Data-driven Modeling Framework Research Articles

Spectral moment segmentation and spectrogram synthesis for simulation of road vehicle vibrations

Vibration testing procedures relying solely on a Fourier profile can introduce only stationary random vibrations. It is in contrast with the non-stationary and non-Gaussian nature of the Road vehicle vibrations (RVV). The following procedure segments the first four spectral moments of the time–frequency domain of RVV, constructs probability density arrays per frequency bins, and perform simulations according to random segment lengths and -root mean squares, yielding more realistic representations of RVV. The distribution of time- and frequency domain moments and normalized spectral entropies are confronted. The Probability-based spectrogram synthesis (PBSS) offers a data-driven stochastic modelling framework for simulating non-stationary RVV.

The contribution of data-driven poverty alleviation funds in achieving mid-21st-Century multidimensional poverty alleviation planning

The first Sustainable Development Goal (SDG) is intended to eradicate multi-dimensional poverty globally. The same multidimensional poverty indices for India and the Middle East/Africa in 2020 indicate that 10–14 years are still required to reach the level of China’s poverty eradication. Using machine learning, spatial statistics, and a scenario analysis, we demonstrate how a Monte Carlo simulation of poverty alleviation funds-guided shared socioeconomic pathways (PAFs-SSPs) in China reveals the necessity to adopt an integrated poverty alleviation strategy. This approach employs multi-dimensional development indicators to reduce wide regional differences. We developed the data-driven model framework of a PAFs-SSPs to analyze the multifaceted and long-term planning needs of poverty alleviation policies, which can be applied to the formulation of poverty alleviation policies in different developing countries. Our findings point to the importance of implementing multidimensional development policies in China to achieve the first SDG worldwide.

Data-driven nonlinear reduced-order modeling of unsteady fluid–structure interactions

A novel data-driven nonlinear reduced-order modeling framework is proposed for unsteady fluid–structure interactions (FSIs). In the proposed framework, a convolutional variational autoencoder model is developed to determine the coordinate transformation from a high-dimensional physical field into a reduced space. This enables the efficient extraction of nonlinear low-dimensional manifolds from the high-dimensional unsteady flow field of the FSIs. The sparse identification of a nonlinear dynamics (SINDy) algorithm is then used to identify the dynamical governing equations of the reduced space and the vibration responses. To investigate and validate the effectiveness of the proposed framework for modeling and predicting unsteady flow fields in FSI problems, the two-dimensional laminar vortex shedding of a fixed cylinder is considered. Furthermore, the proposed data-driven nonlinear reduced-order modeling framework is applied to the three-dimensional vortex-induced vibration of a flexible cylinder. Using the SINDy model to analyze the vibration responses, the dynamics of the flexible cylinder are found to be correlated with the flow wake patterns, revealing the underlying FSI mechanism. The present work is a significant step toward the establishment of machine learning-based nonlinear reduced-order models for complex flow phenomena, the discovery of underlying unsteady FSI physics, and real-time flow control.

Bayesian Methods for Inferring Missing Data in the BATSE Catalog of Short Gamma-Ray Bursts

The knowledge of the redshifts of Short-duration Gamma-Ray Bursts (SGRBs) is essential for constraining their cosmic rates and thereby the rates of related astrophysical phenomena, particularly Gravitational Wave Radiation (GWR) events. Many of the events detected by gamma-ray observatories (e.g., BATSE, Fermi, and Swift) lack experimentally measured redshifts. To remedy this, we present and discuss a generic data-driven probabilistic modeling framework to infer the unknown redshifts of SGRBs in the BATSE catalog. We further explain how the proposed probabilistic modeling technique can be applied to newer catalogs of SGRBs and other astronomical surveys to infer the missing data in the catalogs.

