pith. sign in

arxiv: 2512.13919 · v2 · submitted 2025-12-15 · 💻 cs.LG · cs.NA· math.NA

Adaptive digital twins for predictive decision-making: Online Bayesian learning of transition dynamics

Eugenio Varetti , Matteo Torzoni , Marco Tezzele , Andrea Manzoni This is my paper

Pith reviewed 2026-05-16 21:31 UTC · model grok-4.3

classification 💻 cs.LG cs.NAmath.NA
keywords adaptive digital twinsonline Bayesian learningtransition dynamicsdynamic Bayesian networksstructural health monitoringpredictive maintenanceMarkov decision processesconjugate priors
0
0 comments X

The pith

Digital twins adapt their internal state-transition models online by treating probabilities as random variables with conjugate priors inside dynamic Bayesian networks.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that endowing transition probabilities with conjugate priors inside dynamic Bayesian networks permits hierarchical Bayesian updates that learn dynamics directly from streaming observations. These updates feed into parametric Markov decision processes solved by reinforcement learning, producing policies that evolve with the physical asset. The authors demonstrate the approach on structural health monitoring and maintenance planning for a railway bridge. A sympathetic reader would care because the method promises digital twins that remain accurate and decision-relevant without repeated full retraining or heavy offline computation.

Core claim

By representing state transition probabilities as random variables endowed with conjugate priors within dynamic Bayesian networks, the framework performs effortless hierarchical online Bayesian updates that learn transition dynamics from data. These learned dynamics are then used to solve parametric Markov decision processes via reinforcement learning, yielding dynamic policies that update in response to new observations. The resulting adaptive digital twin therefore supports predictive decision-making with greater personalization, robustness, and cost-effectiveness than static counterparts.

What carries the argument

Dynamic Bayesian networks whose state-transition probabilities are random variables equipped with conjugate priors, enabling closed-form hierarchical online Bayesian updates that feed parametric Markov decision processes solved by reinforcement learning.

If this is right

  • The same conjugate-prior construction applies to a broader family of transition distributions than those treated in prior digital-twin literature.
  • Precision updates to the transition model translate directly into updated optimal policies for maintenance and intervention without re-solving the full decision problem from scratch.
  • The bi-directional physical-virtual loop becomes operational because new sensor readings trigger immediate, low-cost Bayesian revisions rather than batch retraining.
  • Cost-effectiveness follows from reduced need for conservative safety margins once the digital twin tracks the actual evolution of the asset.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same online-update structure could be ported to other asset classes whose degradation can be cast as a Markov chain, such as rotating machinery or power-grid components.
  • If the conjugate priors are elicited from expert knowledge or historical fleets, the method may require far fewer site-specific observations than purely data-driven alternatives.
  • A natural next test is whether the learned transition posteriors can be shared across a population of similar bridges to accelerate learning on new assets.

Load-bearing premise

State transition probabilities in real civil-engineering systems admit conjugate priors that keep the online updates both accurate and computationally tractable without large model mismatch.

What would settle it

In the railway-bridge case study, collect new sensor data and check whether the Bayesian-updated transition model produces maintenance policies whose predicted lifetime cost or reliability differs measurably from policies derived from a static model or from ground-truth long-term observations.

Figures

Figures reproduced from arXiv: 2512.13919 by Andrea Manzoni, Eugenio Varetti, Marco Tezzele, Matteo Torzoni.

Figure 1
Figure 1. Figure 1: Graphical abstraction of the end-to-end information flow enabled by the probabilistic [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Dynamic Bayesian network encoding the asset-twin dynamical system, unrolled over [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Dynamic Bayesian network encoding the asset-twin dynamical system: Bayesian net [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Dynamic Bayesian network used to predict the future evolution of digital states and [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Physical asset and its digital twin - The physical space corresponds to the H¨ornefors [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Exemplary multivariate time series of structural vibrations (standardized displacements [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Digital twin online phase - Comparison of cumulative rewards under control policies [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Digital twin online phase - Simulation under finite-horizon model-based reinforcement [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Digital twin online phase - Future predictions starting at [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Prior and posterior distributions of the transition probabilities associated with the [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗
read the original abstract

