The core problem the researchers tackled is the nonlinear static analysis known in the engineering community as pushover analysis. In a pushover analysis, a finite element model of a building is subjected to a progressively increasing lateral load that mimics the inertial forces of an earthquake, and the response is tracked until key structural members reach their deformation limits. The procedure produces a capacity curve, which is typically simplified into a bilinear acceleration-displacement relationship characterized by an equivalent vibration period, a yield acceleration and an available ductility. These three quantities, evaluated for both the Damage Limitation and Severe Damage limit states and for different lateral load distributions, determine whether a building can survive a design-level earthquake. The difficulty is that each pushover analysis demands an hour or more of computation on a nonlinear finite element model, which makes portfolio-level assessment of millions of buildings an intractable task.

The team’s answer is a hybrid deep neural network that fuses two architectural paradigms: Long Short-Term Memory networks, which excel at modeling sequential dependencies, and Convolutional Neural Networks, which specialize in spatial pattern recognition. The central insight behind the design is that the seismic behavior of a frame building depends simultaneously on how forces propagate vertically through the storeys and on how stiffness and strength are distributed spatially across the floor plan. Neither an LSTM alone nor a CNN alone can capture both aspects, so the researchers constructed an architecture in which fifty parallel LSTM branches process the columns of the building column by column, storey by storey, while a downstream CNN analyzes the spatial arrangement of their outputs.

The input representation is one of the most innovative features of the study. Each building is described by three complementary sources of information: nine global parameters characterizing the structure as a whole, such as the number of storeys and material properties; eight local parameters for each individual reinforced concrete column, including its lateral stiffness, plastic shear capacity and rotational capacities in both plan directions; and a mask that encodes the building’s layout on a 50 by 50 grid spanning a 50 meter square, where each one meter cell records whether a column is present. Because the columns are aligned across storeys, a single mask suffices for the whole building. Every column along a given direction of the plan is represented as a 400-dimensional feature vector, and the stack of these vectors across the storeys forms a sequence whose length equals the number of storeys, allowing the LSTM branches to learn how structural effects accumulate from the foundations upward.

The fifty LSTM outputs are then arranged as columns of a new two-dimensional matrix and combined with the layout mask to form a two-channel image, which is processed by a convolutional network with two convolutional layers and max-pooling stages. The CNN output is concatenated with the nine global building parameters and passed through fully connected layers that produce twenty regression targets: the equivalent period, yield acceleration and kinematic ductility for both limit states, across four pushover directions. The authors emphasize that the mask acts as a spatial filter, telling the network where real columns exist and filtering out the empty background of the grid, which turns out to be critical for performance. A masking layer inside each LSTM branch similarly ignores padded entries, which is what allows the same architecture to process buildings with any number of storeys without modification.

Training the network required a substantial synthetic dataset. The researchers generated 22,576 finite element models of reinforced concrete frame buildings representative of Italian construction practice after World War II, when structures were designed for gravity loads only under outdated codes. Using the OpenSEES-Py framework and high performance computing resources with AMD EPYC processors, they carried out parallel pushover analyses on buildings of two to five storeys, a range that reflects the real Italian building stock, in which two-storey and three-storey structures respectively account for 44 and 26 percent of reinforced concrete buildings. The simulated design procedure reproduced the allowable stress method, typical gravity loads and the single concrete class R’bk 250 characteristic of that era, giving the dataset a realistic grounding in a vulnerable and widespread class of structures.

The trained model, which contains roughly 8.5 million parameters and occupies about 100 megabytes, achieved errors that the authors describe as negligible for practical engineering purposes. On the test set, the mean absolute error for the equivalent period averaged about 0.022 seconds, while the yield acceleration errors averaged around 0.013 to 0.014 meters per second squared, or roughly 0.0014 g, and ductility errors averaged about 0.014. Crucially, the model generalized across building heights despite an intentionally unbalanced dataset: the average error for two-storey buildings was nearly identical to that for five-storey buildings, indicating that the network did not simply overfit the majority class. Because the architecture uses padding and masking rather than a fixed sequence length, the authors note that the design could in principle handle single-storey or taller buildings as well, although accuracy outside the validated two-to-five storey range has not yet been empirically demonstrated.

The study went well beyond a single training run. Repeating the entire training procedure with five additional random seeds produced almost identical results with low standard deviations, demonstrating that the reported performance is not the product of a lucky initialization. A stratified five-fold cross-validation confirmed the stability of the model across different data partitions. An ablation study isolated the contribution of each component: removing the masking mechanism increased the average error from 1.56 to 9.97, removing the LSTM branches raised it to 9.68, and removing the CNN caused the largest degradation, to 14.79, confirming that all three elements are essential to the architecture’s synergy. The hybrid model was also benchmarked against five baselines, including a standard deep neural network, a one-dimensional CNN, a standalone LSTM, XGBoost and a Transformer encoder. It achieved the lowest error on nearly all twenty outputs, outperforming the best baseline by approximately nineteen percent, and differential Wilcoxon signed-rank tests confirmed that the improvements over the CNN, LSTM and Transformer baselines were statistically significant across every output.

