pith. sign in

arxiv: 2605.17608 · v1 · pith:TX3AYJBJnew · submitted 2026-05-17 · 💻 cs.CE · cs.AI

Bayesian-Monte Carlo Schedule Updating for Construction Digital Twins: A Probabilistic Framework for Dynamic Project Forecasting

Atena Khoshkonesh , Mohsen Mohammadagha , Vinayak Kaushal , Navid Ebrahimi This is my paper

Pith reviewed 2026-05-19 22:10 UTC · model grok-4.3

classification 💻 cs.CE cs.AI
keywords construction schedulingBayesian inferenceMonte Carlo simulationdigital twinsprobabilistic forecastingproject managementuncertainty modelingschedule updating
0
0 comments X

The pith

A Bayesian-Monte Carlo framework updates construction schedules by recursively incorporating new observations to produce adaptive probabilistic forecasts.

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

Construction projects face frequent delays from variable productivity, materials, weather, and coordination, yet standard Critical Path Method scheduling treats activity durations as fixed values and therefore under-represents uncertainty. The paper proposes a unified computational framework that first models each activity duration as a lognormal distribution, then applies Bayesian inference to revise those distributions whenever fresh data arrive from BIM reports, drones, IoT sensors, or productivity logs. Monte Carlo simulation propagates the updated distributions through the project network to yield probabilistic completion-time forecasts, delay-risk estimates, and dynamic criticality rankings. Experiments on PSPLIB benchmark networks show that the resulting forecasts are more accurate and better calibrated than those from deterministic CPM or static probabilistic schedules. A sympathetic reader would care because reliable uncertainty quantification directly supports earlier corrective actions and more realistic contingency planning.

Core claim

The Bayesian-Monte Carlo probabilistic schedule updating framework models activity durations with lognormal distributions, performs Bayesian recursive updating as new observations become available, and employs Monte Carlo simulation to propagate the resulting uncertainty through project networks, thereby generating dynamic probabilistic forecasts of completion time, delay risk, and activity criticality that outperform both deterministic CPM and static probabilistic methods on benchmark networks.

What carries the argument

Bayesian recursive updating of lognormal activity-duration distributions followed by Monte Carlo propagation of uncertainty across the project network.

If this is right

  • Probabilistic completion-time forecasts that narrow or widen automatically as new data arrive.
  • Quantitative delay-risk estimates and activity criticality rankings that change with each update cycle.
  • Direct ingestion of heterogeneous data streams (BIM reports, drone imagery, IoT telemetry, productivity logs) into the schedule model.
  • Improved forecasting accuracy relative to both fixed-duration CPM and non-updating probabilistic schedules.

Where Pith is reading between the lines

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

  • The same updating loop could be applied to other domains that maintain network schedules under streaming observations, such as software release planning or manufacturing job-shop scheduling.
  • Digital-twin platforms could use the framework to trigger automated alerts when the probability of missing a milestone exceeds a chosen threshold.
  • Sensitivity analyses could test how robust the forecasts remain when the lognormal assumption is replaced by other common duration distributions.

Load-bearing premise

Activity durations are adequately captured by lognormal distributions and that continuous streams of reliable new observations from BIM, drones, IoT, and logs will be available for the Bayesian updates.

What would settle it

Run the framework on a live construction project or an additional PSPLIB network for which the actual completion date and intermediate milestones are known, then check whether the predicted probability distribution for completion time assigns high probability to the observed outcome.

Figures

Figures reproduced from arXiv: 2605.17608 by Atena Khoshkonesh, Mohsen Mohammadagha, Navid Ebrahimi, Vinayak Kaushal.

