<p>Spiral-bevel gears are important part of many transmission systems, especially aeronautical transmission systems because of their smooth operation and strong carrying capacity. This type of gears have high contact ratio, which makes the diagnosis of even serious defects very difficult. Therefore, spiral-bevel gears have seldomly been used as bench marking for defect diagnosis techniques. Moreover, their vibration signal is very complicated because of varying meshing point, changing number of meshing gear pairs, collision between the tooth during meshing and non-linear and non-stationery behaviour. Therefore, fault detection and diagnosis using conventional vibration analysis techniques is very difficult. In order to overcome these challenges, in this research AI (Artificial Intelligence) techniques have been used for defect diagnosis of spiral-bevel gears. Although Al techniques have gained much success in the field of defect diagnosis, however, mostly these methods use an assumption that training, and testing data are from same operating conditions. Nevertheless, when the operating conditions in which trained model is applied differs from the operating conditions in which model was trained, then the performance of these approaches may drop significantly. Outside laboratory, in real world applications mostly different operating conditions are encountered and to get data for all operating conditions may be exorbitant. Therefore, it is not possible to train the deep learning or machine learning models using data from all possible operating conditions. &nbsp;In order to overcome this limitation and to make the AI techniques suitable for defect diagnosis of spiral bevel gears in different operating conditions, effort has been made to find few fault discriminating features, extracted from raw vibration data, which are lesser sensitive to working conditions but are fault discriminative. Vibration data in different operating conditions (speed and load) was collected from a spiral-bevel gears test rig. The tested spiral-bevel gears have robust design and smooth operation with high contact ratio between 2 and 3. Even One broken tooth do not affect the vibration signal too much, therefore, it is challenging task to diagnose the fault in such gearboxes. Statistical features are extracted from time domain vibration signal and spectral kurtosis. Selected features from one operating condition are used to train ANN and KNN as diagnostic classifiers. After training, predictions were made using trained diagnostic classifiers on testing data from different operating conditions (speed and load). A comparison between the performance of both classifiers with varying hidden neuron count (N) and variable K values was performed in order to check the individual capability and suitability of ANN and KNN for fault diagnosis of spiral bevel gears under different operating conditions.</p>

Intelligent Defect Diagnosis of Spiral Bevel Gears Under Different Operating Conditions Using ANN and KNN Classifiers Defect Diagnostics of Spiral Bevel Gears Under Different Operating Conditions Using ANN and KNN Classifier

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.