System Model Research Articles

Abstract 4134460: Application of Principal Component Analysis to Heterogenous Fontan Registry Data to Identify Independent Contributing Factors to Decline

Introduction: Single ventricle heart disease is a severe and life-threatening illness, and improvements in clinical outcomes of those with Fontan circulation have not yet yielded acceptable survival over the past two decades. Patients are at risk of developing a diverse variety of Fontan-associated comorbidities that may ultimately require heart transplant. Goal: Our study goal was to determine if principal component analysis (PCA), which determines unique directions of data variance, applied to data collected from a pediatric Fontan cohort can predict functional decline (N=140). Methods: Heterogeneous data broadly consisting of measures of cardiac and vascular function, exercise (VO 2max ), lymphatic biomarkers, and blood biomarkers were collected over 9 years at a single site as part of a prospectively applied clinical care pathway; in that time, 16 events occurred that are considered here in a composite adverse outcome measure. After standardization and PCA, principal components (PCs) representing >5% of total variance were thematically labeled based on their constituents and tested for association with the composite outcome. Results: Our main findings suggest that the 6 th PC (PC6), representing 7.1% percent of the total variance in the set, is greatly influenced by albumin, alkaline phosphatase, total protein, BUN and BNP max and is a superior measure of proportional hazard compared to EF; it displayed the greatest accuracy for classifying Fontan patients as determined by AUC. In bivariate hazard analysis, we found that models combining systolic function (EF or PC5) and blood biomarkers (PC6) were most predictive, with the former having the greatest AIC, and the latter having the highest c-statistic. Conclusion: Our findings support the hypothesis that a broad, prospectively collected, multiorgan system model analyzed by way of unsupervised machine learning methods will improve prognosis of adverse outcomes in a Fontan cohort.

Abstract 4136975: Lipoprotein (a) Discovery Project: Assessing Patient and Professional Education Needs

Introduction: It is estimated that 1 in 5 Americans have high lipoprotein (a) [Lp(a)] levels. High levels of Lp(a) are an independent, predominantly inherited, and causal risk factor for atherosclerotic cardiovascular disease (ASCVD). More investigation is needed to better understand the gaps in both professional and patient education on Lp(a) and its associated risk with ASCVD. Goals: The American Heart Association (AHA) is implementing a national 3-year initiative called the Lp(a) Discovery Project to understand system-level practice patterns for patients screened for high Lp(a) and develop models for national education. We aim to gauge an understanding of healthcare professional (HCP) and patient awareness, perceptions, knowledge, and behaviors of Lp(a) to develop a comprehensive educational approach that ultimately encourages patient-provider shared decision-making. Methods: To assess patient education needs, an online survey was conducted from October 11 to November 6, 2023, by Harris Insights&Analytics LLC with a nationally representative sample of 3,006 U.S. adults 18 years+. To assess professional education needs, the AHA engaged 10 champions from diverse health systems with established Lp(a) screening processes and workflows to participate in a Learning Healthcare System model to share best practices, inform on models for Lp(a) screening, and provide input on professional education topics. Results: The patient survey revealed that 34% of adults do not perceive high Lp(a) to pose an increased risk of heart attack, peripheral artery disease or stroke. If told that high Lp(a) does increase this risk, 63% are motivated to have their Lp(a) level checked and 88% would be motivated to reduce their risk for these conditions if they had high Lp(a). Ultimately, HCP recommendation was the most significant driver in getting screened among those who had their Lp(a) measured (61%). AHA-led champion interviews revealed that primary and specialty care HCPs need education on the importance of Lp(a) testing and cascade screening for family members as well as care considerations for patients with high Lp(a). Conclusion: Given that patients are most likely to be screened for Lp(a) if their HCP recommends it, we will focus on 1) educating patients and HCPs on the connection between Lp(a) and risk for cardiovascular disease and stroke and 2) providing HCPs with guidance on care considerations for those with high Lp(a) including cascade screening.

Dynamical system model of gentrification: Exploring a simple rent control strategy

Dynamical system model of gentrification: Exploring a simple rent control strategy

Evaluating Quoddy Region archaeological site vulnerability to sea-level rise and erosion through the integration of geographic information system modeling and surveys

Modeling archaeological site erosion often depends on regional site databases that record sites accurately but with variable precision. This study examines the impact of sea-level rise (SLR) on 10 archaeological sites in the Quoddy Region of Maine through comparing models and field observations. Sites were categorized as low, mid, or high priority for field excavation based on exposure to tides. These model results were compared to field reports of site condition to evaluate the accuracy of modeling SLR as an indicator of erosion and to evaluate the application of models in developing prioritization protocols for site investigations. Models for current sea level scenarios broadly underestimate the degree of erosion reported by field observations because not all site locations were recorded at the precision required for analysis. This study emphasizes the importance of field audits for sites recorded in databases to enable large-scale modeling for the prioritization of urgently threatened sites.

