Multi-dimensional structure damage identification method for convolutional neural network processing of mass vibration transmissibility data

Abstract

The invention discloses a multi-dimensional structure damage identification method for convolutional neural network processing of mass vibration transmissibility data. The method specifically comprises the steps: collecting mass structure acceleration response data through a sensor system; dividing the data into a reference response and a non-reference response, calculating the transmissibility between the reference response and the non-reference response, and forming a mass sample data set based on the vibration transmissibility; establishing a convolutional neural network model, and performing training by using the transmissibility data set; after N times of training, enabling the convolutional neural network to reach convergence, carrying out structural damage identification by using the trained convolutional neural network, and outputting a damage mode of the convolutional neural network. Compared with a traditional method, the built damage identification method integrating the vibration transmissibility and the convolutional neural network can eliminate the excitation interference and the method has the advantages of being high in damage sensitivity, high in damage identification accuracy and high in noise resistance, and can efficiently process massive high-dimensional data.

Description

technical field [0001] The invention relates to a structural damage identification method, in particular to a multi-dimensional structural damage identification method processed by convolutional neural network processing of massive vibration transmissibility data, and belongs to the field of structural damage diagnosis. Background technique [0002] In the long-term service of the structure, due to the degradation of its own materials and the influence of the complex environment, damage will inevitably occur. Structural damage will threaten the safe operation of the structure and even lead to the destruction of the structure. The structural non-destructive testing method based on structural dynamic response signals and artificial intelligence algorithms has developed rapidly in recent years because of its low cost, high efficiency, and the advantages of real-time detection and online monitoring, and is currently being gradually applied to the fields of machinery and civil eng...

AI Technical Summary

Problems solved by technology

Conventional damage identification methods based on structural dynamic response signals and intelligent algorithms generally use traditional time-series-based time-domain signals or Fourier transform-based frequency-domain signals as data for damage identification, using traditional machine learning algorithms such as Artificial neural network is used as an intelligent algorithm for damage identification; traditional time and frequency domain data are not sensitive to damage and are easily disturbed by excitation; traditional intelligent algorithms are difficult to fully extract damage features in data, and the efficiency is low when processing massive high-dimensional data. Requires a large amount of computing resources

Images