ANN-based Model Research Articles

Abstract 12546: Artificial Neural Network Models for Lipid Profile Prediction in the US General Population

Introduction: Hyperlipidemia is a major risk factor for cardiovascular disease. Tests of blood lipid concentrations require a blood sample to be taken in a clinical setting, which is costly, time-consuming, and may preclude measurements in a large number of people at risk. Hypothesis/Objective: This study aimed to develop artificial neural network (ANN) models to predict lipid profiles of the US general population using noninvasive and low-cost diagnostic features. Methods: We used the most recent cycle of the National Health and Nutrition Examination Survey (NHANES) 2017-2018. The participants with missing values in outcomes: total cholesterol (TCH), triglycerides (TG), high-density lipoprotein (HDL), and low-density lipoprotein (LDL), and aged<18 years were excluded. We developed ANN-based prediction models using various combinations of eight features: age, gender, race/ethnicity, body mass index, waist to hip ratio, waist to height ratio, systolic and diastolic blood pressure with three to thirty hidden neurons in one layered feed-forward ANN. We trained ANN models 100 times for seven different input combinations. The performances of these prediction models were evaluated by mean accuracy (ACC), R-squared (R 2 ), mean square error (MSE), root mean square error (RMSE), and mean absolute error (MAE). Results: Of 5176 participants, 51.8% were females with a mean (range) age of 52 (18-80) years. Results show that the ANN models can predict these outcomes with high mean ACC, (TCH 96.69%, HDL 95.11%, LDL 93.47%, TG 91.35%), average R2 (TCH 0.022, HDL 0.186, LDL -0.019, TG 0.158), average MSE (TCH 0.048, HDL 0.056, LDL 0.152, TG 0.268), average RMSE (TCH 0.220, HDL 0.237, LDL 0.389, TG 0.518), and average MAE (TCH 0.171, HDL 0.192, LDL 0.287, TG 0.410). Conclusions: ANN is an effective diagnostic tool for predicting lipid profiles for the US general population, and can be used to predict lipid profiles for clinical diagnosis using noninvasive and low-cost features.

Effect of Various Sequence Descriptors in Predicting Human Proteinprotein Interactions Using ANN-based Prediction Models

Aims: A number of sequence-based descriptors for proteins have been proposed by many researchers. This study aims to evaluate the performance of these descriptors in predicting protein-protein interactions on the benchmark dataset. Background: The behavior of a protein inside or outside the cell is defined by its interaction with the elements present in the surrounding environment, which include small metabolites to the macromolecules such as RNA, DNA, or proteins. Of these, understanding protein-protein interactions (PPIs) is one of the important aspects to investigate the biological role of a protein. The interactions of a protein are determined by how it folds in 3-dimensional space, and this threedimensional folding of a protein largely depends on the linear sequence of amino acids. This information makes it possible to exploit the sequences for proteins to computationally determine the possible interactions among them. Objective: This study aims at studying the efficacy of various sequence-based descriptors in predicting protein-protein interactions. Methods: In this study, we have used the benchmark dataset of interacting and non-interacting protein pairs provided by Pan et al. to build the PPI prediction models using artificial neural networks. We have compared the efficacy of different descriptors on two types of datasets, one with all the protein pairs and the second with proteins having less than 25% identity. Results: The results show that conjoint-triad descriptors performed better than other descriptors in predicting PPIs. The feature selection on the conjoint triad was performed and the effect on the prediction model with reduced features versus all feature sets was studied. Conclusion: The classification model with conjoint-triad descriptors obtained the highest accuracy. The feature ranking for the conjoint triad descriptor was utilized and the model performance was compared with all and selected features. The model with reduced features shows less overfitting.

