基于振动与声音信号深度学习的岩性识别方法

Lithology Identification Method Based on Deep Learning of Vibration and Sound Signals

岩性识别对地质勘查和储层评价具有重要意义,科学有效地开展岩性自动识别的相关研究能够有效地为勘查过程提供指导,减少工作的盲目性和冗杂性。针对常见的砂岩地层,选择三类砂岩,基于室内微钻试验台,设计钻杆转速、钻孔深度和钻孔位置三个变量,检测钻进过程中产生的振动和声音特征信号。将采集的振动和声音信号预处理,提高信噪比,生成数据集。将振动和声音的数据集按6∶2∶2的比例划分为训练集、验证集和测试集,之后分别构建二维卷积神经网络和一维卷积神经网络并使用训练集和验证集训练岩性识别模型,最后运用未经训练的测试集验证模型准确率。模型训练完成后,以频谱图为数据集的振动信号识别模型准确率达到95.19%,以梅尔频率倒谱系数为数据集的声音信号识别模型准确率达到73.58%。研究结果表明,不同岩性在钻进过程中产生的振动和声音信号具有不同信号特征,基于振动和声音信号的岩性自动识别方法可以较好地实现几类砂岩的自动识别,这为地质勘查时的岩性自动识别提供了参考与依据。

Lithology identification is of great significance to geological exploration and reservoir evaluation.Scientifically and effectively carrying out relevant research on automatic identification of lithology can effectively provide guidance for the exploration process and reduce blindness and redundancy of work.For common sandstone formation,three types of sandstone were selected.Based on the indoor micro drill test bench,three variables of drilling rotation speed,drilling depth and drilling position were designed,and then the vibration and sound characteristic signals generated during the drilling process were detected.The signals were preprocessed,the signal to noise ratio was improved,and data sets were generated.Then,the data set was divided into training set,verification set and test set according to the ratio of 6∶2∶2.After constructing a 2D convolutional neural networks and a 1D convolutional neural networks,the training set and verification set were used to train the recognition models,and the test set was used to verify the accuracy of the models.Finally,The accuracy of the vibration signal recognition model with the spectrogram as the data set reaches 95.19%,and the accuracy of the sound signal recognition model with the MFCC(Mel-frequency cepstral coefficients)as the data set reached 73.58%.The research result shows that the vibration and sound signals generated by different lithologies during the drilling process have different signal characteristics.The automatic lithology identification based on vibration and sound signals can better realize the automatic classification of several types of sandstone,and it provides a reference and basis for geological exploration.