Data-augmented turbulence modeling by reconstructing Reynolds stress discrepancies for adverse-pressure-gradient flows

Turbulence modeling based on the Reynolds-averaged Navier–Stokes (RANS) method has been widely applied in industry, but its performance in some complex flows is far from satisfactory. The improvement of turbulence models based on the traditional framework has not made breakthrough progress for decades. In this study, a data-driven turbulence modeling framework based on the reconstruction of Reynolds stress discrepancies is used to aid in the improvement of turbulence models, with the Reynolds stresses of the shear-stress transport model being modified in the eigenspace. The large eddy simulation (LES) dataset of a set of bump cases is used to provide high-fidelity information on adverse-pressure-gradient flows for the modeling framework. First, the Reynolds stress tensors of RANS and LES are compared in terms of amplitude, shape, and orientation. Then, the random forest (RF) algorithm is employed to map the mean flow features to the Reynolds stress discrepancies. The well-trained RF model greatly improves the predictions of Reynolds stresses and other flow variables for the attachment and separation states and enables the numerical simulations to have predictive accuracy close to LES and computation time of the same order of magnitude as RANS.

Assessing the role of geographic context in transportation mode detection from GPS data

The increasing availability of health monitoring devices and smartphones has created an opportunity for researchers to access high-resolution (spatial and temporal) mobility data for understanding travel behavior in cities. Although information from GPS data has been used in several studies to detect transportation modes, there is a research gap in understanding the role of geographic context in transportation mode detection. Integrating the geography in which mobility occurs, provides context clues that may allow models predicting transportation modes to be more generalizable. Our goals are first, to develop a data-driven modeling framework for transportation mode detection using GPS mobility data along with geographic context, and second, to assess how model accuracy and generalizability varies upon adding geographic context. To this extent we extracted features from raw GPS mobility data (speed, altitude, turning angle and net displacement) and integrated context in the form of geographic features to classify active (i.e. walk/bike), public (i.e. bus/train), and private (i.e. car) transportation modes in three different Canadian cities - Montreal, St. Johns, and Vancouver. To assess the role of integrating geographic context in mode detection, we adopted two different modeling approaches – generalized and context-specific, and compared results using random forests, extreme gradient boost, and multilayer perceptron classifiers. Our results indicate that for context-specific models the highest classification accuracy improved by 64% for Montreal, by 74% for St. John's and by 77% for Vancouver compared to the generalized model. We also found that the multilayer perceptron (96%) achieved the highest classification accuracy upon adding contextual variables compared to random forests (94.6%) and extreme gradient boost (93.3%) classifier. Our study highlights that adding contextual information specific to a city's geography can improve the predictive accuracy of transportation mode detection models, however, in case of limited knowledge about the geographic setting of a study area, a generalized model combining GPS data from several cities may still be useful for predicting modes from trip data.

A Data-Driven Framework for Identifying Intensive Care Unit Admissions Colonized With Multidrug-Resistant Organisms.

BackgroundThe rising prevalence of multi-drug resistant organisms (MDROs), such as Methicillin-resistant Staphylococcus aureus (MRSA), Vancomycin-resistant Enterococci (VRE), and Carbapenem-resistant Enterobacteriaceae (CRE), is an increasing concern in healthcare settings.Materials and MethodsLeveraging data from electronic healthcare records and a unique MDRO universal screening program, we developed a data-driven modeling framework to predict MRSA, VRE, and CRE colonization upon intensive care unit (ICU) admission, and identified the associated socio-demographic and clinical factors using logistic regression (LR), random forest (RF), and XGBoost algorithms. We performed threshold optimization for converting predicted probabilities into binary predictions and identified the cut-off maximizing the sum of sensitivity and specificity.ResultsFour thousand six hundred seventy ICU admissions (3,958 patients) were examined. MDRO colonization rate was 17.59% (13.03% VRE, 1.45% CRE, and 7.47% MRSA). Our study achieved the following sensitivity and specificity values with the best performing models, respectively: 80% and 66% for VRE with LR, 73% and 77% for CRE with XGBoost, 76% and 59% for MRSA with RF, and 82% and 83% for MDRO (i.e., VRE or CRE or MRSA) with RF. Further, we identified several predictors of MDRO colonization, including long-term care facility stay, current diagnosis of skin/subcutaneous tissue or infectious/parasitic disease, and recent isolation precaution procedures before ICU admission.ConclusionOur data-driven modeling framework can be used as a clinical decision support tool for timely predictions, characterization and identification of high-risk patients, and selective and timely use of infection control measures in ICUs.

