On generative models as the basis for digital twins

Abstract

A framework is proposed for generative models as a basis for digital twins or mirrors of structures. The proposal is based on the premise that deterministic models cannot account for the uncertainty present in most structural modeling applications. Two different types of generative models are considered here. The first is a physics-based model based on the stochastic finite element (SFE) method, which is widely used when modeling structures that have material and loading uncertainties imposed. Such models can be calibrated according to data from the structure and would be expected to outperform any other model if the modeling accurately captures the true underlying physics of the structure. The potential use of SFE models as digital mirrors is illustrated via application to a linear structure with stochastic material properties. For situations where the physical formulation of such models does not suffice, a data-driven framework is proposed, using machine learning and conditional generative adversarial networks (cGANs). The latter algorithm is used to learn the distribution of the quantity of interest in a structure with material nonlinearities and uncertainties. For the examples considered in this work, the data-driven cGANs model outperforms the physics-based approach. Finally, an example is shown where the two methods are coupled such that a hybrid model approach is demonstrated.

Assessing heterogeneity in spatial data using the HTA index with applications to spatial transcriptomics and imaging

“…Although malicious uses are indeed a concern, generative models could have far reaching positive implications on our lives. Perhaps the most prominent example is the field of healthcare, where generative models have found use in generating digital twins for improved personalized medicine (Bordukova et al 2024;Tsialiamanis et al 2021), medical imaging (Kazerouni et al 2023;Moghadam et al 2023;Pinaya et al 2022) and drug design (Korablyov et al 2024;Chen et al 2023;Hoogeboom et al 2022;Schneuing et al 2022;Corso et al 2022).…”

On robust risk-based active-learning algorithms for enhanced decision support

“…In this context, SHM systems are a necessary component of digital twins. Furthermore, recent work has identified generative models as being a core component of digital twin technology [44]. It follows that, if the context in which SHM system is being applied necessitates a generative model, such as within a digital twin, a discriminative model can be deemed an unsuitable choice of classifier.…”

Digital twinning of self-sensing structures using the statistical finite element method

“…However, even the most advanced data-driven models, which are often based on artificial neural networks, Gaussian process regression, or Bayesian model updating, require a copious amount of real-world training data to make predictions. This challenge has prompted research in the areas of model updating and system identification which have attempted to use both FE and data-driven models to provide better predictions of structural behavior, see, for example, Malekzadeh et al (2015); Pasquier and Smith (2016); Tsialiamanis et al (2021). However, the current modeling approaches seem to be unable to synthesize measurement data with uncertainties and predictions from inherently misspecified FE models in a manner that allows for the generation of continuous predictions as new measurement data become available.…”