Simple Model Research Articles

Pre- and post-surgery: advancements in artificial intelligence and machine learning models for enhancing patient management in infective endocarditis.

Infective endocarditis (IE) is a severe infection of the inner lining of the heart, known as the endocardium. It is characterized by a range of symptoms and has a complicated pattern of occurrence, leading to a significant number of deaths. IE poses significant diagnostic and treatment difficulties. This evaluation examines the utilization of artificial intelligence (AI) and machine learning (ML) models in addressing information extraction (IE) management. It focuses on the most recent advancements and possible applications. Through this paper, we observe that AI/ML can significantly enhance and outperform traditional diagnostic methods leading to more accurate risk stratification, personalized therapies as well and real-time monitoring facilities. For example, early postsurgical mortality prediction models like SYSUPMIE achieved 'very good' area under the curve (AUROC) values exceeding 0.81. Additionally, AI/ML has improved diagnostic accuracy for prosthetic valve endocarditis, with PET-ML models increasing sensitivity from 59% to 72% when integrated into ESC criteria and reaching a high specificity of 83%. Furthermore, inflammatory biomarkers such as IL-15 and CCL4 have been identified as predictive markers, showing 91% accuracy in forecasting mortality, and identifying high-risk patients with specific CRP, IL-15, and CCL4 levels. Even simpler ML models, like Naïve Bayes, demonstrated an excellent accuracy of 92.30% in death rate prediction following valvular surgery for IE patients. Furthermore, this review provides a vital assessment of the advantages and disadvantages of such AI/ML models, such as better-quality decision support approaches like adaptive response systems on one hand, and data privacy threats or ethical concerns on the other hand. In conclusion, Al and ML must continue, through multi-centric and validated research, to advance cardiovascular medicine, and overcome implementation challenges to boost patient outcomes and healthcare delivery.

WONOEP appraisal: Modeling early onset epilepsies.

Epilepsy has a peak incidence during the neonatal to early childhood period. These early onset epilepsies may be severe conditions frequently associated with comorbidities such as developmental deficits and intellectual disability and, in a significant percentage of patients, may be medication-resistant. The use of adult rodent models in the exploration of mechanisms and treatments for early life epilepsies is challenging, as it ignores significant age-specific developmental differences. More recently, models developed in immature animals, such as rodent pups, or in three-dimensional organoids may more closely model aspects of the immature brain and could result in more translatable findings. Although models are not perfect, they may offer a more controlled screening platform in studies of mechanisms and treatments, which cannot be done in pediatric patient cohorts. On the other hand, more simplified models with higher throughput capacities are required to deal with the large number of epilepsy candidate genes and the need for new treatment options. Therefore, a combination of different modeling approaches will be beneficial in addressing the unmet needs of pediatric epilepsy patients. In this review, we summarize the discussions on this topic that occurred during the XVI Workshop on Neurobiology of Epilepsy, organized in 2022 by the Neurobiology Commission of the International League Against Epilepsy. We provide an overview of selected models of early onset epilepsies, discussing their advantages and disadvantages. Heterologous expression models provide initial functional insights, and zebrafish, rodent models, and brain organoids present increasingly complex platforms for modeling and validating epilepsy-related phenomena. Together, these models offer valuable insights into early onset epilepsies and accelerate hypothesis generation and therapy discovery.

A two‐step sizing method for multistage Op Amps based on behavioral initial sizing followed by Spice‐in‐the‐loop refining

