Superior Predictive Capability Research Articles

Experimental and modeling evaluation of droplet size in immiscible liquid-liquid stirred vessel using various impeller designs

Abstract The present study investigates the effects of impeller design and dispersed phase volume ratio on mean drop sizes (d32) in immiscible liquid-liquid stirred vessel through experimental and modeling approaches. Various impeller designs including conventional and new impeller designs were employed to cover both radial and axial flow impellers. The microscopic method associated with image processing tools was used for the drop size analysis. The results showed the hydrofoil impeller produced the largest drop sizes while the double-curved blade turbine produced the smallest drop sizes, corresponding to about 37% difference. Increasing the dispersed phase volume ratio from 1% to 10%) increased the d32 by approximately 20–40%. Adaptive neuro-fuzzy inference system based on fuzzy C–means (ANFIS-FCM) clustering algorithm was used to develop a model to predict drop sizes, and its validation and accuracy were examined by comparing the results to the experimental data. The results also proved the superior prediction capability of the ANFIS-FCM method over the empirical correlations for the most cases.

Using Vine Copulas to Generate Representative System States for Machine Learning

The increasing uncertainty that surrounds electricity system operation renders security assessment a highly challenging task; the range of possible operating states expands, rendering traditional approaches based on heuristic practices and ad hoc analysis obsolete. In turn, machine learning can be used to construct surrogate models approximating the system's security boundary in the region of operation. To this end, past system history can be useful for generating anticipated system states suitable for training. However, inferring the underlying data model, to allow high-density sampling, is problematic due to the large number of variables, their complex marginal probability distributions and the nonlinear dependence structure they exhibit. In this paper, we adopt the C-Vine pair-copula decomposition scheme; clustering and principal component transformation stages are introduced, followed by a truncation to the pairwise dependency modeling, enabling efficient fitting and sampling of large datasets. Using measurements from the French grid, we show that a machine learning training database sampled from the proposed method can produce binary security classifiers with superior predictive capability compared to other approaches.

Modelling and parameter identification of electromechanical systems for energy harvesting and sensing

Abstract Advanced modelling of electro-mechanical systems for energy harvesting (EH) and sensing is important to develop reliable self-powered autonomous electronic devices and for structural health monitoring (SHM). In this perspective, a novel computational approach is here proposed for both real-time and off-line parameter identification (PI). The system response is governed by a set of four partial differential equations (PDE) where the three displacement components and the electrical potential are the unknowns. Firstly, the finite element (FE) method is used to reduce the PDE problem into a set of ordinary differential equations (ODE). Then, a state-space model is derived with the aim to limit the PI problem to a subset of unknowns. After that, an identification error is introduced and the Lyapunov theory is used to derive the PI algorithm. The numerical implementation is based on a sensitivity analysis feedback block. The overall proposed computational strategy is robust and results in an exponential asymptotic convergence. The accuracy of the PI method is demonstrated by analysing the time-domain response of an array of piezoelectric bimorphs subjected to low-frequency structural random vibrations. The selected case-study is an existing cable-stayed bridge, for which an extensive dynamic monitoring campaign has provided the experimental data. Once time histories of the device response are obtained through time-dependent dynamic FE simulations, the PI algorithm is used to determine the unknown lumped coefficients of the state-space model. The comparison between FE method and lumped parameters model in terms of tip displacement and output voltage demonstrates the superior predictive capability of the new PI algorithm. As a result of the sensitivity analysis, guidelines to assess the optimal array configuration are also provided.

Biomass higher heating value prediction analysis by ANFIS, PSO-ANFIS and GA-ANFIS models

<p>In this study, a new model for biomass higher heating value (HHV) prediction based on the Adaptive Neuro-Fuzzy Inference System (ANFIS) approach was proposed. Proximate analysis (volatile matter, fixed carbon and ash content) data for a wide range of various biomass types from the literature were used as input in model studies. Optimization of ANFIS parameters and formation of the model structure were performed by genetic algorithm (GA) and particle swarm optimization (PSO) algorithm in order to achieve optimum prediction capability. The best-fitting model was selected using statistical analysis tools. According to the analysis, PSO-ANFIS model showed a superior prediction capability over ANFIS and GA optimized ANFIS model. The Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Bias Error (MBE) and coefficient of determination (R2) for PSO-ANFIS were determined as 0.3138, 0.2545, -0.00129 and 0.9791 in the training phase and 0.3287, 0.2748, 0.00120 and 0.9759 in the testing phase, respectively. As a result, it can be concluded that the proposed PSO-ANFIS model is an efficient technique and has potential to calculate biomass HHV prediction with high accuracy.</p>