This work shows how adaptivity can enhance value realization of digital twins in civil engineering. We focus on adapting the state transition models within digital twins represented through probabilistic graphical models. The bi-directional interaction between the physical and virtual domains is modeled using dynamic Bayesian networks. By treating state transition probabilities as random variables endowed with conjugate priors, we enable hierarchical online learning of transition dynamics from a state to another through effortless Bayesian updates. We provide the mathematical framework to account for a larger class of distributions with respect to the current literature on digital twins. To compute dynamic policies with precision updates we solve parametric Markov decision processes through reinforcement learning. The proposed adaptive digital twin framework enjoys enhanced personalization, increased robustness, and improved cost-effectiveness. We assess our approach on a case study involving structural health monitoring and maintenance planning of a railway bridge.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper proposes an adaptive digital twin framework for civil engineering applications such as railway bridge maintenance. It models bi-directional physical-virtual interactions via dynamic Bayesian networks, endows state transition probabilities with conjugate priors (e.g., Dirichlet) to enable hierarchical online Bayesian updates of transition dynamics, and derives dynamic policies by solving parametric Markov decision processes with reinforcement learning. The framework is claimed to deliver enhanced personalization, robustness, and cost-effectiveness, with assessment on a structural health monitoring case study.

Significance. If the conjugate-prior updates and resulting policies can be shown to remain accurate under realistic non-stationary degradation, the approach would provide a principled route to online adaptation in digital twins, potentially improving maintenance decisions. The significance is currently difficult to gauge because the manuscript supplies no derivations, quantitative results, error metrics, or baseline comparisons.

major comments (3)
  1. [Mathematical framework (online Bayesian learning section)] The central modeling choice—endowing transition probabilities with conjugate priors inside the dynamic Bayesian network—receives no justification or sensitivity analysis for civil-engineering degradation processes, which are typically non-stationary and path-dependent. Any posterior bias would propagate directly into the parametric MDP policy updates, undermining the asserted robustness and cost-effectiveness gains.
  2. [Case study (railway bridge SHM section)] The case-study assessment is described only at the level of the abstract; no performance metrics, learning curves, cost comparisons against static digital twins, or validation against held-out degradation data appear in the manuscript. Without these, the claims of enhanced personalization and cost-effectiveness cannot be evaluated.
  3. [Parametric MDP and reinforcement learning section] The procedure for solving the parametric MDP with precision updates from the Bayesian posteriors is outlined at a high level but lacks any statement of the RL algorithm, convergence criteria, or computational scaling with the size of the state space.
minor comments (2)
  1. [Abstract] The abstract states that the framework accounts for 'a larger class of distributions' than current literature but neither names the class nor provides the corresponding conjugate-update equations.
  2. [Dynamic Bayesian network definition] Notation for the hierarchical prior parameters and the resulting posterior transition matrices is introduced without a consolidated table or explicit update rule, making the 'effortless Bayesian updates' claim hard to verify.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. We address each major comment point by point below, indicating the revisions we will make to the manuscript.

read point-by-point responses

  1. Referee: The central modeling choice—endowing transition probabilities with conjugate priors inside the dynamic Bayesian network—receives no justification or sensitivity analysis for civil-engineering degradation processes, which are typically non-stationary and path-dependent. Any posterior bias would propagate directly into the parametric MDP policy updates, undermining the asserted robustness and cost-effectiveness gains.

    Authors: Conjugate priors (e.g., Dirichlet) are selected specifically to permit closed-form hierarchical online Bayesian updates, which are central to enabling real-time adaptation without retraining. The dynamic Bayesian network structure is intended to capture path dependence and non-stationarity through sequential updates. We acknowledge that the current manuscript lacks an explicit justification subsection and sensitivity analysis on prior hyperparameters and their effect on posterior bias. We will add this material, including a discussion of how the hierarchical updates mitigate bias propagation into the parametric MDP policies. revision: yes

  2. Referee: The case-study assessment is described only at the level of the abstract; no performance metrics, learning curves, cost comparisons against static digital twins, or validation against held-out degradation data appear in the manuscript. Without these, the claims of enhanced personalization and cost-effectiveness cannot be evaluated.

    Authors: We agree that the case-study section is currently insufficiently detailed. In the revision we will expand it to report quantitative metrics (prediction error on transition probabilities, cumulative maintenance cost, personalization error), learning curves for the online Bayesian updates, direct cost comparisons against a static digital-twin baseline, and validation results on held-out degradation sequences from the railway-bridge dataset. revision: yes

  3. Referee: The procedure for solving the parametric MDP with precision updates from the Bayesian posteriors is outlined at a high level but lacks any statement of the RL algorithm, convergence criteria, or computational scaling with the size of the state space.

