycle between deployment and maintenance, wherein the measurement data is measured by one or more sensors during operation of the vehicle during the individual lifecycle;
obtaining contextual data associated with the operation of the vehicle over the individual lifecycle, wherein the contextual data includes at least one of ambient meteorological data or climate zone data pertaining to operation of the vehicle over the individual lifecycle between deployment and replacement of the component;
clustering the component to a first degradation group from among a plurality of different degradation groups based on a relationship between the measurement data and the contextual data for the individual lifecycle and historical measurement data and historical contextual data associated with a subset of different instances of the component of the first degradation group, wherein the first degradation group comprises a respective subset of prior lifecycles for a respective subset of different instances of the component and the historical measurement data and historical contextual data comprise respective measurement data and respective contextual data characterizing operation of respective instances of the respective subset of different instances of the component for respective prior lifecycles between deployment and maintenance of the respective instance of the component;
determining a value for a degradation metric indicative of an expected lifecycle duration for the component in the future before progressing beyond a threshold as a function of the measurement data for the individual lifecycle and the contextual data over the individual lifecycle using a predictive maintenance model associated with the first degradation group; and
outputting, via a user interface device, a notification of a particular maintenance action recommended to be proactively scheduled and an indication of the expected lifecycle duration for the component before progressing beyond the threshold.

2. The method of claim 1 , wherein clustering the component comprises:
determining, for each degradation group, a respective distance between the measurement data indicative of the performance of component and representative data for the respective degradation group; and
identifying the first degradation group as having a minimum distance between the measurement data indicative of the performance of component and representative data for the first degradation group.

3. The method of claim 1 , further comprising identifying the plurality of different degradation groups based on similarities between respective subsets of the historical measurement data, wherein each degradation group of the plurality of degradation groups comprises a respective subset of prior lifecycles for a respective subset of different instances of the component.

4. The method of claim 3 , further comprising:
obtaining the historical contextual data associated with the respective subset of prior lifecycles for the subset of different instances of the component identified as the first degradation group based on the respective subset of historical measurement data for the respective subset of prior lifecycles identified as the first degradation group; and
determining the predictive maintenance model associated with the first degradation group based at least in part on the historical contextual data associated with the respective subset of prior lifecycles and the respective subset of historical measurement data for the respective subset of prior lifecycles.

5. The method of claim 4 , further comprising obtaining historical maintenance data associated with the respective subset of prior lifecycles, wherein determining the predictive maintenance model comprises determining the predictive maintenance model based at least in part on relationships between the historical maintenance data, the historical contextual data, and the respective subset of historical measurement data.

6. The method of claim 5 , wherein:
determining the predictive maintenance model comprises:
performing association rule mining to identify a contextual variable as being correlative to the historical maintenance data; and
determining the predictive maintenance model as a function of the contextual variable;

the contextual data associated with operation of the vehicle includes recent data for the contextual variable; and
determining the maintenance recommendation comprises determining the maintenance recommendation based at least in part on the recent data for the contextual variable using the predictive maintenance model.

7. The method of claim 5 , wherein clustering the component comprises:
for each degradation group:
determining representative data for the respective degradation group based on the respective subset of the historical measurement data for the respective subset of prior lifecycles for a respective subset of different instances of the component assigned to the respective degradation group; and
determining a respective distance between the measurement data indicative of the performance of component and the representative data for the respective degradation group; and

identifying the first degradation group as having a minimum distance between the measurement data indicative of the performance of component and representative data for the first degradation group.

8. The method of claim 1 , wherein the vehicle comprises an aircraft, the component comprises an auxiliary power unit (APU), and the measurement data comprises at least one of exhaust gas temperature (EGT) measurement data, inlet guide vane position measurement data, fuel pump pressure measurement data, and oil temperature measurement data.

9. The method of claim 8 , further comprising performing clustering to identify the plurality of different degradation groups based on similarities between respective subsets of historical EGT measurement data associated with respective subsets of prior lifecycles for instances of the APU.

10. The method of claim 1 , wherein determining the maintenance recommendation comprises calculating a probability of a maintenance action for the component based on the contextual data using the predictive maintenance model associated with the first degradation group, wherein a maintenance schedule for the component is influenced by the probability of the maintenance action.

11. The method of claim 1 , wherein:
a maintenance recommendation comprises the particular maintenance action and the indication of the expected lifecycle duration; and
a maintenance schedule for the component is influenced by the maintenance recommendation to extend operating time when no fault exists with respect to the component.

12. The method of claim 1 , wherein the individual lifecycle comprises a discrete period of time subdividing a lifetime of the component and the respective prior lifecycles comprise respective discrete periods of time subdividing respective lifetimes of respective instances of the respective subset of different instances of the component.

13. The method of claim 1 , wherein the notification comprises identification of a recommended climate zone that would maximize the expected lifecycle duration or minimize the probability of the particular maintenance action.

14. The method of claim 1 , wherein the component comprises an auxiliary power unit (APU), the vehicle comprises an aircraft, and the particular maintenance action comprises altering an upcoming flight

Assignees

• Honeywell Int Inc

Inventors

• Hawley Megan L
• Griffith Timothy
• Robinson Cavene
• Chandra Mohan Ananda Vel Murugan
• Mylaraswamy Dinkar

Classifications

• B64F5/60 Primary

  Testing or inspecting aircraft components or systems · CPC title

• G06N5/025

  Extracting rules from data · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G07C5/085

  using electronic data carriers · CPC title

• G07C5/0816

  Indicating performance data, e.g. occurrence of a malfunction · CPC title