A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Metamodel-based sensitivity analysis of the optimal outrigger locations for damping outrigger-ATMD systems
https://orcid.org/0000-0002-4017-4617
https://orcid.org/0000-0002-3199-3354
Abstract This paper proposes two metamodels for the optimal outrigger locations (OOLs) of the optimal tuned viscous damper-outrigger-active tuned mass damper (OVD1-ATMD) system under seismic and wind loads. The first metamodel is a probabilistic model of OOLs with direct expressions. For this metamodel, we first investigated the monotonicity of the OOLs with main structural factors by the single-parameter sensitivity analyses. Using the observed monotonicity, three dimensionless parameters were defined. A probabilistic model for the OOLs of the OVD1-ATMD system was constructed and defined as a function of such dimensionless parameters through the dimensionless parameter analyses. Based on the probabilistic mode, we performed the Sobol’ sensitivity analysis to quantify the importance of different input features. The second metamodel is a deep neural network of OOLs. Based on the random analysis using Latin hypercube sampling, we first built a classification neural network for different segments of OOLs. We then built a regression neural network for each segment of OOLs. Using the model-agnostic method, we demonstrated the sensitivity or marginal effect of individual features on the predicted outcome of the regression neural network. Throughout the metamodeling and analysis, incorporating the ATMD into the OVD1 system results in the downward shift of OOLs compared to the pure OVD1 system; under seismic and wind loads, the accuracy of regression neural networks for OOLs is slightly higher than that of probabilistic models; due to more human intervention, probabilistic models may lost some randomness and uncertainty.
Highlights A probabilistic model of OOLs for the OVD1-ATMD system is proposed. Classification and regression neural networks of OOLs for the OVD1-ATMD system are constructed. Sensitivity analyses are performed using the Sobol’ method and model-agnostic method. Addition of the ATMD to the OVD1 system forces the OOLs to head to the lower location.
Metamodel-based sensitivity analysis of the optimal outrigger locations for damping outrigger-ATMD systems
https://orcid.org/0000-0002-4017-4617
https://orcid.org/0000-0002-3199-3354
Abstract This paper proposes two metamodels for the optimal outrigger locations (OOLs) of the optimal tuned viscous damper-outrigger-active tuned mass damper (OVD1-ATMD) system under seismic and wind loads. The first metamodel is a probabilistic model of OOLs with direct expressions. For this metamodel, we first investigated the monotonicity of the OOLs with main structural factors by the single-parameter sensitivity analyses. Using the observed monotonicity, three dimensionless parameters were defined. A probabilistic model for the OOLs of the OVD1-ATMD system was constructed and defined as a function of such dimensionless parameters through the dimensionless parameter analyses. Based on the probabilistic mode, we performed the Sobol’ sensitivity analysis to quantify the importance of different input features. The second metamodel is a deep neural network of OOLs. Based on the random analysis using Latin hypercube sampling, we first built a classification neural network for different segments of OOLs. We then built a regression neural network for each segment of OOLs. Using the model-agnostic method, we demonstrated the sensitivity or marginal effect of individual features on the predicted outcome of the regression neural network. Throughout the metamodeling and analysis, incorporating the ATMD into the OVD1 system results in the downward shift of OOLs compared to the pure OVD1 system; under seismic and wind loads, the accuracy of regression neural networks for OOLs is slightly higher than that of probabilistic models; due to more human intervention, probabilistic models may lost some randomness and uncertainty.
Highlights A probabilistic model of OOLs for the OVD1-ATMD system is proposed. Classification and regression neural networks of OOLs for the OVD1-ATMD system are constructed. Sensitivity analyses are performed using the Sobol’ method and model-agnostic method. Addition of the ATMD to the OVD1 system forces the OOLs to head to the lower location.
Metamodel-based sensitivity analysis of the optimal outrigger locations for damping outrigger-ATMD systems
https://orcid.org/0000-0002-4017-4617
https://orcid.org/0000-0002-3199-3354
Abstract This paper proposes two metamodels for the optimal outrigger locations (OOLs) of the optimal tuned viscous damper-outrigger-active tuned mass damper (OVD1-ATMD) system under seismic and wind loads. The first metamodel is a probabilistic model of OOLs with direct expressions. For this metamodel, we first investigated the monotonicity of the OOLs with main structural factors by the single-parameter sensitivity analyses. Using the observed monotonicity, three dimensionless parameters were defined. A probabilistic model for the OOLs of the OVD1-ATMD system was constructed and defined as a function of such dimensionless parameters through the dimensionless parameter analyses. Based on the probabilistic mode, we performed the Sobol’ sensitivity analysis to quantify the importance of different input features. The second metamodel is a deep neural network of OOLs. Based on the random analysis using Latin hypercube sampling, we first built a classification neural network for different segments of OOLs. We then built a regression neural network for each segment of OOLs. Using the model-agnostic method, we demonstrated the sensitivity or marginal effect of individual features on the predicted outcome of the regression neural network. Throughout the metamodeling and analysis, incorporating the ATMD into the OVD1 system results in the downward shift of OOLs compared to the pure OVD1 system; under seismic and wind loads, the accuracy of regression neural networks for OOLs is slightly higher than that of probabilistic models; due to more human intervention, probabilistic models may lost some randomness and uncertainty.
Highlights A probabilistic model of OOLs for the OVD1-ATMD system is proposed. Classification and regression neural networks of OOLs for the OVD1-ATMD system are constructed. Sensitivity analyses are performed using the Sobol’ method and model-agnostic method. Addition of the ATMD to the OVD1 system forces the OOLs to head to the lower location.
Metamodel-based sensitivity analysis of the optimal outrigger locations for damping outrigger-ATMD systems
Xing, Lili (author) / Song, Ge (author) / Zhou, Ying (author) / Zhang, Peng (author)
2024-01-16
Article (Journal)
Electronic Resource
English
| Springer Verlag | 2023