Statistical Performance Indices Research Articles

Hydrodynamics Modeling of an LSCFB Reactor Using Multigene Genetic Programming Approach: Effect of Particles Size and Shape

AbstractThe multigene genetic programming (MGGP) technique based hydrodynamics models were developed to predict the solids holdups of a liquid-solid circulating fluidized bed (LSCFB) riser. Four different particles were considered to investigate the effects of particle size, shape and density on hydrodynamics behavior of the LSCFB riser. In this regard, two spherical shape glass bead particles (500 and 1200 μm), two irregular shape lava rock particles (500 and 920 μm) were employed as solid phase and water as liquid phase. The MGGP models were developed, relating the solids holdup (${\varepsilon _s}$, output parameter) with eight input parameters. The developed models were first validated by comparing the model predicted and experimental data of solids holdups. The average solids holdups decreased with the increase of net superficial liquid velocity (${U_l} - {U_t}$) and normalized superficial liquid velocity$\left( {\frac{{{U_l}}}{{{U_t}}}} \right)$. Uniform axial solids holdups observed in axial locations (H) except close to the liquid-solid distributor of the riser. The radial non-uniformity of solids holdup observed all radial positions (r/R). In the central region almost flat but increased toward the wall region. The radial profiles of the solid holdup are approximately identical at a fixed average cross-sectional solid holdup for all of the three LSCFB systems of this study. The statistical performance indicators such as the mean absolute percentage error and correlation coefficient are also found to be within acceptable range. All these findings of suggest that the MGGP modeling approach is suitable for predicting effect of particle size and shape on hydrodynamics behavior of the LSCFB system

Developing an innovative soft computing scheme for prediction of air overpressure resulting from mine blasting using GMDH optimized by GA

Air overpressure (AOp) is one of the most important undesirable effects induced by blasting operations in the mining or tunneling projects. Hence, the present precise model for the prediction of AOp would be much beneficial to control the AOp. To this end, the present study proposes a new hybrid of group method of data handling (GMDH) and genetic algorithm (GA). In the other words, the GA is used to optimize the GMDH. The proposed GMDH–GA model was constructed, trained, and tested based on a collection of 84 actual datasets collected from the Shur river dam region. In the modeling, four input parameters were considered: maximum charge per delay, distance between the blasting point and monitoring station, powder factor and rock mass rating. The coefficient of determination (R2), root mean square error (RMSE) and variance account for (VAF), as the statistical performance indices, were used to evaluate the accuracy of the proposed GMDH–GA model. Consequently, the results indicate that the predicted values using the GMDH–GA model are in excellent agreement with the actual data (with the R2 of 0.988), which demonstrate the reliability of the GMDH–GA model.

Evaluation of soil moisture prediction for Gopalkheda sub-catchment, India

ABSTRACT In present study, station-based daily rainfall data and gridded rainfall (0.25° × 0.25°) from India Meteorological Department (IMD) with lumped soil parameters have been used to simulate the soil moisture content within top 90-cm vertical soil profile at an interval depth of 15 cm using 1D numerical unsaturated flow module of coupled MIKE SHE/MIKE 11 for the Gopalkheda sub-catchment, India. The unsaturated flow parameters have been calibrated for the years 1991–1998 and validated for the years 1999–2004. The model has simulated stream flow at sub-catchment outlet and soil moisture content in the sub-catchment using both station-based rainfall and gridded rainfall datasets. The simulated results are compared with observed stream flows and available soil moisture data using statistical performance indices, i.e. Nash-Sutcliffe Efficiency (NSE), Root mean square error (RMSE), coefficient of determination (R 2). The results of calibrated model indicated that performance of hydrological model is highly dependent on nature of rainfall inputs into the model. The overall analysis indicated that gridded rainfall data are promising for stream flow and soil moisture predictions within the sub-catchment. The application of distributed hydrological model would be significant to water resource planner and agronomist in monitoring the water availability and soil moisture within their response sub-catchment.

Numerical and experimental investigation for the effect of permeability of spur dikes on local scour

Abstract The effect of using permeable spur dikes on the produced maximum scour depth compared to that of solid spur dikes is numerically investigated. The numerical model used for such purpose is the Nays-2DH model of the International River Interface Cooperative (iRIC) software package for bed and bank erosion. The model results are verified using the experimental data collected in this study by conducting experiments on five different models of spur dikes having different opening ratios. Using the statistical performance indices, the root mean square error and the coefficient of determination, the results showed an acceptable agreement between the numerical model results for the relative maximum scour depth defined by the ratio of the maximum scour depth to the flow depth and their corresponding observed values. A new empirical equation using nonlinear regression is developed using the experimental data collected in this study and tested with another existing empirical equation available in the literature for their accuracy in determining the relative maximum scour depth.