Investigating spatial impact on indoor personal thermal comfort

Thermal comfort prediction is essential for both maintaining a favorable indoor environment and reducing energy consumption. Predicted Mean Vote (PMV), as the most popular research method, has a limitation in processing various complex parameters and investigating the individual difference in occupants' thermal preference. Therefore, machine learning (ML) method has been utilized in exploring the personal thermal comfort prediction because of its strong self-study ability, high-speed computing ability, and complex problem-solving ability. However, the primary variables considered in previous studies focus on the human body's physiological and psychological aspects, while lack of considering architectural spatial impact, which causes different indoor microclimate. Therefore, the present research proposed a methodology to investigate the impact from spatial parameters on personal thermal comfort prediction model accuracy by developing an ANN-based model and explicitly representing the spatial variables in the model. The spatial parameters were identified and classified into buildings' spatial features, indoor spatial features and individual spatial features. The data required in developing the ANN-based model were collected by various field experiments. A baseline of model prediction accuracy was calculated by using conventional parameters, including personal-dependent parameters and environmental parameters. It was found that the spatial parameters had a noticeable impact on model prediction accuracies. By considering spatial parameters in the ANN-based model development, the prediction accuracies had been increased significantly compared with the conventional models.

Power Estimation and Validation of Embedded Multiplier Based on ANN and Regression Technique

Nowadays, high-end Field-Programmable Gate Arrays (FPGAs) are capable of implementing relatively high-performance systems in the field of Digital Signal Processing (DSP). Due to the abundant application of multipliers, their implementation efficiency and performance have become a critical issue in designing the DSP systems. On the other hand, FPGAs consume a large amount of power due to their complex circuitry. So, the power estimation of FPGA implementations at an early design stage has become a critical design metric. Various models are available in the literature based on Look-up Tables (LUTs), but not much literature is available on speed-optimized multiplier design using DSP slices only. In this paper, an embedded multiplier (12.0 IP core) has been analyzed and customized for different Input/Output (I/O) configurations to estimate the power using Vivado Design Suite (2014.4) targeted to the Zynq-family FPGA device (Zynq evolution and development kit). The embedded multiplier IP has been optimized for performance using two different approaches, i.e., Mults (DSP)-based and LUTs-based. Post-synthesis attributes have been used for formulating the power estimation models based on Artificial Neural Network (ANN) and curve fitting and regression technique. The power values estimated from the proposed models have been authenticated with reference to those assessed from the commercial tool. Based on the results obtained, ANN-based model provides average errors of 0.73% and 0.88% for the LUTs and DSP-based designs, respectively. Whereas, the model based on curve fitting and regression technique provides average errors of 3.61% and 1.59% for the LUTs and DSP-based designs, respectively. The timing analysis has been done to get the design performance and time complexity of the proposed models. Area analysis of the design has also been performed in order to report the resource utilization.

A Machine Learning-Based QSAR Model for Quinolines Tested As Corrosion Inhibitors of Mild Steel in HCl Solution

The investigation of organic compounds as potential inhibitors of metal corrosion using several experimental and/or computational techniques has been well established in literature. Recent advances in the development of corrosion inhibitors has led to the exploration of quantitative structure activity relationship (QSAR) as a worthy alternative to the rigorous, sometimes expensive and time-consuming experimental/computational procedures. QSAR is a robust computational method that generates mathematical models that maps the structural features of organic compounds with their inhibition performances. The obtained mathematical models can then be further used to predict the chemical activity of new and untested organic molecules. Although linear and non-linear methods have been reported in developing QSAR models, only few works have explored the use of state-of-the-art artificial neural network (ANN). This study reports the development of QSAR models for sixty quinolines investigated as corrosion inhibitors for mild steel corrosion in 1 M HCl using both linear and non-linear modelling techniques. Density functional theory (DFT) and Dragon 7 software were used to generate molecular descriptors for model development and were reduced to a significant few using feature selection tools. The significant molecular descriptors were employed in model building using conventional multiple linear regression (MLR) and ANN modelling techniques. The developed models were evaluated using several statistical performance metrics and the ANN-based model was found to produce the best correlation between the predicted and experimental inhibition efficiencies. Furthermore, nine new quinoline molecules were theoretically designed and their inhibition efficiencies were predicted using the best developed ANN model.