    Authors: We will revise this section to name the RL algorithm (value iteration on the parametric MDP with posterior-mean and posterior-precision updates), state the convergence criterion (policy stability measured by maximum value-function change below a threshold), and provide a complexity analysis (linear in the number of states for each precision update, with discussion of scaling for larger state spaces). revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on standard Bayesian and RL methods

full rationale

The paper's core chain endows transition probabilities with conjugate priors (Dirichlet/Beta) inside dynamic Bayesian networks to enable online updates, then solves the resulting parametric MDP via reinforcement learning. These steps invoke well-established techniques from Bayesian statistics and RL literature rather than defining any quantity in terms of itself or renaming a fitted parameter as a prediction. No self-citation forms a load-bearing uniqueness theorem, no ansatz is smuggled, and no equation reduces by construction to an input defined within the paper. The framework is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on standard properties of conjugate priors for Bayesian updates and the suitability of dynamic Bayesian networks for modeling physical-virtual interactions, without introducing new entities or heavily fitted parameters beyond prior hyperparameters.

axioms (2)
  • standard math State transition probabilities can be treated as random variables with conjugate priors to enable closed-form hierarchical online Bayesian updates.
    This is the mathematical basis for effortless learning described in the abstract.
  • domain assumption Dynamic Bayesian networks adequately capture the bi-directional interaction between physical and virtual domains in digital twins.
    Core modeling choice for representing the adaptive digital twin.

pith-pipeline@v0.9.0 · 5444 in / 1257 out tokens · 38744 ms · 2026-05-16T21:31:29.148008+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

87 extracted references · 87 canonical work pages

  1. [1]

    J. D. Achenbach. Structural health monitoring – What is the prescription?Mechanics Research Communications, 36(2):137–142, 2009.doi:10.1016/j.mechrescom.2008.08.011

    work page doi:10.1016/j.mechrescom.2008.08.011 2009

  2. [2]

    American Society of Civil Engineers, Reston, Virginia, 2018.doi:10.1061/ 9780784415191.ch07

    American Society of Civil Engineers.Adaptive Design and Risk Management, chapter 7, pages 173–226. American Society of Civil Engineers, Reston, Virginia, 2018.doi:10.1061/ 9780784415191.ch07

    work page 2018

  3. [3]

    C. P. Andriotis and K. G. Papakonstantinou. Managing engineering systems with large state and action spaces through deep reinforcement learning.Reliability Engineering & System Safety, 191:106483, 2019.doi:10.1016/j.ress.2019.04.036. 21

    work page doi:10.1016/j.ress.2019.04.036 2019

  4. [4]

    Arcieri, C

    G. Arcieri, C. Hoelzl, O. Schwery, D. Straub, K. Papakonstantinou, and E. Chatzi. POMDP inference and robust solution via deep reinforcement learning: An application to railway optimal maintenance.Machine Learning, 2024.doi:10.1007/s10994-024-06559-2

    work page doi:10.1007/s10994-024-06559-2 2024

  5. [5]

    Arcieri, C

    G. Arcieri, C. Hoelzl, O. Schwery, D. Straub, K. G. Papakonstantinou, and E. Chatzi. Bridging pomdps and bayesian decision making for robust maintenance planning under model uncer- tainty: An application to railway systems.Reliability Engineering & System Safety, page 109496, 2023.doi:10.1016/j.ress.2023.109496

    work page doi:10.1016/j.ress.2023.109496 2023

  6. [6]

    D. A. Arcones, M. Weiser, P.-S. Koutsourelakis, and J. F. Unger. Model Bias Identification for Bayesian Calibration of Stochastic Digital Twins of Bridges.Applied Stochastic Models in Business and Industry, special issue article, 2024.doi:10.1002/asmb.2897

    work page doi:10.1002/asmb.2897 2024

  7. [7]

    A comprehensive assessment of America’s infrastructure

    ASCE. A comprehensive assessment of America’s infrastructure. Technical report, World Headquarters: Reston, Virginia, 2021

    work page 2021

  8. [8]

    O. Avci, O. Abdeljaber, S. Kiranyaz, M. Hussein, M. Gabbouj, and D. Inman. A review of vibration-based damage detection in civil structures: From traditional methods to Ma- chine Learning and Deep Learning applications.Mechanical Systems and Signal Processing, 147:107077, 2021.doi:10.1016/j.ymssp.2020.107077

    work page doi:10.1016/j.ymssp.2020.107077 2021

  9. [9]

    Bauer, B

    P. Bauer, B. Stevens, and W. Hazeleger. A digital twin of Earth for the green transition. Nature Climate Change, 11(2):80–83, 2021.doi:10.1038/s41558-021-00986-y

    work page doi:10.1038/s41558-021-00986-y 2021

  10. [10]