A three-quantile bias correction with spatial transfer for the correction of simulated European river runoff to force ocean models

Abstract. In ocean or Earth system model applications, the riverine freshwater inflow is an important flux affecting salinity and marine stratification in coastal areas. However, in climate change studies, the river runoff based on climate model output often has large biases on local, regional, or even basin-wide scales. If these biases are too large, the ocean model forced by the runoff will drift into a different climate state compared to the observed state, which is particularly relevant for semi-enclosed seas such as the Baltic Sea. To achieve low biases in riverine freshwater inflow in large-scale climate applications, a bias correction is required that can be applied in periods where runoff observations are not available and that allows spatial transferability of its correction factors. In order to meet these requirements, we have developed a three-quantile bias correction that includes different correction factors for low-, medium-, and high-percentile ranges of river runoff over Europe. Here, we present an experimental setup using the Hydrological Discharge (HD) model and its high-resolution (1/12°) grid. First, bias correction factors are derived at the locations of the downstream stations with available daily discharge observations for many European rivers. These factors are then transferred to the respective river mouths and mapped to neighbouring grid boxes belonging to ungauged catchments. The results show that the bias correction generally leads to an improved representation of river runoff. Especially over northern Europe, where many rivers are regulated, the three-quantile bias correction provides an advantage compared to a bias correction that only corrects the mean bias of the river runoff. Evaluating two NEMO (Nucleus for European Modelling of the Ocean) model simulations in the German Bight indicated that the use of the bias-corrected discharges as forcing leads to an improved simulation of sea surface salinity in coastal areas. Although the bias correction is tailored to the high-resolution HD model grid over Europe in the present study, the methodology is suitable for any high-resolution model region with a sufficiently high coverage of river runoff observations. It is also noted that the methodology is applicable to river runoff based on climate hindcasts, as well as on historical climate simulations where the sequence of weather events does not match the actual observed history. Therefore, it may also be applied in climate change simulations.

Abstract 4145826: Atrial Fibrillation Screening During Sinus Rhythm Periods by Interrelated Systems Dynamics Analysis

Background: Previous studies have shown that AF screening in at-risk populations can reduce stroke incidence. However, non-targeted screening approaches often result in high false positive rates, placing an unnecessary burden on the healthcare system. In contrast, artificial intelligence-guided screening has been demonstrated to increase diagnostic yield in large prospective clinical trials. This approach, however, requires recording an ECG and a large-scale dataset for model training. Heart rate variability (HRV) analysis has proven effective in deciphering key heart dynamics. By analyzing HRV as interrelated dynamic systems, it may be possible to facilitate targeted AF screening using wearable devices that measure heart rate. The Koopman operator, used for data-driven modeling of interrelated dynamic systems, has been shown to accurately predict complex phenomena in chaotic systems such as climate forecasting and drug adverse reaction prediction. This is achieved by utilizing common characteristics of the systems for most model parameters, with only a small fraction of the parameters being specific to a certain system. Methods: Long (>10 hour) records from 361 individuals (AFDB, LTAFDB) and healthy individuals' datasets from PhysioNet and THEW were analyzed for inter-beat intervals. The unified dataset was then split into 94 training, 17 validation, and 250 test set patients. Recordings from the training set were used to train both the common and specific parts of the interrelated dynamic systems model for each patient, along with a shared small neural network classifying patients into low and high risk for AF based on the unique (not shared between patients) singular values of the dynamic system model. Patient-specific dynamic system models were then fitted for the validation and test sets to calculate the patient dynamic singular values, which were used to classify patients into low and high risk for AF groups. Results: Atrial fibrillation occurred in 48 of 202 (23%) patients classified as low risk and 35 of 48 (72.9%) patients classified as high risk (odds ratio 8.63, 95% CI 4.23-17.64), yielding 72.9% sensitivity with 76.2% specificity. Conclusion: In this retrospective analysis, classification of the dynamic system model singular values identified patients at high risk for atrial fibrillation from sinus rhythm period.