Modeling the Shear Strength of FRP-Strengthened Rc Beams Using Artificial Neural Networks

Externally strengthening reinforced concrete (RC) structures has been a desired practice in both research community and the industry over the last few decades. This application entails bonding composites to the surfaces of RC members to upgrade their strength, stiffness, and ductility. This study will attempt to use Machine Learning (ML) techniques to study and predict the shear strength of RC beams strengthened in shear with externally-bonded fiber-reinforced polymer (EB-FRP) laminates. An extensive database consisting of 511 tested specimens and 17 test parameters were collected. An appropriate artificial neural network (ANN)-based model was used to predict the shear capacities (Vu) of the FRP-strengthened beams, including the specific contributions of the EB-FRP (Vf) to these shear capacities. The obtained results indicate that the ANN-based model provided reasonable predictions for both and Vu and Vf.

Automated constitutive modeling of isotropic hyperelasticity based on artificial neural networks

Herein, an artificial neural network (ANN)-based approach for the efficient automated modeling and simulation of isotropic hyperelastic solids is presented. Starting from a large data set comprising deformations and corresponding stresses, a simple, physically based reduction of the problem’s dimensionality is performed in a data processing step. More specifically, three deformation type invariants serve as the input instead of the deformation tensor itself. In the same way, three corresponding stress coefficients replace the stress tensor in the output layer. These initially unknown values are calculated from a linear least square optimization problem for each data tuple. Using the reduced data set, an ANN-based constitutive model is trained by using standard machine learning methods. Furthermore, in order to ensure thermodynamic consistency, the previously trained network is modified by constructing a pseudo-potential within an integration step and a subsequent derivation which leads to a further ANN-based model. In the second part of this work, the proposed method is exemplarily used for the description of a highly nonlinear Ogden type material. Thereby, the necessary data set is collected from virtual experiments of discs with holes in pure plane stress modes, where influences of different loading types and specimen geometries on the resulting data sets are investigated. Afterwards, the collected data are used for the ANN training within the reduced data space, whereby an excellent approximation quality could be achieved with only one hidden layer comprising a low number of neurons. Finally, the application of the trained constitutive ANN for the simulation of two three-dimensional samples is shown. Thereby, a rather high accuracy could be achieved, although the occurring stresses are fully three-dimensional whereas the training data are taken from pure two-dimensional plane stress states.

Comparative analysis of the optimized ANN, SVM, and tree ensemble models using Bayesian optimization for predicting GSHP COP

This study analyzes the performances of three machine learning (ML) methods for predicting the ground source heat pump (GSHP) coefficient of performance (COP) by applying Bayesian optimization to maximize the ML performances. Predicting accurate COP is a prerequisite for the efficient operation of a GSHP system with energy saving potentials, and ML application has been proven to be an ideal solution for predicting COP. ML performances can be further enhanced by tuning each factor affecting the learning process, yielding the overall efficient operation of the GSHP system. To derive a compelling COP prediction model for GSHP system control, a residential building with a GSHP was modeled using TRNSYS 18, and data were acquired for ML based on temperature variables and flow rate scenarios. Using different ML methods, i.e., an artificial neural network (ANN), a tree ensemble, and a support vector machine (SVM), three COP prediction models were developed in MATLAB. The performance of each model was optimized using Bayesian optimization to determine the optimal combination of the hyperparameters, which dramatically affect the ML methods. The R-squared (R2) and coefficient of variation of the root mean square error (Cv(RMSE)) results indicate that the ANN model exhibited the highest prediction accuracy, followed by the tree ensemble and SVM models. Particularly, the R2 of the SVM model did not meet the standards recommended by ASHRAE. Additionally, the ANN model exhibited the lowest maximum error (−0.025 ≤ x < 0.025), demonstrating high predictive ability. Thus, the proposed ANN-based prediction model can be employed in the control algorithm of GSHP systems to promote energy efficiency by determining the system variables affording the highest COP.

Applied Deep Learning for Slender Marine Structure Dynamic Analysis

Abstract Nonlinear finite element models (FEMs) are commonly used to perform analysis in the time domain to simulate a limited number of stochastic loading scenarios that a slender marine structure may undergo, requiring a high computational time effort. Analytical equations and frequency domain analysis can be used to speed up these simulations, but they are not a convenient choice when high nonlinearities are present in the dynamic system. Alternative models can be developed to reduce the simulation time while maintaining a good accuracy level of the system’s response. This work proposes different strategies to develop artificial neural network (ANN) architectures, based on deep learning (DL) algorithms, which can predict multiple structural node responses at once, in time and space, significantly reducing the total training time when a great number of structural nodes are considered. A novel classification concept of ANN-based models is introduced for this application: the NodeNet and the LengthNet class types. In the first approach, the model predictor focuses on a single structural node, while in the latter the model focuses on a length (segment) comprising many structural nodes. The work also extends the response predictions of such marine structures from the top region down to the touchdown zone (TDZ).