    R. Bellman. A Markovian Decision Process.Journal of Mathematics and Mechanics, 6(5):679– 684, 1957

    work page 1957

  11. [11]

    Bellman.Dynamic Programming

    R. Bellman.Dynamic Programming. Princeton University Press, Princeton, New Jersey, 1957

    work page 1957

  12. [12]

    Bianchi and F

    S. Bianchi and F. Biondini. Life-Cycle Assessment of Deteriorating RC Bridges Using Artificial Neural Networks.Journal of Infrastructure Systems, 28(2), 2022.doi:10.1061/(ASCE)IS. 1943-555X.0000672

    work page doi:10.1061/(asce)is 2022

  13. [13]

    Biondini and D

    F. Biondini and D. M. Frangopol. Life-Cycle Performance of Deteriorating Structural Systems under Uncertainty: Review.Journal of Structural Engineering, 142(9):F4016001, 2016.doi: 10.1061/(ASCE)ST.1943-541X.0001544

    work page doi:10.1061/(asce)st.1943-541x.0001544 2016

  14. [14]

    Biondini and D

    F. Biondini and D. M. Frangopol. Time-variant redundancy and failure times of deteriorating concrete structures considering multiple limit states.Structure and Infrastructure Engineering, 13(1):94–106, 2017.doi:10.1080/15732479.2016.1198403

    work page doi:10.1080/15732479.2016.1198403 2017

  15. [15]

    C. M. Bishop.Pattern Recognition and Machine Learning. Information Science and Statistics. Springer, New York, New York, 2006

    work page 2006

  16. [16]

    J. M. Brownjohn. Structural health monitoring of civil infrastructure.Philosophical Transac- tions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 365(1851):589– 622, 2007.doi:10.1098/rsta.2006.1925

    work page doi:10.1098/rsta.2006.1925 2007

  17. [17]

    Chaudhuri, G

    A. Chaudhuri, G. Pash, D. A. Hormuth, G. Lorenzo, M. Kapteyn, C. Wu, E. A. Lima, T. E. Yankeelov, K. Willcox, et al. Predictive digital twin for optimizing patient-specific radiother- apy regimens under uncertainty in high-grade gliomas.Frontiers in Artificial Intelligence, 6, 2023.doi:10.3389/frai.2023.1222612

    work page doi:10.3389/frai.2023.1222612 2023

  18. [18]

    Cherifi, V

    K. Cherifi, V. Mehrmann, and P. Schulze. Simulations in a Digital Twin of an Electrical Machine. arXiv preprint arXiv:2207.02171, 2022.doi:10.48550/arXiv.2207.02171

    work page doi:10.48550/arxiv.2207.02171 2022

  19. [19]

    Chinesta, A

    F. Chinesta, A. Huerta, G. Rozza, and K. Willcox. Model Reduction Methods. In E. Stein, R. de Borst, and T. J. R. Hughes, editors,Encyclopedia of Computational Mechanics, Second Edition, pages 1–36. John Wiley & Sons, 2017

    work page 2017

  20. [20]

    Corral-Acero, F

    J. Corral-Acero, F. Margara, M. Marciniak, C. Rodero, F. Loncaric, Y. Feng, A. Gilbert, J. F. Fernandes, H. A. Bukhari, A. Wajdan, et al. The ‘Digital Twin’ to enable the vision of precision cardiology.European Heart Journal, 41(48):4556–4564, 2020.doi:10.1093/ eurheartj/ehaa159. 22

    work page 2020

  21. [21]

    F. Darema. Dynamic Data Driven Applications Systems: A New Paradigm for Application Simulations and Measurements. In M. Bubak, G. D. van Albada, P. M. A. Sloot, and J. Don- garra, editors,International Conference on Computational Science, pages 662–669, Berlin, Germany, 2004. Springer.doi:10.1007/978-3-540-24688-6\_86

    work page doi:10.1007/978-3-540-24688-6 2004

  22. [22]

    Eftekhar Azam and S

    S. Eftekhar Azam and S. Mariani. Online damage detection in structural systems via dynamic inverse analysis: A recursive Bayesian approach.Engineering Structures, 159:28–45, 2018. doi:10.1016/j.engstruct.2017.12.031

    work page doi:10.1016/j.engstruct.2017.12.031 2018

  23. [23]

    Fabris, M

    L. Fabris, M. Torzoni, and M. Tezzele. pgmtwin library on github.https://github.com/ pgmtwin/pgmtwin, 2025

    work page 2025

  24. [24]