Evaluation of RSM for Simulating Dispersion of CO2 Cloud in Flat and Urban Terrains

Accurate and reliable computational fluid dynamics (CFD) simulation of pollutant dispersion is essential for protecting human health, and the choice of turbulence model is an important parameter determining the accuracy of simulation results. This paper evaluates the ability of Reyonds stress model (RSM) to predict dispersion of carbon dioxide (CO2) cloud, which is a typical type of heavy gas and similar to some particulate pollutants, in flat and urban terrains. The RSM simulation is conducted with stress-ω model, whereas SST k-ω two-equation model is selected as the benchmark. The simulation results are compared with the available wind tunnel measurements, and statistical performance indicators are used to obtain a comprehensive and quantitative evaluation of the performances of the two turbulence models. The results reveal that stress-ω model exhibits different capacities in flat terrain and urban terrain. Specifically, stress-ω model can present better results than SST k-ω model in flat terrain, and it performs better in the far-field region than in the near-field region. Although SST k-ω model can describe CO2 dispersion more accurately in urban terrain, the concentration distribution reproduced by stress-ω model is still within acceptable range.

A statistical method based on the ensemble probability density function for the prediction of “Wind Days”

Numerical Weather Prediction (NWP) models are often used to predict meteorological events in a deterministic way. In recent years, operational Ensemble Prediction Systems are able to take into account some of the errors affecting the NWP models, and allow to estimate the probability of occurrence. In the traditional approach, this probability is given by the percentage of ensemble members predicting the event. In this study, we propose an alternative method to estimate the probability of occurrence, based on the ensemble probability density function (PDF), which takes into account only random errors unavoidably affecting the model. To estimate its reliability, we compare this method with classical categorical and probabilistic approaches by using different global models: ECMWF, GFS, and GEFS.In particular, we focus on wind speed forecasts in the area around the city of Taranto, located in Apulia region (southeastern Italy), to simulate the events called “Wind Days”, i.e. northwesterly wind above 7 m/s for 3 consecutive hours. Our analysis concerns 34 case studies covering 2016, opportunely chosen to have a balanced dataset of WD and no WD, the latter category mainly including cases that are very difficult to predict, at the border of the two categories. The results show that the probabilistic approaches have a better skill than the categorical ones. Among the probabilistic approaches, the best result (accuracy of 82%) is obtained using the method proposed here, with the control run of GEFS used to estimate the true value and the gamma distribution to model the error distribution.To reduce the systematic error, we test different thresholds and numbers of consecutive hours when the definition of WD is applied to model outputs. All the models show remarkably better performances after these parameters are changed. In particular, our method shows the best performance, with an accuracy of 94%. The analysis on test (leave-one-out strategy in 2016) and validation datasets (66 cases in 2017) confirms the previous outcomes. We test our procedures considering the forecast time intervals of 49–72 and 25–48 h, where similar performances are found. In conclusion, our analysis show that the proposed method presents better performances compared to the traditional approaches for different statistical performance indicators.

Distribution Grids Fault Location employing ST based Optimized Machine Learning Approach

Precise information of fault location plays a vital role in expediting the restoration process, after being subjected to any kind of fault in power distribution grids. This paper proposed the Stockwell transform (ST) based optimized machine learning approach, to locate the faults and to identify the faulty sections in the distribution grids. This research employed the ST to extract useful features from the recorded three-phase current signals and fetches them as inputs to different machine learning tools (MLT), including the multilayer perceptron neural networks (MLP-NN), support vector machines (SVM), and extreme learning machines (ELM). The proposed approach employed the constriction-factor particle swarm optimization (CF-PSO) technique, to optimize the parameters of the SVM and ELM for their better generalization performance. Hence, it compared the obtained results of the test datasets in terms of the selected statistical performance indices, including the root mean squared error (RMSE), mean absolute percentage error (MAPE), percent bias (PBIAS), RMSE-observations to standard deviation ratio (RSR), coefficient of determination (R2), Willmott’s index of agreement (WIA), and Nash–Sutcliffe model efficiency coefficient (NSEC) to confirm the effectiveness of the developed fault location scheme. The satisfactory values of the statistical performance indices, indicated the superiority of the optimized machine learning tools over the non-optimized tools in locating faults. In addition, this research confirmed the efficacy of the faulty section identification scheme based on overall accuracy. Furthermore, the presented results validated the robustness of the developed approach against the measurement noise and uncertainties associated with pre-fault loading condition, fault resistance, and inception angle.

