System and method for evaluating residual life of components made of composite materials

Patent metadata

Field	Value
Publication number	US-12584887-B2
Application number	US-202217934187-A
Country	US
Kind code	B2
Filing date	Sep 21, 2022
Priority date	Sep 28, 2021
Publication date	Mar 24, 2026
Grant date	Mar 24, 2026

Abstract

This disclosure generally relates to the field of structural health monitoring, and, more particularly, to a method and system for evaluating residual life of components made of composite materials. Existing methods require performing computational methods such as Finite Element Analysis (FEA) on the results of Non-Destructive Testing (NDT) every time a component is inspected. This makes the process expensive and time-consuming. Thus, embodiments of present disclosure provide a method wherein NDT is performed using different sensing methods such as ultrasound, ultrasound pulse echo, thermography to determine type of defect, location of defect and depth of defect in a test component which are then fed into a pre-trained machine learning model to predict residual life of the component. Testing time is greatly reduced since the pre-trained machine learning model is trained offline using results of the computational methods.

First claim

What is claimed is:

1 . A processor implemented method comprising:
transmitting, via an ultrasound sensor controlled by one or more hardware processors, a plurality of ultrasound signals towards a test component comprising a defect, and receiving the plurality of ultrasound signals reflected by the test component; extracting, via the one or more hardware processors, a plurality of features comprising time-frequency and statistical features from each of the plurality of received ultrasound signals; determining, via the one or more hardware processors, type of the defect in the test component using a pre-trained classifier based on the plurality of features, wherein training of the pre-trained classifier comprises:
performing ultrasound test on a component without defect and a plurality of components with known defects, wherein the ultrasound test comprises transmitting a plurality of ultrasound signals to the component and recording the plurality of ultrasound signals reflected by the component;
extracting a plurality of features from the results of the ultrasound test;
correlating the plurality of features with the known defects of the plurality of components; and
training the classifier using the plurality of features and corresponding defects;
determining, via the one or more hardware processors, a location of the defect comprised in the test component based on time of flight analysis of the plurality of ultrasound signals, wherein the time of flight of the ultrasound signal is time taken by the ultrasound signal to travel a certain distance through the test component, wherein the location of the defect is determined by minimizing an objective function J=Σ j−1 N f(j) 2 , wherein

f
⁡
(
j
)

=

(

L

S
T

-
D

V

+

L

D
-

S
j

V

)

-

L

S
T

-

S
j

V

-

Δ
⁢

t

T
-
j

=
0

,

wherein L S T −D is distance between transmitter of the ultrasound sensor and the defect which is calculated as √{square root over (x D 2 +y D 2 )} wherein (x D , y D ) is the location of the defect, V is velocity of an ultrasound signal among the plurality of ultrasound signals, L D−S j is distance between the defect and a receiver (j) among one or more receivers of the ultrasound sensor which is calculated as √{square root over ((x D −x j ) 2 +(y D −y j ) 2 )} wherein (x j , y j ) is location of the receiver (j), L S T −S j is distance between the transmitter and the receiver (j) which is calculated as √{square root over (x j 2 +y j 2 )}, and Δt T−j is difference in time of flight of the ultrasound signal travelled via a direct path and an indirect path from the transmitter to the receiver;
scanning, via a pulse echo ultrasound sensor controlled by the one or more hardware processors, the location of the defect in the test component using a pulse echo ultrasound signal to determine depth of the defect;
scanning, via a thermal camera controlled by the one or more hardware processors, the location of the defect in the test component to estimate a dimension of the defect; and
predicting, via the one or more hardware processors, residual life of the test component using a pre-trained machine learning model based on the type of the defect, the location of the defect, the depth of the defect and the dimension of the defect, wherein training of the pre-trained machine learning model comprises:
obtaining residual properties of a plurality of components with known defects, wherein the residual properties comprise stiffness and load bearing capacity;
determining residual life of each of the plurality of components using a material degradation mechanism and a computational method based on the defect in the component, one or more predefined loading conditions, and material properties of the component; and
training a machine learning model to generate the pre-trained machine learning model using the residual life of the plurality of components, features of defects in the plurality of components for the one or more predefined loading conditions and the material properties of the plurality of components, wherein the features of defects comprise type of the defect, location of the defect, depth of the defect and dimension of the defect.

2 . The method of claim 1 , wherein the plurality of features comprises box-pierce statistic of Discrete Wavelet Transform (DWT), mean of windowed box-pierce statistic of DWT, Hjorth complexity time-domain and standard deviation of windowed box-pierce stat of DWT, resonance peak locations, amplitude of resonance peaks, width of resonance peaks and adjacent resonance peak to peak distance, wherein the pre-trained classifier is ensemble adaptive boost classifier.

3 . A system comprising:
a memory storing instructions; one or more communication interfaces; an ultrasound sensor; a pulse echo ultrasound sensor; a thermal camera; and one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to:
transmit, via the ultrasound sensor, a plurality of ultrasound signals towards a test component comprising a defect and receiving the plurality of ultrasound signals reflected by the test component;
extract a plurality of features comprising time-frequency and statistical features from each of the plurality of received ultrasound signals;
determine type of the defect in the test component using a pre-trained classifier based on the plurality of features, wherein training of the pre-trained classifier comprises:
performing ultrasound test on a component without defect and a plurality of components with known defects, wherein the ultrasound tes

Assignees

• Tata Consultancy Services Ltd

Inventors

• Kalariya Yagnik Pravinchandra
• Gorey Abhijeet
• Salvi Amit Gangadhar
• Basu Subhadeep
• Gain Supriya
• Chakravarty Tapas
• Bhaumik Chirabrata
• Pal Arpan
• Senguttuvan Sooriyan
• Sinharay Arijit

Classifications

• G01N2291/0258

  Structural degradation, e.g. fatigue of composites, ageing of oils · CPC title

• G01N29/043 Primary

  in the interior, e.g. by shear waves · CPC title

• G01N29/4481

  Neural networks · CPC title

• G01N2291/044

  Internal reflections (echoes), e.g. on walls or defects · CPC title

• G01N29/069 Primary

  Defect imaging, localisation and sizing using, e.g. time of flight diffraction [TOFD], synthetic aperture focusing technique [SAFT], Amplituden-Laufzeit-Ortskurven [ALOK] technique · CPC title