    Farrar and K

    C. Farrar and K. Worden.Structural Health Monitoring: A Machine Learning Perspective. John Wiley & Sons, Chichester, United Kingdom, 2013.doi:10.1002/9781118443118

    work page doi:10.1002/9781118443118 2013

  25. [25]

    C. R. Farrar and K. Worden. An Introduction to Structural Health Monitoring.Philosoph- ical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 365(1851):303–315, 2007.doi:10.1098/rsta.2006.1928

    work page doi:10.1098/rsta.2006.1928 2007

  26. [26]

    Masset, R

    A. Fernandez-Navamuel, D. Zamora-S´ anchez,´Angel J. Omella, D. Pardo, D. Garcia-Sanchez, and F. Magalh˜ aes. Supervised Deep Learning with Finite Element simulations for dam- age identification in bridges.Engineering Structures, 257:114016, 2022.doi:10.1016/j. engstruct.2022.114016

    work page doi:10.1016/j 2022

  27. [27]

    Ferrari and K

    A. Ferrari and K. Willcox. Digital twins in mechanical and aerospace engineering.Nature Computational Science, 4:178–183, 2024.doi:10.1038/s43588-024-00613-8

    work page doi:10.1038/s43588-024-00613-8 2024

  28. [28]

    N. R. Franco, D. Fraulin, A. Manzoni, and P. Zunino. On the latent dimension of deep autoencoders for reduced order modeling of PDEs parametrized by random fields.Advances in Computational Mathematics, 50:96, 2024.doi:10.1007/s10444-024-10189-6

    work page doi:10.1007/s10444-024-10189-6 2024

  29. [29]

    Garayalde, L

    G. Garayalde, L. Rosafalco, M. Torzoni, and A. Corigliano. Mastering Truss Structure Optimization With Tree Search.Journal of Mechanical Design, 147(10):101702, 2025. doi:10.1115/1.4068300

    work page doi:10.1115/1.4068300 2025

  30. [30]

    Garc´ ıa-Mac´ ıas and F

    E. Garc´ ıa-Mac´ ıas and F. Ubertini. Integrated SHM Systems: Damage Detection Through Unsupervised Learning and Data Fusion. In A. Cury, D. Ribeiro, F. Ubertini, and M. D. Todd, editors,Structural Health Monitoring Based on Data Science Techniques, pages 247–268. Springer International Publishing, Cham, Switzerland, 2022.doi:10.1007/ 978-3-030-81716-9\_12

    work page 2022

  31. [31]

    Gelman, J.B

    A. Gelman, J. B. Carlin, H. S. Stern, D. B. Dunson, A. Vehtari, and D. B. Rubin.Bayesian Data Analysis. CRC Press, Boca Raton, FL, 3rd edition, 2013.doi:10.1201/b16018

    work page doi:10.1201/b16018 2013

  32. [33]

    H.-Y. Guo, Y. Dong, and X.-L. Gu. Two-step translation method for time-dependent relia- bility of structures subject to both continuous deterioration and sudden events.Engineering Structures, 225:111291, 2020.doi:10.1016/j.engstruct.2020.111291

    work page doi:10.1016/j.engstruct.2020.111291 2020

  33. [34]

    Henao-Garcia, M

    S. Henao-Garcia, M. Kapteyn, K. E. Willcox, M. Tezzele, M. Castroviejo-Fernandez, T. Kim, M. Ambrosino, I. Kolmanovsky, H. Basu, P. Jirwankar, and R. Sanfelice. Digital-Twin- Enabled Multi-Spacecraft On-Orbit Operations. InAIAA SCITECH 2025 Forum. American Institute of Aeronautics and Astronautics, Inc., 2025.doi:10.2514/6.2025-1432

    work page doi:10.2514/6.2025-1432 2025

  34. [35]

    Hern´ andez-Gonz´ alez and E

    I. Hern´ andez-Gonz´ alez and E. Garc´ ıa-Mac´ ıas. Towards a comprehensive damage identification of structures through populations of competing models.Engineering with Computers, pages 85–91, 2024.doi:10.1007/s00366-024-01972-6

    work page doi:10.1007/s00366-024-01972-6 2024

  35. [36]

    R. A. Howard.Dynamic programming and Markov processes. John Wiley, Oxford, England, 1960. 23

    work page 1960

  36. [37]

    T. J. Hughes.The Finite Element Method: Linear Static and Dynamic Finite Element Anal- ysis. Dover Civil and Mechanical Engineering. Dover Publications, 2000

    work page 2000

  37. [38]