Komparasi Algoritma Naive bayes dan SVM Untuk Memprediksi Keberhasilan Imunoterapi Pada Penyakit Kutil

Warts is a skin health problem that is generally characterized by the appearance of small, rough-textured lumps on the skin surface caused by a virus that is human papilloma virus (HPV). One technique of treatment of wart disease is immunotherapy, this method is a treatment by boosting the immune system to overcome the disease of warts. Naive bayes and Support Vector Machine (SVM) is a method of data mining algorithm used to classify. The aim of this study was to compare the Naive bayes algorithm with Support Vector Machine (SVM) in predicting the success of immunotherapy treatment method in the treatment of wart disease. Tests conducted using the method of Naive bayes and Support Vector Machine (SVM) using the R programming language, then the results are used to do the comparison. The results of this study revealed that the Naive bayes method has superior prediction capability compared to Support Vector Machine (SVM) because Naive bayes can predict all class instances correctly with the accuracy level of 1.

Abstract 095: Comparison of Various Anthropometric Indices for the Identification of a Predictor of Incident Hypertension

We compared the predictive capability of weight, waist circumference (WC), waist-to-height ratio (WHtR), waist-to-hip ratio (WHR), body mass index (BMI), body roundness index (BRI), and a body shape index (ABSI) to identify incident hypertension, and to determine whether any of these indices may be used as a better single predictor of incident hypertension. A total of 1,718 populations aged 39-72 years were collected in a longitudinal study. Logistic regression models were used to evaluate various anthropometric indices as significant predictors of hypertension. During 2.8 years of follow-up, 185 new cases of hypertension (10.8%) were reported. The BRI and ABSI were significantly higher in the participants who had developed hypertension than in those who had not (4.15±1.01 vs. 3.57±1.03, 0.80±0.04 vs. 0.78±0.05; respectively, p <0.001). After adjusting for confounding variables, logistic regression analysis indicated that participants within the highest quartile of WC and WHtR were 4.79 and 4.51 times more likely to have hypertension than those within the lowest quartile (OR 4.79, 95% CI 2.49-9.20 vs. OR 4.51, 95% CI 2.41-8.43, respectively, p<0.0001); in contrast, no such correlation was found for BMI, WHR, BRI, and ABSI. WC (AUC: 0.672) showed a more powerful predictive ability for hypertension (p<0.0001) than BMI (AUC: 0.623), and an equal predictive power for hypertension as WHtR (AUC: 0.662) and BRI (AUC: 0.662) in the general population. We concluded that WC and/or WHtR but not BMI, showed superior prediction capability compared to WHR, BRI and ABSI, for determining the incidence of hypertension in a community-based prospective study.

External Validation of a Survival Nomogram for Non-Small Cell Lung Cancer Using the National Cancer Database

Survival nomograms offer individualized predictions using a more diverse set of factors than traditional staging measures, including the American Joint Committee on Cancer Tumor Node Metastasis (AJCC TNM) Staging System. A nomogram predicting overall survival (OS) for resected, non-metastatic non-small cell lung cancer (NSCLC) has been previously derived from Asian patients. The present study aims to determine the nomogram's predictive capability in the US using the National Cancer Database (NCDB). This was a retrospective review of adults with resected, non-metastatic NSCLC entered into the NCDB between 2004 and 2012. Concordance indices and calibration plots analyzed discrimination and calibration, respectively. Multivariate analysis was also used. A total of 57,313 patients were included in this study. The predominant histologies were adenocarcinoma (48.2%) and squamous cell carcinoma (31.3%), and patients were diagnosed with stage I-A (38.3%), stage I-B (22.7%), stage II-A (14.2%), stage II-B (11.5%), and stage III-A (13.3%). Median OS was 74 months. 1-, 3- and 5-year OS rates were 89.8% [95% confidence interval (CI) 89.5-90.0%], 71.1% (95% CI 70.7-71.6%), and 55.7% (95% CI 54.7-56.6%), respectively. The nomogram's concordance index (C-index) was 0.804 (95% CI 0.792-0.817). AJCC TNM staging demonstrated higher discrimination (C-index 0.833, 95% CI 0.821-0.840). The nomogram's individualized estimates accurately predicted survival in this patient collective, demonstrating higher discrimination in this population than in the developer's cohorts. However, the generalized survival estimates provided by traditional staging demonstrated superior predictive capability; therefore, AJCC TNM staging should remain the gold standard for the prognostication of resected NSCLC in the US.

