.png){kind=icon}
.png){kind=icon}
Return
An NGauss-translation model for spatially variable material properties
.png){kind=icon}
PRE
AI
P
Z
Q
G
Y
DOI：10.1016/j.probengmech.2026.103984.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Most material properties are nonnegative and simulated by non-Gaussian random processes, which are generally generated via the conventional Gauss-translation methodology from a Gaussian process. However, translating a Gaussian random process directly to a non-Gaussian random process distorts the correlation structure. To mitigate this kind of distortion, this study proposed an NGauss-translation model for generating non-Gaussian random processes, where the underlying random process is replaced by the Gamma random process. In contrast to the Gaussian distribution, the Gamma distribution has a higher flexibility in its distribution function. As such, the underlying Gamma distribution function can be determined by adjusting its statistical parameters so that its shape is close to that of the target non-Gaussian distribution function. In this manner, the distortion in correlation from Gamma to the target non-Gaussian process can be reduced to a negligible level. On the other hand, many non-Gaussian distributions are special cases of the Gamma distribution, such as Chi-square distribution and Exponential distribution. For those cases, no distortion occurs in correlation structure. The capabilities of the methodology are also demonstrated by numerical examples. The NGauss-translation model was employed to simulate the soil parameters of the sludge disposal site and was combined with numerical simulation for calculating the foundation bearing capacity. The results demonstrate that the proposed method exhibits practical engineering application value.
Journal
IF:
3.5
Papers:
1.7K
Citations:
4.1K
Organization
Cited Papers