    Jans-Singh, K

    M. Jans-Singh, K. Leeming, R. Choudhary, and M. Girolami. Digital twin of an urban- integrated hydroponic farm.Data-Centric Engineering, 1:20, 2020.doi:10.1017/dce.2020. 21

    work page doi:10.1017/dce.2020 2020

  38. [39]

    Jones, T

    C. Jones, T. Lethco, J. Curley, and D. Zackin. A preventable crisis: The economic and human costs of a hudson river rail tunnel shutdown. Technical report, Regional Plan Association, 2019

    work page 2019

  39. [40]

    M. G. Kapteyn, D. J. Knezevic, D. Huynh, M. Tran, and K. E. Willcox. Data-driven physics- based digital twins via a library of component-based reduced-order models.International Journal for Numerical Methods in Engineering, 123(13):2986–3003, 2020.doi:10.1002/nme. 6423

    work page doi:10.1002/nme 2020

  40. [41]

    M. G. Kapteyn, J. V. Pretorius, and K. E. Willcox. A probabilistic graphical model foundation for enabling predictive digital twins at scale.Nature Computational Science, 1(5):337–347, 2021.doi:10.1038/s43588-021-00069-0

    work page doi:10.1038/s43588-021-00069-0 2021

  41. [42]

    P. M. Karve, Y. Guo, B. Kapusuzoglu, S. Mahadevan, and M. A. Haile. Digital twin approach for damage-tolerant mission planning under uncertainty.Engineering Fracture Mechanics, 225:106766, 2020.doi:10.1016/j.engfracmech.2019.106766

    work page doi:10.1016/j.engfracmech.2019.106766 2020

  42. [43]

    Koller and N

    D. Koller and N. Friedman.Probabilistic Graphical Models: Principles and Techniques. MIT Press, Cambridge, Massachusetts, 2009

    work page 2009

  43. [44]

    Kritzinger, M

    W. Kritzinger, M. Karner, G. Traar, J. Henjes, and W. Sihn. Digital Twin in manufacturing: A categorical literature review and classification.IFAC-PapersOnLine, 51(11):1016–1022,

    work page

  44. [45]

    16th IFAC Symposium on Information Control Problems in Manufacturing INCOM 2018.doi:10.1016/j.ifacol.2018.08.474

    work page doi:10.1016/j.ifacol.2018.08.474 2018

  45. [46]

    C. Li, S. Mahadevan, Y. Ling, S. Choze, and L. Wang. Dynamic Bayesian Network for Aircraft Wing Health Monitoring Digital Twin.AIAA Journal, 55(3):930–941, 2017.doi: 10.2514/1.J055201

    work page doi:10.2514/1.j055201 2017

  46. [47]

    D. V. Lindley.Understanding Uncertainty. John Wiley & Sons, 2013

    work page 2013

  47. [48]

    L¨ ohner, F

    R. L¨ ohner, F. Airaudo, H. Antil, R. W¨ uchner, F. Meister, and S. Warnakulasuriya. High- fidelity digital twins: Detecting and localizing weaknesses in structures.International Journal for Numerical Methods in Engineering, 125(21):e7568, 2024.doi:10.1002/nme.7568

    work page doi:10.1002/nme.7568 2024

  48. [49]

    Mariniello, T

    G. Mariniello, T. Pastore, C. Menna, P. Festa, and D. Asprone. Structural damage detection and localization using decision tree ensemble and vibration data.Computer-Aided Civil and Infrastructure Engineering, 36(9):1129–1149, 2021.doi:10.1111/mice.12633

    work page doi:10.1111/mice.12633 2021

  49. [50]

    PhD thesis, Politecnico di Milano, 2024

    Matteo Torzoni.Model-based and data-driven methodologies toward predictive digital twins of structures. PhD thesis, Politecnico di Milano, 2024

    work page 2024

  50. [51]

    McClellan, J

    A. McClellan, J. Lorenzetti, M. Pavone, and C. Farhat. A Physics-Based Digital Twin for Model Predictive Control of Autonomous Unmanned Aerial Vehicle Landing.Philosophi- cal Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 380(2229):20210204, 2022.doi:10.1098/rsta.2021.020417

    work page doi:10.1098/rsta.2021.020417 2022

  51. [52]

    G. Morcous. Performance Prediction of Bridge Deck Systems Using Markov Chains.Jour- nal of Performance of Constructed Facilities, 20(2):146–155, 2006.doi:10.1061/(ASCE) 0887-3828(2006)20:2(146)

    work page doi:10.1061/(asce 2006

  52. [53]

    J. Morgan. Risk metrics – technical document. Technical report, JP Morgan and Reuters, New York, 1996

    work page 1996

  53. [54]