Projection-tree reduced-order modeling for fast N-body computations

This work presents a data-driven reduced-order modeling framework to accelerate the computations of nonlocal and N-body methods that model dynamical systems. The proposed framework differs from traditional acceleration methods, like the Barnes–Hut method, which requires online tree building of the state space, or the fast-multipole method, which requires rigorous a priori analysis of governing kernels and online tree building. Our approach combines Barnes–Hut hierarchical decomposition, projection-based reduced-order modeling via the least-squares Petrov–Galerkin (LSPG) projection, and hyper-reduction by way of the Gauss–Newton with approximated tensor (GNAT) approach. The resulting projection-tree reduced-order model (PTROM) enables a drastic reduction in operational count complexity by constructing sparse hyper-reduced pairwise interactions of the non-compact N-body dynamical system. As a result, the presented framework is capable of achieving an operational count complexity that is independent of N, the number of bodies in the numerical domain. Capabilities of the PTROM method are demonstrated on the two-dimensional fluid-dynamic Biot–Savart kernel within a parametric and reproductive setting. Results show the PTROM is capable of achieving over 2000× wall-time speed-up with respect to the full-order model, where the speed-up increases with N. The resulting solution delivers quantities of interest with errors that are less than 0.1% with respect to full-order model.

A Data-Driven Model Framework Based on Deep Learning for Estimating the States of Lithium-Ion Batteries

The accurate estimation of state of charge (SOC) and state of health (SOH) of lithium-ion battery is crucial to ensure the safe and stable operation of the battery. In this paper, a data-driven model framework based on deep learning for estimating SOC and SOH is proposed, which mainly consists of long short-term memory (LSTM) neural network and back propagation (BP) neural network. The switch between SOC estimation model and SOH estimation model can be realized by adjusting the output mode of LSTM. When estimating SOC, the LSTM is set to have corresponding output at each input. The model takes 10 consecutive voltage sampling points as input and the estimated value of SOC at the last sampling moment as output. When estimating SOH, the LSTM is set to have a corresponding output only at the last input. The model takes the sequence of 150 sampling points on the charging voltage curve as input and the SOH value at the current cycle as output. Experiments are carried out on the Oxford battery degradation dataset, and the results show that the proposed model framework can obtain accurate and reliable estimates of SOC and SOH.

Data-Driven Modeling of Linear Dynamical Systems with Quadratic Output in the AAA Framework

We extend the Adaptive Antoulas-Anderson (AAA) algorithm to develop a data-driven modeling framework for linear systems with quadratic output (LQO). Such systems are characterized by two transfer functions: one corresponding to the linear part of the output and another one to the quadratic part. We first establish the joint barycentric representations and the interpolation theory for the two transfer functions of LQO systems. This analysis leads to the proposed AAA-LQO algorithm. We show that by interpolating the transfer function values on a subset of samples together with imposing a least-squares minimization on the rest, we construct reliable data-driven LQO models. Two numerical test cases illustrate the efficiency of the proposed method.

A framework for developing generalisable discrete event simulation models of hospital emergency departments

Discrete event simulation (DES) is routinely used to model hospital emergency departments (EDs), primarily due to its ability to represent complex patient flow processes and investigate improvement strategies. Despite this, it is clear from published studies that many DES models are not subsequently implemented in hospitals or reused for other sites. This research addresses a gap in the literature by presenting a new data-driven modelling framework ‘GE-DES’, which outlines an approach to the design and development of generalisable ED models. The nature of the framework means that it is sufficiently flexible (i) for use across multiple EDs, and (ii) for investigating hospital-specific problems through data-driven customisation. The primary aim of GE-DES is to support model reuse and implementation. The framework is demonstrated through application to a case study ED in Australia.