Figure 1
Figure 1. Figure 1: Evolution of Scheduling Methods Toward Dynamic Probabilistic Forecasting [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Bayesian–Monte Carlo Framework for Dynamic Probabilistic Schedule Forecasting within Construction Digital Twins [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Effect of Bayesian Updating Strategies on Project Completion Time Probability DistributionsDistribution [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
read the original abstract

Construction projects frequently experience schedule delays and forecasting uncertainty due to variability in labor productivity, material availability, weather conditions, and project coordination. Conventional deterministic scheduling methods such as the Critical Path Method (CPM) assume fixed activity durations and therefore cannot adequately represent dynamic project uncertainty. This study presents a Bayesian-Monte Carlo probabilistic schedule updating framework for construction digital twin environments. The proposed methodology integrates stochastic activity-duration modeling, Bayesian recursive updating, Monte Carlo simulation, and uncertainty propagation within a unified computational framework for adaptive schedule forecasting. Activity durations are modeled using lognormal probability distributions and continuously updated through Bayesian inference as new project observations become available. Monte Carlo simulation is then used to propagate updated uncertainty throughout project networks and generate probabilistic completion-time forecasts, delay-risk estimates, and activity criticality measures. Simulation experiments using PSPLIB benchmark project networks demonstrate that the proposed framework improves forecasting accuracy and uncertainty representation compared with deterministic CPM and static probabilistic scheduling approaches. The framework further supports adaptive project forecasting through integration of BIM reports, drone observations, IoT telemetry, productivity logs, and site monitoring data.

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

2 major / 2 minor

Summary. The manuscript presents a Bayesian-Monte Carlo probabilistic schedule updating framework for construction digital twins. Activity durations are modeled as lognormal distributions and recursively updated via Bayesian inference using streams of observations from BIM, drones, IoT, and logs. Monte Carlo simulation then propagates the updated uncertainties through project networks to produce probabilistic completion-time forecasts, delay risks, and criticality measures. The central claim is that experiments on PSPLIB benchmark networks demonstrate improved forecasting accuracy and uncertainty representation relative to deterministic CPM and static probabilistic schedulers.

Significance. If the empirical results can be substantiated with quantitative metrics and proper controls, the framework offers a coherent way to integrate real-time data into adaptive project forecasting. The combination of Bayesian updating with Monte Carlo propagation in a digital-twin setting addresses a practical need in construction management and could support more reliable risk assessment when continuous observation streams are available.

major comments (2)
  1. [Abstract and simulation-experiments section] Abstract and simulation-experiments section: the claim that PSPLIB experiments 'demonstrate that the proposed framework improves forecasting accuracy and uncertainty representation' is unsupported because no quantitative metrics (MAE, CRPS, interval calibration, or similar), error bars, baseline comparisons, or statistical tests are reported. This directly undermines the central empirical validation.
  2. [Simulation-experiments section] Simulation-experiments section: the protocol for generating synthetic observations used in the Bayesian update step, the number of Monte Carlo replications, the exact static probabilistic baseline (with identical lognormal parameters), and any significance testing are not described. Without these elements the reported accuracy gains cannot be reproduced or assessed for robustness.
minor comments (2)
  1. [Methodology] The initialization of the lognormal distribution parameters is mentioned but not given an explicit functional form or default values; adding this would improve reproducibility.
  2. [Bayesian updating subsection] Notation for the Bayesian update equations could be clarified by explicitly distinguishing prior, likelihood, and posterior parameters in a single display equation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. The feedback correctly identifies areas where the empirical validation and experimental details require strengthening to better support the central claims. We will revise the manuscript to incorporate quantitative metrics, detailed protocols, and reproducibility information as outlined below.

read point-by-point responses

  1. Referee: [Abstract and simulation-experiments section] Abstract and simulation-experiments section: the claim that PSPLIB experiments 'demonstrate that the proposed framework improves forecasting accuracy and uncertainty representation' is unsupported because no quantitative metrics (MAE, CRPS, interval calibration, or similar), error bars, baseline comparisons, or statistical tests are reported. This directly undermines the central empirical validation.

    Authors: We agree that the current manuscript does not report specific quantitative metrics such as MAE, CRPS, interval calibration, error bars, or statistical tests to substantiate the claimed improvements. In the revised version, we will add these elements, including direct numerical comparisons against deterministic CPM and the static probabilistic baseline, along with appropriate statistical significance testing to rigorously demonstrate gains in forecasting accuracy and uncertainty representation. revision: yes

  2. Referee: [Simulation-experiments section] Simulation-experiments section: the protocol for generating synthetic observations used in the Bayesian update step, the number of Monte Carlo replications, the exact static probabilistic baseline (with identical lognormal parameters), and any significance testing are not described. Without these elements the reported accuracy gains cannot be reproduced or assessed for robustness.

