A unified probabilistic assessment framework combining Bayesian updating and nested Monte Carlo simulations can track the progressive structural deterioration of corroded steel-concrete composite beams across diverse environmental settings over a century-long operational service life. Developed by civil engineering researchers Duy-Duan Nguyen, Viet-Chuong Ho, Kieu-Vinh Nguyen Thi, Thanh-Tung Nguyen Thi, and Trong-Ha Nguyen, the analytical modeling methodology integrates time-variant empirical metal loss formulations with Bayesian parameter calibration to continuously refine probability distributions of uncertain corrosion rates and structural resistance capacities. Using simulated structural configurations benchmarked against standard design procedures from Eurocode 3 and Eurocode 4, the investigators demonstrated that traditional first-order and second-order structural reliability assessments progressively underestimate long-term structural failure risks when deterioration mechanisms induce pronounced non-linearities.
Steel-concrete composite beams serve as fundamental structural components in bridge superstructures, high-rise buildings, and critical civic infrastructure because they efficiently unite the high compressive strength of concrete slabs with the tensile resistance and structural efficiency of structural steel sections. Atmospheric corrosion degrades the beam by causing severe cross-sectional area loss, surface roughening, elastic modulus reduction, and stress-raising pit formations that impair both local stability and overall bending capacity. While corrosion-induced section loss accounts for a substantial fraction of infrastructural failures worldwide, conventional reliability analyses frequently treat structural capacity deterministically or use static reliability formulations that require complete computational restarts whenever updated inspection measurements become available.
To model this ongoing physical degradation accurately over extended operational spans, the researchers adopted an atmospheric corrosion depth model based on power-law exponential formulations originally characterized by Komp. In this formulation, corrosion depth over time is governed by multiplicative rate and kinetic curve-shape parameters, designated as variables A and B, which are treated as random variables possessing distinct statistical distributions. Because corrosion depth must remain strictly non-negative while reflecting extensive environmental variability, parameter A is assigned a lognormal prior distribution, whereas parameter B is modeled using a normal distribution. Non-uniform degradation patterns observed in empirical field surveys are captured through localized position coefficients alongside lognormally distributed spatial multipliers, explicitly representing intensified material loss across lower flanges and lower web zones susceptible to moisture retention and contaminant accumulation.
The structural load-bearing capacity of the composite section is formulated mathematically by evaluating plastic bending moments and vertical shear resistance in strict accordance with the limit states defined in Eurocode 4. Depending on structural geometry and component dimensions, the plastic neutral axis of the composite section can locate within the concrete slab, the upper steel flange, or the steel web. As atmospheric exposure progressively reduces the effective cross-sectional thickness of the steel beam according to local position coefficients, structural resistance parameters—including plastic section modulus, shear area, and second moment of area—are continuously recalculated over discrete operational intervals spanning from baseline conditions up to one hundred years.
The mathematical evaluation of structural safety is conducted across three defined limit state functions representing bending capacity, shear resistance, and serviceability deflection thresholds. Under this multi-criteria formulation, failure probability is defined through a series system assumption where structural failure occurs if external applied load effects exceed available capacity for any single limit state. The researchers implemented an iterative updating protocol wherein synthetic five-point thickness measurement records taken across web and flange locations calibrate the prior distributions of corrosion parameters via Bayes theorem. The resulting posterior parameter samples are subsequently propagated through nested inner-loop Monte Carlo simulations comprising ten thousand sampling iterations to quantify time-dependent failure probabilities and compute standard reliability indices.
When applied to an illustrative composite girder configuration consisting of a standardized H400 by 400 by 8 by 12 structural steel section integrated with a reinforced concrete slab measuring one hundred and twenty millimeters in thickness and one thousand millimeters in effective width, the framework revealed stark performance contrasts across environmental classifications. In relatively benign rural environments, baseline failure probability centered narrowly around 4.8 percent before rising moderately to approximately 9.4 percent after twenty-five years, with a ninety-five percent confidence interval ranging between 6.55 percent and 14.75 percent. After one hundred years of continuous rural exposure, the mean failure probability reached 10.44 percent, although the upper boundary of the confidence interval expanded to 20.10 percent, illustrating how parameter uncertainty gradually widens long-term predictive intervals.
Following twenty-five years of simulated urban exposure, lower flange thickness declined to an average range between 9.5 and 9.7 millimeters, reflecting twelve to fifteen percent greater section loss than rural baselines, which maintained thicknesses between 10.8 and 11.0 millimeters. Consequently, the mean failure probability under urban conditions rose to approximately 16.6 percent at year twenty-five, reaching an average of 20.5 percent at one hundred years, with an upper confidence bound exceeding fifty percent. At fifty years, the nested Monte Carlo approach predicted an 18.2 percent failure probability, while traditional first-order and second-order reliability methods yielded lower estimates of 16.6 percent and 17.2 percent, respectively, failing to fully capture nonlinear limit-state behavior.
Simulated coastal marine conditions produced the most severe structural degradation due to the intense corrosive action of chloride deposition. Marine exposure caused rapid early-stage deterioration, driving mean failure probabilities from an initial 4.82 percent to 23.45 percent within the first twenty-five years of service. By one hundred years, severe section loss reduced average lower flange thicknesses to between 4 and 5 millimeters—representing more than sixty-five percent total thickness loss—and lowered lower web dimensions to between 3 and 4 millimeters. Mean failure probability reached 32.50 percent, corresponding to a structural reliability index reduction of roughly 0.45 to 0.50, while the ninety-five percent confidence interval expanded dramatically to span between 11.5 percent and 85.7 percent, representing sixty-five percent greater dispersion than observed in urban conditions.
To isolate which engineering parameters exert the greatest influence over system reliability across time, the authors conducted a global variance-based Sobol sensitivity analysis, computing first-order and total sensitivity indices. During early service phases spanning the first decade, structural reliability was predominantly governed by initial component dimensions and steel yield strength, with initial flange thickness yielding first-order indices between 0.18 and 0.23 across rural settings. Over longer operating intervals exceeding fifty years, initial manufacturing geometry lost significance, and corrosion kinetic parameters emerged as the dominant sources of variance; the shape coefficient B achieved first-order indices near 0.25 and total indices reaching 0.60 to 0.65, underscoring the critical necessity of field-updating deterioration parameters.
The authors concluded that while simplified first-order and second-order reliability approximations remain computationally efficient and sufficiently accurate during the initial fifty years of service life, advanced simulation frameworks are essential for capturing extreme distribution tails under heavy deterioration. Because urban and marine exposures produce significant degradation gradients, the researchers recommend shortening structural inspection intervals to ten to fifteen years in coastal environments, with prioritized monitoring directed specifically toward the lower web and lower flange surfaces most exposed to water accumulation and aggressive salt deposits.
The analytical framework evaluates structural safety through a limit-state formulation defined as the numerical margin between structural resistance and applied load effects. System failure is characterized using a series system configuration where failure occurs as soon as any single limit state governing plastic bending capacity, shear resistance, or allowable serviceability deflection is reached. Importantly, the authors noted that potential physical interactions among these competing limit states were omitted from the analysis, meaning the resulting system failure probabilities serve as an approximation under a non-interacting operational assumption. To ensure rigorous computational precision during posterior sampling and limit-state evaluations, the investigators enforced a convergence criterion that restricted the simulation error to 0.0025.
From an asset management perspective, the study outlines practical maintenance criteria by recommending proactive interventions whenever the estimated cumulative failure probability exceeds approximately 15 percent. To keep structural performance within safe operating boundaries, the researchers emphasized that coastal structures benefit from regular inspection intervals scheduled every ten to fifteen years, complemented by localized evaluations every two to five years after the first operational decade. During these maintenance cycles, engineers can apply protective surface mitigation measures, such as epoxy coatings, cold galvanizing, or polyurethane treatments, to maintain a target reliability level above 0.90 and keep overall failure probabilities beneath 0.10 over the service life.
The findings also emphasize the contrasting structural consequences of distinct deterioration mechanisms, distinguishing between generalized uniform section thinning and aggressive localized pitting. Although localized pitting affects smaller surface zones, it generates concentrated stress points and accelerated loss of cross-sectional area near lower flanges and web junctions. By updating five inspection measurement points across key cross-sectional positions over simulated service intervals from five to one hundred years, the Bayesian protocol continuously refines the kinetic deterioration parameters. This calibration confirms that environmental degradation eventually overshadows initial geometric tolerances, reinforcing the value of adaptive monitoring for aging composite structures.
Subject of Research: Time-dependent reliability assessment of corroded steel-concrete composite beams using Bayesian updating and Monte Carlo simulations
Article Title: Time-dependent reliability evaluation of corroded steel-concrete composite beams using Bayesian updating and Monte-Carlo simulation
Article References: Nguyen, D.-D., Ho, V.-C., Thi, K.-V. N., Thi, T.-T. N., & Nguyen, T.-H. (2026). Time-dependent reliability evaluation of corroded steel-concrete composite beams using Bayesian updating and Monte-Carlo simulation. Journal of Infrastructure Preservation and Resilience, 7(1), Article 29. https://doi.org/10.1186/s43065-026-00187-x
Image Credits: AI Generated
DOI: 10.1186/s43065-026-00187-x
Keywords: steel-concrete composite beams, structural reliability, Bayesian updating, Monte Carlo simulation, corrosion modeling, time-dependent reliability, Sobol sensitivity analysis, structural deterioration, marine corrosion, urban corrosion, rural corrosion, limit state functions, failure probability, Eurocode 4, infrastructure preservation, structural maintenance
Cite Scienmag News
Reid Dalton. (November 19, 2026). Bayesian Monte Carlo Framework Predicts Time-Dependent Reliability for Corroded Composite Beams. Scienmag. https://scienmag.com/bayesian-monte-carlo-framework-predicts-time-dependent-reliability-for-corroded-composite-beams/
Reid Dalton. "Bayesian Monte Carlo Framework Predicts Time-Dependent Reliability for Corroded Composite Beams." Scienmag, 19 November 2026, https://scienmag.com/bayesian-monte-carlo-framework-predicts-time-dependent-reliability-for-corroded-composite-beams/. Accessed 19 November 2026.
Reid Dalton. "Bayesian Monte Carlo Framework Predicts Time-Dependent Reliability for Corroded Composite Beams." Scienmag. November 19, 2026. https://scienmag.com/bayesian-monte-carlo-framework-predicts-time-dependent-reliability-for-corroded-composite-beams/








