地下水封洞库单裂隙花岗岩纵波速度变化规律与预测模型

Prediction model and variation law of P-wave velocity of single fracture granite in an underground water-sealed storage cavern

揭示裂隙岩体纵波速度变化规律对工程岩体质量分级与稳定性评价具有重要意义。以某地下水封洞库无充填型单裂隙花岗岩为研究对象,基于钻孔电视成像、水压致裂法地应力测试与声波全波列测井,获取了384组单裂隙花岗岩的几何特性、受力状态与纵波速度,构建起了预测单裂隙花岗岩纵波速度的进化-神经网络模型,分析了关键指标影响下单裂隙花岗岩纵波速度的变化规律。研究表明:该水封洞库单裂隙花岗岩纵波速度分布于4300~5330 m/s之间,82.3%的纵波速度在4700~5200 m/s之间;选取裂隙法向应力、平均张开度与倾角作为单裂隙花岗岩纵波速度的预测指标是合理可行的;将现场测试数据分为训练样本与测试样本,基于遗传算法优化神经网络权值、阈值的进化-神经网络模型构建出单裂隙花岗岩纵波速度预测模型,其测试误差最大仅为2.9%,85%的样本测试误差不超过1.5%,预测模型精度较高。分析了纵波速度变化规律,发现单裂隙花岗岩纵波速度随裂隙法向应力增大而增大,但当法向应力增至5 MPa后的纵波速度增大速率逐渐减小,纵波速度随裂隙张开度增大而逐渐减小,纵波速度在裂隙倾角小于40°时无明显变化,此后纵波速度随倾角增大而增大。

Revealing the variation behavior of P-wave velocity in fractured rock masses is of great significance for the quality grading and stability evaluation of rock masses for engineering purposes.The nonfilling single fracture granite of an under ground water sealed storage cavern was taken as the research object.Based on borehole television images,hydraulic fracturing geostress tests,and ultrasonic full waveform logging,the geometric characteristics,stress state,and P-wave velocity of 384 groups of single fracture granites were obtained.An evolutionary neural network model for the prediction of the P-wave velocity of granite with a single fracture was constructed,and the variation behavior of the key indexes affecting the P-wave velocity of granite with a single fracture was analyzed.The study shows that the P-wave velocity of a single fracture granite in the water-sealed storage cavern is distributed around 4300-5330 m/s,and 82.3%of the P-wave velocity varies within 4700-5200 m/s.It is reasonable and feasible to select the fracture normal stress,average aperture,and dip angle as prediction indexes of the P-wave velocity of granite with a single fracture.The field test data sets are divided into training samples and test samples.The P-wave velocity prediction model of granite with a single fracture,based on the evolutionary neural network model,is constructed.The neural network weight and threshold are optimized by the genetic algorithm.The maximum test error of the prediction model is only 2.9%,and the test error of 85%of the samples is less than 1.5%.The prediction model thus yields high accuracy.The variation feature of the P-wave velocity revealed that the P-wave velocity of granite with a single fracture increases with increasing normal stress on the fracture.However,the increase in the P-wave velocity decreases gradually when the normal stress increases to 5 MPa.The P-wave velocity decreases with an increasing fracture aperture.The P-wave velocity increases with an increasing dip angle.However,no difference occurred considering that the fracture dip angle is less than 40°.