Assessment and Comparison of Six Machine Learning Models in Estimating Evapotranspiration over Croplands Using Remote Sensing and Meteorological Factors

Accurate estimates of evapotranspiration (ET) over croplands on a regional scale can provide useful information for agricultural management. The hybrid ET model that combines the physical framework, namely the Penman-Monteith equation and machine learning (ML) algorithms, have proven to be effective in ET estimates. However, few studies compared the performances in estimating ET between multiple hybrid model versions using different ML algorithms. In this study, we constructed six different hybrid ET models based on six classical ML algorithms, namely the K nearest neighbor algorithm, random forest, support vector machine, extreme gradient boosting algorithm, artificial neural network (ANN) and long short-term memory (LSTM), using observed data of 17 eddy covariance flux sites of cropland over the globe. Each hybrid model was assessed to estimate ET with ten different input data combinations. In each hybrid model, the ML algorithm was used to model the stomatal conductance (Gs), and then ET was estimated using the Penman-Monteith equation, along with the ML-based Gs. The results showed that all hybrid models can reasonably reproduce ET of cropland with the models using two or more remote sensing (RS) factors. The results also showed that although including RS factors can remarkably contribute to improving ET estimates, hybrid models except for LSTM using three or more RS factors were only marginally better than those using two RS factors. We also evidenced that the ANN-based model exhibits the optimal performance among all ML-based models in modeling daily ET, as indicated by the lower root-mean-square error (RMSE, 18.67–21.23 W m−2) and higher correlations coefficient (r, 0.90–0.94). ANN are more suitable for modeling Gs as compared to other ML algorithms under investigation, being able to provide methodological support for accurate estimation of cropland ET on a regional scale.

A novel classifier for improving wetland mapping by integrating image fusion techniques and ensemble machine learning classifiers

Among all the natural features, wetlands are the most difficult to map and monitor because of their complicated morphology, i.e., their diverse shapes and sizes, ranging from open waterlogged regions with sparse vegetation to thickly wooded areas with a distant distribution pattern. Wetland mapping has been performed with optical remote sensing data for a long time; but the accuracy varies with time and space owing to the complexity of wetland systems. Therefore, the objective of this work is to examine the efficacy of a novel and unique approach to automated complex wetland mapping in Bangladesh's Sunamganj District, using combined image fusion with machine learning techniques. The image fusion technique was used to fuse the multispectral bands (MS) of Landsat 8 OLI (bands 5, 4, 3) with the panchromatic band of Landsat 8, (band 8). The quality of the fused images was determined by calculating the correlation between the spectral values of the original multispectral and fused images, as well as through edge detection techniques. In addition to this, four advanced machine learning classifiers, i.e. random forest (RF), support vector machine (SVM), k-nearest neighbor (KNN), and decision tree (DT), were used to classify the complex wetland image. Furthermore, a hybrid classifier was created by applying an artificial neural network (ANN) to previously identified models and spectral bands of satellite images. To validate the identified wetland maps, the Kappa coefficient and root mean square error (RMSE) have been used. The result shows that in wetland mapping, the Brovey image fusion approach surpassed the Gram-Schmidt image fusion technique. In addition, the accuracy of all four machine learning classifiers was higher on the fused images than on the MS image. The proposed fused ANN-based hybrid model outperformed all fused and MS classification models in terms of accuracy (kappa: 89.4% and RMSE: 0.13). According to the findings of this study, image-fusion based machine learning classifiers could be used for wetland mapping in other regions of the world, as well as for mapping river dynamics, forest cover, urban growth, and croplands.