Abstract 4136995: Veterans Affairs (VA) Lipid Optimization Reimagined Quality Improvement Program (VALOR-QI) in Veterans with Atherosclerotic Cardiovascular Disease (ASCVD) at the San Juan VA

ASCVD is the principal cause of morbidity and mortality for Veterans in the VA Healthcare System. San Juan VA is one of 50 VA sites participating in a national quality improvement initiative called the VALOR QI: VA Lipid Optimization Reimagined Quality Improvement Project. Is a collaborative project between the U.S. Department of Veterans Affairs (VA) and the American Heart Association (AHA) with the goal of positively impacting Veterans’ cardiovascular (CV) health. VA sites work with an AHA QI consultant to develop and deploy a local quality improvement plan to help overcome site specific barriers preventing Veterans from achieving optimal cholesterol levels. The program’s objectives include improving clinical pathways and processes to impact ASCVD management through a Learning Healthcare System model of quality improvement designed to improve the management of dyslipidemia among those with high-risk ASCVD. The hypothesis is that patients who are part of the VALOR QI program will be more successful in reaching LDL-C target. This will help to reduce the risk of ASCVD and mortality for the patients, promoting better management of lipid levels through medications, diet and regular monitoring. As part of the methodology, the patients are enrolled to the program using the criteria of Veterans > 18 years of age with diagnosis of ASCVD and LDL-C > 70 mg/dl or non-HDL > 100 mg/dl. The Healthcare Coach has a full track of their lipid medications, assists the patients with refills, provides educational materials, follows up with their clinical appointments, and emphasizes the importance of having a good compliance with therapy, appointments and diets to maintain the target of LDL-C. Once the patient starts with a new lipid medication, they are followed in 3 months with labs. After reaching the target the patient is followed up closely in 6 months with labs to prevent relapse. The VALOR QI Program started at VA San Juan in June 2023 and, as of May 30, 2024, had directly engaged 282 Hispanic patients. 111 (39.4%) achieved target LDL-C; 268 (95%) have any statin use; 117 (41.5%) are on any non-statin lipid-lowering therapy (LLT); 116 (41%) are on statin plus LLT; 25 (8.9%) are on PCSK9, and 133 (47.2%) received educational materials. In conclusion, this initiative led to target LDL levels in 39% of patients which represents a significant population. As this program is further implemented in our institution, we expect improvement in patient outcomes and prognosis.

Frequency-Voltage-var Function for Active Front-end VFD on Oil and Gas Platforms with Offshore Wind Generation

Renewable energy resources emerge as a sustainable alternative to augmenting the energy supply of floating production storage and offloading (FPSO) platforms. However, the increased generation at FPSO based on converter-interfaced energy decreases the system-equivalent inertia constant, which becomes more susceptible to frequency deviations. This paper proposes and evaluates the combined frequency-voltage-var control performance to mitigate frequency variation in a typical FPSO unit with penetration of floating wind energy generation. The control functions are communication-free and embedded in the active front-end variable frequency drives (AFE-VFDs), which are installed on the FPSO and have the primary function of controlling the speed of water injection pumps. The FPSO electrical power system model is developed in MATLAB/Simulink®. Comparative results obtained from the AFE-VFD equipped with volt-var, freq-var, and combined freq-volt-var functions are shown to highlight the proposed solution merits. The results have shown a conflicting behavior with the frequency and voltage deviation improvement associated with absorption and injection of reactive power, respectively. Accordingly, the frequency-volt-var prioritizes frequency deviations during heavy transient events and voltage deviations during regular operation.

Generalized stability landscape of the Atlantic meridional overturning circulation

Abstract. The Atlantic meridional overturning circulation (AMOC) plays a crucial role in shaping climate conditions over the North Atlantic region and beyond, and its future stability is a matter of concern. While the AMOC stability when faced with surface freshwater forcing (FWF) has been thoroughly investigated, its equilibrium response to changing CO2 remains largely unexplored, precluding a comprehensive understanding of its stability under global warming. Here we use an Earth system model to explore the stability of the AMOC when faced with combined changes in FWF in the North Atlantic and atmospheric CO2 concentrations between 180 and 560 ppm. We find four different AMOC states associated with qualitatively different convection patterns. Apart from an “Off” AMOC state with no North Atlantic deep-water formation and a “Modern”-like AMOC with deep water forming in the Labrador and Nordic seas as observed at present, we find a “Weak” AMOC state with convection occurring south of 55° N and a “Strong” AMOC state characterized by deep-water formation extending into the Arctic. The Off and Weak states are stable for the entire range of CO2 but only for positive FWF. The Modern state is stable under higher than pre-industrial CO2 for a range of positive FWF and for lower CO2 only for negative FWF. Finally, the Strong state is stable only for CO2 above 280 ppm and FWF < 0.1 Sv. Generally, the strength of the AMOC increases with increasing CO2 and decreases with increasing FWF. Our AMOC stability landscape helps to explain AMOC instability in colder climates, and although it is not directly applicable to the fundamentally transient AMOC response to global warming on a centennial timescale, it can provide useful information about the possible long-term fate of the AMOC. For instance, while under pre-industrial conditions the AMOC is monostable in the model, the Off state also becomes stable for CO2 concentrations above ∼ 400 ppm, suggesting that an AMOC shutdown in a warmer climate might be irreversible.