    B. Moya, I. Alfaro, D. Gonzalez, F. Chinesta, and E. Cueto. Physically sound, self-learning digital twins for sloshing fluids.PLoS One, 15(6):e0234569, 2020.doi:10.1371/journal. pone.0234569. 24

    work page doi:10.1371/journal 2020

  54. [55]

    The Na- tional Academies Press, Washington, DC, 2024.doi:10.17226/26894

    National Academy of Engineering and National Academies of Sciences, Engineering, and Medicine.Foundational Research Gaps and Future Directions for Digital Twins. The Na- tional Academies Press, Washington, DC, 2024.doi:10.17226/26894

    work page doi:10.17226/26894 2024

  55. [56]

    F. Negri. redbKIT, version 2.2, 2016.http://redbkit.github.io/redbKIT

    work page 2016

  56. [57]

    Pandey and M

    A. Pandey and M. Biswas. Damage Detection in Structures Using Changes in Flexibility. Journal of Sound and Vibration, 169(1):3–17, 1994.doi:10.1006/jsvi.1994.1002

    work page doi:10.1006/jsvi.1994.1002 1994

  57. [58]

    Papakonstantinou, C

    K. Papakonstantinou, C. Andriotis, and M. Shinozuka. Point-based POMDP solvers for life- cycle cost minimization of deteriorating structures. In J. Bakker, D. M. Frangopol, and K. van Breugel, editors,Life-Cycle of Engineering Systems: Emphasis on Sustainable Civil Infrastructure, pages 427–434. CRC Press, London, United Kingdom, 2016

    work page 2016

  58. [59]

    K. G. Papakonstantinou and M. Shinozuka. Planning structural inspection and maintenance policies via dynamic programming and Markov processes. Part II: POMDP implementation. Reliability Engineering & System Safety, 130:214–224, 2014.doi:10.1016/j.ress.2014.04. 006

    work page doi:10.1016/j.ress.2014.04 2014

  59. [60]

    J. Pearl. Reverend Bayes on Inference Engines: A Distributed Hierarchical Approach.Prob- abilistic and Causal Inference: The Works of Judea Pearl, pages 129–138, 1982

    work page 1982

  60. [61]

    A. Phua, C. Davies, and G. Delaney. A digital twin hierarchy for metal additive manufacturing. Computers in Industry, 140:103667, 2022.doi:10.1016/j.compind.2022.103667

    work page doi:10.1016/j.compind.2022.103667 2022

  61. [62]

    Quarteroni, A

    A. Quarteroni, A. Manzoni, and F. Negri.Reduced Basis Methods for Partial Differential Equations: An Introduction, volume 92. Springer, Cham, Switzerland, 2015.doi:10.1007/ 978-3-319-15431-2

    work page 2015

  62. [63]

    R. T. Rockafellar and S. Uryasev. Optimization of conditional value-at-risk.Journal of Risk, 2:21–42, 2000.doi:10.21314/JOR.2000.038

    work page doi:10.21314/jor.2000.038 2000

  63. [64]

    Rosafalco, M

    L. Rosafalco, M. Torzoni, A. Manzoni, S. Mariani, and A. Corigliano. Online structural health monitoring by model order reduction and deep learning algorithms.Computers & Structures, 255:106604, 2021.doi:10.1016/j.compstruc.2021.106604

    work page doi:10.1016/j.compstruc.2021.106604 2021

  64. [65]

    S. Ross, J. Pineau, B. Chaib-draa, and P. Kreitmann. A Bayesian Approach for Learning and Planning in Partially Observable Markov Decision Processes.Journal of Machine Learning Research, 12(48):1729–1770, 2011

    work page 2011

  65. [66]

    S. J. Russell.Artificial Intelligence: A Modern Approach. Pearson Education, London, United Kingdom, 2020

    work page 2020

  66. [67]

    Sanchez-Silva, G.-A

    M. Sanchez-Silva, G.-A. Klutke, and D. V. Rosowsky. Life-cycle performance of structures subject to multiple deterioration mechanisms.Structural Safety, 33(3):206–217, 2011.doi: 10.1016/j.strusafe.2011.03.003

    work page doi:10.1016/j.strusafe.2011.03.003 2011

  67. [68]

    Seventekidis, D

    P. Seventekidis, D. Giagopoulos, A. Arailopoulos, and O. Markogiannaki. Structural Health Monitoring using deep learning with optimal finite element model generated data.Mechanical Systems and Signal Processing, 145:106972, 2020.doi:10.1016/j.ymssp.2020.106972

    work page doi:10.1016/j.ymssp.2020.106972 2020

  68. [69]