ANN Model-Based Simulation of the Runoff Variation in Response to Climate Change on the Qinghai-Tibet Plateau, China

Precisely quantitative assessments of stream flow response to climatic change and permafrost thawing are highly challenging and urgent in cold regions. However, due to the notably harsh environmental conditions, there is little field monitoring data of runoff in permafrost regions, which has limited the development of physically based models in these regions. To identify the impacts of climate change in the runoff process in the Three-River Headwater Region (TRHR) on the Qinghai-Tibet Plateau, two artificial neural network (ANN) models, one with three input variables (previous runoff, air temperature, and precipitation) and another with two input variables (air temperature and precipitation only), were developed to simulate and predict the runoff variation in the TRHR. The results show that the three-input variable ANN model has a superior real-time prediction capability and performs well in the simulation and forecasting of the runoff variation in the TRHR. Under the different scenarios conditions, the forecasting results of ANN model indicated that climate change has a great effect on the runoff processes in the TRHR. The results of this study are of practical significance for water resources management and the evaluation of the impacts of climatic change on the hydrological regime in long-term considerations.

Simulation and uncertainty analysis for groundwater levels using radial basis function neural network and support vector machine models

Accurate measurement of groundwater levels is often difficult and involves great uncertainty. Therefore, simulating and predicting the fluctuating behavior of groundwater levels is necessary for water resource planning and management. In this study, radial basis function (RBF) neural networks and support vector machines (SVM) were employed to simulate groundwater level fluctuations. The time series data of precipitation, evaporation, and temperature were used as model inputs. Groundwater level data from the first 10 years, from 2003 to 2014, were used as the training dataset, while data from the last 2 years were used as the test dataset. Uncertainties caused by errors in the measurements of the variables or in outputs were estimated at 95% confidence intervals. The results showed that the SVM model had a superior simulation and prediction capability according to four statistical standards. The comparisons of the outputs and the confidence intervals of the two models showed that the SVM model was more accurate and had less uncertainty. The conclusions suggest that SVM is an effective method for simulating groundwater levels and analyzing model uncertainties using confidence intervals and can be used to facilitate sustainable groundwater management strategies.

Modeling of atmospheric OH reaction rates using newly developed variable distance weighted zero order connectivity index

A new variable distance weighted zero order connectivity index was used for development of structure-activity relationship for modeling reactivity of OH radical with alkanes and non-conjugated alkenes in the atmosphere. The proposed model is based on the assumptions that the total reaction rate can be obtained by summing all partial reaction rates and that all reaction sites are interrelated by influencing each other. The results suggest that these assumptions are justified. The model was compared with the EPA implemented model in the studied application domain and showed superior prediction capabilities. Further, optimized values of the weights that were used in our model permit some insight into mechanisms that govern the reaction OH + alkane/alkene. The most important conclusion is that the branching degree of the forming radical seems to play a major role in site specific reaction rates. Relative qualitative structural interpretation is possible, e.g. allylic site is suggested to be much more reactive than even tertiary sp3 carbon.Novel modeling software MACI, which was developed in our lab and is now available for research purposes, was used for calculations. Various variable topological indices that are again starting to be recognized because of their great potentials in simplicity, fast calculations, very good correlations and structural information, were implemented in the program.