Forecasting solar-thermal systems performance under transient operation using a data-driven machine learning approach based on the deep operator network architecture

Modeling and prediction of the dynamic behavior of thermal systems operating under intermittent energy input and variable load requirements represent one of the greatest challenges in the development of efficient and reliable renewable-based power generation technologies. In this work, a data-driven machine learning modeling framework was developed based on a modified version of the Deep Operator Network architecture where the time coordinate in the trunk net is replaced with historical data of the predicting quantity. The modeling framework can be used to accurately predict the performance of renewable-based energy conversion technologies including wind- and solar-based power plants. This novel framework was applied on a solar-thermal system that consists of a solar collection loop using a flat plate collector, a power generation loop comprising an Organic Rankine Cycle, and a thermal energy storage tank connecting both loops. Variable solar irradiance, air temperature, and power load profiles were used by the Deep Operator Network to predict the State-of-Charge and the efficiency of the thermal system for several days. The results were compared with the State-of-Charge and efficiency functions calculated using a physics-based model. For a simple operation scenario, characterized by a clear sky solar irradiance profile and constant load, the standard deviation in the State-of-Charge prediction by Deep Operator Network is below 0.9% during a seven-day prediction time horizon. For the most realistic operation scenario that considers real solar irradiance and a rough load profile, the maximum standard deviation in the predictions for the State-of-Charge and efficiency are below 6.8% and 2.5%, respectively. A comparison between Deep Operator Network and Long Short Term Memory network was also performed. In general, both networks predict very well the State-of-Charge for different data density conditions; however, a higher accuracy, with a standard deviation below 2.0%, is obtained by the Deep Operator Network during three and half days using sparser training data of 20-minute points. The same accuracy for the State-of-Charge prediction with the Long Short Term Memory network is achieved only for 14 h. Average standard deviations for the State-of-Charge prediction of 1.1% with the Deep Operator Network and 1.5% with the Long Short Term Memory network are obtained for a four-day prediction time using a denser training data of 5-minute points.

COVID-19: Data-Driven optimal allocation of ventilator supply under uncertainty and risk

This study presents a new risk-averse multi-stage stochastic epidemics-ventilator-logistics compartmental model to address the resource allocation challenges of mitigating COVID-19. This epidemiological logistics model involves the uncertainty of untested asymptomatic infections and incorporates short-term human migration. Disease transmission is also forecasted through a new formulation of transmission rates that evolve over space and time with respect to various non-pharmaceutical interventions, such as wearing masks, social distancing, and lockdown. The proposed multi-stage stochastic model overviews different scenarios on the number of asymptomatic individuals while optimizing the distribution of resources, such as ventilators, to minimize the total expected number of newly infected and deceased people. The Conditional Value at Risk (CVaR) is also incorporated into the multi-stage mean-risk model to allow for a trade-off between the weighted expected loss due to the outbreak and the expected risks associated with experiencing disastrous pandemic scenarios. We apply our multi-stage mean-risk epidemics-ventilator-logistics model to the case of controlling COVID-19 in highly-impacted counties of New York and New Jersey. We calibrate, validate, and test our model using actual infection, population, and migration data. We also define a new region-based sub-problem and bounds on the problem and then show their computational benefits in terms of the optimality and relaxation gaps. The computational results indicate that short-term migration influences the transmission of the disease significantly. The optimal number of ventilators allocated to each region depends on various factors, including the number of initial infections, disease transmission rates, initial ICU capacity, the population of a geographical location, and the availability of ventilator supply. Our data-driven modeling framework can be adapted to study the disease transmission dynamics and logistics of other similar epidemics and pandemics.