The researchers are careful to frame the tool’s role. It is not intended to replace finite element analysis of an individual structure, and practitioners will still rely on professional nonlinear analysis software when designing retrofit interventions for a specific building. What the network offers instead is a fast, interpretable second opinion, capable of checking the order of magnitude of pushover results and, more importantly, enabling vulnerability assessments at a scale that traditional methods cannot reach. Because the outputs are the same bilinear capacity parameters used in code-based performance assessment, the predictions feed directly into regional risk frameworks without requiring qualitative guesswork.

The implications extend well beyond Italy. Most of the world’s seismically exposed cities contain large stocks of pre-modern-code reinforced concrete buildings, from Los Angeles to Istanbul to Kathmandu, and the ability to triage thousands of structures rapidly could transform how governments prioritize retrofit funding before the next major earthquake. The Salerno team suggests that extending the dataset to a broader range of structural configurations and validating out-of-distribution generalization are the next steps, but the study makes a persuasive case that hybrid LSTM-CNN architectures, guided by simple geometric masks, can compress an hour of nonlinear mechanics into a fraction of a second, bringing large-scale seismic resilience assessment within practical reach.

Subject of Research: A mask-guided hybrid LSTM-CNN neural network for rapid seismic assessment of existing reinforced concrete frame buildings

Article Title: A mask-guided multi-LSTM-CNN architecture for rapid seismic analysis of existing RC frames

Article References: D’Angelo, G., Nigro, F., & Martinelli, E. (2026). A mask-guided multi-LSTM-CNN architecture for rapid seismic analysis of existing RC frames. Neural Computing and Applications, 38(18), Article 741. https://doi.org/10.1007/s00521-026-12458-5

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12458-5

Keywords: seismic assessment, reinforced concrete frames, pushover analysis, LSTM networks, convolutional neural networks, deep learning, structural engineering, finite element method, earthquake engineering, neural network architecture, seismic retrofitting, machine learning

The research, conducted by Resat Oyguc and Evrim Oyguc of the Institute of Disaster Management, Earthquake Engineering Department at Istanbul Technical University, focuses on a class of buildings that keeps structural engineers awake at night: non-ductile reinforced concrete frames. These are structures designed before modern seismic codes demanded the careful reinforcement detailing that allows a building to bend and deform without collapsing. Many such buildings across seismically active regions, from Türkiye to California, remain in service today, and retrofitting them has become a global engineering priority.

Two of the most common retrofit strategies anchor the study. The first involves wrapping columns with bonded carbon fiber-reinforced polymer, or CFRP, jackets—thin, immensely strong composite sheets that confine the concrete and delay brittle failure. The second deploys steel buckling-restrained braces, ingenious devices in which a steel core designed to yield in both tension and compression is housed within a tube that prevents it from buckling, allowing the brace to dissipate seismic energy stably through hundreds of loading cycles. Both approaches can dramatically improve a frame’s performance, but they do so in fundamentally different ways, and this difference has long frustrated attempts at unified assessment.

The problem, as the authors explain, lies in the very design of existing damage indices. Scalar indices calibrated against one intervention family record how much damage has occurred rather than which mechanism produced it. An index tuned to the behavior of CFRP-retrofitted frames transfers poorly to braced frames, and neither can sound an alarm when a brittle mode—such as joint panel shear failure or sliding—is about to take control of the response. In a severe earthquake, that distinction can mean the difference between a repairable structure and a catastrophic collapse.

The researchers’ answer is the Mechanism Evolution Index, or MEI, a framework that condenses six independently measurable response quantities into a bounded scalar on the unit interval, while simultaneously decomposing the structural state into a ternary coordinate that separates demand into three interpretable channels: ductility, shear, and connection. The six constituents include quantities familiar to any experimentalist: the normalized hysteretic energy dissipated during a run, the ratio of peak base shear to design-level capacity, the peak interstorey drift ratio, the residual drift left after shaking stops, the dynamic amplification between the shaking table and the structure above it, and the period softening ratio that captures how much the structure’s fundamental frequency has degraded from one excitation to the next.

What makes the framework remarkable is its economy of calibration. The six constituents are combined through a single frozen weight set, anchored once at the joint-panel cracking and sliding thresholds of a three-specimen calibration series. Each constituent passes through a sigmoid normalization function whose half-saturation value is fixed to the measured response of the bare specimen at design intensity. Once these anchors are set, the weights are frozen and applied without any re-tuning across retrofit type and shaking intensity. This is a deliberate departure from the prevailing practice, in which every new structural configuration demands its own bespoke index calibration, rendering comparisons across interventions nearly impossible.