AbstractAnalog integrated circuit sizing is a laborious process that requires many times of iteration with Spice simulations. Even by applying the method, it still requires iterations and test assignment of device values (and biasings) in each iteration until reaching a satisfactory sizing result. When facing a design of multiple‐stage circuits (such as multiple‐stage operational amplifier [Op Amp]), sizing becomes more challenging due to the requirement on pole‐zero placement in order to achieve a better high‐frequency performance. Simulation‐in‐the‐loop method has been the dominating means taken by a great many of existing circuit sizer, but they suffer from intolerable runtime while lacking the possibility of offering insight or design knowledge acquisition. On the other hand, behavioral‐level synthesis methods, although intuitive and fast, suffer from accuracy loss due to the adoption of simplified device models and significantly condensed design equations. However, a proper combination of these two types of methods could lead to a lucrative research territory, yet it demands a methodological development for efficient deployment in practice, namely, implementation easy, less runtime, and guaranteed sizing quality. In this paper, we propose a two‐step method that takes the advantage of behavioral synthesis (in the first step) that is capable of fast and broader coverage of design space then makes correction (in the second step) on sizing refining by Spice simulation. Due to the profiling knowledge acquired on the design space in the first step, it can avoid blindness of the optimization landscape that is faced by many simulation‐centric methods which could easily get trapped in local optima. Conducted experiments over eight three‐stage Op Amps with different compensation configurations, including nested Miller capacitor (NMC), NMC with feedforward path (NMCF), NMC with feedforward path and nulling resistor (NMCFNR), NMC with nulling resistor (NMCNR), double pole‐zero cancellation (DPZC), transconductance with capacitances feedback compensation (TCFC), impedance adapting compensation (IAC), active feedback frequency compensation (AFFC), have validated that the proposed method can successfully generate qualified sizing results, offering an opportunity to compare fairly what merit a specific compensation strategy could possibly have.

Probabilistic ductile deformation limit state prediction of monolithic exterior shear keys based on quantile regression machine learning techniques

Shear strength and ductile deformation capacity of monolithic exterior shear keys play important roles on the seismic behavior of highway bridges. Improving the ductility of the shear key through reasonable design could effectively restrict superstructure displacement and minimize damage to pier columns. However, most existing studies primarily focus on evaluating the shear strength and damage mechanisms of shear keys, with limited attention given to analytical models for assessing ductile deformation capacity. To this end, this study proposes a method that combines quantile regression and machine learning techniques to predict the ductile deformation limit states of shear keys. The proposed numerical modeling method for shear keys is validated using the experimental data on lateral load-displacement curves. On this basis, Latin hypercubic sampling and joint simulation are utilized to establish the ductile deformation limit state database. The ductile deformation limit states of shear keys are characterized by their lateral displacements in three distinct damage states. The main influencing factors of each limit states are investigated by correlation analysis. Three quantile regression machine learning models are trained and constructed to conduct point and interval prediction for the ductile deformation limit states. All three models, particularly the Gaussian process quantile regression model, are proven to demonstrate exceptional prediction accuracy and uncertainty quantification. Experimental data validation confirms that these models outperform the prediction accuracy of existing simplified models.

What factors preventing the older adults in China from living longer: a machine learning study

BackgroundThe fact that most older people do not live long means that they do not have more time to pursue self-actualization and contribute value to society. Although there are many studies on the longevity of the elderly, the limitations of traditional statistics lack the good ability to study together the important influencing factors and build a simple and effective prediction model.MethodsBased on the the data of Chinese Longitudinal Healthy Longevity Survey (CLHLS), 2008–2018 cohort and 2014–2018 cohort were selected and 16 features were filtered and integrated. Five machine learning algorithms, Elastic-Net Regression (ENR), Decision Tree (DT), Random Forest (RF), K-Nearest Neighbor (KNN), and eXtreme Gradient Boosting (XGBoost), were used to develop models and assessed by internal validation with CLHLS 2008–2018 cohort and temporal validation with CLHLS 2014–2018 cohort. Besides, the best performing model was explained and according to the variable importance results, simpler models would be developed.ResultsThe results showed that the model developed by XGBoost algorithm had the best performance with AUC of 0.788 in internal validation and 0.806 in temporal validation. Instrumental activity of daily living (IADL), leisure activity, marital status, sex, activity of daily living (ADL), cognitive function, overall plant-based diet index (PDI) and psychological resilience, 8 features were more important in the model. Finally, with these 8 features simpler models were developed, it was found that the model performance did not decrease in both internal and temporal validation.ConclusionsThe study indicated that the importance of these 8 factors for predicting the death of elderly people in China and built a simple machine learning model with good predictive performance. It can inspire future key research directions to promote longevity of the elderly, as well as in practical life to make the elderly healthy longevity, or timely end-of-life care for the elderly, and can use predictive model to aid decision-making.

