一种用于声发射数据分析的新型贝叶斯方法。

A novel Bayesian approach to acoustic emission data analysis.

作者信息

Agletdinov E, Pomponi E, Merson D, Vinogradov A

机构信息

Institute of Advanced Technologies, Togliatti State University, 445667, Russia.

Department of Data Science and Analytic, CGnal s.p.a., Via Carducci 38, 20122 Milano, Italy.

出版信息

Ultrasonics. 2016 Dec;72:89-94. doi: 10.1016/j.ultras.2016.07.014. Epub 2016 Jul 25.

DOI:10.1016/j.ultras.2016.07.014

PMID:27494093

Abstract

Acoustic emission (AE) technique is a popular tool for materials characterization and non-destructive testing. Originating from the stochastic motion of defects in solids, AE is a random process by nature. The challenging problem arises whenever an attempt is made to identify specific points corresponding to the changes in the trends in the fluctuating AE time series. A general Bayesian framework is proposed for the analysis of AE time series, aiming at automated finding the breakpoints signaling a crossover in the dynamics of underlying AE sources.

摘要

声发射（AE）技术是用于材料表征和无损检测的常用工具。AE源于固体中缺陷的随机运动，本质上是一个随机过程。每当试图识别与波动的AE时间序列趋势变化相对应的特定点时，就会出现具有挑战性的问题。本文提出了一个用于分析AE时间序列的通用贝叶斯框架，旨在自动找到指示潜在AE源动态变化的断点。