Fast computation of the Jovian-moon three-body flyby map based on artificial neural networks

Multiple Jovian moon flybys are common and essential for orbital transfers in Jupiter system exploration missions. To improve the preliminary design accuracy and find low-cost transfer trajectories, the restricted three-body model is used to describe the motion during the flybys. However, the restricted three-body problem has no analytical solution and requires numerical integration for computation, which requires a long computational time for the design of complex Jupiter missions. Therefore, an artificial neural network (ANN)-based model is proposed to improve the computing efficiency. The dynamics of the Jovian-moon flybys are analyzed to facilitate the design of the ANNs. A computational efficient map between the states before and after a flyby is constructed by the ANNs. Numerical simulations verify that the trained ANN-based model can achieve high accuracy and fast computational speed.

Artificial neural network model in predicting yield of mechanically transplanted rice from transplanting parameters in Bangladesh

The yield of rice largely depends on transplanting techniques. Mechanical transplanting is gaining popularity as a cost-saving and on-time operation with less labor orientation in rice cultivation. This experiment was performed to investigate the relation of rice yield (g m−2) with four mechanical rice transplanting parameters (i.e., seedling density in the tray (nos.cm−2), missing hill percentage, floating hill percentage, and seedling number per hill), and to develop an Artificial Neural Networks (ANN) model to predict the yield from the transplanting parameters. A regression analysis was also conducted to validate the accuracy of the trained ANN model. This study reveals that the dependency of yield on these four parameters is not ignorable and significant. The ANN model performed an accurate match in predicting yield from transplanting parameters with an R2 value of 0.994 and adjusted R2 of 0.993. The ANN model possessed an RMSE of 4.577 in predicting yield which lied in the allowable range of 10 %. The ANN model showed better accuracy in predicting yield than the regression model and pretended to be an alternative to numerical models in predicting yield. The findings of this study showed that ANN-based models would be an alternative to the regression model and a more accurate method of yield prediction based on the transplanting field parameters.

Prediction of Surface Water Quality by Artificial Neural Network Model Using Probabilistic Weather Forecasting

We developed an artificial neural network (ANN)-based water quality prediction model and evaluated the applicability of the model using regional probability forecasts provided by the Korea Meteorological Administration as the input data of the model. The ANN-based water quality prediction model was constructed by reflecting the actual meteorological observation data and the water quality factors classified using an exploratory factor analysis (EFA) for each unit watershed in Nam River. To apply spatial refinement of meteorological factors for each unit watershed, we used the data of the Sancheong meteorological station for Namgang A and B, and the data of the Jinju meteorological station for Namgang C, D, and E. The predicted water quality variables were dissolved oxygen (DO), biochemical oxygen demand (BOD), chemical oxygen demand (COD), total organic carbon (TOC), total phosphorus (T-P), and suspended solids (SS). The ANN evaluation results reveal that the Namgang E unit watershed has a higher model accuracy than the other unit watersheds. Furthermore, compared with Namgang C and D, Namgang E has a high correlation with water quality due to meteorological effects. The results of this study will help establish a water quality forecasting system based on probabilistic weather forecasting in the long term.

Explore artificial neural networks for solving complex hydrocarbon chemistry in turbulent reactive flows

Global warming caused by the use of fossil fuels is a common concern of the world today. It is of practical importance to conduct in-depth fundamental research and optimal design for modern engine combustors through high-fidelity computational fluid dynamics (CFD), so as to achieve energy conservation and emission reduction. However, complex hydrocarbon chemistry, an indispensable component for predictive modeling, is computationally demanding. Its application in simulation-based design optimization, although desirable, is quite limited. To address this challenge, we propose a methodology for representing complex chemistry with artificial neural networks (ANNs), which are trained with a comprehensive sample dataset generated by the Latin hypercube sampling (LHS) method. With a given chemical kinetic mechanism, the thermochemical sample data is able to cover the whole accessible pressure/temperature/species space in various turbulent flames. The ANN-based model consists of two different layers: the self-organizing map (SOM) and the back-propagation neural network (BPNN). The methodology is demonstrated to represent a 30-species methane chemical mechanism. The obtained ANN model is applied to simulate both a non-premixed turbulent flame (DLR_A) and a partially premixed turbulent flame (Flame D) to validate its applicability for different flames. Results show that the ANN-based chemical kinetics can reduce the computational cost by about two orders of magnitude without loss of accuracy. The proposed methodology can successfully construct an ANN-based chemical mechanism with significant efficiency gain and a broad scope of applicability, and thus holds a great potential for complex hydrocarbon fuels.

