Research On Acoustic Emission Signal Identification Algorithm Of Rock Fracture Based On Deep Learning

The acoustic emission phenomenon of rock can reflect a lot of information contained in the process of rock fracture.By analyzing the acoustic emission characteristics of rock fracture process,we can not only understand the rock fracture mechanism,but also provide theoretical basis for rock acoustic emission monitoring technology.At present,rock fracturing technology is the main technical means of oil and gas energy development,and a lot of useful information contained in rock fracturing process can be reflected by acoustic emission signals generated during fracturing.Therefore,this paper is mainly based on the acoustic emission signals generated by indoor rock fracturing,the sparrow search algorithm and Variational Mode Decomposition(VMD)algorithm optimized by chaos are used to filter the collected acoustic emission signals,and the acoustic emission signals released by rock fracture with different brittle mineral content are identified.The specific research work and achievements of this topic include the following four aspects:1.Conduct rock fracturing experiments.Four representative rock samples with brittle minerals content of 0%,10%,30% and 50% were prepared and subjected to rock fracturing tests.In the laboratory,acoustic emission monitoring system and loading system were used to collect acoustic emission signals released by rock fracture under the content of four brittle minerals.AEWIN visualization software is used to display the signal waveform,which provides the data basis for the following experimental research.2.In view of the problem that the AE signal is nonlinear and nonstationarity and the VMD algorithm is hardly to determine the decomposition parameter K and penalty factor αwhen denoising rock acoustic emission signal.In this paper,the VMD algorithm is applied to the AE signal filtering of rock fracture,and the parameters K and α of VMD algorithm are optimized by the sparrow search algorithm optimized by Tent chaos.Then,the decomposition parameter and penalty factor of VMD are determined according to the search results,and the optimized VMD is used to filter the AE signal of rock fracture.3.In this paper,a method based on acoustic emission technology is proposed to identify acoustic emission signals released by rock fracture with different brittle mineral content.Aiming at the related interference characteristics of acoustic emission signal data released by rock mass fracture,such as noise and so on,in this paper,a multi-scale one-dimensional convolutional neural network embedded with super efficient channel attention(ECA)module is constructed,and using multi-scale convolution kernel to extract features with different fineness.After that,the AE signal data is fully extracted by multi-scale One-Dimensional Convolutional Neural Network(1DCNN)and Bidirectional Long-Short Term Memory(BLSTM)networks,including the time-series correlation characteristics,local spatial non-correlation characteristics and weak periodicity laws.Finally,the AE signals released by rocks with different brittle mineral contents are accurately recognized by softmax function,And the recognition accuracy can more than 90%.