On monitoring fretting fatigue damage in solid railway axles by acoustic emission with unsupervised machine learning and comparison to non-destructive testing techniques.
Railway axles are safety-critical components of the rolling stock and the consequences of possible in-service failures can have dramatic effects. Although this element is traditionally designed against such failures, the initiation and propagation of service cracks are still occasionally observed, r...
| Autores: | , |
|---|---|
| Formato: | artículo |
| Fecha de publicación: | 2023 |
| País: | España |
| Recursos: | Universidad Francisco de Vitoria |
| Repositorio: | DDFV. Repositorio Institucional de la Universidad Francisco de Vitoria |
| Idioma: | inglés |
| OAI Identifier: | oai:ddfv.ufv.es:10641/5594 |
| Acesso em linha: | https://hdl.handle.net/10641/5594 |
| Access Level: | acceso abierto |
| Palavra-chave: | Fretting fatigue Solid railway axle Non-destructive testing Ultrasonic phased array testing Structural health monitoring Acoustic emission Unsupervised machine learning |
| Resumo: | Railway axles are safety-critical components of the rolling stock and the consequences of possible in-service failures can have dramatic effects. Although this element is traditionally designed against such failures, the initiation and propagation of service cracks are still occasionally observed, requiring an effective application of non-destructive testing and structural health monitoring approaches. This paper investigates the application of structural health monitoring by acoustic emission to the case of solid railway axles subject to fretting fatigue damage. A full-scale test was performed on a specimen in which artificial notches were suitably manufactured in order to cause the initiation and evolution of fretting fatigue damage up to the stage of relevant propagating fatigue cracks. During the test, both periodical phased array ultrasonic inspections and continuous acquisition of acoustic emission data have been carried out. Moreover, at the end of the test, the specimen was inspected, analyzed and evaluated by visual inspection and magnetic particles testing, while acoustic emission raw data were post-processed by a special unsupervised machine learning algorithm based on an Artificial Neural Network. It is demonstrated that the proposed methodology is very effective to detect the onset of crack initiation in a non-invasive and safe way. |
|---|