Gear Fault Diagnostics Using Extended Phase Space Topology

T. Haj Mohamad and C. Nataraj
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_17_014.pdf8.65 MBSeptember 6, 2017 - 9:41am

Maintenance cost contributes as a major part of operating costs in any industry. This has motivated industries to adopt best machine condition monitoring techniques so that costs can be reduced and productivity can be increased. In any machinery system, gears are one of the most important transmission elements and the availability of the entire system depends on the proper function of gears. The demand is increasing day by day for enhancement of the performance and service life of gears.

Gear fault detection is a critical task especially in the case of complex machines. Vibration and acoustics methods are most widely used for fault diagnosis of gears. Many of the used techniques such as frequency analysis, statistical analysis and time frequency analysis are able to detect different faults. However, most of them suffer from low detection quality and or slow response time. These limitations have restricted their effectiveness in online or real time fault detection applications.

This paper applies a novel feature extraction method called the Extended Phase Space Topology (EPST) for gear diagnostics. The EPST method is based on characterizing the topology of the density distribution of the vibration data. The density distribution of the vibration signal is approximated using Legendre polynomials. Then the coefficients of the orthogonal polynomials are used as features. Finally, an artificial neural network is used as a classifier to distinguish between the different gear conditions.

The proposed method has been applied to vibration data of a helicopter gear box mock-up of system (16 ft long). For this study, multiple test gears with different health conditions (such as a healthy gear and defective gears with root crack on one tooth, multiple cracks on five teeth and missing tooth) are studied. The vibrational signals are recorded using a triaxial accelerometer installed on the test.

We show that the density distribution provides rich information about the status of the health of the gear. Furthermore, results show that the innovative EPST procedure has an outstanding performance in gear fault detection and classification with minimum knowledge about the dynamic response of the system. The EPST is robust to noise, fast to execute, and requires no feature ranking or selection, and therefore, can easily be applied in an automated process.

Publication Year: 
2017
Publication Volume: 
8
Publication Control Number: 
014
Submission Keywords: 
Machinery diagnostics
gear fault detection
feature extraction
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed