Seismic Failure Mechanism Shift in RC Buildings Revealed by NDT-Supported, Field-Calibrated BIM-Based Models

Abstract

This study proposes a field-calibrated, NDT-integrated BIM modeling framework to improve the reliability of post-earthquake assessment for reinforced concrete (RC) buildings. The approach combines destructive and nondestructive testing (NDT) data—including core drilling, Schmidt hammer, ultrasonic pulse velocity (UPV), and Windsor probe—through a site-specific WinSonReb regression model. The calibrated material properties (average compressive strength ≈ 18.6 MPa, CoV > 20%) were embedded into a Building Information Modeling (BIM) environment, producing an as-is, NDT-calibrated BIM model representing a Level-2 static digital twin of the structure. Nonlinear static pushover analyses performed in accordance with TBDY-2018 and ASCE 41-17 showed that the calibrated model exhibits a fundamental period of 0.85 s—approximately 18% longer than the uncalibrated BIM model. This elongation increased displacement demand and caused a shift in performance classification: while the uncalibrated model indicated Life Safety (LS), the calibrated model predicted behavior approaching Collapse Prevention (CP) in the Y direction. Furthermore, calibration reversed the predicted damage hierarchy, from ductile beam hinging to brittle column- and wall-controlled failure near elevator openings, consistent with post-event observations from the 2023 Kahramanmaraş earthquakes. These results demonstrate that integrating field-calibrated NDT data into BIM-based seismic models fundamentally alters both strength estimation and failure-mechanism prediction, reducing epistemic uncertainty and providing a more conservative basis for retrofit prioritization. Although demonstrated on a single case study, the proposed workflow offers a realistic and scalable pathway for NDT-supported seismic performance assessment of existing RC buildings.

1. Introduction

Earthquakes remain among the most critical natural hazards affecting reinforced concrete (RC) building stocks worldwide. The devastating 6 February 2023 Kahramanmaraş earthquakes in Türkiye again revealed the high vulnerability of existing building inventories, with more than 500,000 damaged or collapsed structures [1]. Field investigations consistently reported deficiencies such as soft-story mechanisms, short columns, and poor construction quality—factors that similarly contributed to significant failures in previous events, including the 2010 Maule earthquake in Chile [2] and the 2018 Palu earthquake in Indonesia [3]. Regional seismic risk assessments, such as those conducted for the Kathmandu Valley, further indicate that upgrading existing RC buildings could reduce fatalities by more than 50% and direct economic losses by about 14% [4]. These observations highlight the urgent need for reliable, field-calibrated post-earthquake performance assessment methods for RC buildings.

Although modern seismic design codes have significantly improved life-safety performance, post-event investigations from the 1994 Northridge and 1995 Kobe earthquakes revealed that many RC buildings formally meeting design provisions nonetheless sustained substantial structural and nonstructural damage, resulting in long downtime and severe economic disruption [5]. These limitations motivated the development of the Performance-Based Earthquake Engineering (PBEE) framework, which adopts a probabilistic, multi-dimensional view of structural performance across hazard, structural, damage, and loss analyses [5,6]. International standards and technical guidelines—including Eurocode 8 [7], ASCE/SEI-41 [8], and the PEER-Tall Buildings Initiative (PEER-TBI) guidelines [9]—have increasingly incorporated performance-based earthquake engineering (PBEE) concepts. Recent analytical studies on RC buildings further underscore the importance of PBEE frameworks for next-generation seismic risk evaluation and decision-making [10]. PBEE thus provides a comprehensive basis not only for assessing safety, but also for quantifying resilience, sustainability, and functional recovery performance under earthquake demands.

Realizing the full potential of PBEE in practice requires advanced digital platforms capable of integrating heterogeneous, multi-scale datasets and supporting rapid post-earthquake decision-making. Building Information Modeling (BIM) has emerged as a central tool for generating structured digital representations of existing assets and facilitating multi-disciplinary coordination throughout the building lifecycle [11]. When extended into a Digital Twin (DT) framework, BIM can incorporate updated field information to better reflect the evolving state of physical assets [12]. In this study, the term “Digital Twin” refers to a Level-2 static DT architecture, characterized by one-directional semantic updates derived from field measurements rather than real-time bi-directional data streaming. Although digital twin maturity concepts have been widely discussed across different sectors, recent studies emphasize that DT maturity frameworks remain largely context-dependent and nonstandardized, particularly for civil and structural engineering applications, necessitating pragmatic and application-oriented classifications rather than rigid taxonomies [13,14]. This definition, therefore, reflects a deliberate engineering abstraction aligned with the scope of post-earthquake structural assessment.

Research on BIM–DT integration continues to expand, particularly in areas such as energy management, predictive maintenance, and occupant-centric control when combined with IoT infrastructures, XR technologies, and AI-driven analytics [15,16]. Bibliometric analyses identify increasing convergence among BIM, DT, artificial intelligence, and blockchain technologies, while also highlighting persistent challenges such as interoperability constraints, cybersecurity risks, and the absence of standardized maturity frameworks [17]. Despite this rapid development, most BIM–DT applications focus on geometric fidelity, sensor integration, or operational monitoring, with comparatively limited attention given to embedding calibrated mechanical properties—particularly in-situ concrete strength—into DT environments [18,19]. Yet, for RC buildings, material heterogeneity is among the most influential parameters governing seismic performance and failure mechanisms.

Previous studies have explicitly examined the influence of in-situ concrete strength variability on the seismic response of existing reinforced concrete buildings. For example, ref. [20] investigated irregular RC structures by systematically varying concrete compressive strength both globally and floor-wise, demonstrating that material heterogeneity can significantly affect nonlinear seismic response, displacement demand, and damage localization. Their findings underscore the sensitivity of performance assessment to strength characterization and analysis methodology for existing buildings. However, such investigations are generally conducted within conventional numerical modeling frameworks and parametric sensitivity analyses, without embedding field-calibrated material properties into BIM-based digital twin environments. As a result, while the importance of concrete strength variability is well recognized, its direct integration into semantic BIM–DT models—and its implications for predicted failure hierarchy and performance classification—remain insufficiently explored. In particular, the transition from parametric sensitivity studies to field-calibrated, semantically enriched digital twin models remains largely unexplored.

Reliable estimation of in-situ concrete strength remains a cornerstone of post-earthquake assessment. Nondestructive testing (NDT) techniques, such as the Schmidt hammer, ultrasonic pulse velocity (UPV), and Windsor probe, are widely used for their practicality, but when applied individually, they often yield inconsistent or scattered estimates in heterogeneous or deteriorated concrete [21]. To mitigate this variability, combined methods, such as SonReb and WinSonReb, incorporate multiple NDT measurements and calibrate them against destructive core tests to improve predictive accuracy [22,23]. Recent studies further demonstrate that AI-based regression models and response surface methodologies (RSMs) can reduce estimation errors to below 10% [24]. Beyond strength estimation, NDT approaches have been used to detect corrosion in prestressed systems [25] and to evaluate durability-related degradation mechanisms under service conditions [26,27]. However, despite these advances, NDT outputs are frequently presented as isolated datasets and are rarely fused with BIM or DT platforms for holistic structural performance assessment. Recent physics-guided and self-supervised learning studies further indicate that embedding physical constraints into ultrasonic NDT interpretation substantially improves robustness and interpretability under heterogeneous concrete conditions, particularly when labeled field data are limited [28].

Moreover, while previous studies have demonstrated the value of NDT-enhanced material characterization and BIM-based modeling, the direct impact of field calibration on governing seismic failure mechanisms and performance classification remains insufficiently explored. In particular, explicit comparisons demonstrating how calibration-induced updates to material properties may alter predicted collapse modes or shift performance levels are still scarce in the literature, limiting the interpretability of digital assessment workflows for post-earthquake decision-making.

To address this gap, the present study develops a methodology that integrates NDT and destructive testing data into a BIM-based DT model for post-earthquake performance assessment of RC buildings. Field data—including Schmidt hammer, UPV, Windsor probe, and electromagnetic reinforcement scanning—were combined with core sampling to construct a calibrated material property database embedded directly within the DT environment. The methodology was validated using a heavily damaged RC building in Malatya, Türkiye, which experienced significant damage during the 6 February 2023 Kahramanmaraş earthquakes.

Structural performance analyses were conducted using the BIM-integrated platform ProtaStructure, enabling direct calibration of analytical models with field-measured parameters. Embedding heterogeneous test results within a static DT environment allows for more realistic, uncertainty-aware performance predictions compared to nominal-code approaches. Although this study focuses on a single case, the adopted approach is consistent with established performance-based earthquake engineering practice, where representative benchmark buildings are frequently used to derive mechanism-oriented insights into seismic behavior. Previous benchmark studies have demonstrated that detailed single-building analyses may reveal multiple collapse mechanisms and that modeling-related uncertainties may influence collapse predictions to a degree comparable to record-to-record ground motion variability [29]. Accordingly, the proposed workflow emphasizes mechanism-based interpretation rather than case-specific generalization.

In summary, this paper presents a generalizable BIM→DT workflow that integrates site-calibrated NDT and core data as semantic property sets, enabling analyses that more faithfully represent in-situ material heterogeneity. The approach reveals failure mechanisms and damage hierarchies that remain hidden under nominal design assumptions and provides measurable, decision-relevant insights for PBEE, as validated through a heavily damaged RC building.

2. Materials and Methods

2.1. Case Study Building and Field Investigations

The case study focuses on a six-story RC residential building located in Malatya, Türkiye, which sustained severe structural damage during the 6 February 2023 Kahramanmaraş earthquakes. Constructed in the late 1990s under outdated seismic provisions, the building exhibited typical deficiencies of that era, including insufficient transverse confinement, weak-column–strong-beam proportions, and inadequate reinforcement detailing. Post-earthquake inspections revealed pronounced shear cracking and spalling in the elevator core wall, along with widespread deterioration in primary vertical load-bearing elements, indicating a predominantly brittle damage pattern.

To characterize the in-situ material properties while minimizing epistemic uncertainty, an integrated field-testing campaign was conducted, combining destructive and nondestructive techniques. Core samples extracted from beams, columns, and shear walls provided direct measurements of compressive strength in accordance with ASTM C42/C42M [30]. In parallel, Schmidt rebound hammer, UPV, and Windsor probe tests were performed following ASTM C805/C805M [31], ASTM C597 [32], and ASTM C803/C803M [33], respectively, at corresponding locations across multiple floor levels to ensure spatial consistency between destructive and nondestructive measurements. The resulting dataset was compiled and statistically evaluated, as summarized in

Table 1. Comparison of destructive (core) and nondestructive test results for case study building.

Test Method	N (Samples)	Mean (MPa)	Std. Dev. (MPa)	CoV (%)	Remarks
Core (Destructive)	18	18.6	3.9	21.0	Baseline (ASTM C42)
Schmidt Hammer (NDT)	18	23.4	4.5	19.2	ASTM C805
UPV (NDT)	18	22.6	3.7	16.4	ASTM C597
Windsor Probe (NDT)	18	21.9	4.2	19.1	ASTM C803
SonReb (Combined)	18	22.99	3.2	13.9	Rebound + UPV
WinSonReb (Combined)	18	21.33	2.8	13.1	Rebound + UPV + Windsor

Note: Average in-situ compressive strength from core = 18.6 MPa, SonReb = 22.99 MPa, WinSonReb = 21.33 MPa; CoV > 20% indicates high material heterogeneity across elements.

.

The average compressive strength obtained from core tests was approximately 18.6 MPa, reflecting relatively low material quality consistent with construction practices of the period. In contrast, uncalibrated NDT-based estimates yielded systematically higher strength values (SonReb: 22.99 MPa; WinSonReb: 21.33 MPa), in agreement with previous findings indicating that NDT methods, when applied without site-specific calibration, tend to overestimate in-situ concrete strength [34,35,36]. The coefficient of variation (CoV) exceeding 20% across core results highlights substantial material heterogeneity among structural elements, a factor known to strongly influence seismic response and potential failure localization in existing RC buildings.

The combined use of destructive testing and multi-parameter NDT thus provided a robust and internally consistent basis for subsequent digital modeling. Rather than relying on probabilistic sampling or large inventories, this study adopts a deterministic, field-calibrated material assignment strategy, prioritizing physical representativeness over statistical generalization. Calibrated material properties were directly embedded into the BIM environment to enable element-level differentiation in structural analysis.

As illustrated in Figure 1, the spatial distribution of facade damage and test locations demonstrates a clear correspondence between observed physical deterioration and the measurement points used for calibration. This spatial linkage forms the foundation for the digital twin-based performance analyses presented in the following sections, which detail the testing methodologies, regression framework, workflow integration, and BIM-enabled digital twin modeling strategy.

2.2. Testing Methodologies and Calibration Framework

All tests were conducted in accordance with internationally recognized standards. Core compressive strength tests were performed following ASTM C42/C42M [30] and EN 12504-1 [37]; Schmidt rebound hammer tests complied with ASTM C805; UPV measurements followed ASTM C597; and Windsor probe testing was carried out in accordance with ASTM C803.

When applied individually, NDT techniques are known to exhibit significant scatter and sensitivity to surface conditions, moisture content, aggregate type, and internal heterogeneity of concrete [38,39]. To improve predictive reliability, combined NDT approaches have, therefore, been extensively investigated in the literature. Among these, the SonReb method—originally introduced under RILEM recommendations—integrates rebound number and ultrasonic pulse velocity measurements through a regression-based formulation expressed as follows:

f_c = a + bR + cV

(1)

where R denotes the Schmidt rebound number and V represents the ultrasonic pulse velocity (km/s).

In the present study, regression coefficients derived from paired core–NDT measurements yielded a coefficient of determination of R² = 0.87, indicating a strong correlation between predicted and measured in-situ compressive strengths. The extended WinSonReb formulation further incorporates Windsor probe measurements, enhancing predictive accuracy and reducing relative estimation error to below 10% [40,41]. Consistent with prior investigations, these hybrid correlations systematically outperform single-parameter NDT approaches across diverse material qualities and environmental conditions [42].

Although exponential and nonlinear SonReb formulations have been proposed in the literature, the present study adopts a linear regression model as a deliberate engineering abstraction tailored to post-earthquake field conditions. This choice is motivated by three primary considerations: (i) the limited size and heterogeneous nature of post-disaster datasets, which can lead to unstable parameter estimation in nonlinear models; (ii) the need for transparent, interpretable calibration procedures compatible with deterministic performance-based seismic assessment frameworks; and (iii) consistency with code-oriented practices, where material properties are incorporated through explicit coefficients rather than data-driven black-box models. Within the calibrated strength range of the investigated building, the linear formulation provided statistically robust performance without introducing artificial sensitivity to regression form.

Recent advances in machine-learning-based concrete strength prediction have reported coefficients of determination as high as 0.98 when extensive, high-quality training datasets are available [43]. However, such data-intensive models typically require large, homogeneous datasets and controlled testing conditions—requirements that are rarely satisfied in rapid post-earthquake field investigations. Accordingly, this study prioritizes site-calibrated SonReb and WinSonReb correlations as a pragmatic and transparent compromise between predictive accuracy and field applicability, rather than pursuing maximum statistical performance at the expense of robustness and interpretability.

From a reliability perspective, the adopted calibration strategy primarily reduces epistemic uncertainty associated with model bias and systematic strength overestimation, while the remaining intra-element variability reflects the inherent aleatory randomness of concrete material properties. This distinction is particularly relevant for seismic performance evaluation, as calibration improves confidence in assigned strength values without artificially suppressing spatial heterogeneity, which governs local damage concentration and failure initiation in existing reinforced concrete structures.

Table 2. Literature comparison of NDT, SonReb, WinSonReb, and AI-based prediction accuracy levels.

Method	Input Parameters	Typical R2 Range	Mean Absolute Error (MAE, %)	Reference
Rebound Hammer (NDT)	R	0.50–0.70	15–25	[38]
UPV (NDT)	V	0.60–0.75	12–20	[39]
SonReb	R, V	0.80–0.90	8–12	[40]
WinSonReb	R, V, W	0.85–0.93	5–10	[41]
AI-based Models	R, V, W (+features)	0.90–0.98	<5	[43]

summarizes typical accuracy ranges reported in the literature for individual NDT techniques, combined regression-based methods, and data-driven approaches, providing a comparative context for the selected calibration framework.

2.3. Data Integration Workflow

A structured four-stage workflow was formulated to integrate heterogeneous field and analytical datasets into a coherent digital environment suitable for post-earthquake structural assessment:

• Data Collection and Pre-Processing. Field measurements, including both destructive and NDT data, were digitized, cleaned, and organized together with descriptive metadata to ensure traceability. Recorded attributes included structural element type, test location, measurement conditions, and associated uncertainties.
• Model Development in BIM. A detailed building information model was developed using a BIM-based structural modeling platform (ProtaStructure), in which geometry, member dimensions, and reinforcement layouts were represented as parametric entities. The BIM model served as the digital backbone for subsequent integration of calibrated material properties and analytical modeling.
• Data Embedding and Calibration. NDT-derived strength estimates were calibrated against core test results using the regression coefficients defined in Section 2.2. The calibrated material parameters were then embedded into the BIM model as custom property sets (Psets) assigned at the element level. This approach ensured that material properties reflected field-measured characteristics rather than nominal design assumptions, while preserving spatial variability among structural components.
• Digital Twin Synchronization. The BIM model was subsequently extended into a Level-2 static digital twin, enabling periodic semantic updates based on post-earthquake field observations and calibrated test results. Although real-time bi-directional data exchange was not implemented, the adopted configuration provides a traceable and sufficiently responsive framework for post-earthquake engineering assessments that require rapid yet controlled model updates.

This workflow ensures high fidelity in representing the as-is condition of the structure while remaining adaptable for future data incorporation. Figure 2 illustrates the complete data flow, from field investigations through BIM integration and digital twin synchronization, highlighting the interaction between physical measurements and their digital representation.

2.4. Digital Twin Modeling and Seismic Performance Analysis

The calibrated BIM model was exported to a digital twin environment using the IFC4 schema to ensure interoperability between modeling and analytical platforms. This process enabled the consistent transfer of geometry, reinforcement layouts, and calibrated material properties, ensuring that the digital twin accurately represented the as-is condition of the building and could accommodate additional inspection data if required [44]. Within this framework, the Level-2 static digital twin served as the analytical baseline for seismic performance evaluation.

Seismic analyses were performed in accordance with TBDY-2018 [45] and ASCE/SEI 41-17 [8] provisions, which define three principal performance levels: Immediate Occupancy (IO), Life Safety (LS), and Collapse Prevention (CP). Nonlinear static (pushover) analyses were employed to capture global inelastic deformation capacity and progressive stiffness degradation, while modal response spectrum analyses were conducted to verify dynamic characteristics, including fundamental period elongation [46].

Site-specific seismic demand parameters were defined using local geotechnical data, including soil class ZC and a peak ground acceleration (PGA) of 0.44 g. Elastic design spectra corresponding to the Design-Basis Earthquake (DBE) and Maximum-Considered Earthquake (MCE) levels were developed in accordance with relevant guidelines [47]. Equivalent lateral load patterns were based on modal participation factors, and vertical seismic effects were explicitly considered following TBDY-2018 requirements.

Crucially, site-calibrated SonReb and WinSonReb correlations were embedded directly into structural elements within the digital twin environment, enabling element-level differentiation of concrete strength and stiffness. This calibration process primarily reduced epistemic uncertainty associated with systematic strength overestimation, while preserving the inherent spatial variability observed in field measurements [48,49]. As a result, the calibrated digital twin provides a more realistic representation of stiffness distribution and local strength degradation compared to nominal-code-based models.

Performance acceptance criteria were harmonized across ASCE/SEI 41-17 and TBDY-2018 and included the following:

• Plastic hinge rotations for primary and secondary members;
• Interstory drift limits of approximately 0.7%, 2.5%, and 4.0% for IO, LS, and CP, respectively [45];
• Shear and axial capacity ratios for critical vertical elements;
• Global stability verification for P–Δ effects and torsional irregularities [50].

Figure 3 presents a representative pushover curve obtained from the calibrated digital twin, illustrating the evolution of structural response across IO, LS, and CP performance thresholds. Comparison with the uncalibrated BIM-based model indicates increased fundamental period and reduced base shear capacity following calibration, consistent with observed field damage and material degradation. These shifts reflect changes in governing failure mechanisms induced by field-calibrated material properties rather than alterations to structural geometry or loading assumptions.

2.5. Interoperability Challenges in BIM–DT Data Exchange

Although IFC standards provide a common foundation for cross-platform data exchange, interoperability challenges persist in practice. Semantic ambiguities may lead different platforms to interpret identical components inconsistently, affecting geometry representation and material assignment [51]. In particular, advanced structural parameters—such as nonlinear hinge definitions and calibrated NDT-derived material properties—are not fully supported in current IFC schemas, limiting their direct and lossless transfer between BIM and analytical environments [52].

In the present study, partial information loss was observed during the export of custom Psets containing SonReb coefficients and calibrated concrete strengths from a BIM-based structural modeling platform to the IFC4 schema. While geometric information was preserved, some reinforcement attributes and stiffness modifiers associated with element-level calibration were incompletely mapped, necessitating targeted manual verification to maintain analytical fidelity.

Proposed solutions in the literature include ontology-based semantic enrichment [53], discipline-specific model view definitions (MVDs) [54], and open-domain data exchange frameworks aimed at extending IFC capabilities for performance-oriented applications [55]. Figure 4 illustrates representative geometry and property mismatches encountered during cross-platform exchange, underscoring the need for more robust and semantically explicit interoperability pipelines when calibrated material data are integrated into BIM–DT workflows.

3. Results

This section presents the outcomes of destructive and nondestructive field investigations, the development of the calibrated digital twin model, and the subsequent nonlinear seismic performance analyses. Comparative results between the calibrated and uncalibrated models are discussed to examine the influence of field-based calibration on seismic performance evaluation.

3.1. Field Investigation Findings

The results of the destructive and nondestructive field tests are summarized in

Table 3. Statistical summary of field test results.

Test Method	N (Samples)	Mean (MPa)	Std. Dev. (MPa)	CoV (%)	R2	p-Value
Core (Destructive)	18	18.60	3.9	21.0	–	–
Schmidt Hammer (NDT)	18	23.40	4.5	19.2	–	–
UPV (NDT)	18	22.60	3.7	16.4	–	–
Windsor Probe (NDT)	18	21.90	4.2	19.1	–	–
SonReb (Combined)	18	22.99	3.2	13.9	0.87	<0.05
WinSonReb (Combined)	18	21.33	2.8	13.1	0.87	<0.05

Note: Table 3 presents a statistical comparison of destructive and nondestructive test results, including sample size (N), mean compressive strength, standard deviation, CoV, and, where applicable, regression correlation coefficients (R²) and significance levels (p-values).

. Core drilling yielded an average compressive strength of 18.6 MPa, with a CoV exceeding 20%, indicating pronounced heterogeneity in concrete quality across the structure.

By contrast, Schmidt hammer, UPV, and Windsor probe measurements produced systematically higher strength estimates—up to approximately 15–20% above the core-based baseline. This tendency is consistent with previously reported biases in existing RC buildings when NDT measurements are applied without site-specific calibration [56,57,58].

The variability in the strength estimates is further illustrated in Figure 5, which presents box plots of compressive strength distributions obtained from core, SonReb, and WinSonReb measurements. The wider scatter observed in the core test results reflects the intrinsic heterogeneity of in-situ concrete and element-to-element material variability in existing RC buildings, consistent with prior assessments of aging RC structures [59,60].

Schmidt hammer and UPV results are not plotted separately in Figure 5, as their individual estimates exhibit systematic bias and higher scatter when considered independently, and are, therefore, discussed primarily in the context of combined regression-based calibration.

To reduce discrepancies between destructive and nondestructive measurements, a site-specific WinSonReb regression model was developed by combining rebound hammer and UPV data. The resulting relationship, given in Equation (2), exhibits a strong fit (R² = 0.87, p < 0.05), confirming the internal consistency of the calibrated model. This outcome aligns with prior studies emphasizing the advantages of multi-parameter regression and hybrid calibration strategies in heterogeneous concrete materials [61,62].

$$fc=7.65+0.42R+3.91V$$

(2)

Here, R represents the Schmidt rebound number, V is the ultrasonic pulse velocity (km/s), and f_c is the estimated in-situ compressive strength (MPa).

Taken together, Table 3, Figure 5, and Equation (2) underline the necessity of applying site-specific calibration. Without such adjustments, compressive strength would likely be overestimated, leading to unconservative and potentially unsafe performance predictions in subsequent seismic assessments.

3.2. Calibrated Digital Twin Model

The field-calibrated material properties obtained from Section 3.1 were incorporated into ProtaStructure to develop an as-is DT model of the investigated building. While the nominal BIM model, which assumed an idealized concrete strength of 20 MPa (C20), the calibrated DT integrates the field-measured average compressive strength of 18.6 MPa, its associated variability, and regression-based correction factors derived from the combined NDT dataset. Figure 6 shows the resulting three-dimensional DT model, in which the geometry and reinforcement layout are fully represented in digital form.

A comparative summary of nominal design parameters versus calibrated in-situ values is presented in

Table 4. Comparison of design vs. calibrated material parameters.

Parameter	Design Model (Nominal)	Calibrated Digital Twin (Field-Based)	Difference (%)
Concrete Strength (fck, MPa)	20.0	18.6	–7.0
Reinforcement Yield Strength (fyk, MPa)	420	415	–1.2
Modulus of Elasticity (E, MPa)	30,000	27,500	–8.3
Concrete Class (EN 206 [65])	C20/25	C16/20	—
Coefficient of Variation (CoV, %)	—	>20	—

Note: The calibrated compressive strength (18.6 MPa) corresponds to C16 concrete class per EN 206 [65] equivalency; the nominal design class was C20 (20 MPa).

. The discrepancies between design assumptions and field-derived material characteristics are consistent with findings from previous post-earthquake investigations of RC buildings and highlight the influence of field-derived material properties on analytical modeling outcomes [63,64].

Material heterogeneity integrated into the DT model is visualized in Figure 7, where element-level compressive strength values are displayed using a color-coded strength map. Weaker columns and wall segments are concentrated near the elevator core, indicating localized zones of reduced capacity. Similar spatial mapping techniques have been used in prior studies to identify critical weaknesses and to guide targeted retrofit strategies in earthquake-damaged RC buildings [66,67].

Dynamic analysis of the calibrated DT indicated a fundamental period of approximately 0.85 s—about 18% longer than that of the uncalibrated BIM model (0.72 s). This period of elongation is attributed to a reduction in global stiffness due to decreased concrete strength and pre-existing structural damage. The first mode shape, illustrated in Figure 8, exhibits noticeable torsional components arising from asymmetric stiffness and mass distribution. As emphasized in the literature, mode-shape validation is essential to ensure the dynamic reliability and interpretive value of DT-based seismic evaluations [68].

3.3. Seismic Performance Analysis Results

Nonlinear static pushover analyses were performed on the calibrated DT in both principal directions (X and Y), in accordance with TBDY-2018 [45] and ASCE 41-17 [8]. The resulting capacity curves are given in Figure 9, where performance thresholds corresponding to IO, LS, and CP are marked.

The maximum base shear capacities were found to be 2050 kN in the X direction and 1980 kN in the Y direction, corresponding to roof displacements of approximately 0.19 m and 0.21 m, respectively. The performance point in the X direction was located between the IO and LS thresholds, whereas in the Y direction, it was positioned close to the LS–CP boundary, indicating reduced deformation capacity in the weaker horizontal axis. This behavior is consistent with previous observations from post-earthquake evaluations of substandard RC frames with limited ductility [69,70].

Interstory drift ratios corresponding to the performance points are summarized in

Table 5. Interstory drift ratios at the performance point compared with ASCE 41-17 limits.

Story	Direction	Drift Ratio (%)	ASCE 41-17 Limit (%)	Performance Level
6	X	1.8	2.5	LS OK
6	Y	3.4	2.5	>LS
5	X	2.1	2.5	LS OK
5	Y	2.8	2.5	Slight Exceedance
4–1	X	≤2.0	2.5	LS OK
4–1	Y	≤2.5	2.5	LS OK

Note: LS limit = 2.5%, CP limit = 4.0%.

. While lower stories satisfied the LS drift limit of 2.5% in both directions, upper stories—particularly in the Y direction—slightly exceeded this threshold, indicating localized drift concentration at upper stories. Such drift patterns are commonly reported in mid-rise RC buildings that have undergone significant earthquake-induced damage [71,72].

Plastic hinge development patterns are presented in Figure 10, showing column base hinging, flexural yielding in beams, and localized hinges around elevator wall openings. These patterns closely match damage mechanisms observed in the 2023 Kahramanmaraş earthquakes, demonstrating that the calibrated DT is capable of capturing key damage mechanisms observed in real earthquake-damaged RC buildings [73,74].

Key performance indicators obtained from the analyses are listed in

Table 6. Key performance indicators of the calibrated model.

Parameter	Direction X	Direction Y
Maximum Base Shear (kN)	2050	1980
Roof Displacement (m)	0.19	0.21
Performance Point Displacement (m)	0.14	0.16
Performance Level	LS	Near-CP
Fundamental Period (s)	0.85	0.85
Dominant Mode Type	Translational	Torsional

, including base shear capacity, roof displacement, and final performance classification. The calibrated DT results indicate a LS performance level in the X direction and a response approaching the CP threshold in the Y direction. These outcomes bridge the gap between nominal design-based expectations and actual field-measured behavior, highlighting the importance of calibration for realistic performance assessment [75,76,77].

Overall, the analyses indicate a substantial reduction in functional performance under DBE-level ground motion, accompanied by elevated deformation demands in critical structural components. These findings emphasize the role of field-calibrated DT models in providing realistic inputs for post-earthquake performance evaluation [78,79].

3.4. Comparison with Uncalibrated Model

The calibrated DT model was systematically compared with an uncalibrated baseline configuration adopting nominal C20 concrete strength and original design-based reinforcement detailing. The uncalibrated model represents a conventional as-designed analytical approach and does not explicitly incorporate field-measured material degradation or in-situ heterogeneity. Comparative pushover analyses reveal distinct differences in stiffness, capacity, and performance classification between the two models, consistent with the modeling framework described in Section 2.4.

The uncalibrated model exhibited a steeper initial stiffness, corresponding to a shorter fundamental period (0.72 s), and approximately 15–20% higher base shear capacity, while underestimating roof displacements by about 22%. These trends align with prior research indicating that nominal material assumptions in as-designed RC models may lead to unconservative performance evaluations when material degradation and stiffness reduction are not explicitly considered [80,81].

A quantitative comparison is provided in

Table 7. Comparison of calibrated vs. uncalibrated performance indicators.

Parameter	Uncalibrated Model	Calibrated DT	Difference (%)
Fundamental Period (s)	0.72	0.85	+18.1
Max. Base Shear (kN)	2280	1980	−13.2
Roof Displacement (m)	0.13	0.16	+22.0
Performance Level	LS	Near-CP	—

. While the uncalibrated model satisfied the LS performance criteria in both directions, the calibrated DT response in the Y direction was positioned closer to the CP threshold. This divergence reflects the influence of field-measured material degradation and realistic stiffness reduction on performance classification.

In addition, the plastic hinge distribution differed noticeably between the two models. The uncalibrated configuration predominantly exhibited beam-dominated hinging, whereas the calibrated DT showed increased hinge concentrations in columns and shear walls, particularly near the elevator core. These patterns closely resemble damage mechanisms documented during the 2023 Kahramanmaraş earthquakes, indicating that field calibration influences not only global response metrics but also the representation of governing failure mechanisms [82,83,84].

Overall, the comparison demonstrates that calibration can substantially influence the interpretation of structural response in seismic performance assessment. While uncalibrated, design-oriented models provide a useful initial approximation, field-calibrated DT approaches yield performance predictions that more closely reflect in-situ material conditions. Recent studies similarly report that integrated workflows combining NDT, BIM, and DT technologies can reduce epistemic uncertainty and improve the reliability of seismic safety evaluations for existing RC buildings [85,86].

4. Discussion

4.1. Principal Findings

The principal finding of this study is that the calibrated DT predicts a systematically lower seismic performance level compared with the uncalibrated, design-based analytical model. While the nominal model indicates that the structure satisfies the LS performance level, the calibrated DT response in the Y direction approaches the CP threshold.

This divergence highlights the limitations of assuming that nominal design parameters remain representative of structural behavior several decades after construction. The results suggest that reliance on design-based material properties may lead to overestimation of stiffness and strength, potentially resulting in unconservative performance classifications for existing RC buildings.

Directional differences observed between the X and Y axes—primarily associated with torsional asymmetry and nonuniform material degradation—further emphasize the importance of incorporating field-measured data into performance-based seismic assessment workflows. This issue is particularly relevant for irregular RC buildings, where geometric eccentricities can amplify the effects of material heterogeneity on global structural response.

Figure 11 presents a synthesized overview of the analytical workflow, illustrating how progressive calibration of material properties enhances model fidelity, influences predicted failure mechanisms, and leads to shifts in performance classification.

4.2. Comparison with Existing Literature

The findings of this study are consistent with previous research demonstrating that uncalibrated NDT methods tend to overestimate in-situ compressive strength in RC buildings [56,57,58]. The site-specific WinSonReb regression model developed herein supports this established understanding by quantifying the magnitude of overestimation and highlighting the role of multi-parameter, field-derived calibration in improving strength estimation reliability. These observations are in line with the recommendations of Musella et al. [44], who emphasized the importance of regression-based refinement for robust in-situ concrete strength assessment.

Beyond material strength estimation, the present study extends prior work by demonstrating that calibration influences not only predicted strength values but also the representation of governing failure mechanisms. While uncalibrated models predominantly indicated beam-dominated hinging, the calibrated DT captured a shift toward column- and wall-controlled failure modes—particularly around the elevator core—consistent with field observations reported after the 2023 Kahramanmaraş earthquakes [73,74].

This combined contribution—confirming known NDT overestimation tendencies while illustrating the behavioral implications of field calibration—represents a focused addition to the existing body of knowledge. In parallel, the proposed BIM→DT workflow aligns with the emerging literature emphasizing interoperability, semantic enrichment, and empirical validation in seismic performance modeling [41]. Importantly, the present study contributes one of the relatively limited number of documented field-calibrated case studies involving a severely damaged RC building with an elevator shear wall core, a structural configuration that remains underrepresented in current research. A comparative summary of the present study with selected key literature, highlighting methodological focus and distinctive contributions, is provided in

Table 8. Comparison of this study with selected key literature.

Reference	Focus	Key Findings	Distinction of This Study
[56,57,58]	NDT overestimation in RC structures	Uncalibrated NDT yields 15–25% higher predicted strength	Confirms and quantifies bias through WinSonReb calibration
[44]	Regression calibration post-earthquake	Regression improves correlation but limited to strength refinement	Extends calibration to behavioral prediction (failure mode shift)
[73,74]	Field-observed RC damage patterns	Column and wall hinging near cores	Replicated via DT-based calibrated analysis
[86]	BIM–DT integration	Conceptual frameworks for data interoperability	Provides validated, field-calibrated empirical workflow

.

4.3. Implications of the Study

4.3.1. Practical Implications

The findings hold significant implications for engineers, local authorities, and policymakers responsible for post-earthquake safety evaluations. Rapid but uncalibrated assessments risk misclassifying damaged structures as safe, potentially resulting in life-threatening consequences. By contrast, the calibrated DT workflow—while requiring additional field measurements and computational effort—provides conservative and reliable performance classifications that prioritize safety and inform evidence-based retrofit decisions.

4.3.2. Theoretical Implications

From a theoretical standpoint, the study contributes to the BIM-to-DT paradigm by demonstrating that field calibration transforms a design-based BIM model into an as-is digital representation capable of capturing both material heterogeneity and modal behavior. This dual capability bridges the gap between empirical field investigations and analytical seismic performance evaluation, enriching the methodological tools available to researchers and practitioners.

Figure 12 presents a visual summary of these dual implications, highlighting the intersection between practical safety assurance and theoretical advancement.

4.4. Limitations of the Study

A formal sensitivity analysis on the regression coefficients was not conducted within the scope of this study. However, the observed proximity of the calibrated response to the LS–CP boundary, together with the consistently high coefficient of variation (CoV > 20%) obtained from core tests, suggests that moderate variations in regression parameters would not qualitatively alter the identified performance shift. The primary outcome of the study is, therefore, not dependent on a specific regression coefficient value, but rather on the systematic incorporation of field-measured material heterogeneity into the analytical model.

An additional limitation pertains to data interoperability during BIM-to-DT transfer. While the IFC4 schema was used to ensure platform independence, certain custom Psets containing calibrated NDT-based material parameters were not fully preserved during export. These attributes required manual verification and reassignment to maintain analytical fidelity, following a predefined and traceable reassignment protocol. This limitation reflects current constraints in IFC-based semantic enrichment rather than a deficiency of the proposed workflow, and highlights the need for more robust interoperability standards for performance-oriented digital twin applications.

Several limitations should be acknowledged.