Estimation of Daily Potential Evapotranspiration in Real-Time from GK2A/AMI Data Using Artificial Neural Network for the Korean Peninsula

Evapotranspiration (ET) is a fundamental factor in energy and hydrologic cycles. Although highly precise in-situ ET monitoring is possible, such data are not always available due to the high spatiotemporal variability in ET. This study estimates daily potential ET (PET) in real-time for the Korean Peninsula, via an artificial neural network (ANN), using data from the GEO-KOMPSAT 2A satellite, which is equipped with an Advanced Meteorological Imager (GK2A/AMI). We also used passive microwave data, numerical weather prediction (NWP) model data, and static data. The ANN-based PET model was trained using data for the period 25 July 2019 to 24 July 2020, and was tested by comparing with in-situ PET for the period 25 July 2020 to 31 July 2021. In terms of accuracy, the PET model performed well, with root-mean-square error (RMSE), bias, and Pearson’s correlation coefficient (R) of 0.649 mm day−1, −0.134 mm day−1, and 0.954, respectively. To examine the efficiency of the GK2A/AMI-derived PET data, we compared it with in-situ ET measured at flux towers and with MODIS PET data. The accuracy of the GK2A/AMI-derived PET, in comparison with the flux tower-measured ET, showed RMSE, bias, and Pearson’s R of 1.730 mm day−1, 1.212 mm day−1, and 0.809, respectively. In comparison with the in-situ PET, the ANN model produced more accurate estimates than the MODIS data, indicating that it is more locally optimized for the Korean Peninsula than MODIS. This study advances the field by applying an ANN approach using GK2A/AMI data and could play an important role in examining hydrologic energy for air-land interactions.

Examination of artificial neural networks to predict wind-induced displacements of cable net roofs

This paper discusses the implementation of an Artificial Neural Network (ANN) model for predicting the wind-induced, vertical displacements of cable net roofs with a hyperbolic paraboloid shape. The ANN model was calibrated by training and comparing three preliminary ANN topologies. The first one was trained using both experimental pressure coefficients on similar building geometries and wind-induced vertical displacements estimated by Finite Element Method (FEM). This first set was subsequently used to calibrate a second ANN model, which was based on surrogate modeling of pressure coefficients, calculated by polynomial fitting of the experimental data. Further generalization of the pressure coefficients was obtained through parametrization of the polynomial fitting as a function of geometrical properties. Pressure loads were used to calibrate the third ANN, which was employed to synthetically generate vertical displacements of a varied geometrical sample. The ANN-based models exhibit remarkable results; the coefficient of determination between physically-estimated flexible roof displacements and approximated values was usually above 0.90 for all the tested geometries, suggesting that this type of surrogate model is capable of replicating complex, geometrically nonlinear structural behavior.

Estimated Artificial Neural Network Modeling of Maximal Oxygen Uptake Based on Multistage 10-m Shuttle Run Test in Healthy Adults.

We aimed to develop an artificial neural network (ANN) model to estimate the maximal oxygen uptake (VO2max) based on a multistage 10 m shuttle run test (SRT) in healthy adults. For ANN-based VO2max estimation, 118 healthy Korean adults (59 men and 59 women) in their twenties and fifties (38.3 ± 11.8 years, men aged 37.8 ± 12.1 years, and women aged 38.8 ± 11.6 years) participated in this study; data included age, sex, blood pressure (systolic blood pressure (SBP), diastolic blood pressure (DBP)), waist circumference, hip circumference, waist-to-hip ratio (WHR), body composition (weight, height, body mass index (BMI), percent skeletal muscle, and percent body), 10 m SRT parameters (number of round trips and final speed), and VO2max by graded exercise test (GXT) using a treadmill. The best estimation results (R2 = 0.8206, adjusted R2 = 0.7010, root mean square error; RMSE = 3.1301) were obtained in case 3 (using age, sex, height, weight, BMI, waist circumference, hip circumference, WHR, SBP, DBP, number of round trips in 10 m SRT, and final speed in 10 m SRT), while the worst results (R2 = 0.7765, adjusted R2 = 0.7206, RMSE = 3.494) were obtained for case 1 (using age, sex, height, weight, BMI, number of round trips in 10 m SRT, and final speed in 10 m SRT). The estimation results of case 2 (using age, sex, height, weight, BMI, waist circumference, hip circumference, WHR, number of round trips in 10 m SRT, and final speed in 10 m SRT) were lower (R2 = 0.7909, adjusted R2 = 0.7072, RMSE = 3.3798) than those of case 3 and higher than those of case 1. However, all cases showed high performance (R2) in the estimation results. This brief report developed an ANN-based estimation model to predict the VO2max of healthy adults, and the model’s performance was confirmed to be excellent.

