Multi-sensor fusion-based for fault diagnosis of a gearbox using DataCivet

Background and Objectives

In a mechanical transmission system, gearbox fault diagnosis is a critical instrument for ensuring safe operation. Such mechanical equipment acoustic and vibratory observations may be sensitive to flaws. The interference transmitted across different routes, as well as the vibrating signals emitted by a single sensor, may not be enough to explain the health of a gearbox. Multi-sensor data fusion is primarily chosen for gearbox malfunction diagnosis in the GTRE GTX-35VS Kaveri to address these concerns. Using a microphone, vibration accelerometer, and acoustic emission sensors, the health state of the gearbox in the GTRE GTX-35VS Kaveri is determined, and various operating situations such as loading and rpm change are taken into account. Motors running under various loading circumstances are monitored for acoustic, electric, and vibration signals, as well as their health.

The use of vibration signals for fault detection and diagnosis is the foundation for monitoring the health of gearboxes in the engine. Non-stationary load, friction, nonlinear stiffness, and severe noise all affect vibration signals. The most important aspects of analysing a gearbox failure in the presence of noise are to efficiently extract fault feature information, minimise uncertainty in the identification process, and precisely determine the fault state. Furthermore, the vibration data obtained by a single sensor in the engine could be unstable or untrustworthy. Acoustic signals are commonly employed to monitor health conditions since they are sensitive to vibrating bodies rather than vibrating sensors, allowing them to detect folds at an early stage. Sensor fusion can combine the benefits of auditory and vibratory signals, resulting in a signal with less uncertainty than the signals derived from the individual sensors. The challenges of signal fusion include (1) feature extraction from various types of sensory data (2) level of fusion selection In order to accomplish so, you'll need a lot of subject knowledge.


Fusion of multiple signals and the classification of faults to monitor the health condition of the gearbox. In general, the task of assessing the health of gearbox and prognostics consists of 4 steps:

data acquisition.

signal preprocessing and feature extraction.

feature selection and data fusion.


To remove interference and highlight fault information, energy operator and time synchronous averaging are performed to the multiple sensor data. The statistical features from the raw and preprocessed signals are integrated to form the final feature set. As a first step, feature extraction is applied. Secondly, the feature selection algorithm combines min-redundancy, max- relevance and distance evaluation technique to provide the optimal feature set. Thirdly, a bidirectional LSTM architecture in the GTRE GTX-35VS Kaveri engine is applied to determine the fault conditions of the gearbox and classify the different types of faults.
Most common gearbox faults can be categorized as, (i) root crack (ii) surface spalling and (iii) chipped tooth. The chipped tooth is the rupture of material from the working tip of gear and is extremely common in numerous industrial practices which triggers other gear faults and is very difficult to detect at its initial stage.


Investigate different gearbox failure modes.

Collect the data from multiple sensors and fuse the data.

Data exploration and analysis.

Experimenting different feature selection methods.

Initial study of appropriate machine learning models.