The ternary coordinate is where the index earns its diagnostic power. The three components—ductility, shear, and connection—sum to one, placing every response state as a point within a triangular coordinate space whose corners represent pure ductile, pure shear, and pure connection demand. As a frame is shaken progressively harder, its point traces a trajectory through this space, and the leading axis at any moment names the mechanism that governs. The framework also defines a dominance margin, the difference between the two largest coordinates, which quantifies how decisively one mechanism leads over the others—a measure of confidence in the diagnosis itself.

To build and test the framework, the researchers drew on a one-third-scale specimen series: a bare non-ductile frame, the same frame retrofitted with a three-ply CFRP jacket, and a version strengthened with an innovative tube-in-tube buckling-restrained brace. All three configurations were reproduced by a single calibrated OpenSees finite element model, whose joint hysteresis was tuned to match measured displacement traces, base shear histories, hysteretic energy, and the fundamental periods identified from white-noise excitation before and after each strong-motion run. The calibration objective weighted all four error measures simultaneously, ensuring the simulation platform captured not just peak responses but the degradation pathways that unfold between them.

The findings tell a story that should concern anyone responsible for retrofit decisions. Under design-level intensity, the bare frame became connection-dominated, its ternary point migrating toward the joint sliding threshold exactly as its unreinforced beam-column panels began to fail. The CFRP jacket, perhaps counterintuitively, suppressed joint engagement through the design band but then collapsed through the very same connection channel when pushed to the ultimate level—the strengthening shifted the brittle mode upward in intensity rather than eliminating it. Only the braced frame remained on the ductility axis throughout, converting the failure mode into a stable, energy-dissipating one. The connection component of the index thus serves as an indicative retrofit-screening signal, flagging interventions that leave a residual brittle pathway even when headline performance metrics look satisfactory.

Transferability—the acid test for any index claiming generality—was examined on six external cases lying outside the calibration set, four bare and two retrofitted. In every labelled case, the mechanism reported by the experimental or analytical source landed on the leading ternary axis of the index. The classical scalars, by contrast, tracked damage severity faithfully but could not name the mechanism in any of them. This separation between knowing how much and knowing what kind is, in the authors’ framing, the essential contribution of the work.

The practical implications extend well beyond the laboratory. Post-earthquake reconnaissance following recent sequences in Türkiye has repeatedly documented the lethal consequences of brittle joint failures in non-ductile frames, and retrofit campaigns guided only by severity-based indices may inadvertently favor interventions that mask rather than remove the underlying vulnerability. A screening tool that can be applied identically across bare, jacketed, and braced configurations, using response quantities measurable from instrumentation or extracted from a calibrated simulation, offers engineers a common currency for comparing retrofit options on the dimension that matters most in a collapsing building: which mechanism will win.

The authors are careful to delineate the scope of their framework. It applies to regular planar systems under single-direction motion, and the frozen weight set is anchored to joint-panel thresholds specific to the calibration series used. Extending the approach to three-dimensional response, torsionally irregular buildings, and bidirectional ground motion remains open. The data supporting the study are available from the corresponding author upon reasonable request, and the authors report no external funding or competing interests. Even within these bounds, however, the Mechanism Evolution Index represents a meaningful conceptual advance: a damage measure that evolves alongside the structure, tracing not just the accumulation of harm but the shifting identity of the mechanism inflicting it.

As cities with vast stocks of pre-modern-code concrete buildings confront escalating seismic risk, the difference between a retrofit that redirects failure into a ductile pathway and one that merely postpones a brittle one is not an academic distinction—it is a matter of life and death. By giving engineers a single bounded number paired with an interpretable mechanism diagnosis, the Istanbul Technical University team has provided a tool that could make that distinction visible before the next earthquake makes it tragically apparent.

Subject of Research: A mechanism evolution index for the seismic assessment of retrofitted non-ductile reinforced concrete frames

Article Title: A mechanism evolution index for the seismic assessment of retrofitted non-ductile RC frames

Article References: Oyguc, R., & Oyguc, E. (2026). A mechanism evolution index for the seismic assessment of retrofitted non-ductile RC frames. Bulletin of Earthquake Engineering. https://doi.org/10.1007/s10518-026-02679-3

Image Credits: AI Generated

DOI: 10.1007/s10518-026-02679-3

Keywords: seismic retrofitting, reinforced concrete, damage index, ductility, CFRP, buckling-restrained brace, OpenSees, non-ductile frames, beam-column joints, shake table testing, earthquake engineering, structural assessment