Space fractional kinetic model for different types of suspension profiles in turbulent flows with a neural network-based estimation of fractional orders

Estimation of particle concentration distribution profile is important to understand the inherent mixing process, changes of river morphology, and nature of transportation of particles. In this study, the effects of non-locality are investigated to analyze, distinguish and predict different types of profiles of the suspended concentration distribution in turbulent flows through open channels, pipes, and mixing boxes. The effects of non-locality are captured through fractional derivatives and integrals. Starting from the fractional Liouville equation and considering the power-law jumps of particles in the carrier fluid, the fractional kinetic equation is derived through mathematical analysis. The kinetic model is solved and an analytical model is proposed using the Laplace transformation. The proposed solution contains the two parameters (as α the order of fractional model and μ the shape parameter) Mittag-Leffler function and is a more general one that contains several previous models as special cases. The proposed model can predict four different types of suspension concentration distributions namely type-I, type-II, plume-like, and convex type profiles under different conditions. Further, the model is validated with existing experimental data of dilute and dense flow in open channels and pipes, flows in square pipes, slurry pipelines, and fluid mud gravity currents. Satisfactory results are obtained. A detailed non-linear multivariate regression analysis is carried out with a wide range of selected data and models are proposed to compute model parmeters. To increase the efficiency of the model, an artificial neural network (ANN) model is proposed to predict the model parameters. The ANN-based models show better results than the regression analysis. Validation results show that non-locality has significant effects on the structure of the concentration profiles. From the analysis of the model parameters, it is found that type-I profile occurs under the subdiffusion process where α⩽1 and μ=1 and other three types (type-II, plume-like and convex) of profiles are associated with the superdiffusion process when α>1 and μ>1.

Predicting the survivals and favorable neurologic outcomes after targeted temperature management by artificial neural networks

To identify the outcome-associated predictors and develop predictive models for patients receiving targeted temperature management (TTM) by artificial neural network (ANN). The derived cohort consisted of 580 patients with cardiac arrest and ROSC treated with TTM between January 2014 and August 2019. We evaluated the predictive value of parameters associated with survival and favorable neurologic outcome. ANN were applied for developing outcome prediction models. The generalizability of the models was assessed through 5-fold cross-validation. The performance of the models was assessed according to the accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). The parameters associated with survival were age, duration of cardiopulmonary resuscitation, history of diabetes mellitus (DM), heart failure, end-stage renal disease (ESRD), systolic blood pressure (BP), diastolic BP, body temperature, motor response after ROSC, emergent coronary angiography or percutaneous coronary intervention (PCI), and the cooling methods. The parameters associated with the favorable neurologic outcomes were age, sex, DM, chronic obstructive pulmonary disease, ESRD, stroke, pre-arrest cerebral-performance category, BP, body temperature, motor response after ROSC, emergent coronary angiography or PCI, and cooling methods. After adequate training, ANN Model 1 to predict survival achieved an AUC of 0.80. Accuracy, sensitivity, and specificity were 75.9%, 71.6%, and 79.3%, respectively. ANN Model 4 to predict the favorable neurologic outcome achieved an AUC of 0.87, with accuracy, sensitivity, and specificity of 86.7%, 77.7%, and 88.0%, respectively. The ANN-based models achieved good performance to predict the survival and favorable neurologic outcomes after TTM. The models proposed have clinical value to assist in decision-making.