IGA Using Offset-based Overlapping Domain Parameterizations
Authors:
Somayeh Kargaran,
Bert Jüttler,
Thomas Takacs
Abstract:
Isogeometric analysis (IGA) is a numerical method that connects computer-aided design (CAD) with finite element analysis (FEA). In CAD the computational domain is usually represented by B-spline or NURBS patches. Given a NURBS parameterization of the domain, an isogeometric discretization is defined on the domain using the same NURBS basis as for the domain parameterization. Ideally, such an isoge…
▽ More
Isogeometric analysis (IGA) is a numerical method that connects computer-aided design (CAD) with finite element analysis (FEA). In CAD the computational domain is usually represented by B-spline or NURBS patches. Given a NURBS parameterization of the domain, an isogeometric discretization is defined on the domain using the same NURBS basis as for the domain parameterization. Ideally, such an isogeometric discretization allows an exact representation of the underlying CAD model.
CAD models usually represent only the boundary of the object, thus, for planar domains, it is given as a collection of curves. Finding a suitable parameterization of the interior is one of the major issues in IGA, similar to the mesh generation process in FEA. The objective of this parameterization problem is to obtain a set of patches, which exactly represent the boundary of the domain and which are parameterized regularly and without self-intersections. This can be achieved by segmenting the domain into patches which are matching along interfaces, or by covering the domain with overlapping patches. In this paper we follow the second approach.
To construct from a given boundary a planar parameterization suitable for IGA, we propose an offset-based domain parameterization algorithm. Given a boundary curve, we obtain an inner curve by generalized offsetting. Those two curves define a ring-shaped patch, which has a hole that can be covered by a multi-cell domain. Consequently, the domain is represented as a union of two overlapping subdomains which are both regularly parameterized. On such a configuration, one can employ the overlapping multi-patch method introduced in (Kargaran, Jüttler, Kleiss, Mantzaflaris, Takacs; CMAME, 2019), to solve PDEs on the given domain. The performance of the proposed method is reported in several numerical examples, considering different shapes of the domain.
△ Less
Submitted 3 October, 2023;
originally announced October 2023.
Automated machine learning for borehole resistivity measurements
Authors:
M. Shahriari,
D. Pardo,
S. Kargaran,
T. Teijeiro
Abstract:
Deep neural networks (DNNs) offer a real-time solution for the inversion of borehole resistivity measurements to approximate forward and inverse operators. It is possible to use extremely large DNNs to approximate the operators, but it demands a considerable training time. Moreover, evaluating the network after training also requires a significant amount of memory and processing power. In addition…
▽ More
Deep neural networks (DNNs) offer a real-time solution for the inversion of borehole resistivity measurements to approximate forward and inverse operators. It is possible to use extremely large DNNs to approximate the operators, but it demands a considerable training time. Moreover, evaluating the network after training also requires a significant amount of memory and processing power. In addition, we may overfit the model. In this work, we propose a scoring function that accounts for the accuracy and size of the DNNs compared to a reference DNN that provides a good approximation for the operators. Using this scoring function, we use DNN architecture search algorithms to obtain a quasi-optimal DNN smaller than the reference network; hence, it requires less computational effort during training and evaluation. The quasi-optimal DNN delivers comparable accuracy to the original large DNN.
△ Less
Submitted 20 July, 2022;
originally announced July 2022.