    Sisson, P

    W. Sisson, P. Karve, and S. Mahadevan. Digital Twin Approach for Component Health- Informed Rotorcraft Flight Parameter Optimization.AIAA Journal, 60(3):1923–1936, 2022. doi:10.2514/1.J060770

    work page doi:10.2514/1.j060770 1923

  69. [70]

    R. S. Sutton and A. G. Barto.Reinforcement Learning: An Introduction. MIT Press, Cam- bridge, Massachusetts, 2018

    work page 2018

  70. [71]

    Teughels, J

    A. Teughels, J. Maeck, and G. De Roeck. Damage assessment by FE model updating us- ing damage functions.Computers & Structures, 80(25):1869–1879, 2002.doi:10.1016/ S0045-7949(02)00217-1

    work page 2002

  71. [72]

    Tezzele, S

    M. Tezzele, S. Carr, U. Topcu, and K. E. Willcox. Adaptive planning for risk-aware predictive digital twins. arXiv preprint arXiv:2407.20490v2, 2024.doi:10.48550/arXiv.2407.20490. 25

    work page doi:10.48550/arxiv.2407.20490 2024

  72. [73]

    Thelen, X

    A. Thelen, X. Zhang, O. Fink, Y. Lu, S. Ghosh, B. D. Youn, M. D. Todd, S. Mahadevan, C. Hu, and Z. Hu. A comprehensive review of digital twin – part 1: modeling and twinning enabling technologies.Structural and Multidisciplinary Optimization, 65:354, 2022.doi: 10.1007/s00158-022-03425-4

    work page doi:10.1007/s00158-022-03425-4 2022

  73. [74]

    Thoft-Christensen

    P. Thoft-Christensen. Life-cycle cost-benefit (LCCB) analysis of bridges from a user and social point of view.Structures and Infrastructure Engineering, 5:49–57, 2009.doi:10.1080/ 15732470701322818

    work page 2009

  74. [75]

    P. D. Thompson and M. B. Johnson. Markovian bridge deterioration: developing models from historical data.Structure and Infrastructure Engineering, 1(1):85–91, 2005.doi:10.1080/ 15732470412331289332

    work page 2005

  75. [76]

    G. G. Tong, C. A. S. Long, and D. E. Schiavazzi. InVAErt networks: A data-driven framework for model synthesis and identifiability analysis.Computer Methods in Applied Mechanics and Engineering, 423:116846, 2024.doi:10.1016/j.cma.2024.116846

    work page doi:10.1016/j.cma.2024.116846 2024

  76. [77]

    Torzoni, D

    M. Torzoni, D. Maisto, A. Manzoni, F. Donnarumma, G. Pezzulo, and A. Corigliano. Active Digital Twins via Active Inference. arXiv preprint arXiv:2506.14453, 2025.doi:10.48550/ arXiv.2506.14453

    work page arXiv 2025

  77. [78]

    Torzoni, A

    M. Torzoni, A. Manzoni, and S. Mariani. Structural health monitoring of civil structures: A diagnostic framework powered by deep metric learning.Computers & Structures, 271:106858, 2022.doi:10.1016/j.compstruc.2022.106858

    work page doi:10.1016/j.compstruc.2022.106858 2022

  78. [79]

    Torzoni, A

    M. Torzoni, A. Manzoni, and S. Mariani. A Deep Neural Network, Multi-fidelity Surrogate Model Approach for Bayesian Model Updating in SHM. In P. Rizzo and A. Milazzo, editors, European Workshop on Structural Health Monitoring, pages 1076–1086. Springer International Publishing, Cham, Switzerland, 2023.doi:10.1007/978-3-031-07258-1\_108

    work page doi:10.1007/978-3-031-07258-1 2023

  79. [80]

    Mechanical Systems and Signal Processing197(2023) https://doi.org/ 10.1016/j.ymssp.2023.110376

    M. Torzoni, A. Manzoni, and S. Mariani. A multi-fidelity surrogate model for structural health monitoring exploiting model order reduction and artificial neural networks.Mechanical Systems and Signal Processing, 197:110376, 2023.doi:10.1016/j.ymssp.2023.110376

    work page doi:10.1016/j.ymssp.2023.110376 2023

  80. [81]

    Torzoni, A

    M. Torzoni, A. Manzoni, and S. Mariani. Enhancing Bayesian model updating in structural health monitoring via learnable mapping.Engineering with Computers, December, 2025. doi:10.1007/s00366-025-02224-x

    work page doi:10.1007/s00366-025-02224-x 2025

Showing first 80 references.