Structural damage identification method and system based on deep extreme learning machine

Abstract

The invention discloses a structural damage identification method and system based on a deep extreme learning machine, and the method combines an advanced signal processing technology, the deep extreme learning machine and a particle swarm optimization algorithm, and builds a structural damage identification method which can achieve the self-adaptive hyper-parameter optimization, is high in anti-noise performance and is high in damage identification robustness. Advanced signal processing technology Hilbert-Huang transform is adopted to reduce the influence of measurement noise on vibration signals, and the noise immunity of a damage identification algorithm is improved. And a particle swarm optimization algorithm is adopted to optimize hyper-parameters based on a deep extreme learning machine, so that the accuracy of structural damage identification is improved. And finally, the performance of the damage identification algorithm is tested by using ten-fold cross experiment verification, so that the robustness of the damage identification algorithm is ensured. Meanwhile, the structure damage data monitored on line is continuously used for updating the structure damage identification data set, so that the damage identification precision is continuously improved.

Description

technical field [0001] The invention belongs to the field of structural damage identification, and specifically relates to a structural damage identification method and system based on a deep extreme learning machine. Background technique [0002] Large-scale architectural structures are the carrier of cities, and the healthy service of architectural structures is the basis for people's normal life. With the aging of materials, changes in loads and the continuous impact of natural disasters during the use of the structure, the building structure will be damaged to varying degrees, and with the deepening of the damage, the safety performance of the building structure cannot be guaranteed. Therefore, it is of great significance to conduct health monitoring and damage identification for building structures and key components of building structures in service. Structural health monitoring and damage identification, specifically, refers to arranging sensors on the structure, col...

AI Technical Summary

Problems solved by technology

However, usually in structural damage measurement, there is a problem of high environmental noise, which sometimes leads to poor identification of structural damage

[0004] Shallow machine learning models have also been widely used in damage identification methods based on vibration signals, but because shallow learning models cannot establish an accurate mapping relationship between damage data and damage degrees, the accuracy of damage identification is often poor. Sometimes it is not even as good as the recognition accuracy of damage identification methods based on damage indicators

Images