A probabilistic method for multiaxial HCF based on highly loaded regions below the threshold depth

For High Cycle Fatigue (HCF) assessment of structural components, a variety of evaluation methods that take important factors into account are available. The application of probabilistic evaluation methods is not always straightforward. Investigations show that the methods have limited predictive capability and most often poor transferability of fatigue test results for more complex geometries. Here, a new probabilistic HCF model based on the concept of highly loaded region is proposed. A depth parameter is taking on an important role in the proposed model. Different local stress criteria in combination with the new model and two other evaluation methods are studied. The purpose is to determine the combination of criteria and evaluation method that gives the best transferability. The results show that the new model has superior predictive capability than other models regardless the choice of fatigue stress criteria.

Simulation and prediction of suprapermafrost groundwater level variation in response to climate change using a neural network model

Suprapermafrost groundwater has an important role in the hydrologic cycle of the permafrost region. However, due to the notably harsh environmental conditions, there is little field monitoring data of groundwater systems, which has limited our understanding of permafrost groundwater dynamics. There is still no effective mathematical method and theory to be used for modeling and forecasting the variation in the permafrost groundwater. Two ANN models, one with three input variables (previous groundwater level, temperature and precipitation) and another with two input variables (temperature and precipitation only), were developed to simulate and predict the site-specific suprapermafrost groundwater level on the slope scale. The results indicate that the three input variable ANN model has superior real-time site-specific prediction capability and produces excellent accuracy performance in the simulation and forecasting of the variation in the suprapermafrost groundwater level. However, if there are no field observations of the suprapermafrost groundwater level, the ANN model developed using only the two input variables of the accessible climate data also has good accuracy and high validity in simulating and forecasting the suprapermafrost groundwater level variation to overcome the data limitations and parameter uncertainty. Under scenarios of the temperature increasing by 0.5 or 1.0 degrees C per 10 years, the suprapermafrost groundwater level is predicted to increase by 1.2-1.4% or 2.5-2.6% per year with precipitation increases of 10-20%, respectively. There were spatial variations in the responses of the suprapermafrost groundwater level to climate change on the slope scale. The variation ratio and the amplitude of the suprapermafrost groundwater level downslope are larger than those on the upper slope under climate warming. The obvious vulnerability and spatial variability of the suprapermafrost groundwater to climate change will impose intensive effects on the water cycle and alpine ecosystems in the permafrost region. (C) 2015 Elsevier B.V. All rights reserved.

Reasonable incomplete revascularisation after percutaneous coronary intervention: the SYNTAX Revascularisation Index.

Incomplete revascularisation is common after percutaneous coronary intervention (PCI). While the absolute amount of residual coronary artery disease (CAD) after PCI has been shown to be associated with worse outcomes, whether the proportion of treated CAD is prognostically important remains to be determined. We sought to quantify the proportion of CAD burden treated by PCI and to evaluate its impact on outcomes using a new prognostic instrument - the SYNTAX Revascularisation Index (SRI). The baseline SYNTAX score (bSS) and residual SYNTAX score (rSS) were determined from 2,618 angiograms of patients enrolled in the prospective ACUITY trial. The SRI was then calculated for each patient using the following formula: SRI=(1-[rSS/bSS])×100. Outcomes were examined according to three SRI groups (SRI=100% [complete revascularisation], 50-99%, and <50%). The median bSS was nine (IQR 5, 16), and after PCI the median rSS was one (IQR 0, 6). The median SRI was 85% (IQR 50, 100), and was 100% in 1,079 patients (41.2%), 50-99% in 907 patients (34.6%), and <50% in 632 patients (24.1%). One-year adverse outcomes, including death, were inversely proportional to the SRI. An SRI cut-off of <80% (present in 1,189 [45.4%] patients after PCI) had the best prognostic accuracy for prediction of death (area under the curve 0.60, 95% confidence interval [CI]: 0.53-0.67, p<0.0001). By multivariable analysis, SRI was an independent predictor of one-year mortality (hazard ratio [HR] 2.17, 95% CI: 1.05-4.35, p=0.03). However, when compared to other scores, the rSS showed superior accuracy and predictive capability for one-year mortality. The SRI is a newly described method for quantifying the proportion of CAD burden treated by PCI. Given its correlation with mortality, and pending external validation, the SRI may be useful in assessing the degree of revascularisation after PCI, with SRI ≥80% representing a reasonable goal. However, the rSS showed superior predictive capability for one-year mortality.

