A vibration-based method for contact pattern assessment in straight bevel gears. (1st April 2019)

• Record Type: Journal Article
• Authors: Buzzoni, Marco; D'Elia, Gianluca; Mucchi, Emiliano; Dalpiaz, Giorgio
• Is Part Of: Mechanical systems and signal processing. Volume 120(2019)
• Page Start: 693
• Page End: 707
• Publication Date: 2019-04-01
• Subjects: Straight bevel gears -- Contact pattern -- Vibration analysis -- Naive Bayes -- k-Nearest Neighbors; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2018.10.002
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: A methodology for the assessment of bevel gear contact is proposed. A classifier is developed combining Naive Bayes and k-Nearest Neighbors methods. An extended and dedicated experimental campaign is presented. The method is verified through a comprehensive validation. Abstract: So far, the study of gear contacts in lightly loaded gears by means of vibration analysis has not been sufficiently addressed in the literature. Indeed, the complex nature of the physical phenomena involved makes the vibration analysis extremely challenging. This paper deals with the development and the validation of an approach for the contact pattern assessment in straight bevel gears within a pass/fail decision process. The proposed methodology is based on blending vibration-based condition indicators with classification algorithms in order to discriminate proper contact patterns from improper ones. Specifically, three different classification algorithms have been investigated: the Naive Bayes classifier, the weighted k-Nearest Neighbors classifier and a novel classifier proposed by the authors. The classifier accuracies are evaluated with a MC cross-validation that includes an extended experimental campaign consisting of more than one hundred different straight bevel gear pairs. The results show that the proposed classifier is superior to the other considered classifiers in terms of average accuracy. Finally, this manuscript proposes an original methodology that provides a reliable and quick assessment of the contact pattern in straight bevel gears considering different speeds, gear parameters and surface finish.