Determining a health condition of a structure

• US 10,228,278 B2
• Filed: 11/22/2013
• Issued: 03/12/2019
• Est. Priority Date: 05/01/2013
• Status: Active Grant

• Current Assignee: National ICT Australia Limited
• Original Assignee: National ICT Australia Limited
• Inventors: Chen, Fang, Zhang, Matt, Li, Zhidong, Wang, Yang
• Primary Examiner(s): Alkafawi, Eman
• Application Number:
  US14/787,912
• Publication Number:
  US 20160238438A1
• Time in Patent Office: 1,936 Days
• Field of Search: 702 33
• CPC Class Codes: G01H 1/00 Measuring characteristics of vibrations in solids by using direct conduction to the detector G01H9/00, G01H11/00 take precedence
• CPC Class Codes: G01H 13/00 Measuring resonant frequency
• CPC Class Codes: G01M 5/0008 of bridges
• CPC Class Codes: G01M 5/0033 by determining damage, crack or wear
• CPC Class Codes: G01M 5/0066 by exciting or detecting vibration or acceleration vibration testing of structures G01M7/00
• CPC Class Codes: G01M 7/00 Vibration-testing of structures; Shock-testing of structures G01M9/00 takes precedence
• CPC Class Codes: G01M 7/025 Measuring arrangements

Abstract

The disclosure relates to structural health monitoring (SHM). In particular determining a health condition of a structure, such as a bridge, based on vibration data measured of the bridge. Measured vibration data is calibrated (410-450). Features are then extracted from the calibrated data (610-630) and a support vector machine classifier is then applied (720) to the extracted features to determine (730) the health condition of a part of the structure. Training of the support vector machine classifier by a machine learning process (910) is also described.

Claims

1. A computer-implemented method, comprising:

monitoring a building structure by constantly or regularly performing the steps of;

receiving or accessing vibration data of a part of the structure, the vibration data being measured by two or more accelerometers and, the vibration data representing acceleration of the part of the structure substantially caused by an external vibration source;

transforming the vibration data by aligning an axis direction of the vibration data from each of the two or more accelerometers with a direction corresponding to maximum variance of the vibration data to create calibrated vibration data;

extracting an amplitude for a frequency of the calibrated vibration data based on frequency analysis of the calibrated vibration data; and

determining a health condition of the part of the structure by a support vector machine classifier based on the amplitude for the frequency.

2. The method according to claim 1, wherein the vibration data comprises data in rest and data in movement, wherein the data in rest represents acceleration of the part of the structure substantially caused by ambient factors, and the data in movement represents the acceleration of the part of the structure substantially caused by an external vibration source.

3. The method according to claim 2, wherein the method further comprises:

calibrating coordinate axis directions of the two or more accelerometers based on the vibration data of the part of the structure by principal component analysis to calibrate the vibration data.

4. The method according to claim 3, wherein calibrating coordinate axis directions of the two or more accelerometers comprises aligning an axis direction of the at least one accelerometer with a direction along which the external vibration source moves based on the data in movement.

5. The method according to claim 1, wherein the vibration data is stored in a datastore and the determined health condition is stored in the same or different datastore.

6. The method according to claim 1, wherein the frequency analysis comprises fast Fourier transform of the vibration data.

7. The method according to claim 6, wherein extracting the amplitude for a frequency comprises extracting an amplitude spectrum of the fast Fourier transform of the calibrated vibration data.

8. The method according to claim 1, wherein the support vector machine classifier comprises a classification function g taking a form of g(E)=wE−

b, where E represents feature vectors of the vibration data, w and b are parameters determined by a machine learning process.

9. The method according to claim 8, wherein the machine learning process comprises an unsupervised machine learning process.

10. A non-transitory computer-readable medium, including computer-executable instructions stored thereon that when executed by a processor causes the processor to perform the method of claim 1.

11. A system for determining a health condition of a part of a structure, the system comprising:

two or more accelerometers providing vibration data,

a processor configured to monitor a building structure by constantly or regularly performing the steps of;

receiving or accessing the vibration data of the part of the structure, the vibration data being measured by the two or more accelerometers, the vibration data representing acceleration of the part of the structure substantially caused by an external vibration source;

transforming the vibration data by aligning an axis direction of the vibration data from each of the two or more accelerometers with a direction corresponding to maximum variance of the vibration data to create calibrated vibration data;

extracting an amplitude for a frequency of the calibrated vibration data based on frequency analysis of the calibrated vibration data; and

determining the health condition of the part of the structure by a support vector machine classifier based on the amplitude for the frequency.

12. A computer-implemented method for training a support vector machine classifier, the method comprising:

monitoring a building structure by constantly or regularly performing the steps of;

receiving or accessing vibration data of one or more parts of one or more structures measured by two or more accelerometers, the vibration data representing acceleration of the part of the structure substantially caused by an external vibration source;

transforming the vibration data by aligning an axis direction of the vibration data from each of the two or more accelerometers with a direction corresponding to maximum variance of the vibration data to create calibrated vibration data;

extracting two or more amplitudes for a frequency of the calibrated vibration data based on frequency analysis of the calibrated vibration data; and

applying a machine learning process to determine the support vector machine classifier based on the two or more amplitudes for a frequency.

13. A non-transitory computer-readable medium, including computer-executable instructions, stored thereon that when executed by a processor, causes the processor to perform the method of claim 12.

14. A system for training a support vector machine classifier for use in determining a health condition of a part of a structure, the system comprising:

two or more accelerometers providing vibration data of the part of the structure,

a processor configured to monitor a building structure by constantly or regularly performing the steps of;

receiving or accessing the vibration data of the part of the structure, the vibration data being measured by the two or more accelerometers, the vibration data representing acceleration of the part of the structure substantially caused by an external vibration source;

transforming the vibration data by aligning an axis direction of the vibration data from each of the two or more accelerometers with a direction corresponding to maximum variance of the vibration data to create calibrated vibration data;

extracting two or more amplitudes for a frequency of the calibrated vibration data based on frequency analysis of the calibrated vibration data; and

applying a machine learning process to determine the support vector machine classifier based on the two or more amplitudes for a frequency.