Application of classification neural networks for identification of damage stages of degraded low alloy steel based on acoustic emission data analysis

Abstract

AbstractThe paper presents the influence of low alloy steel degradation on the acoustic emission (AE) generated during static tension of notched specimen. The material was cut from a technological pipeline long-term operated in the oil refinery industry. Comparative analysis of AE activity generated by damage process of degraded and new material has been carried out. The different AE parameters were used to detect different stages of fracture process of low alloy steel under quasi-static tensile test. Neural networks with three layers were created with Broyden–Fletcher–Goldfarb–Shanno learning algorithm for a database analysis. The different AE parameters were included in the input layer. Classification neural networks were created in order to determine the stages of material degradation. The results obtained from the carried out studies will be used as the basis for new methodology development of the assessment of the structural condition of in-service equipment.