The invention discloses a flaw state diagnosis method based on improved acoustic emission density clustering, which comprises the following steps: carrying out fusion and dimensionality reduction on acoustic emission features based on random neighborhood embedding, carrying out clustering diagnosis on the fused features through a density clustering algorithm, and jointly optimizing parameters of the two features based on a genetic algorithm; and finally, diagnosing the crack signal contained in the acoustic emission feature set and the corresponding damage state. An unsupervised algorithm is adopted, the internal distribution rule and essential features of acoustic emission signals in the steel degradation process are explored. A dimension reduction feature space is constructed through a random neighborhood embedding technology in manifold learning, a density clustering algorithm is improved by a proposed internal evaluation target. A selection process of sensitive parameters is autonomously optimized, and parameter robustness and anti-interference performance of the algorithm are enhanced, so that automatic clustering of acoustic emission samples in different damage stages is realized.
本发明公开了一种基于改进声发射密度聚类的伤损状态诊断方法,所述方法基于随机邻域嵌入对声发射特征融合降维,再经由密度聚类算法对融合特征聚类诊断,并基于遗传算法联合优化二者的参数,最终诊断出声发射特征集中包含的裂纹信号及其对应的伤损状态。本发明采用无监督算法,发掘钢材劣化过程中声发射信号的内在分布规律与本质特征,包括通过流形学习中的随机邻域嵌入技术构建降维特征空间,并通过所提出的内部评估目标对密度聚类算法进行改进,自主优化其敏感参数的选取过程并增强算法自身的参数鲁棒性与抗干扰性,从而实现对不同伤损阶段的声发射样本自动分簇。
Damage state diagnosis method based on improved acoustic emission density clustering
一种基于改进声发射密度聚类的伤损状态诊断方法
2021-07-23
Patent
Electronic Resource
Chinese
Diagnosis using acoustic emission
Online Contents | 2012
|An improved density-based spatial clustering of application with noise
British Library Online Contents | 2018
|Damage Monitoring of Bridge Cable through Acoustic Emission Technique
British Library Conference Proceedings | 2010
|Fatigue Design Model Based on Damage Mechanisms Revealed by Acoustic Emission Measurements
British Library Online Contents | 1994
|