Quantifying uncertainty in neural network predictions of forced vibrations

AbstractThe prediction of forced vibrations in nonlinear systems is a common task in science and engineering, which can be tackled using various methodologies. A classical approach is based on solving differential (algebraic) equations derived from physical laws ('first principles'). Alternatively, Artificial Neural Networks (ANNs) may be applied, which rely on learning the dynamics of a system from given data. However, a fundamental limitation of ANNs is their lack of transparency, making it difficult to understand and trust the model's predictions. In this contribution, we follow a hybrid modelling approach combining a data‐based prediction using a stabilised Autoregressive Neural Network (s‐ARNN) with a priori knowledge from first principles. Moreover, aleatoric and epistemic uncertainty is quantified by a combination of mean‐variance estimation (MVE) and deep ensembles. Validating this approach for a classical Duffing oscillator suggests that the MVE ensemble is the most accurate and reliable method for prediction accuracy and uncertainty quantification. These findings underscore the significance of understanding uncertainties in deep ANNs and the potential of our method in improving the reliability of predictive nonlinear system modelling. We also demonstrate that including partially known dynamics can further increase accuracy, highlighting the importance of combining ANNs and physical laws.

Projecting atmospheric N2O rise until the end of the 21st century: an Earth System Model study

Abstract Nitrous Oxide (N2O) is a potent greenhouse gas with a centennial-scale lifetime that contributes significantly to global warming. It is emitted from natural and anthropogenic sources. In nature, N2O is released mainly from nitrification and denitrification from the ocean and terrestrial systems. The use of agricultural fertilizers has significantly increased the emission of N2O in the past century. Here we present, to our knowledge, the first coupled ocean and terrestrial N2O modules within an Earth System Model. The coupled modules were used to simulate the six Shared Socioeconomic Pathways (SSPs) scenarios with available nitrogen fertilizer inputs. Our results are compared to projections of atmospheric N2O concentrations used for SSPs scenario experiments. Additionally, an extra set of simulations were prescribed with emulated N2O concentrations available as input in Shared Socioeconomic Pathways scenarios. We report four main drivers for terrestrial N2O uncertainties: atmospheric temperature, agricultural fertilizer input, soil denitrification and agricultural model dynamics. We project an atmospheric N2O concentration range from 401 to 418 ppb in six SSPs simulations with a robust lack of sensitivity to equilibrium climate sensitivity. We found a large difference between our low emission scenarios N2O concentrations by 2100 compared to the concentration provided for SSPs experiments. This divergence is likely explained by strong mitigation assumptions that were not accounted for in this study, which would require a substantial decrease of agricultural N2O emissions. The coupled model and the simulations prescribed with N2O concentrations showed a difference between −0.02 and 0.09 ∘C by 2100. Our model simulation shows a lack of sensitivity to climate mitigation efforts projecting similar N2O concentration in low and high mitigation scenarios, that could indicate the need of further development of agricultural model dynamics. Further improvements in Earth system models should focus on the impact of oxygen decline on N2O dynamics in the ocean and the representation of anaerobic soils and agricultural dynamics on land, including mitigation methods on nitrogen fertilizers.

Applying circular business models in the Chilean construction sector: A system dynamics perspective

AbstractThe construction sector remains a significant contributor to global climate change, accounting for one‐third of global greenhouse gas emissions as well as a substantial portion of solid waste. In response to this pressing crisis, the adoption of circular business models (CBMs) emerges as a way to avoid negative environmental impacts. This study sought to model the interconnected factors influencing future CBM implementation in the Chilean construction industry using a system dynamics modeling approach. By engaging 20 sector experts in a system modeling workshop, it was possible to model and analyze these interconnected factors as a CBM System. The structural analysis of the resulting CBM system model revealed the triumvirate importance of information dissemination, certification of projects and products, and professional training and education. Recommendations for policy and practice that successfully leverage these mechanisms center on a three‐pronged approach of combining effective stakeholder collaboration, technological innovation, and thoughtful product certification frameworks. Overall, a systems thinking and modeling approach, such as the one applied here, is essential for identifying the complex interactions between enablers and barriers that promote or inhibit sustainable transformation in the construction industry in Chile and beyond.