Estimating the Lengthy Missing Log Interval Using Group Method of Data Handling (GMDH) Technique

Estimation of well log properties is crucial in identifying the trends and the properties in oil and gas industry, which will enable firms to avoid problems during operation procedures. In this paper, Group Method of Data Handling (GMDH) technique is utilized to generate a generic model with superior prediction capabilities. NeuraLogTM software is used in order to convert the scanned image logs into digit one that can be used by MATLAB code. Group Method of Data Handling is utilized to predict the same missing interval. An aggregate of 601 field data sets were utilized to create the model. These information sets were separated into training, cross validation and testing sets in the degree of 2:1:1. Trend analyses as well as graphical and statistical tools have been utilized in order to assess the model performance.

Isotopic correlation for 242Pu composition prediction: Multivariate regresssion approach

Multivariate regression calibration using multiple linear regression (MLR), principle component regression (PCR) and partial least squares regression (PLSR) algorithm was performed on 238Pu, 239Pu, 240Pu and 241Pu atom% abundances to predict 242Pu isotopic abundance. The MLR algorithm was found to be the best among these three algorithms. The effect of 238Pu composition on the 242Pu abundance prediction was found to be small but significant especially for achieving high accuracy of <0.5%. PCR and PLSR generated nearly identical results and were inferior to the MLR results. A comparison of MLR results with those obtained by employing seven previously reported empirical methods revealed far superior prediction capability of MLR model. Among the seven empirical models, the best prediction capability was found for Bignan correlation containing 238Pu isotopic data. The study clearly demonstrates that the production of 238Pu and 242Pu has some small correlation and the use of 238Pu in isotopic correlation for 242Pu prediction is important to get accurate results.

Flash-point estimation for binary partially miscible mixtures of flammable solvents by UNIFAC group contribution methods

The flash points of binary partially miscible mixtures of flammable solvents were estimated using the flash point prediction model of Liaw et al. (2010) [18] handling liquid phase non-ideality through UNIFAC-type models, which do not need experimentally regressed binary parameters. This predictive model was validated by the experimental data over the entire composition range of seven mixtures. The model described the experimental flash point data well, with the exception of the region close to the span of the two-liquid-phase region nearing the high boiling-point compound of the mixtures because the estimation error in such a span resulted from poor predictive capability of the modified LLE UNIFAC parameters when using VLLE models to estimate the activity coefficient. Because of nonexistence of the disadvantage of VLLE model and superior predictive capability, a model based upon the VLE UNIFAC parameters provided accurate flash point predictions for binary partially miscible mixtures of flammable solvents, as revealed by comparison between predicted and experimental data. In addition, the flash point behaviors for two different types of partially miscible mixtures of flammable solvents are discussed.

Intelligent prediction of fan rotation stall in power plants based on pressure sensor data measured in-situ.

Blower and exhaust fans consume over 30% of electricity in a thermal power plant, and faults of these fans due to rotation stalls are one of the most frequent reasons for power plant outage failures. To accurately predict the occurrence of fan rotation stalls, we propose a support vector regression machine (SVRM) model that predicts the fan internal pressures during operation, leaving ample time for rotation stall detection. We train the SVRM model using experimental data samples, and perform pressure data prediction using the trained SVRM model. To prove the feasibility of using the SVRM model for rotation stall prediction, we further process the predicted pressure data via wavelet-transform-based stall detection. By comparison of the detection results from the predicted and measured pressure data, we demonstrate that the SVRM model can accurately predict the fan pressure and guarantee reliable stall detection with a time advance of up to 0.0625 s. This superior pressure data prediction capability leaves significant time for effective control and prevention of fan rotation stall faults. This model has great potential for use in intelligent fan systems with stall prevention capability, which will ensure safe operation and improve the energy efficiency of power plants.