A Data-Driven Memory-Dependent Modeling Framework for Anomalous Rheology: Application to Urinary Bladder Tissue

We introduce a data-driven fractional modeling framework for complex materials, and particularly bio-tissues. From multi-step relaxation experiments of distinct anatomical locations of porcine urinary bladder, we identify an anomalous relaxation character, with two power-law-like behaviors for short/long long times, and nonlinearity for strains greater than 25%. The first component of our framework is an existence study, to determine admissible fractional viscoelastic models that qualitatively describe linear relaxation. After the linear viscoelastic model is selected, the second stage adds large-strain effects to the framework through a fractional quasi-linear viscoelastic approach for the nonlinear elastic response of the bio-tissue of interest. From single-step relaxation data of the urinary bladder, a fractional Maxwell model captures both short/long-term behaviors with two fractional orders, being the most suitable model for small strains at the first stage. For the second stage, multi-step relaxation data under large strains were employed to calibrate a four-parameter fractional quasi-linear viscoelastic model, that combines a Scott-Blair relaxation function and an exponential instantaneous stress response, to describe the elastin/collagen phases of bladder rheology. Our obtained results demonstrate that the employed fractional quasi-linear model, with a single fractional order in the range α = 0.25–0.30, is suitable for the porcine urinary bladder, producing errors below 2% without need for recalibration over subsequent applied strains. We conclude that fractional models are attractive tools to capture the bladder tissue behavior under small-to-large strains and multiple time scales, therefore being potential alternatives to describe multiple stages of bladder functionality.

Relocating operational and damaged bikes in free-floating systems: A data-driven modeling framework for level of service enhancement

Free-floating bike sharing is an innovative and sustainable travel mode, where shared bikes can be picked up and returned at any proper place on the streets and not just at docking stations. Nevertheless, in these systems, two major problems arise. One is the imbalance of free-floating shared bikes (FFSB) between zones due to one-way trips, the other is the damaged bikes that must be brought for repair. In this study, a modeling framework for dynamic relocating operational and damaged bikes is proposed that starts with predicting the number and location of shared bikes using deep learning algorithms. The demand forecasting model adopts the Encoder-Decoder architecture embedded with the attention mechanism to further enhance the model’s prediction ability and flexibility. Then, a data-driven optimization model for FFSB relocations is presented, where the multi-period optimization is applied to dynamically plan the relocation activities throughout the day. A new hybrid metaheuristic algorithm that incorporates variable neighborhood search (VNS) and enhanced simulated annealing (ESA) algorithm is developed for solving the relocating problem, in which satisfactory performance is observed from the numerical example. We test the proposed framework with the real-world FFSB data from Beijing, China. The results show that relocating both operational and damaged bikes timely decreases the probability of users finding damaged bikes in the system, but leads to higher relocation costs. For peak-hours, considering only the operational bikes for relocation is the most effective strategy given the limited relocation resources. It is urgent at those times of the day to focus on providing bikes to clients where they are undersupplied.

Short-term forecasting for supercharged boiler safety performance based on advanced data-driven modelling framework

The safety of boilers is directly related to the operation of steam power plants. Consequently, accurate short-term forecasting model is needed for the robust control of important safety performance parameters. In this study, an advanced data-driven modelling framework based on neural networks is proposed to establish short-term forecasting (STF) models for three safety indicators of a supercharged boiler: main steam temperature (MST), drum pressure (DP) and drum level (DL). To overcome the challenges of external noise and non-stationarity, locally weighted regression (LWR) and differencing method are introduced to improve the model performance. Convolutional neural network (CNN) and long short-term memory network (LSTM) are proved as practical options for STF modelling, while CNN has a lower computational cost. The models show excellent performance on a real dataset containing multiple dynamic processes: the mean absolute percentage errors of the three targets are 0.0041%, 0.0355% and 0.0221% respectively and the forecasting horizon is future 60 timesteps. The sensitivity of STF models to outliers is analyzed. The results show that the models are not sensitive to eroded input. The proposed framework and methods can be a preferential solution to the forecasting modelling of power systems.