A two-tier feature selection method for predicting mortality risk in ICU patients with acute kidney injury

Acute kidney injury (AKI) is one of the most important lethal factors for patients admitted to intensive care units (ICUs), and timely high-risk prognostic assessment and intervention are essential to improving patient prognosis. In this study, a stacking model using the MIMIC-III dataset with a two-tier feature selection approach was developed to predict the risk of in-hospital mortality in ICU patients admitted for AKI. External validation was performed using separate MIMIC-IV and eICU-CRD. The area under the curve (AUC) was calculated using the stacking model, and features were selected using the Boruta and XGBoost feature selection methods. This study compares the performance of a stacking model using two-tier feature selection with a model using single-tier feature selection (XGBoost: 85; Boruta: 83; two-tier: 0.91). The predictive effectiveness of the stacking model was further validated by using different datasets (Validation 1: 0.83; Validation 2: 0.85) and comparing it with a simpler model and traditional clinical scores (SOFA: 0.65; APACH IV: 0.61). In addition, this study combined interpretable techniques and causal inference to analyze the causal relationship between features and predicted outcomes.

Three-dimensional vibration analysis of multilayered composite and functionally graded piezoelectric plates and shells

In the present paper, a coupled 3D exact electro-elastic shell model for Functionally Graded (FG) and composite piezoelectric structures is proposed. Primary variables of the electro-elastic model are the electric potential and the three displacement components. The model allows to evaluate the piezoelectric effect in terms of frequencies and vibration modes. Both closed and open circuit configurations are analyzed and compared. The 3D equilibrium equations and the 3D divergence electric displacement equation for spherical shells give the set of partial differential equations for the electro-elastic problem. The proposed model for spherical shells automatically degenerates into simpler models for plates and cylindrical shells via properly considerations on radii of curvature along in-plane directions. The orthogonal mixed curvilinear coordinates α, β and z are employed. The partial differential governing equations have constant coefficients considering fictitious layers and they are solved using the Navier harmonic form and the exponential matrix method. These features lead to an exact solution for simply-supported boundary conditions. Free vibration analyses are conducted and circular frequencies for the first three thickness vibration modes are computed. After a global assessment phase to verify the correctness of the developed model, new benchmarks are proposed: different thickness ratios and material configurations are investigated. The present work can be intended as a reference general solution for those scientists interested in the study of piezoelectric structures via 2D analytical and numerical formulations.

Best-estimate water hammer simulations to avoid the calculation of unrealistically high loads or unphysical pressure and force peaks

For safety-related systems fluid loads due to fluid transients have to be quantified for subsequent structural analyses to ensure their integrity or function, as required. Usually transient fluid loads in pipe systems are determined with one-dimensional water hammer software. For single-phase liquid flow, the method of characteristics (MOC) is often used that gives in this case appropriate results. For the consideration of local vapor bubbles, the MOC is combined with the discrete vapor cavity model (DVC). The DVC model may generate unrealistic pressure spikes due to the calculation of the collapse of multi-cavities in scenarios, where only one vapor bubble should actually occur. The application of a two-phase code may improve the calculation results. One requirement for the latter codes is the ability to calculate the propagation of steep gradients without suffering from numerical diffusion to exclude the underestimation of fluid loads. This is commonly attained by applying higher-order numerical schemes. However, the application of a numerical method of pure 2nd order leads to the calculation of unphysical oscillations at steep gradients causing severe problems during the solution. To exclude this, numerical methods with flux limiters can be used. With their application, the calculation of unrealistically high loads due to numerical deficiencies can be minimized. In addition, the consideration of further physical effects, that lead to the reduction of loads during transient flow processes, allows for a more realistic calculation of the loads. These are unsteady friction, widening of the pipe caused by pressure increase, fluid-structure interaction at junctions and due to friction, degassing of gas that is initially dissolved physically in a liquid and thermodynamic non-equilibrium during vapor bubble collapse. The in-house code DYVRO applies a second-order accurate scheme with flux limiters based on the Godunov method and can account for the above-described physical phenomena. It is shown by comparison of calculation results obtained by DYVRO with experimental data from literature that with modeling of these physical effects the loads can be calculated more realistically. Generally, these loads are lower than the results calculated by simplified models, which do not account for these effects. Considering that these loads are applied in subsequent structural analyses, cost-intensive oversizing of pipes and their supports can be avoided, by ensuring the necessary safety.