    Authors: We acknowledge that the experimental protocol details are currently insufficient for full reproducibility. The revised manuscript will explicitly describe the procedure for generating synthetic observations from the simulated data streams, the number of Monte Carlo replications employed, the precise parameterization of the static probabilistic baseline using identical lognormal distributions, and the statistical tests applied to evaluate the significance of observed accuracy improvements. revision: yes

Circularity Check

0 steps flagged

No circularity: framework uses external observations for Bayesian updates and independent PSPLIB simulations

full rationale

The derivation chain begins with lognormal modeling of activity durations, proceeds to Bayesian recursive updating driven by incoming external data streams (BIM, drones, IoT, logs), then applies Monte Carlo simulation to propagate uncertainty and produce forecasts. These forecasts are compared against deterministic CPM and static probabilistic baselines on PSPLIB networks. No equation or step reduces a claimed prediction to a quantity defined solely by parameters fitted inside the same model; the updates explicitly require new observations outside the initial model. No self-citations are invoked as load-bearing uniqueness theorems, and the simulation experiments are presented as external validation rather than tautological renaming or self-definition. The framework therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard probabilistic modeling choices and the availability of external data streams; no new entities are postulated and the only free parameters are the initial distribution parameters that are subsequently updated.

free parameters (1)
  • Initial lognormal distribution parameters
    Activity durations are modeled as lognormal; initial shape and scale parameters must be chosen or fitted from historical data before Bayesian updating begins.
axioms (2)
  • domain assumption Activity durations follow lognormal distributions
    Stated directly as the modeling choice for stochastic activity-duration modeling in the abstract.
  • domain assumption Continuous reliable observations are available for updating
    The framework assumes ongoing input from BIM reports, drone observations, IoT telemetry, and productivity logs.

pith-pipeline@v0.9.0 · 5738 in / 1396 out tokens · 50075 ms · 2026-05-19T22:10:27.479538+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

50 extracted references · 50 canonical work pages

  1. [1]

    Love, D.J

    P.E.D. Love, D.J. Edwards, Z. Irani, Moving beyond optimism bias and strategic misrepresentation: An explanation for social infrastructure project cost overruns, IEEE Trans. Eng. Manage. 59 (2012) 560– 571

    work page 2012

  2. [2]

    Flyvbjerg, What you should know about megaprojects and why: An overview, Proj

    M. Flyvbjerg, What you should know about megaprojects and why: An overview, Proj. Manage. J. 45 (2014) 6–19

    work page 2014

  3. [3]

    Chitkara, Construction Project Management: Planning, Scheduling and Controlling, McGraw- Hill, 2014

    K.K. Chitkara, Construction Project Management: Planning, Scheduling and Controlling, McGraw- Hill, 2014

    work page 2014

  4. [4]

    Kerzner, Project Management: A Systems Approach to Planning, Scheduling, and Controlling, Wiley, 2017

    H. Kerzner, Project Management: A Systems Approach to Planning, Scheduling, and Controlling, Wiley, 2017

    work page 2017

  5. [5]

    Moder, C.R

    J.J. Moder, C.R. Phillips, E.W. Davis, Project Management with CPM, PERT and Precedence Diagramming, Van Nostrand Reinhold, 1983

    work page 1983

  6. [6]

    Pinto, O.P

    J.K. Pinto, O.P. Kharbanda, How to fail in project management (without really trying), Bus. Horiz. 38 (1995) 45–53

    work page 1995

  7. [7]

    Kolisch, S

    R. Kolisch, S. Hartmann, Experimental investigation of heuristics for resource-constrained project scheduling: An update, Eur. J. Oper. Res. 174 (2006) 23–37

    work page 2006

  8. [8]

    Hajdu, Effects of the application of activity calendars on the distribution of project duration in PERT networks, Autom

    M. Hajdu, Effects of the application of activity calendars on the distribution of project duration in PERT networks, Autom. Constr. 42 (2014) 65–73

    work page 2014

  9. [9]

    Schuyler, Risk and Decision Analysis in Projects, Project Management Institute, 2001

    R. Schuyler, Risk and Decision Analysis in Projects, Project Management Institute, 2001

    work page 2001

  10. [10]