Modular Modeling of a Half-Vehicle System Using Generalized Receptance Coupling and Frequency-Based Substructuring (GRCFBS)

This paper presents an advanced modular modeling approach for vertical vibration analysis of dynamic systems using the Generalized Receptance Coupling and Frequency-Based Substructuring (GRCFBS) method. The focus is on a four-DoF half-vehicle model comprising three key subsystems: front suspension, rear suspension, and the vehicle’s trimmed body. The proposed technique is designed to predict dynamic responses in reconfigurable systems across various applications, including automotive, robotics, mechanical machinery, and aerospace structures. By coupling the receptance matrices (FRFs) of individual vehicle modules, the overall system receptance matrix is efficiently derived in a disassembled configuration. Two generalized coupling methods, originally developed by Jetmundsen and D.D. Klerk, are employed to determine the complete vehicle’s receptance matrix from its subsystems. Validation is achieved by comparing the results with established methods, such as direct solution and modal analysis, demonstrating high accuracy and reliability for complex dynamic systems. This modular approach allows for the creation of reduced-order models focused on key measurement points without the need for detailed system representation. The method offers significant advantages in early-stage vehicle development, providing critical insights into system vibration behavior.

System Identification for Robust Control of an Electrode Positioning System of an Industrial Electric Arc Melting Furnace

Through system identification for robust control methods and utilizing real-time experimental field data, a comprehensive mathematical model is derived that represents the dynamic performance of a single electrode positioning system (EPS) in an industrial electric arc melting furnace (EAF). This EPS is characterized by large, time-varying dynamic parameters, which fluctuate based on operating conditions, specifically as the electrode weight changes within its operational range. The system identification methodology for robust control is developed in four main steps, progressing from experimental design to model validation. This approach yields a nominal model of the actual system and provides a trustworthy estimate of the region of uncertainty of the model, bounded by models of the real system under maximum and minimum electrode weight conditions (limit operating models). The methodology generates three fourth-order time-delay models using an ARMAX structure. The results are promising, as system identification for robust control enables the derivation of mathematical models specifically tailored for designing robust controllers. These controllers significantly enhance the EPS control system’s performance and substantially reduce energy consumption and environmental emissions.

CRUMPLE ZONE MODELLING OF PASSENGER VEHICLE USING MULTIPLE KELVIN MODEL AND OPTIMIZED WITH PARTICLE SWARM OPTIMIZATION

This paper presents the mathematical modelling of a vehicle crash system using a mass-spring-damper approach to simulate the behavior of a real vehicle during a frontal impact. A Multiple Kelvin model is developed to represent a real vehicle crash situation where the impact is introduced on the frontal vehicle body. The modelling process of Multiple Kelvin model is based on a single Kelvin model that developed into a set of seven mass-spring-damper systems representing the front crumple zone of a real vehicle. The model is then optimized for the parameters k and c using an optimization method namely Particle Swarm Optimization (PSO) algorithm in MATLAB-Simulink. The simulation results demonstrate deformation and acceleration responses closely follow the previous experimental results where the parameters namely Ni, Np, and iw are varied to enhance the model's precision through the calculation of error of 12.15 using parameters Ni = 80, Np = 40 and iw = 0.9.

Teaching deep networks to see shape: Lessons from a simplified visual world

Deep neural networks have been remarkably successful as models of the primate visual system. One crucial problem is that they fail to account for the strong shape-dependence of primate vision. Whereas humans base their judgements of category membership to a large extent on shape, deep networks rely much more strongly on other features such as color and texture. While this problem has been widely documented, the underlying reasons remain unclear. We design simple, artificial image datasets in which shape, color, and texture features can be used to predict the image class. By training networks from scratch to classify images with single features and feature combinations, we show that some network architectures are unable to learn to use shape features, whereas others are able to use shape in principle but are biased towards the other features. We show that the bias can be explained by the interactions between the weight updates for many images in mini-batch gradient descent. This suggests that different learning algorithms with sparser, more local weight changes are required to make networks more sensitive to shape and improve their capability to describe human vision.

