Research On Damage Trend Prediction Of Metal Materials Based On CNN-LSTM Network

Metal materials have the advantages of high mechanical strength,good heat resistance and durability,and are widely used in various practical projects.However,due to the many accidents caused by metal material damage,and the occurrence of metal material damage usually has no warning,it is very important to be able to predict the trend of metal material damage in time.Acoustic emission testing technology is a dynamic non-destructive testing technology.Since the acoustic emission signal collected by the acoustic emission detection technology comes from the defects of the material itself,and the detection of the change of the material damage is relatively sensitive,the acoustic emission detection technology can monitor the occurrence and development of the damage in real time,and has a higher reliability monitoring of the damage process.Time series is a special series.Time series is not a variable series.It contains certain historical laws.Using the laws to predict its future trend is an important subject and has practical application value.However,traditional time series forecasting methods have some shortcomings.In recent years,the development of artificial intelligence has been very rapid,of which artificial neural networks have attracted the most attention.Researchers use artificial neural network models to deal with some of the problems in traditional time series forecasting methods.This article first introduces the theoretical basis of acoustic emission detection technology,and then introduces the acoustic emission characteristics of the metal material damage process.Based on the acoustic emission signal data collected in the static load test of metal materials,the acoustic emission signal trend of metal material damage is obtained from the time series of the acoustic emission signal during the test.Combined with the damage mechanism of metal materials,the acoustic emission signal trend is analyzed,and the acoustic emission signal in the metal damage process is divided into stages.Based on the microscopic mechanism of metal plastic deformation,the acoustic emission signal trend corresponds to the metal material damage trend.According to the change trend of RMS characteristic parameters of acoustic emission signal with time,combined with the related research of artificial neural network,a metal material damage trend prediction model based on CNN-LSTM network is designed.This paper inputs the processed acoustic emission signal time series into the CNN network model for feature extraction,and then inputs the extracted feature vectors into the LSTM network model to predict the trend of metal materials.The test results show that the CNN-LSTM network has smaller prediction errors and better results than the ARIMA model in the traditional time prediction method and the NARX neural network and LSTM neural network in the modern time series prediction method.