Feasibility region analysis of active systems: investigating the lateral performance envelope of sports cars

Active systems are gaining significant importance because of their ability of enhancing handling performance, stability, and driver engagement, especially in sports cars. The upfront study of the achievable handling characteristics with a specific actuator set is crucial, given the involved development costs and time. In this context, although several recent studies have compared the cornering response associated with different chassis actuations, none of them has analysed in detail the resulting performance envelope, i.e. the feasibility region, based on the operating limit of the available hardware. This study targets the identified gap and presents a methodology for obtaining the feasibility region of chassis actuation systems, defined as the locus of the achievable understeer characteristics. The analysis considers front and rear active camber (FAC and RAC), rear-wheel-steering (RWS) and variable roll moment distribution (RMD). Simplified vehicle models with reduced number of degrees of freedom (DoFs) and other features to accurately obtain the feasibility region and the relevant dynamic response indicators in the early design stages of the vehicle are compared with a high-fidelity model based on VI-CarRealTime. The results highlight the favourable trade-off between complexity and accuracy of a 3-DoF nonlinear model accounting for suspension kinematics and compliances.

Identifying the influence of airbag structure on driver injury during a crash using a dummy model

This study undertakes the analysis of collision scenario using a car model with a dummy and airbags, in the event of a direct collision with a hard wall, one of the necessary studies of passive safety. To describe in detail, the input conditions, a simulation problem of the driver's seat displacements was performed and this displacements data was exported as boundary conditions for the collision simulation. The results simulation crash show that the calculated energy values and simulation results are approximately the same (7.381E+07 and 7.367E+07), energy is converted from kinetic energy into internal energy of the elements. The airbag deployment simulation results are similar to NHTSA's previous research, both in terms of graph shape and maximum value. The impact of the collision incident on the driver is not excessively large, as evidenced by surveys on head (HIC 300), thigh (F 2.8 kN), and neck (F3,098 kN; T 190 Nm) injuries. However, the study proceeds to further analyze and assess the airbag's structure, examining its influence on these metrics, concluding that changes in the exhaust valve size (increase from 1000 mm2 to 2000 mm2) lead to a reduction in the evaluated parameters. These results suggest changes to the airbag structure to enhance driver safety, as well as a simpler simulation model to save analysis time

An exploratory study on just-in-time multi-programming-language bug prediction

Context:An increasing number of software systems are written in multiple programming languages (PLs), which are called multi-programming-language (MPL) systems. MPL bugs (MPLBs) refers to the bugs whose resolution involves multiple PLs. Despite high complexity of MPLB resolution, there lacks MPLB prediction methods. Objective:This work aims to construct just-in-time (JIT) MPLB prediction models with selected prediction metrics, analyze the significance of the metrics, and then evaluate the performance of cross-project JIT MPLB prediction. Methods:We develop JIT MPLB prediction models with the selected metrics using machine learning algorithms and evaluate the models in within-project and cross-project contexts with our constructed dataset based on 18 Apache MPL projects. Results:Random Forest is appropriate for JIT MPLB prediction. Changed LOC of all files, added LOC of all files, and the total number of lines of all files of the project currently are the most crucial metrics in JIT MPLB prediction. The prediction models can be simplified using a few top-ranked metrics. Training on the dataset from multiple projects can yield significantly higherAUC than training on the dataset from a single project for cross-project JIT MPLB prediction. Conclusions:JIT MPLB prediction models can be constructed with the selected set of metrics, which can be reduced to build simplified JIT MPLB prediction models, and cross-project JIT MPLB prediction is feasible.