Análise do desempenho de reservatório de uso múltiplo: estudo de caso na sub-bacia Billings

O presente estudo analisou o desempenho do Reservatório Billings, inserido na bacia do Alto Tietê - São Paulo, quanto à sua capacidade de atender as demandas de abastecimento urbano e de geração de energia elétrica concomitantemente. Para isso, foram considerados três cenários de operação para o Reservatório: no primeiro, adotou-se diferentes valores de transferência de água para o Reservatório, além de demanda consuntiva fixa; no segundo, a transferência de águas foi suprimida e considerou-se diferentes demandas consuntivas; no terceiro, empregou-se o reúso potável indireto da água como auxílio para recarga do Reservatório, e considerou-se diferentes demandas consuntivas. Os indicadores de desempenho estatísticos confiabilidade, resiliência e vulnerabilidade e o Índice de Sustentabilidade Hídrica foram calculados. Os resultados obtidos para o Cenário 1, pelo cruzamento entre os fatores critérios de desempenho versus bombeamento do reservatório, indicaram que a confiabilidade, a resiliência e o Índice de Sustentabilidade Hídrica aumentaram, ao passo que a vulnerabilidade do sistema diminuiu significativamente. No Cenário 2, observou-se que o aumento da demanda hídrica implica em reduções significativas da confiabilidade, da resiliência e da sustentabilidade hídrica e um aumento da vulnerabilidade do sistema. Por último, o Cenário 3 demonstrou que a recarga do Reservatório por meio da água de reúso proporciona melhorias no seu desempenho se comparado ao Cenário 2, e beneficia a redução dos impactos ambientais provindos do lançamento dos esgotos na Billings. Concluiu-se que o planejamento de um sistema de Reservatório é um fator essencial para que ele possa satisfazer as múltiplas demandas e que, como uma solução parcial a fim de aumentar o aporte de água na Billings, a inserção da prática do reúso da água é uma ideia importante a ser considerada, pois poderá auxiliar na alimentação hídrica do Reservatório até que a transferência de águas dos rios Tietê e Pinheiros volte a ser viável, como foi projetado inicialmente.

Bio-inspired computational heuristics for Sisko fluid flow and heat transfer models

A novel approach of integrated meta-heuristic computing is designed for Sisko fluid dynamics (SFD) system using artificial neural networks (ANN) based differential equation models being optimized with exploration capabilities of Genetic Algorithms (GAs) and efficient convergence of active-set algorithm (ASA) for different scenarios based on variation of magnetic interaction parameter, stretching parameter, generalized Prandtl and Biot numbers. Governing partial differential equations (PDEs) of SFD system are converted into coupled nonlinear ordinary differential equations (ODEs) through similarity variables and then ANN based models are constructed for the transformed problem. These networks are trained globally using GAs with fine tuning by ASA for speedy optimization. Accurate and consistent convergence of this scheme is established by comparison of results with standard Adams numerical approach and further validated through statistical performance indices.

Mobile Soft Switch Traffic Prediction using Polynomial Neural Networks

This work investigates busy hour traffic demand pattern of mobile soft switch (MSS) over a period of two years and propose the application of Group Method of Data Handling (GMDH) polynomial neural network for seven-days to three-month ahead busy hour (BH) traffic forecasting for effective optimization of network resources. Busy hour call attempt (BHCA) utilization and A-interface utilization key performance indicators (KPI) are used as inputs into GMDH prediction model and BH traffic as model target. The performance of the model was evaluated based on three statistical performance indices: mean absolute percentage error (MAPE), root mean square percentage error (RMSPE) and goodness of fit (R2) values. Experimental results show that R2 value as high as 96% was achieved with the proposed model for both short-term and mid-term forecasting. The GMDH model proves an effective tool for accurate prediction of traffic demand and hence, proper optimization of GSM/GPRS MSS network resources.

Mobile Soft Switch Traffic Prediction using Polynomial Neural Networks

This work investigates busy hour traffic demand pattern of mobile soft switch (MSS) over a period of two years and propose the application of Group Method of Data Handling (GMDH) polynomial neural network for seven-days to three-month ahead busy hour (BH) traffic forecasting for effective optimization of network resources. Busy hour call attempt (BHCA) utilization and A-interface utilization key performance indicators (KPI) are used as inputs into GMDH prediction model and BH traffic as model target. The performance of the model was evaluated based on three statistical performance indices: mean absolute percentage error (MAPE), root mean square percentage error (RMSPE) and goodness of fit (R2) values. Experimental results show that R2 value as high as 96% was achieved with the proposed model for both short-term and mid-term forecasting. The GMDH model proves an effective tool for accurate prediction of traffic demand and hence, proper optimization of GSM/GPRS MSS network resources.