A dynamic microsimulation model for epidemics

A large evidence base demonstrates that the outcomes of COVID-19 and national and local interventions are not distributed equally across different communities. The need to inform policies and mitigation measures aimed at reducing the spread of COVID-19 highlights the need to understand the complex links between our daily activities and COVID-19 transmission that reflect the characteristics of British society. As a result of a partnership between academic and private sector researchers, we introduce a novel data driven modelling framework together with a computationally efficient approach to running complex simulation models of this type. We demonstrate the power and spatial flexibility of the framework to assess the effects of different interventions in a case study where the effects of the first UK national lockdown are estimated for the county of Devon. Here we find that an earlier lockdown is estimated to result in a lower peak in COVID-19 cases and 47% fewer infections overall during the initial COVID-19 outbreak. The framework we outline here will be crucial in gaining a greater understanding of the effects of policy interventions in different areas and within different populations.

An intelligent hybrid feature subset selection and production pattern recognition method for modeling ethylene plant

A data-driven model framework integrating Feature Subset Selection (FSS), production pattern clustering analysis and prediction was proposed for predicting ethylene yield of ethylene plant by using the massive sensing data recorded by the Distributed Control System (DCS) of petrochemical enterprises. Firstly, an Ensemble-Filter FSS model based on three different metrics is designed to initially filter all the steam cracking furnace features, and then a Wrapper FSS model based on GA-SVR is used to obtain the optimal subset of features affecting ethylene yield. The steam cracking furnace was identified based on the Density Peak Clustering (DPC) algorithm based on the production patterns embedded in the data. Ethylene yield prediction models were separately developed for each production pattern to summarize the final prediction results. The proposed model was validated against historical data from an industrial steam cracking furnace in northwest China. Results show that the number of features have a 93.4% reduction in the FSS stage. and a 40.6% reduction in predicted MSE. Compared with the benchmark ANN model, the proposed DPNN model has a 56.6% reduction in MSE based on the optimal cluster result. What’s more, the proposed framework has a strong generalization ability and with a modular structure which is easy to modify., which is expected to be used to guide the ethylene plant operating in reasonable intervals.

Condition monitoring and performance forecasting of wind turbines based on denoising autoencoder and novel convolutional neural networks

With the proportion of wind power in the grid increasing, the monitoring and maintenance of wind turbines is becoming more and more important for the reliability of the grid. In this study, a data-driven modelling framework based on deep convolutional neural networks is constructed for wind turbines condition monitoring (CM) and performance forecasting (PF). For CM, a robust denoising autoencoder (DAE) model is introduced to output the reconstruction error (RE) of raw signals. The RE is processed to a state indicator by exponentially weighted moving average (EWMA) and monitored on a control chart. For PF, two multi-steps ahead forecasting models are constructed for the forecasting of generator bearing and transformer temperature. To prevent overfitting caused by abundant features, the marginal effect analysis based on random forests is implemented to measure the importance of features. Besides, novel residual attention module (RAM) and training strategies are used improve their representation power of DAE and PF models. Experiments on a real wind turbine dataset prove the effectiveness of the proposed models and methods.

Prediction of land surface temperature of major coastal cities of India using bidirectional LSTM neural networks

Abstract Surface Temperature (ST) is important in terms of surface energy and terrestrial water balances affecting urban ecosystems. In this study, to process the nonlinear changes of climatological variables by leveraging the distinct advantages of Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (BiLSTM), we propose an LSTM-BiLSTM hybrid deep learning model which extracts multi-dimension features of inputs, i.e., backward (future to past) or forward (past to future) to predict ST. This study assessed the climatological variables, i.e., wind speed, wind direction, relative humidity, dew point temperature, and atmospheric pressure impact on ST using five major coastal cities of India: Chennai, Mangalore, Visakhapatnam, Cuddalore, and Cochin. The Recurrent Neural Networks (RNN) and hybrid LSTM-BiLSTM models have effectively predicted ST and outperformed the standalone Artificial Neural Networks (ANN), LSTM, and BiLSTM models. The RNN and LSTM-BiLSTM models have performed better in predicting ST for Mangalore (Nash-Sutcliffe efficiency (NSE)=0.91), followed by Cochin (NSE=0.89), Chennai (NSE=0.88), Cuddalore (NSE=0.88), and Vishakhapatnam (NSE=0.81). The hybrid data-driven modeling framework indicated that coupling the LSTM and BiLSTM models was proven effective in predicting the ST of coastal cities.