國立虎尾科技大學 |

Beyond Finite Element : = Physics Informed Neural Network for Stress Prediction.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Beyond Finite Element :/
Reminder of title:	Physics Informed Neural Network for Stress Prediction.
Author:	Bolandi, Hamed.
Description:	1 online resource (98 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-04, Section: B.
Contained By:	Dissertations Abstracts International85-04B.
Subject:	Computer engineering. -
Online resource:	click for full text (PQDT)
ISBN:	9798380604048

Beyond Finite Element : = Physics Informed Neural Network for Stress Prediction.
Bolandi, Hamed.

Beyond Finite Element :Physics Informed Neural Network for Stress Prediction. - 1 online resource (98 pages)

Source: Dissertations Abstracts International, Volume: 85-04, Section: B.

Thesis (Ph.D.)--Michigan State University, 2023.

Includes bibliographical references

This Multidisciplinary research proposes deep neural networks to bypass the Finite Element Analysis (FEA) and predict high-resolution stress distributions on loaded steel plates with variable loading, geometries, and boundary conditions. FEA for structures has been broadly used to conduct stress analysis of various civil and mechanical engineering structures. Conventional methods, such as FEA, provide high-fidelity solutions but require solving large linear systems that can be computationally intensive. The existing workflow for FEM applications includes: (i) modeling the geometry and its components, (ii) specifying material properties, boundary conditions, and loading, (iii) Applying mesh strategy, and (iv) stress analysis which may be time-consuming based on the complexity of the model. Instead, Deep learning (DL) techniques can generate solutions significantly faster than conventional run-time analysis. This can prove extremely valuable in real-time structural assessment applications. In this work, The Convolutional Neural network (CNN) was designed and trained to use the geometry, boundary conditions, and static load as input to predict the stress contours in intact steel plates. Furthermore, we predict high-resolution stress distributions on damaged steel plates using CNNs augmented with custom loss functions that use physics rules to bypass the need for Finite Element Analysis. We embedded physics constraints into the loss function to enforce the model training, precisely capturing stress concentrations around the tips of various structural damage configurations. The proposed technique's performance was compared to Finite-Element simulations using partial differential equation (PDE) solvers. Neuro-DynaStress is also proposed to predict the entire sequence of stress distribution based on Finite Element simulations using a partial differential equation (PDE) solver. More specifically, CNN, along with the multi-head attention transformer and feature alignment, is used to extract features and capture the data's temporal dependence. The model was designed and trained to use the geometry, boundary conditions, and sequence of loads as input and predict the sequences of high-resolution von Mises stress contours. Moreover, to increase the accuracy of dynamic stress prediction, we propose a Physics Informed Neural Network (PINN). The PINN-Stress model can predict the entire sequence of stress distribution based on finite element simulations using a PDE solver. In order to force our model to learn the physical constraints, we minimize the violation of the equation of motion and also minimize the boundary condition violation to fully enforce the underlying PDE. The PINN-Stress model can predict the sequence of normal and shear stress distribution in almost real-time and can generalize better than the model without PINN. Our model is also able to predict von Mises stress using the von Mises equation.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798380604048Subjects--Topical Terms:

569006
Computer engineering.
Subjects--Index Terms:

Deep learningIndex Terms--Genre/Form:

554714
Electronic books.

Beyond Finite Element : = Physics Informed Neural Network for Stress Prediction.
LDR:04406ntm a22004097 4500 001 1152597
005 20241209114605.5
006 m o d
007 cr mn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798380604048
035 $a (MiAaPQ)AAI30575306
035 $a AAI30575306
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Bolandi, Hamed. $3 1479638
245 1 0 $a Beyond Finite Element : $b Physics Informed Neural Network for Stress Prediction.
264 0 $c 2023
300 $a 1 online resource (98 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-04, Section: B.
500 $a Advisor: Lajnef, Nizar;Boddeti, Vishnu.
502 $a Thesis (Ph.D.)--Michigan State University, 2023.
504 $a Includes bibliographical references
520 $a This Multidisciplinary research proposes deep neural networks to bypass the Finite Element Analysis (FEA) and predict high-resolution stress distributions on loaded steel plates with variable loading, geometries, and boundary conditions. FEA for structures has been broadly used to conduct stress analysis of various civil and mechanical engineering structures. Conventional methods, such as FEA, provide high-fidelity solutions but require solving large linear systems that can be computationally intensive. The existing workflow for FEM applications includes: (i) modeling the geometry and its components, (ii) specifying material properties, boundary conditions, and loading, (iii) Applying mesh strategy, and (iv) stress analysis which may be time-consuming based on the complexity of the model. Instead, Deep learning (DL) techniques can generate solutions significantly faster than conventional run-time analysis. This can prove extremely valuable in real-time structural assessment applications. In this work, The Convolutional Neural network (CNN) was designed and trained to use the geometry, boundary conditions, and static load as input to predict the stress contours in intact steel plates. Furthermore, we predict high-resolution stress distributions on damaged steel plates using CNNs augmented with custom loss functions that use physics rules to bypass the need for Finite Element Analysis. We embedded physics constraints into the loss function to enforce the model training, precisely capturing stress concentrations around the tips of various structural damage configurations. The proposed technique's performance was compared to Finite-Element simulations using partial differential equation (PDE) solvers. Neuro-DynaStress is also proposed to predict the entire sequence of stress distribution based on Finite Element simulations using a partial differential equation (PDE) solver. More specifically, CNN, along with the multi-head attention transformer and feature alignment, is used to extract features and capture the data's temporal dependence. The model was designed and trained to use the geometry, boundary conditions, and sequence of loads as input and predict the sequences of high-resolution von Mises stress contours. Moreover, to increase the accuracy of dynamic stress prediction, we propose a Physics Informed Neural Network (PINN). The PINN-Stress model can predict the entire sequence of stress distribution based on finite element simulations using a PDE solver. In order to force our model to learn the physical constraints, we minimize the violation of the equation of motion and also minimize the boundary condition violation to fully enforce the underlying PDE. The PINN-Stress model can predict the sequence of normal and shear stress distribution in almost real-time and can generalize better than the model without PINN. Our model is also able to predict von Mises stress using the von Mises equation.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Computer engineering. $3 569006
650 4 $a Computer science. $3 573171
650 4 $a Civil engineering. $3 561339
653 $a Deep learning
653 $a Finite Element Analysis
653 $a Machine learning
653 $a Partial differential equation
653 $a Physics Informed Neural Network
653 $a Stress prediction
655 7 $a Electronic books. $2 local $3 554714
690 $a 0543
690 $a 0984
690 $a 0464
710 2 $a Michigan State University. $b Civil Engineering - Doctor of Philosophy. $3 1465368
710 2 $a ProQuest Information and Learning Co. $3 1178819
773 0 $t Dissertations Abstracts International $g 85-04B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30575306 $z click for full text (PQDT)