First, the proposed workflow was demonstrated on a single mid-rise RC building with an elevator shear wall core, which may limit the direct generalizability of the numerical results to other structural typologies. However, the primary objective of the study is to demonstrate the feasibility and analytical implications of a field-calibrated BIM–DT workflow rather than to provide statistically generalized performance metrics.

Second, the testing campaign utilized a restricted subset of available NDT techniques—specifically, the Schmidt hammer, UPV, and Windsor probe. Incorporating additional diagnostics such as ground-penetrating radar (GPR), infrared thermography, or acoustic tomography could enhance the depth and robustness of the calibration process.

Finally, AI-driven automated calibration and data-fusion techniques were not implemented, reflecting a deliberate scope decision aimed at maintaining methodological transparency. These limitations do not undermine the practical or scientific merit of the proposed workflow but should be considered when extending the framework to broader applications.

4.5. Future Research Directions

Future research should aim to validate and extend the proposed workflow across a broader range of structural typologies, including high-rise RC cores, masonry-infilled frames, and precast systems. Expanding the range of NDT techniques and integrating AI- or ML-based algorithms for automated calibration and model updating may greatly enhance efficiency and enable near-real-time post-earthquake assessment.

At a regional scale, federated and scalable digital twin frameworks could be integrated into urban resilience platforms, forming dynamic “resilience dashboards” to support emergency preparedness, rapid assessment, and resource allocation. Extending digital twin concepts beyond individual structures to interconnected urban systems could fundamentally reshape how cities prepare for, respond to, and recover from seismic events, providing a foundation for data-driven and adaptive seismic resilience planning in the built environment.

Beyond seismic damage, the proposed field-calibrated BIM–DT framework is also well suited for extension to fire-damaged reinforced concrete structures. Experimental studies on full-scale RC columns exposed to elevated temperatures have demonstrated that thermal damage induces significant microstructural transformations, including portlandite decomposition, C–S–H degradation, crack formation, and increased porosity, which collectively lead to severe reductions in residual mechanical properties [87]. Such physicochemical changes are strongly depth-dependent and often remain undetected by visual inspection alone, necessitating advanced diagnostic interpretation.

In this context, NDT techniques such as UPV and rebound hammer testing—when properly calibrated—can provide indirect indicators of thermally induced stiffness loss and material degradation, while core-based investigations may capture localized strength deterioration associated with fire exposure. Integrating these post-fire diagnostics into a Level-2 digital twin framework would allow thermally induced material heterogeneity to be explicitly embedded into seismic performance models, enabling coupled fire–earthquake damage assessments. This multi-hazard extension represents a promising pathway for post-event structural forensics and resilience-oriented decision-making.

5. Conclusions

This study proposed and demonstrated an integrated, field-calibrated BIM–DT framework for post-earthquake seismic performance assessment of RC buildings. Destructive core tests and complementary NDT methods—namely Schmidt rebound hammer, UPV, and Windsor probe measurements—were systematically combined to develop a site-specific WinSonReb calibration model. The calibrated material properties were embedded into a BIM environment to generate an as-is, Level-2 static digital twin that explicitly represents the in-situ material condition and heterogeneity of an earthquake-damaged structure.

The primary conclusion of the study is that field-based calibration fundamentally alters seismic performance predictions. While the uncalibrated, design-oriented model indicated that the case study building would satisfy the LS performance level under DBE loading, the calibrated digital twin revealed behavior approaching CP, particularly in the Y direction. This divergence highlights the inherent risk of relying on nominal material properties and idealized design assumptions when evaluating existing RC buildings subjected to material degradation and construction-era deficiencies. Without calibration, seismic performance may be systematically overestimated, leading to unconservative safety classifications.

From a practical perspective, the findings demonstrate that integrating field-calibrated material data—despite requiring additional testing effort and computational processing—yields more conservative and, therefore, more reliable performance assessments. Such calibrated DT-based evaluations offer clear value for engineers, local authorities, and post-earthquake decision-makers by improving the realism of safety evaluations and supporting evidence-based retrofit or occupancy decisions. The results further confirm that calibrated digital twins can reproduce observed damage mechanisms and deformation patterns with high fidelity.

From a theoretical standpoint, this study advances the BIM-to-DT paradigm by showing that calibration influences not only material strength estimation but also global structural response characteristics, including stiffness degradation, fundamental period elongation, torsional effects, and shifts in predicted failure mechanisms. By embedding empirically derived material properties into the analytical model, the proposed framework reduces epistemic uncertainty and strengthens the linkage between field investigations and performance-based seismic assessment.

Several limitations should be acknowledged. The methodology was demonstrated on a single mid-rise RC building with an elevator shear wall core, and the testing campaign relied on a limited subset of NDT techniques without AI-driven automation. Nevertheless, the objective of the study was to validate the analytical implications and feasibility of a field-calibrated BIM–DT workflow rather than to provide statistically generalized conclusions. Future research may extend the framework to diverse structural typologies, incorporate additional diagnostic tools, and explore automated calibration and data-fusion strategies using machine-learning approaches. At a broader scale, federated digital twin implementations could support urban resilience planning by linking building-level diagnostics with regional seismic risk management systems.

In summary, the integration of destructive and nondestructive field measurements with BIM-based digital twin modeling provides a robust, field-informed pathway for post-earthquake assessment of RC buildings. By bridging the gap between theoretical design assumptions and actual in-situ behavior, the proposed methodology enhances the credibility of seismic safety evaluations and contributes to the evolving body of knowledge in digital construction, structural diagnostics, and earthquake engineering.