Granted Patent B2

US 12,576,990 · App. 17/378,678 · Granted Mar 17, 2026

Inventors: Robert Michael Leitch (Vancouver, CA); Yikan Wang (Burnaby, CA)

Assignee: The Boeing Company

B64F5/40B64F5/60G05B23/0216G05B23/0235G05B23/024G05B23/0281G05B23/0283G06N5/04G06N20/00G06Q10/20

Abstract

Claims (78)

1 . A data processing system for generating predictive maintenance models, comprising:

one or more processors;

a memory including one or more digital storage devices; and

a plurality of instructions stored in the memory and executable by the one or more processors to:

receive a historical dataset relating to each system of a plurality of systems,

wherein the plurality of systems includes a fleet of aircraft, and

wherein the historical dataset includes maintenance data and operational data associated with the fleet of aircraft;

receive a rule set for processing a first attribute of the operational data;

calculate a custom data feature from the historical dataset according to the received rule set;

calculate the custom data feature from a first sample of labelled data and a second sample of labelled data according to the received rule set,

wherein the first sample of labelled data is randomly selected from flights having a first label, and the second sample of labelled data is randomly selected from flights having a second label;

display a comparison of the custom data feature for the first sample and the second sample,

wherein the first sample of labelled data and the second sample of labelled data are, respectively, subsets of the operational data of the historical dataset, labelled according to a corresponding subset of maintenance data of the historical dataset;

generate a predictive maintenance model using the custom data feature according to a machine learning method based on receiving approval from the display;

receive additional operational data;

input the additional operational data to the predictive maintenance model;

implement the predictive maintenance model; and

cause an action to occur based on implementing the predictive maintenance model,

wherein the action includes one or more of: generate an alert for detected anomalies or preventative maintenance action.

2 . The data processing system of claim 1 , wherein the plurality of instructions are further executable by the one or more processors to:

calculate a preview of the custom data feature from a subset of the operational data of the historical dataset according to the received rule set, and

display the preview of the custom data feature in a graphical user interface.

3 . The data processing system of claim 2 , wherein the plurality of instructions are further executable by the one or more processors to receive a selection of the subset of the operational data.

4 . The data processing system of claim 1 , wherein the first sample of labelled data and the second sample of labelled data are labelled according to a time relationship to maintenance events of the corresponding subset of maintenance data.

5 . The data processing system of claim 1 , wherein the plurality of instructions are further executable by the one or more processors to:

receive a selection of maintenance events of the maintenance data of the historical dataset, and

create the first sample of labelled data and the second sample of labelled data, the first sample of labelled data and second sample of labelled data being labelled according to a time relationship to the selected maintenance events.

6 . The data processing system of claim 1 , wherein the data of the first sample of labelled data are labelled with a first qualitative label, and the data of the second sample of labelled data are labelled with a second qualitative label different from the first qualitative label.

7 . The data processing system of claim 1 , wherein the rule set includes one or more rules defining a constraint on a value of the first-attribute.

8 . The data processing system of claim 1 , wherein the rule set includes one or more rules defining a constraint on a value of the first attribute, and

wherein the rule set is a Boolean combination of a plurality of constraints on values of attributes of the operational data including the first attribute.

9 . The data processing system of claim 7 , wherein the rule set further includes one or more aggregating statistical functions.

10 . The data processing system of claim 1 , wherein the rule set is also for processing at least a second attribute of the operational data, and the custom data feature is an aggregate data feature.

11 . The data processing system of claim 10 , wherein the rule set includes a combination of a first rule for processing the first attribute and a second rule for processing the second attribute.

12 . The data processing system of claim 1 , wherein the custom data feature is a differential comparison between or an algebraic combination of at least two sensors of each system.

13 . The data processing system of claim 1 , wherein the operational data is timelabeled, and the rule set is time-dependent.

14 . The data processing system of claim 13 , wherein the rule set includes one or more rules defining a time range from which to extract operational data.

15 . The data processing system of claim 1 , wherein the plurality of instructions are further executable by the one or more processors to:

calculate a plurality of pre-determined aggregate data features from the operational data.

16 . The data processing system of claim 15 , wherein the plurality of systems have multiple operational phases, and the pre-determined aggregate data features are dependent on an operational phase of the multiple operational phases.

17 . The data processing system of claim 1 , wherein the plurality of systems are a fleet of aircraft.

18 . A computer implemented method of generating a predictive maintenance model, comprising:

receiving, by a device, a historical dataset relating to each system of a plurality of systems,

wherein the plurality of systems includes a fleet of aircraft, and

wherein the historical dataset includes maintenance data and operational data;

receiving, by the device, a rule set for processing a first attribute of the operational data;

calculating, by the device, a custom data feature from the historical dataset according to the received rule set;

calculating, by the device, the custom data feature from a first sample of labelled data and a second sample of labelled data according to the received rule set,

wherein the first sample of labelled data is randomly selected from flights having a first label, and the second sample of labelled data is randomly selected from flights having a second label;

displaying, by the device, a comparison of the custom data feature for the first sample of labelled data and the second sample of labelled data,

wherein the first sample of labelled data and second sample of labelled data are subsets of the operational data of the historical dataset, labelled according to a corresponding subset of maintenance data of the historical dataset;

generating, by the device, a predictive maintenance model using the custom feature according to a machine learning method;

receiving, by the device, additional operational data;

inputting, by the device, the additional operational data to the predictive maintenance model;

implementing, by the device, the predictive maintenance model; and

causing, by the device, a preventative maintenance action to occur based on implementing the predictive maintenance model.

19 . A computer program product for generating predictive maintenance models, the computer program product comprising:

a non-transitory computer-readable storage medium having computer-readable program code embodied in the storage medium, the computer-readable program code configured to cause a data processing system to generate a predictive maintenance model, the computer-readable program code configured to:

receive a historical dataset relating to each system of a plurality of systems,

wherein the plurality of systems includes a fleet of aircraft, and

wherein the historical dataset includes maintenance data and operational data;

receive a rule set for processing a first attribute of the operational data;

calculate a custom data feature from the historical dataset according to the received rule set;

calculate the custom data feature from a first sample of labelled data and a second sample of labelled data according to the received rule set,

wherein the first sample of labelled data is randomly selected from flights having a first label, and the second sample of labelled data is randomly selected from flights having a second label;

display a comparison of the custom data feature for the first sample of labelled data and the second sample of labelled data,

wherein the first sample of labelled data and second sample of labelled data are subsets of the operational data of the historical dataset, labelled according to a corresponding subset of maintenance data of the historical dataset;

generate a predictive maintenance model using the custom feature according to a machine learning method;

receive additional operational data;

input the additional operational data to the predictive maintenance model;

implement the predictive maintenance model; and

cause preventative maintenance action based on implementing the predictive maintenance model,

wherein the preventative maintenance action includes one or more of:

an inspection of equipment,

testing of the equipment,

repair of the equipment, or

replacement of the equipment.

20 . The data processing system of claim 1 , wherein the first sample of labelled data and the second sample of labelled data include a same number of flights.

Assignments (1)

Continuity (2)

References Cited (78)