Non-Ideal Push–Pull Converter Model: Trade-Off between Complexity and Practical Feasibility in Terms of Topology, Power and Operating Frequency

Power converters are the basic elements of any power electronics system in many areas and applications. Among them, the push–pull converter topology is one of the most widespread due to its high efficiency, versatility, galvanic isolation, reduced number of switching devices and the possibility of implementing high conversion ratios with respect to non-isolated topologies. Optimal design and control requires very accurate models that consider all the non-idealities associated with the actual converter. However, this leads to the use of high-order models, which are impractical for the design of model-based controllers in real-time applications. To obtain a trade-off model that combines the criteria of simplicity and accuracy, it is appropriate to assess whether it is necessary to consider all non-idealities to accurately model the dynamic response of the converter. For this purpose, this paper proposes a methodology based on a sensitivity analysis that allows quantifying the impact of each non-ideality on the converter behaviour response as a function of the converter topology, power and frequency. As a result of the study, practical models that combine the trade-off between precision and simplicity are obtained. The behaviour of the simplified models for each topology was evaluated and validated by simulation against the most complete and accurate non-ideal model found in the literature. The results have been excellent, with an error rate of less than 5% in all cases.

Structural literacy in architectural studio learning

ABSTRACT Architectural education encompasses a multidisciplinary curriculum including design, theory, history, technology, environmental considerations, urban planning, social aspects, and professional practice. This study aims to investigate the extent to which students can effectively integrate structural concepts into architectural design. This examination takes place within the framework of a third-year architectural design studio course, where students advance their design processes with input from architectural design instructors and experts in structural concepts. To facilitate this endeavour, a simple visual screening form was introduced to aid in the learning process. The study utilises a newly generated survey, namely ‘Structural System Control Form’, which is inspired by the rapid visual screening forms, to assess the students’ comprehension of structural concepts during a 14-week architectural design studio. Employing a descriptive and correlational research approach, the study assesses the effectiveness of this integrated strategy in enhancing the students’ understanding of structural principles. The findings revealed a notable development among students who participated in the integrated approach, highlighting the value of feedback and evaluation from instructors versed in construction. Furthermore, the use of simplified structural models in students’ projects improved their understanding of structural concepts and their ability to incorporate them into their architectural representations.

Drug-drug interactions and synergy: from pharmacological models to clinical application.

This review explores the concept of synergy in pharmacology, emphasizing its importance in optimizing treatment outcomes through the combination of drugs with different mechanisms of action. Synergy, defined as an effect greater than the expected additive effect elicited by individual agents according to specific predictive models, offers a promising approach to enhance therapeutic efficacy while minimizing adverse events. The historical evolution of synergy research, from ancient civilizations to modern pharmacology, highlights the ongoing quest to understand and harness synergistic interactions. Key concepts such as concentration-response curves, additive effects, and predictive models are discussed in detail, emphasizing the need for accurate assessment methods throughout translational drug development. While various mathematical models exist for synergy analysis, selecting the appropriate model and software tools remains a challenge, necessitating careful consideration of experimental design and data interpretation. Furthermore, this review addresses practical considerations in synergy assessment, including preclinical and clinical approaches, mechanism of action, and statistical analysis. Optimizing synergy requires attention to concentration/dose ratios, target site localization, and timing of drug administration, ensuring that the benefits of combination therapy detected at bench-side are translatable into clinical practice. Overall, the review advocates for a systematic approach to synergy assessment, incorporating robust statistical analysis, effective and simplified predictive models, and collaborative efforts across pivotal sectors such as academic institutions, pharmaceutical companies, and regulatory agencies. By overcoming critical challenges and maximizing therapeutic potential, effective synergy assessment in drug development holds promise for advancing patient care. Significance Statement Combining drugs with different mechanisms of action for synergistic interactions optimizes treatment efficacy and safety. Accurate interpretation of synergy requires the identification of the expected additive effect. Despite innovative models to predict the additive effect, consensus in drug interaction research is lacking, hindering the bench-to-bedside development of combination therapies. Collaboration among science, industry, and regulation is crucial for advancing combination therapy development, ensuring rigorous application of predictive models in clinical settings.