USLC: Universal self‐learning control via physical performance policy‐optimization neural network

AbstractThis article proposes an online universal self‐learning control (USLC) algorithm based on a physical performance policy‐optimization neural network, which aims to solve the problem of universal self‐learning optimal control laws for nonlinear systems with various uncertain dynamics. As a key system characterization, this algorithm predicts the discrepancy between the optimal and current control laws by evaluating overall performance in each iterative learning cycle, leveraging an offline‐trained universal policy network. This approach is universal, as it does not rely on an exact system model and can adaptively control performance preferences across various tasks by customizing the physical performance cost weights. Using the established control law‐performance surface and contraction Lyapunov function, the necessary assumptions and proofs for the stable convergence of the system within a three‐dimensional manifold space are provided. To demonstrate the universality of USLC, simulation experiments are conducted on two different systems: a low‐order circuit system and a high‐order variable‐span aircraft attitude control system. The stable control achieved under varying initial values and boundary conditions in each system illustrates the effectiveness of the proposed method. Finally, the limitations of this study are discussed.

Data Reconciliation-Based Hierarchical Fusion of Machine Learning Models

In the context of hierarchical system modeling, ensuring constraints between different hierarchy levels are met, so, for instance, ensuring the aggregation constraints are satisfied, is essential. However, modelling and forecasting each element of the hierarchy independently introduce errors. To mitigate this balance error, it is recommended to employ an optimal data reconciliation technique with an emphasis on measurement and modeling errors. In this study, three different machine learning methods for development were investigated. The first method involves no data reconciliation, relying solely on machine learning models built independently at each hierarchical level. The second approach incorporates measurement errors by adjusting the measured data to satisfy each constraint, and the machine learning model is developed based on this dataset. The third method is based on directly fine-tuning the machine learning predictions based on the prediction errors of each model. The three methods were compared using three case studies with different complexities, namely mineral composition estimation with 9 elements, forecasting of retail sales with 14 elements, and waste deposition forecasting with more than 3000 elements. From the results of this study, the conclusion can be drawn that the third method performs the best, and reliable machine learning models can be developed.

Computationally Efficient Inference via Time-Aware Modular Control Systems

Control in multi-agent decision-making systems is an important issue with a wide variety of existing approaches. In this work, we offer a new comprehensive framework for distributed control. The main contributions of this paper are summarized as follows. First, we propose PHIMEC (physics-informed meta control)—an architecture for learning optimal control by employing a physics-informed neural network when the state space is too large for reward-based learning. Second, we offer a way to leverage impulse response as a tool for system modeling and control. We propose IMPULSTM, a novel approach for incorporating time awareness into recurrent neural networks designed to accommodate irregular sampling rates in the signal. Third, we propose DIMAS, a modular approach to increasing computational efficiency in distributed control systems via domain-knowledge integration. We analyze the performance of the first two contributions on a set of corresponding benchmarks and then showcase their combined performance as a domain-informed distributed control system. The proposed approaches show satisfactory performance both individually in their respective applications and as a connected system.

Analyzing students’ systems thinking in-situ through screencasts in the context of computational modeling: a case study

AbstractIn our interconnected world, Systems Thinking (ST) is increasingly being recognized as a key learning goal for science education to help students make sense of complex phenomena. To support students in mastering ST, educators are advocating for using computational modeling programs. However, studies suggest that students often have challenges with using ST in the context of computational modeling. While previous studies have suggested that students have challenges modeling change over time through collector and flow structures and representing iterative processes through feedback loops, most of these studies investigated student ST through pre and post tests or through interviews. As such there is a gap in the literature regarding how student ST approaches develop and change throughout a computational modeling unit. In this case study, we aimed to determine which aspects of ST students found challenging during a computational modeling unit, how their approaches to ST changed over time, and how the learning environment was supporting students with ST. Building on prior frameworks, we developed a seven-category analysis tool that enabled us to use a mixture of student discourse, writing, and screen actions to categorize seven ST behaviors in real time. Through using this semi-quantitative tool and subsequent narrative analysis, we found evidence for all seven behavior categories, but not all categories were equally represented. Meanwhile our results suggest that opportunities for students to engage in discourse with both their peers and their teacher supported them with ST. Overall, this study demonstrates how student discourse and student writing can be important evidence of ST and serve as a potential factor to evaluate ST application as part of students’ learning progression. The case study also provides evidence for the positive impact that the implementation of a social constructivist approach has in the context of constructing computational system models.