Machine fault modelling

Patent metadata

Field	Value
Publication number	US-12566430-B2
Application number	US-202217891943-A
Country	US
Kind code	B2
Filing date	Aug 19, 2022
Priority date	Dec 16, 2016
Publication date	Mar 3, 2026
Grant date	Mar 3, 2026

Abstract

Sensor logs corresponding to a first machine are accessed. Each sensor log spans at least a first period. First computer readable logs corresponding to the first machine are accessed. Each computer readable log spans at least the first period, the computer readable logs include a maintenance log including maintenance task objects, each maintenance task object includes a time and a maintenance task type. A set of statistical metrics are derived from the sensor logs. A set of log metrics are derived from the computer readable logs. Using a risk model that receives the statistical metrics and log metrics as inputs, fault probabilities or risk scores indicative of one or more fault types occurring in the first machine within a second period are determined.

First claim

The invention claimed is:

1 . A computer-implemented method performed using one or more processors or special-purpose computing hardware, the method comprising:
accessing one or more data sources to receive updates at the data sources corresponding to a machine, the one or more data sources comprising logs, the logs comprising sensor logs, and any of maintenance logs, fault logs, and message logs recording information related to any of locations, durations, types, and resolutions of respective historical faults that have occurred on the machine, the sensor logs comprising time-series data, the one or more data sources comprising availability information of facilities and materials used in performing one or more fault addressing mechanisms, wherein at least a portion of the logs comprises free text; extracting or deriving one or more metrics from the logs using natural language processing based on one or more semantic rules, using keyword searching to determine frequencies of occurrence of particular words, and using determination of one or more patterns of free-text information; based on the one or more metrics, determining weights for a data model used to predict a probability of a fault; based on the data model, predicting a probability of a fault in the machine or one or more sub-systems of the machine; and proactively implementing a fault addressing mechanism on the machine or on a sub-system of the machine based on a schedule associated with the machine, the availability information, and the probability of the fault, wherein proactively implementing a fault addressing mechanism comprises:
receiving, at a robotic system, the probability of the fault; and
based on the probability of the fault, performing, by the robotic system, a physical task to replace or purge one or more machine components.

2 . The computer-implemented method according to claim 1 , wherein the logs further comprise warped sensor logs, and the metrics are extracted or derived from the warped sensor logs.

3 . The computer-implemented method of claim 1 , wherein the proactively implementing of the fault addressing mechanism comprises selecting the fault addressing mechanism corresponding to a fault type of a highest probability.

4 . The computer-implemented method of claim 1 , wherein the fault addressing mechanism comprises a physical action.

5 . The computer-implemented method of claim 1 , further comprising generating the data model to predict the fault addressing mechanism, wherein the generating of the data model comprises:
preparing a training set, the training set comprising, for one or more types of the historical faults, first candidate fault addressing mechanisms that were previously implemented and failed to address the respective historical faults and second candidate fault addressing mechanisms that were previously implemented and successfully addressed the respective historical faults; and the generating of the data model is based on the training set.

6 . The computer-implemented method of claim 1 , further comprising:
transforming, according to a schema map and an ontology, the one or more data sources to generate a representation of the one or more data sources; receiving a modification to the representation of the one or more data sources; and synchronizing the modification to the information within the one or more data sources.

7 . The computer-implemented method of claim 1 , wherein the extracting the one or more metrics comprises:
identifying schemas defining structures of the data sources; based on a schema map that defines how elements of the schemas map to an ontology, wherein the elements correspond to data items of the data sources, mapping the data items to ontology elements, wherein the ontology elements comprise any of an object type, a relationship, a property type, an attribute of the object type, or an attribute of the property type.

8 . The system of claim 1 , wherein the physical task comprises a priority maintenance task.

9 . The system of claim 1 , wherein
the physical task causes one or more engine parameters of the machine to change to a non-fault status.

10 . The system of claim 1 , wherein the instructions that, when executed by the one or more processors, cause the system to perform:
dynamically updating a fault probability pane on a display interface based on a changed interval length input over which the probability of the fault is evaluated.

11 . The system of claim 1 , wherein the data model comprises a first data model; and the instructions that, when executed by the one or more processors, cause the system to perform:
predicting an updated probability in an event that the fault addressing mechanism is carried out based on rerunning the first data model and a second data model, wherein rerunning the first data model and the second data model is based on a modified maintenance log and an additional maintenance task object corresponding to the fault addressing mechanism.

12 . The system of claim 1 , where the physical task comprises purging a hydraulic system, replacing a bearing, replacing a valve, replacing a sensor, or draining or replacing coolant.

13 . A system comprising:
one or more processors; and a memory storing instructions that, when executed by the one or more processors, cause the system to perform:
accessing one or more data sources to receive updates at the data sources corresponding to a machine, the one or more data sources comprising logs, the logs comprising sensor logs, and any of maintenance logs, fault logs, and message logs recording information related to any of locations, durations, types, and resolutions of respective historical faults that have occurred on the machine, the sensor logs comprising time-series data, the one or more data sources comprising availability information of facilities and materials used in performing one or more fault addressing mechanisms, wherein at least a portion of the logs comprises free text;
extracting or deriving one or more metrics from the logs using natural language processing based on one or more semantic rules, using keyword searching to determine frequencies of occurrence of particular words, and using determination of one or more patterns of free-text information;
based on the one or more metrics, determining weights for a data model used to predict a probability of a fault;
based on the data model, predicting a probability of a fault in the machine or one or more sub-systems of the machine; and
proactively implementing a fault addressing mechanism on the machine or on a sub-system of the machine based on a schedule associated with the machine, the availability information, and the probability of the fault, wherein proactively implementing a fault addressing mechanism comprises:
receiving, at a robotic system, the probability of the fault; and
based on the probability of the fault, performing, by the robotic system, a physical task to replace or purge one or more machine components.

Assignees

• Palantir Technologies Inc

Inventors

• Spiro Ezra
• Menck Andre Frederico Cavalheiro
• Prasad Anshuman
• Thouzeau Arthur
• Henry Caroline
• Shepherd Charles
• Peller Joanna
• Yip Jennifer
• Diciolla Marco
• Todd Matthew
• Maag Peter
• Tank Spencer
• Powell Thomas

Classifications

• G01D9/00

  Recording measured values · CPC title

• G06N7/01

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

• G06F2218/16

  by matching signal segments · CPC title

• G06F18/40

  Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor · CPC title

• G06F30/20

  Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30) · CPC title