Performance of Simplified Damage-Based Concrete Models in Seismic Applications

Performance of Simplified Damage-Based Concrete Models in Seismic Applications

Comparison of Transcranial Magnetic Stimulation Dosimetry between Structured and Unstructured Grids Using Different Solvers.

In recent years, the interest in transcranial magnetic stimulation (TMS) has surged, necessitating deeper understanding, development, and use of low-frequency (LF) numerical dosimetry for TMS studies. While various ad hoc dosimetric models exist, commercial software tools like SimNIBS v4.0 and Sim4Life v7.2.4 are preferred for their user-friendliness and versatility. SimNIBS utilizes unstructured tetrahedral mesh models, while Sim4Life employs voxel-based models on a structured grid, both evaluating induced electric fields using the finite element method (FEM) with different numerical solvers. Past studies primarily focused on uniform exposures and voxelized models, lacking realism. Our study compares these LF solvers across simplified and realistic anatomical models to assess their accuracy in evaluating induced electric fields. We examined three scenarios: a single-shell sphere, a sphere with an orthogonal slab, and a MRI-derived head model. The comparison revealed small discrepancies in induced electric fields, mainly in regions of low field intensity. Overall, the differences were contained (below 2% for spherical models and below 12% for the head model), showcasing the potential of computational tools in advancing exposure assessment required for TMS protocols in different bio-medical applications.

Insight into adsorption behaviors of shale oil in kerogen slit by molecular dynamics

Shale oil is predominantly stored in organic nanopores, with kerogen playing a pivotal role in its adsorption. This study utilizes molecular dynamics (MD) simulations to quantitatively investigate the adsorption behavior of multi-component shale oil within kerogen slits. Unlike previous studies that focused on single-component hydrocarbon fluids and simplified kerogen models to graphene, this research employs an authentic kerogen structure and a multi-component shale oil model, incorporating the effects of surface roughness. The research analyzes the interaction mechanisms and distribution patterns of light, medium, and heavy components within the pores. Additionally, it examines the effects of temperature, pore size, and wall surface physical and chemical properties on shale oil adsorption. Results indicate that strong adsorption between asphaltene and kerogen significantly reduces the width of the kerogen slits, inhibiting the adsorption of lighter components. The study introduces the concept of effective radius for kerogen slits and provides a quantitative analysis of the adsorption capacity of kerogen surfaces. Different kerogen surface roughness and chemical properties were compared to understand their influence on shale oil adsorption. Key findings show that with increasing temperature, the distribution of hydrocarbons becomes more uniform, with medium components showing the most significant response. Surface roughness of kerogen affects the adsorption capacity, although the component proportions remain consistent, while different chemical property surfaces influence selective adsorption and the overall adsorption amount of shale oil components. This research offers a deeper understanding of shale oil adsorption mechanisms and provides valuable insights for optimizing shale oil extraction and utilization.