    V ose, Risk Analysis: A Quantitative Guide, Wiley, 2008

    D. V ose, Risk Analysis: A Quantitative Guide, Wiley, 2008

    work page 2008

  11. [11]

    Acebes, J

    A. Acebes, J. Pajares, J.M. Galán, A. López-Paredes, A new approach for project control under uncertainty: The earned schedule, Int. J. Proj. Manage. 32 (2014) 123–135

    work page 2014

  12. [12]

    Nasir, M

    H. Nasir, M. McCabe, G. Hartono, Evaluating risk in construction-schedule model, J. Constr. Eng. Manage. 129 (2003) 518–527

    work page 2003

  13. [13]

    Sakka, A

    S. Sakka, A. Bouchlaghem, H. Anumba, Simulation of uncertainty in construction projects, Constr. Innov. 16 (2016) 299–315

    work page 2016

  14. [14]

    Elshaer, Impact of sensitivity information on the prediction of project duration using Monte Carlo simulation, Int

    A. Elshaer, Impact of sensitivity information on the prediction of project duration using Monte Carlo simulation, Int. J. Proj. Manage. 31 (2013) 579–588

    work page 2013

  15. [15]

    Grieves, Digital Twin: Manufacturing Excellence through Virtual Factory Replication, 2014

    M. Grieves, Digital Twin: Manufacturing Excellence through Virtual Factory Replication, 2014

    work page 2014

  16. [16]

    F. Tao, H. Zhang, A. Liu, A. Nee, Digital twin in industry: State-of-the-art, IEEE Trans. Ind. Inform. 15 (2019) 2405–2415

    work page 2019

  17. [17]

    Q. Qi, F. Tao, Digital twin and big data towards smart manufacturing and Industry 4.0, Engineering 5 (2019) 653–661

    work page 2019

  18. [18]

    B. Boje, A. Guerriero, S. Kubicki, Y . Rezgui, Towards a semantic construction digital twin: Directions for future research, Autom. Constr. 114 (2020) 103179

    work page 2020

  19. [19]

    Sacks, R

    M. Sacks, R. Brilakis, P. Pikas, X. Xie, O. Girolami, Construction with digital twin information systems, Data-Centric Eng. 1 (2020) e14

    work page 2020

  20. [20]

    Y . Deng, Q. Cheng, C. Anumba, Mapping between BIM and digital twins: A systematic review, Autom. Constr. 120 (2020) 103412

    work page 2020

  21. [21]

    Opoku, S

    A. Opoku, S. Perera, S. Osei-Kyei, Digital twin application in construction: A review, Sustainability 13 (2021) 10615

    work page 2021

  22. [22]

    Gelman, J.B

    A. Gelman, J.B. Carlin, H.S. Stern, D.B. Dunson, A. Vehtari, D.B. Rubin, Bayesian Data Analysis, CRC Press, 2013

    work page 2013

  23. [23]

    Bishop, Pattern Recognition and Machine Learning, Springer, 2006

    C.M. Bishop, Pattern Recognition and Machine Learning, Springer, 2006

    work page 2006

  24. [24]

    Murphy, Machine Learning: A Probabilistic Perspective, MIT Press, 2012

    K.P. Murphy, Machine Learning: A Probabilistic Perspective, MIT Press, 2012

    work page 2012

  25. [25]

    D. Blei, A. Kucukelbir, J. McAuliffe, Variational inference: A review, J. Am. Stat. Assoc. 112 (2017) 859–877

    work page 2017

  26. [26]

    R. M. van Slyke, Monte Carlo methods and the PERT problem, Oper. Res. 11 (1963) 839–860

    work page 1963

  27. [27]

    G. R. R. Lakshminarayanan, Simulation-based schedule risk analysis, J. Constr. Eng. Manage. 135 (2009) 408–417

    work page 2009

  28. [28]

    P. K. K. Chan, Monte Carlo simulation for construction scheduling risk, Autom. Constr. 29 (2013) 198–205

    work page 2013

  29. [29]

    Barraza, Probabilistic control of project performance, J

    A. Barraza, Probabilistic control of project performance, J. Constr. Eng. Manage. 130 (2004) 528– 533

    work page 2004

  30. [30]

    Hulett, Integrated schedule risk analysis, PM World J

    M. Hulett, Integrated schedule risk analysis, PM World J. 1 (2012) 1–16

    work page 2012

  31. [31]

