Reliability Analysis of Suspension Bridge Using Gaussian Process Based Response Surface Method

Zhao, Wei; Su, Guoshao; Hu, Li

doi:10.4028/www.scientific.net/amr.860-863.2970

Cited by 1 publication

(2 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Given the specified training set D = ( X N , T N ) (In the conception, the structure parameter input of X N and T N are the vibration frequency of X N corresponding to the output), the maximum probability distribution

p ((), t_{N + 1} {|x)}_{N + 1}, D) = \frac{p ((),, T_{N}, t_{N + 1})}{p ((), T_{N})}

associated with x N + 1 and target values t N + 1 is also subject to the Gaussian distribution 10 . The maximum probability distribution of the target value t N + 1 is

p ((), t_{N + 1} (|, x_{N + 1}, D)) ~ italicGP ((),, {\overset{t}{true}}_{N + 1}, σ_{{\overset{false^}{t}}_{N + 1}}^{2})

…”

Section: The Gaussian Process Model Prediction Methods Based On the Dymentioning

confidence: 99%

“…Because of the disadvantage of over‐reliance on a large number of finite element samples by iterative method, Box et al 3 first proposed the experimental design of the response surface method. Many authors have used the response surface method for finite element model correction and for proposing improved methods 4–10 . The traditional response surface method mainly uses regression analysis to establish the response surface model instead of the complex finite element model; then, it approximates the target value.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Prediction method of bridge static deformation based on dynamic test

Pan

Hong

et al. 2020

Structural Concrete

View full text Add to dashboard Cite

To solve the high cost of bridge load tests, considerable influence on traffic, and damage to bridge structure in the process of problem detection and maintenance of existing bridges, we use the method of improved Gaussian process response surface to explore the prediction of the rigidity and unit weight of existing bridges to produce high‐precision predictions of the existing bridge static test. The Gaussian process response surface method in Bayesian principle is adopted to optimize and correct the model. Because Gaussian process is a non‐parameter model with high sensitivity, reliable accuracy, and small calculation, it avoids repeatedly using finite element model. This makes the efficiency of finite element correction improve significantly. A continuous beam bridge is adopted as a study case. In sample selection, an improved uniform design method based on weighting is designed to improve the learning efficiency and prediction accuracy of the model while using ANSYS to analyze and predict the typical bridge. The analysis of the date of the bridge dynamic test is used to correct the finite element model, obtain the actual parameters of the bridge structure, and predict the measured data of the bridge static load test based on the modified finite element model.

show abstract

p ((), t_{N + 1} {|x)}_{N + 1}, D) = \frac{p ((),, T_{N}, t_{N + 1})}{p ((), T_{N})}

associated with x N + 1 and target values t N + 1 is also subject to the Gaussian distribution 10 . The maximum probability distribution of the target value t N + 1 is