The role of the foundation flexibility on the seismic response of a modern tall building: Vertically incident plane waves

The soil-structure interaction (SSI) is an integral part of the seismic response of structures. The knowledge about its effects is based largely on numerical simulations involving simplified models. In this paper, we examine the consequence of several assumptions in SSI models of buildings, with emphasis on the rigid foundation assumption. To this end, we analyze several variants of a linear three-dimensional finite element model of the structure and soil corresponding to a uniquely instrumented 50-story skyscraper, with a basement on a pile foundation, located in southwest China. The models are excited by triaxial motion due to a vertically incident plane wave. The most realistic model (with basement and piles, both with bond contact with the soil), predicted an 8 % reduction of the apparent frequency and a 30 %–50 % increase in the apparent damping ratio of the 1st mode of vibration, relative to the model with fixed basement. The model with a rigid foundation (basement) embedded in the soil underestimated the reduction in apparent frequency by a factor of two. The SSI effects were the most pronounced for the model with flexible basement and no piles (16 %–19 % reduction in apparent frequency and 73 %–86 % increase in apparent damping ratio). The basement-structure interaction reduced the apparent frequency by 5 %–8 % depending on the direction. It is concluded that, for such a building, a detailed model of the structure and its foundation is needed to predict reliably the SSI effects.

Seismic behavior of high-rise modular buildings with simplified models of inter-module connections

Inter-module connections serve as crucial components, horizontally and vertically connecting individual modules to form modular buildings. This paper presents a component-based model that divides inter-module connections into vertical and horizontal connection components, facilitating the simplification of shear and axial behaviors for input into the global numerical model. Initially, experimental studies and numerical simulations were conducted to investigate the shear and axial behavior of these connection components, following which simplified connection component models were derived. Subsequently, a component-based model of inter-module connections was established, in which the shear and axial properties of the vertical and horizontal components were represented by nonlinear connector elements. Finally, the proposed component-based model was incorporated directly into the global frame models of high-rise steel modular buildings, and the seismic behavior was evaluated. The accuracy of the proposed component-based model was investigated by comparing the natural vibration characteristics and global building responses of high-rise steel modular buildings with traditional empirical models established based on assumptions of rigid or pinned fixities. The results reveal that traditional empirical models fail to accurately represent the discontinuous properties of inter-module connections, resulting in an overestimation of structural lateral stiffness. In contrast, the component-based model accounts for the actual load-transfer mechanisms of vertical and horizontal components, providing a more accurate understanding of the natural vibration characteristics and seismic response of high-rise steel modular buildings. The proposed component-based model establishes a critical link between isolated connection behavior and structural response assessment for high-rise steel modular buildings, making it well-suited for accurately predicting global building responses.

Interpretable machine learning models for displacement demand prediction in reinforced concrete buildings under pulse-like earthquakes

This work proposes a novel procedure to guide the development of machine learning models for estimating the seismic demand in existing reinforced concrete (RC) buildings. The proposed approach is organized across two scales. A large-scale (nonparametric) machine learning model is first obtained by means of Gaussian Process Regression (GPR) using all candidate building attributes and intensity measures. SHapley Additive exPlanations (SHAP) values are utilized to facilitate its interpretation and to assist the rational selection of a small subset of intensity measures, which is finally employed to develop a (symbolic) reduced-scale machine learning model by means of Genetic Programming (GP). Simplified models of archetype buildings are adopted to develop machine learning techniques at both scales, in such a way to alleviate the simulation time for preparing large datasets. Refined models representative of actual buildings are instead considered for the unbiased final assessment.The proposed approach is applied to develop predictive machine learning models for the maximum inter-storey drift in bare frames, pilotis frames and frames with infills under pulse-like seismic ground motions. Consequently, the critical examination of the SHAP values revealed the most significant intensity measures and unfolded interesting patterns depending on the occupancy rate of the infills. Moreover, the final assessment demonstrates that this approach allows the management of a non-homogeneous building stock consisting of very diverse structural systems (i.e., spanning from existing buildings designed against gravity loads only to buildings that comply with outdated seismic codes) while providing satisfactory predictions of the seismic demand with minimum computational effort.