Abstract:Digital twinning offers a capability of effective real-time monitoring and control, which are vital for cost-intensive experimental facilities, particularly the ones where extreme conditions exist. Sparse experimental measurements collected by various diagnostic sensors are usually the only source of information available during the course of a physical experiment. Consequently, in order to enable monitoring and control of the experiment (digital twinning), the ability to perform inverse analysis, facilitating the full field solution reconstruction from the sparse experimental data in real time, is crucial. This paper shows for the first time that it is possible to directly solve inverse problems, such as solution reconstruction, where some or all boundary conditions (BCs) are unknown, by purely using a finite-element (FE) approach, without needing to employ any traditional inverse analysis techniques or any machine learning models, as is normally done in the field. This novel and efficient FE-based inverse analysis framework employs a conventional FE discretisation, splits the loading vector into two parts corresponding to the known and unknown BCs, and then defines a loss function based on that split. In spite of the loading vector split, the loss function preserves the element connectivity. This function is minimised using a gradient-based optimisation. Furthermore, this paper presents a novel modification of this approach, which allows it to generate a range of different solutions satisfying given requirements in a controlled manner. Controlled multiple solution generation in the context of inverse problems and their intrinsic ill-posedness is a novel notion, which has not been explored before. This is done in order to potentially introduce the capability of semi-autonomous system control with intermittent human intervention to the workflow.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is how to reconstruct the complete temperature field in real - time from sparse experimental measurement data in the context of digital twins. Specifically, the author proposes a new framework based on the finite element method (FEM) to directly solve the inverse problem, without the need to use traditional inverse analysis techniques or machine - learning models even when some or all of the boundary conditions are unknown. ### Core problems of the paper 1. **Real - time reconstruction of the temperature field**: - In many engineering applications, especially in experimental facilities involving extreme conditions, sparse experimental measurement data is the only available information source. In order to achieve effective real - time monitoring and control, it must be possible to quickly reconstruct the complete temperature field from these sparse data. - For example, in some cases, the control system may need to know the maximum temperature inside the test piece, and this parameter cannot be directly obtained from the sparse measurement data. 2. **Generate multiple feasible solutions**: - Inverse problems are usually ill - posed, that is, the solutions are not unique. The paper takes advantage of this characteristic and proposes a method to generate multiple feasible solutions that meet specific requirements. This helps to introduce human intervention in semi - autonomous system control and provides multiple options to assist decision - making. ### Overview of the solution - **Inverse analysis framework based on the finite element method**: By dividing the load vector into two parts of known and unknown boundary conditions and defining a loss function to minimize the error, real - time temperature field reconstruction is achieved. - **Multiple - solution generation**: By modifying the loss function, the framework can generate multiple different solutions under controlled conditions to meet specific energy criteria or other requirements. ### Key points of the method - **Loss function design**: The loss function includes a residual term, a measurement term, and a regularization term. Among them, the regularization term is used to prevent physically unreasonable temperature changes and ensure the smoothness of the solutions. - **Optimization algorithm**: Use the nonlinear conjugate gradient method and the Gauss - Newton method for iterative optimization, combined with automatic differentiation techniques to improve computational efficiency. - **GPU acceleration**: Use the PyTorch library to perform matrix operations and automatic differentiation on the GPU to ensure efficient computational performance. ### Application background - **Digital twins**: The application of this framework aims to enhance the real - time monitoring and control capabilities of digital twin systems, especially in complex engineering environments where experimental conditions need to be precisely controlled. ### Summary This paper proposes an innovative finite element method framework that can reconstruct the complete temperature field from sparse experimental data in real - time and generate multiple feasible solutions to assist decision - making. This method not only improves the practicality and flexibility of digital twin systems but also provides new ideas for solving other similar inverse problems.

A novel, finite-element-based framework for sparse data solution reconstruction and multiple choices

A Model-Based Regularized Inverse Method for Ultrasonic B-scan Image Reconstruction

Physics Constrained Learning for Data-driven Inverse Modeling from Sparse Observations

An Algebraic Framework for the Real-Time Solution of Inverse Problems on Embedded Systems

Efficient Inverse Design Optimization through Multi-fidelity Simulations, Machine Learning, and Search Space Reduction Strategies

Data-Driven Sparse Sensor Placement for Reconstruction

Novel Algorithm for Sparse Solutions to Linear Inverse Problems with Multiple Measurements

A novel formulation and sequential solution strategy with time-space adaptive mesh refinement for efficient reconstruction of local boundary heat flux

A Bi-fidelity Ensemble Kalman Method for PDE-constrained Inverse Problems in Computational Mechanics

Solving Inverse Problems with Model Mismatch using Untrained Neural Networks within Model-based Architectures

A Learning-Based Method for Solving Ill-Posed Nonlinear Inverse Problems: A Simulation Study of Lung EIT

Fast numerical techniques for FE simulations in electrical capacitance tomography

A Fast and Scalable Method for A-Optimal Design of Experiments for Infinite-dimensional Bayesian Nonlinear Inverse Problems

Finite element interpolated neural networks for solving forward and inverse problems

Finite Element Method-enhanced Neural Network for Forward and Inverse Problems

Advances In Sparse Solver For Finite Element Analysis

A Novel Objective Function for the Inverse Problem of Simultaneous Identification of Unknown Youngs Modulus and Boundary Conditions with Noisy and Partial Observation

Physics-informed and graph neural networks for enhanced inverse analysis

Effective Solution of Nonlinear Subsurface Flow Inverse Problems in Sparse Bases

Inverse-free distributed neurodynamic optimization algorithms for sparse reconstruction

Finite Operator Learning: Bridging Neural Operators and Numerical Methods for Efficient Parametric Solution and Optimization of PDEs