Experimental Investigation of Flow Condensation in Microgravity

Future manned space missions are expected to greatly increase the space vehicle's size, weight, and heat dissipation requirements. An effective means to reducing both size and weight is to replace single-phase thermal management systems with two-phase counterparts that capitalize upon both latent and sensible heat of the coolant rather than sensible heat alone. This shift is expected to yield orders of magnitude enhancements in flow boiling and condensation heat transfer coefficients. A major challenge to this shift is a lack of reliable tools for accurate prediction of two-phase pressure drop and heat transfer coefficient in reduced gravity. Developing such tools will require a sophisticated experimental facility to enable investigators to perform both flow boiling and condensation experiments in microgravity in pursuit of reliable databases. This study will discuss the development of the Flow Boiling and Condensation Experiment (FBCE) for the International Space Station (ISS), which was initiated in 2012 in collaboration between Purdue University and NASA Glenn Research Center. This facility was recently tested in parabolic flight to acquire condensation data for FC-72 in microgravity, aided by high-speed video analysis of interfacial structure of the condensation film. The condensation is achieved by rejecting heat to a counter flow of water, and experiments were performed at different mass velocities of FC-72 and water and different FC-72 inlet qualities. It is shown that the film flow varies from smooth-laminar to wavy-laminar and ultimately turbulent with increasing FC-72 mass velocity. The heat transfer coefficient is highest near the inlet of the condensation tube, where the film is thinnest, and decreases monotonically along the tube, except for high FC-72 mass velocities, where the heat transfer coefficient is enhanced downstream. This enhancement is attributed to both turbulence and increased interfacial waviness. One-ge correlations are shown to predict the average condensation heat transfer coefficient with varying degrees of success, and a recent correlation is identified for its superior predictive capability, evidenced by a mean absolute error of 21.7%.

Modeling Skewness in Vulnerability Discovery

A vulnerability discovery model attempts to model the rate at which the vulnerabilities are discovered in a software product. Recent studies have shown that the S‐shaped Alhazmi–Malaiya Logistic (AML) vulnerability discovery model often fits better than other models and demonstrates superior prediction capabilities for several major software systems. However, the AML model is based on the logistic distribution, which assumes a symmetrical discovery process with a peak in the center. Hence, it can be expected that when the discovery process does not follow a symmetrical pattern, an asymmetrical distribution based discovery model might perform better. Here, the relationship between performance of S‐shaped vulnerability discovery models and the skewness in target vulnerability datasets is examined. To study the possible dependence on the skew, alternative S‐shaped models based on the Weibull, Beta, Gamma and Normal distributions are introduced and evaluated. The models are fitted to data from eight major software systems. The applicability of the models is examined using two separate approaches: goodness of fit test to see how well the models track the data, and prediction capability using average error and average bias measures. It is observed that an excellent goodness of fit does not necessarily result in a superior prediction capability. The results show that when the prediction capability is considered, all the right skewed datasets are represented better with the Gamma distribution‐based model. The symmetrical models tend to predict better for left skewed datasets; the AML model is found to be the best among them. Copyright © 2013 John Wiley &amp; Sons, Ltd.

Crop Residue Modeling and Mapping Using Landsat, ALI, Hyperion and Airborne Remote Sensing Data

Various studies have demonstrated that spectral indices derived from remotely sensed data can be used to quantify crop residue cover, if adequately calibrated using in situ data. This study evaluates the capability of the Normalized Difference Tillage Index (NDTI) derived from Advance Land Imager (ALI) relative to that of Landsat Thematic Mapper (TM) and the performance of the Cellulose Absorption Index (CAI) derived from Hyperion and airborne hyperspectral data acquired over central Indiana watersheds. A framework based on Cumulative Distribution Function (CDF) matching is also proposed to leverage the superior predictive capability of hyperspectral based indices to improve predictions of multispectral based indices over extended regions. ALI data consistently yielded crop residue models with lower root mean square error (RMSE) values than those developed using Landsat TM data. Hyperspectral based indices were generally superior in predictive capability to the NDTI based predictions. Observation operators derived from the CDF matching method were successful in scaling multiple data sets to achieve models with lower RMSE and improved predictive capability over the entire range of index values.