A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
State Parameter Based Liquefaction Probability Evaluation
https://orcid.org/http://orcid.org/0009-0007-6599-0948
The relative density and effective stress of the soil influence the cyclic stress or liquefaction behaviour of granular soil significantly. The state parameter (ѱ) which accounts for relative density and effective stress is utilised in this work to quantify soil liquefaction probability of failure (PL) using the First order second moment (FOSM) method. It is observed that the state parameter-based cyclic resistance ratio (CRR) model used to analyse PL is a complex function of cone penetration test (CPT) parameters since ѱ itself is based on empirical relations which include several derived parameters and other factors, like the coefficient of earth pressure at rest (K0) and soil compression parameters. In order to overcome this complexity in liquefaction probability evaluation, three Machine learning (ML) models, namely Simple Recurrent Neural Network (Simple RNN), Convolutional Neural Network (CNN) and, Deep Neural Network (DNN) are developed and recommended according to their performance in predicting PL. To evaluate the effectiveness of these models, nine statistical performance parameters are calculated. In order to find the model with the highest performance, additional plots such as regression plots, Taylor’s diagrams, error matrix, rank analysis, and regression error characteristic curves are presented. The present study shows that the three ML models based on state parameter perform well in predicting PL. Among these ML models, the CNN model demonstrates highest performance. The findings of this work will aid in expanding the application of state parameter-based ML models and offer risk evaluations for geotechnical engineering design.
State Parameter Based Liquefaction Probability Evaluation
https://orcid.org/http://orcid.org/0009-0007-6599-0948
The relative density and effective stress of the soil influence the cyclic stress or liquefaction behaviour of granular soil significantly. The state parameter (ѱ) which accounts for relative density and effective stress is utilised in this work to quantify soil liquefaction probability of failure (PL) using the First order second moment (FOSM) method. It is observed that the state parameter-based cyclic resistance ratio (CRR) model used to analyse PL is a complex function of cone penetration test (CPT) parameters since ѱ itself is based on empirical relations which include several derived parameters and other factors, like the coefficient of earth pressure at rest (K0) and soil compression parameters. In order to overcome this complexity in liquefaction probability evaluation, three Machine learning (ML) models, namely Simple Recurrent Neural Network (Simple RNN), Convolutional Neural Network (CNN) and, Deep Neural Network (DNN) are developed and recommended according to their performance in predicting PL. To evaluate the effectiveness of these models, nine statistical performance parameters are calculated. In order to find the model with the highest performance, additional plots such as regression plots, Taylor’s diagrams, error matrix, rank analysis, and regression error characteristic curves are presented. The present study shows that the three ML models based on state parameter perform well in predicting PL. Among these ML models, the CNN model demonstrates highest performance. The findings of this work will aid in expanding the application of state parameter-based ML models and offer risk evaluations for geotechnical engineering design.
State Parameter Based Liquefaction Probability Evaluation
https://orcid.org/http://orcid.org/0009-0007-6599-0948
The relative density and effective stress of the soil influence the cyclic stress or liquefaction behaviour of granular soil significantly. The state parameter (ѱ) which accounts for relative density and effective stress is utilised in this work to quantify soil liquefaction probability of failure (PL) using the First order second moment (FOSM) method. It is observed that the state parameter-based cyclic resistance ratio (CRR) model used to analyse PL is a complex function of cone penetration test (CPT) parameters since ѱ itself is based on empirical relations which include several derived parameters and other factors, like the coefficient of earth pressure at rest (K0) and soil compression parameters. In order to overcome this complexity in liquefaction probability evaluation, three Machine learning (ML) models, namely Simple Recurrent Neural Network (Simple RNN), Convolutional Neural Network (CNN) and, Deep Neural Network (DNN) are developed and recommended according to their performance in predicting PL. To evaluate the effectiveness of these models, nine statistical performance parameters are calculated. In order to find the model with the highest performance, additional plots such as regression plots, Taylor’s diagrams, error matrix, rank analysis, and regression error characteristic curves are presented. The present study shows that the three ML models based on state parameter perform well in predicting PL. Among these ML models, the CNN model demonstrates highest performance. The findings of this work will aid in expanding the application of state parameter-based ML models and offer risk evaluations for geotechnical engineering design.
State Parameter Based Liquefaction Probability Evaluation
Int. J. of Geosynth. and Ground Eng.
Kumar, Kishan (author) / Samui, Pijush (author) / Choudhary, S. S. (author)
2023-12-01
Article (Journal)
Electronic Resource
English
Assessing Probability-based Methods for Liquefaction Potential Evaluation
| Online Contents | 2002
Assessing Probability-based Methods for Liquefaction Potential Evaluation
| British Library Online Contents | 2002
Probability-based assessment of liquefaction risk potential
| British Library Conference Proceedings | 2003
The fuzzy probability method for soil liquefaction hazard evaluation
| British Library Conference Proceedings | 1996