首页|基于BP神经网络的海底浅层超软黏土不排水抗剪强度预测

基于BP神经网络的海底浅层超软黏土不排水抗剪强度预测

扫码查看
准确评估海底浅层超软黏土的不排水抗剪强度对海洋资源开发、灾害评估、海洋工程建设等具有重要的意义.世界多个海域超软黏土数据分析对比表明,现有不排水抗剪强度评估经验公式存在受地域限制的问题,计算精度较低.因此,以超软黏土的6个物性指标为输入变量,不排水抗剪强度为输出变量,建立BP神经网络模型进行海底浅层超软黏土不排水抗剪强度预测;根据建立的超软黏土不排水抗剪强度与物性指标之间的非线性映射网络,以权积法求解超软黏土不排水抗剪强度对各指标的敏感度系数,定量分析各指标对不排水抗剪强度的影响程度.结果表明:考虑多因素影响的BP神经网络模型在超软黏土不排水抗剪强度预测方面具有普遍适应性,其预测结果非常接近落锥试验实测结果,均方差0.021 39,R2值达到0.987 4,预测精度远高于现有经验公式;采用权积法计算得到的对数流动性指标对不排水抗剪强度最为敏感;据此建立了以对数流动性指标为参数的黏土不排水抗剪强度预测经验公式,该公式在计算中国海域超软黏土不排水剪切强度方面具有较高精度,但由于数据量有限,依旧存在一定局限性.本研究为海底浅层超软黏土不排水抗剪强度的计算提供了参考.
Prediction of undrained shear strength of shallow seabed ultra soft clay based on BP neural network
Accurately evaluating the undrained shear strength of shallow ultra soft clay on the seabed is of great significance for marine resource development,disaster assessment,and marine engineering construction.The analysis and comparison of data on ultra soft clay in several sea areas around the world show that the existing empirical formulas for undrained shear strength assessment are limited by geographical regions and have low calculation accuracy.Therefore,taking six physical properties of ultra soft clay as input variables and undrained shear strength as output variables,a BP neural network model is established to predict the undrained shear strength of shallow ultra soft clay on the seabed.Based on the established nonlinear mapping network between the undrained shear strength and physical properties of ultra soft clay,the sensitivity coefficients of the undrained shear strength of ultra soft clay to each index are solved using the weight product method,and the influence of each index on the undrained shear strength is quantitatively analyzed.The results show that the BP neural network model considering multiple factors has universal adaptability in predicting the undrained shear strength of ultra soft clay.Its prediction results are very close to the measured results of the cone drop test,with a mean square error of 0.021 39 and R2 reaching 0.987 4.The prediction accuracy is much higher than existing empirical formulas.The undrained shear strength is most sensitive to the logarithmic liquidity index calculated by the weight product method.Based on this,an empirical formula for predicting the undrained shear strength of clay using logarithmic fluidity index as a parameter was established.This formula has high accuracy in calculating the undrained shear strength of ultra soft clay in Chinese waters.Due to the limited data,there are still certain limitations.This study provides a method reference for calculating the undrained shear strength of ultra soft clay in shallow seabed layers.

shallow seabedultra soft clayundrained shear strengthBP neural networkweight product methodphysical property indicatorssensitivity

孙耀、李飒、李怀亮、甘惠良、赵福臣

展开 >

天津大学建筑工程学院 天津 300350

海洋石油工程股份有限公司 天津 300461

海底浅层 超软黏土 不排水抗剪强度 BP神经网络 权积法 物性指标 敏感度

国家自然科学基金

51890911

2024

中国海上油气
中海石油研究中心

中国海上油气

CSTPCD北大核心
影响因子:1.266
ISSN:1673-1506
年,卷(期):2024.36(1)
  • 12