    Heckerman, A tutorial on learning with Bayesian networks, Innov

    D. Heckerman, A tutorial on learning with Bayesian networks, Innov. Bayesian Netw. (1998) 33–82

    work page 1998

  32. [32]

    Neapolitan, Learning Bayesian Networks, Pearson, 2004

    R. Neapolitan, Learning Bayesian Networks, Pearson, 2004

    work page 2004

  33. [33]

    Pearl, Probabilistic Reasoning in Intelligent Systems, Morgan Kaufmann, 1988

    J. Pearl, Probabilistic Reasoning in Intelligent Systems, Morgan Kaufmann, 1988

    work page 1988

  34. [34]

    McCabe, Bayesian methods for construction productivity forecasting, J

    S. McCabe, Bayesian methods for construction productivity forecasting, J. Constr. Eng. Manage. 128 (2002) 19–27

    work page 2002

  35. [35]

    Wang, Bayesian networks in construction risk analysis, Autom

    H. Wang, Bayesian networks in construction risk analysis, Autom. Constr. 19 (2010) 909–917

    work page 2010

  36. [36]

    Kritzinger, Digital Twin in manufacturing: A categorical literature review, IF AC-PapersOnLine 51 (2018) 1016–1022

    A. Kritzinger, Digital Twin in manufacturing: A categorical literature review, IF AC-PapersOnLine 51 (2018) 1016–1022

    work page 2018

  37. [37]

    Bolton, The digital twin: Realizing the cyber-physical connection, Comput

    M. Bolton, The digital twin: Realizing the cyber-physical connection, Comput. Ind. 135 (2022) 103568

    work page 2022

  38. [38]

    Tao, Digital twin-driven smart manufacturing, Rob

    F. Tao, Digital twin-driven smart manufacturing, Rob. Comput.-Integr. Manuf. 61 (2020) 101837

    work page 2020

  39. [39]

    Pan, Digital twin in construction: Review and future directions, Autom

    Y . Pan, Digital twin in construction: Review and future directions, Autom. Constr. 122 (2021) 103517

    work page 2021

  40. [40]

    Lu, Smart construction using digital twins, J

    Z. Lu, Smart construction using digital twins, J. Constr. Eng. Manage. 147 (2021) 04021068

    work page 2021

  41. [41]

    K. A. Bowers, Risk analysis of construction projects, Int. J. Proj. Manage. 16 (1998) 55–63

    work page 1998

  42. [42]

    R. L. Chapman, Risk management in construction, Int. J. Proj. Manage. 19 (2001) 97–105

    work page 2001

  43. [43]

    Dikmen, Risk management in construction projects, Build

    A. Dikmen, Risk management in construction projects, Build. Environ. 42 (2007) 275–283

    work page 2007

  44. [44]

    El-Sayegh, Risk assessment in construction scheduling, J

    S. El-Sayegh, Risk assessment in construction scheduling, J. Constr. Eng. Manage. 134 (2008) 285– 293

    work page 2008

  45. [45]

    P. T. Williams, Critical path variability in project scheduling, Eur. J. Oper. Res. 128 (2001) 31–45

    work page 2001

  46. [46]

    Kim, Simulation-based project scheduling optimization, Autom

    M. Kim, Simulation-based project scheduling optimization, Autom. Constr. 50 (2015) 71–81

    work page 2015

  47. [47]

    Wang, Hybrid simulation for construction planning, Autom

    J. Wang, Hybrid simulation for construction planning, Autom. Constr. 37 (2014) 1–10

    work page 2014

  48. [48]

    AbouRizk, Simulation in construction engineering, J

    S. AbouRizk, Simulation in construction engineering, J. Constr. Eng. Manage. 136 (2010) 1–10

    work page 2010

  49. [49]

    Al-Sudairi, Simulation-based scheduling improvement, Build

    A. Al-Sudairi, Simulation-based scheduling improvement, Build. Res. Inf. 35 (2007) 85–98

    work page 2007

  50. [50]

    Odeh, Simulation modeling in project scheduling, Autom

    H. Odeh, Simulation modeling in project scheduling, Autom. Constr. 12 (2003) 635–644

    work page 2003