Method and device for predicting the condition of a component or system, computer program product

Patent metadata

Field	Value
Publication number	US-9449274-B2
Application number	US-201314089145-A
Country	US
Kind code	B2
Filing date	Nov 25, 2013
Priority date	May 31, 2011
Publication date	Sep 20, 2016
Grant date	Sep 20, 2016

Abstract

A device method and computer program product are disclosed for trend prediction of the course of a time-dependent series of data points of a component or system, particularly for an aircraft or spacecraft, including: providing an optimised decision tree, the input node of which is provided for inputting an input vector, the nodes of which contain the data points of a respective input vector and the leaves of which each contain an extrapolation function; iteratively calculating future data points by a respective time-dependent series of data points being inputted into the decision tree as an input vector and the decision tree calculating therefrom a data point subsequent to the last data point of the input vector in an automated manner, the calculated subsequent data point being added to the time-dependent series of data points in order to be used as a new input vector for the next iteration step.

First claim

What is claimed is:

1. A computer-implemented method for predicting the condition of a component or system, in particular for an aircraft or spacecraft, by trend prediction of the course of a time-dependent series of data points determined at the component or system by measuring, the method comprising:
providing an optimised decision tree, the input node of which is provided for inputting an input vector, the nodes of which contain the attributes of the respective input vector and the leaves of which each contain an extrapolation function provided for trend prediction;
iteratively calculating future data points by an input vector being derived from a time-dependent series of data points and the decision tree calculating a data point subsequent to the last data point of the time-dependent series of data points from the input vector in an automated manner, the calculated subsequent data point being added to the time-dependent series of data points in order to be used for forming a new input vector for the next iteration step.

2. The method according to claim 1 , wherein at least one input vector is initially provided before the iterative calculation, which input vector is derived from a time-dependent series of data points of a predetermined length, the data points representing a measured physical value of the component or system.

3. The method according to claim 1 , wherein the oldest data point is removed during iterative calculation each time a newly calculated data point is added to the time-dependent series of data points.

4. The method according to claim 1 , wherein the iterative calculation of new data points is terminated after a predetermined number of iteration steps and/or if an error determined during calculation of a subsequent data point exceeds a predetermined threshold.

5. The method according to claim 1 , wherein an extrapolation function of this type of the decision tree is used during iterative calculation of subsequent data points for the newly calculated data point for which a newly calculated data point has the highest value in relation to the input vector used.

6. The method according to claim 1 , wherein an evaluation of the determined subsequent data points is carried out after the iterative calculation, in particular by the determined subsequent data points being compared with a critical threshold value.

7. The method according to claim 1 , wherein the temperature of the component or system and/or an oscillation generated by the component or system is measured as a physical parameter, the system being in particular an air-conditioning system in an aircraft or spacecraft.

8. The method according to claim 1 , wherein for the provision of an optimised decision tree, a feature vector is initially calculated which is determined from training data and a parameter set.

9. The method according to claim 8 , wherein the parameter set preferably contains parameters of the type which are required for creating the decision tree and which in particular comprise at least one of the following parameters:
a number of data points of a series of data which are used for trend prediction;
a calculation of the gradient of the series of data which is used for the trend prediction;
a calculation of the minimum and maximum values of the data points of a series of data which are used for the trend prediction;
the available extrapolation function of a series of data.

10. The method according to claim 8 , wherein a physical value of the component or system is measured for the provision of the training data and a predetermined number of data points are generated from the thus determined data signal by sampling, which data points form the time-dependent series of data points.

11. The method according to claim 8 , wherein random values are determined for the parameter set at the start, and wherein, using the training data and the parameter set, an increasingly optimised feature vector and decision tree is iteratively calculated until an error of the optimised decision tree falls below a predetermined threshold value.

12. The method according to claim 1 , wherein the decision tree is calculated using an ID3 algorithm or a C4.5 algorithm, the feature vector, optimised in the context of the optimisation of the parameter set for determining the decision tree, being provided as an input vector.

13. The method according to claim 1 , wherein in addition to the respectively calculated future data point, during iterative calculation of future data points an error value is also calculated showing how far the calculated data point deviates from an optimum value.

14. A device for predicting the condition of a component or system, in particular for an aircraft or spacecraft, in particular using a method according to claim 1 , the device comprising:
a measuring device which is configured so as to measure a physical measured value at the component or system for generating a time-dependent series of data points;
a memory in which at least one time-dependent series of data points can be stored; and
a calculation and evaluation device which is configured so as to generate an optimised decision tree with reference to at least one of the time-dependent series of data points and to carry out a trend prediction with reference to the thus generated decision tree.

15. A non-transitory computer-readable storage medium with an executable program stored thereon, wherein the program instructs a microprocessor to perform a method according to claim 1 .

Assignees

• Airbus Operations Gmbh

Inventors

• Randerath Bernhard
• Gerdes Mike
• Scholz Dieter

Classifications

• G05B23/024

  Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06N99/005

  Physics · mapped topic

• G06N5/02 Primary

  Knowledge representation; Symbolic representation · CPC title

• G05B23/0232

  based on qualitative trend analysis, e.g. system evolution · CPC title