Parameter estimation for Hammerstein control autoregressive systems using differential evolution

In the present study, strength of stochastic computational paradigms is investigated for parameter estimation of Hammerstein control autoregressive (HCAR) model by exploiting differential evolution, genetic algorithms and pattern search methods. Multidimensional and nonlinear nature of the problem emerging in digital signal systems along with noise makes it a challenging optimization task, which is dealt with robustness and effectiveness of stochastic solvers to ensure convergence and avoid trapping in local minima. The performance of meta-heuristic approaches is validated through statistical performance indices based on absolute error, weight deviations and mean squared error. Comparative studies of HCAR system identification established efficacy of the designed methodology based on differential evolution over its counterparts.

Frequency Analysis of Storm-Surge-Induced Flooding for the Huangpu River in Shanghai, China

Shanghai, as a coastal city, is vulnerable to various types of flooding. The floodwalls along the Huangpu River provide protection against typhoon-induced flooding. However, there is limited insight into the actual safety level of the flood defences in Shanghai, and recent failures have highlighted their vulnerability. Therefore, the aims of this paper are to derive a series of new flood frequency curves for the Huangpu River, and to evaluate the level of protection of the floodwall system. This paper analysed over 100 years (1912–2013) of annual maximum water levels for three stations at near-sea, mid-stream and inland locations along the Huangpu River. Best-fit curves were determined for a number of selected probability distributions using statistical performance indicators. As a result, new flood frequency curves of the water levels for different storm surge return periods were produced. The results showed that generalised extreme value (GEV) was identified as the most suitable probability distribution for the datasets. Analysis showed that the current design water levels correspond to exceedance probabilities of 1/500 per year at the near-sea and mid-stream stations, and no more than 1/50 per year at the inland station of the Huangpu River, whereas the intended safety standard is 1/1000 per year. A comparison of the findings with a dataset of the crest heights of the floodwalls showed that the current protection level of the floodwalls along the Huangpu River is expected to be around 1/50 per year in terms of overtopping for the lowest sections. The results of this study can be utilized to provide future recommendations for flood risk management in Shanghai.

Performance evaluation of adaptive neuro-fuzzy inference system, artificial neural network and response surface methodology in modeling biodiesel synthesis from palm kernel oil by transesterification

ABSTRACTModeling capabilities of an adaptive neuro-fuzzy inference system (ANFIS), artificial neural network (ANN) and response surface methodology (RSM) were assessed in the transesterification of esterified palm kernel oil (PKO) with methanol in the presence of KOH as a catalyst. A central composite rotatable design (CCRD) of RSM was applied using methanol/oil ratio (0.25–0.50 v/v), catalyst loading (0.75–2.00 w/v) and reaction time (30–70 min) as the independent variables and palm kernel oil biodiesel (PKOB) yield as the response. Statistical performance indicators showed ANFIS (coefficient of determination, R2 = 0.99, mean absolute error, MAE = 0.21 and mean relative percentage deviation, MRPD = 0.22%) and ANN (R2 = 0.99, MAE = 0.23, MRPD = 0.24%) models describe the process with higher precision and accuracy compared to RSM (R2 = 0.79, MAE = 1.05, MRPD = 1.12%). To maximize the PKOB yield, the process input variables investigated were optimized using an RSM optimization tool and genetic algorithm (GA) coupled with the developed ANFIS, ANN and RSM models. The best combination of the estimated process input variables (methanol/oil ratio 0.48 v/v, catalyst loading 0.86 w/v and reaction time 71.5 min) with highest PKOB yield (99.5 wt.%) was given by ANFIS-GA.

A Multigene Genetic Programming approach for modeling effect of particle size in a liquid–solid circulating fluidized bed reactor

This communication presents the application of Multigene Genetic Programming, a new soft computing technique to investigate the effects of particle size on hydrodynamics behavior of a liquid–solid circulating fluidized bed (LSCFB) riser. The Multigene Genetic Programming based model is developed/trained based on experimental data collected from a pilot scale LSCFB reactor using two different size glass beads (500 & 1200μm) as solid phase and water as liquid phase. The trained Genetic Programming model successfully predicted experimental phase holdups of the LSCFB riser under different operating parameters. It is observed that the model predicted cross-sectional average of solids holdups in the axial directions and radial flow structure are well agreement with the experimental values. The statistical performance indicators including the mean absolute error (∼5.89%) and the correlation coefficient (∼0.982) also show favorable indications of the suitability of Genetic Programming modeling approach in predicting the solids holdup of the LSCFB system.

Multigene genetic programming for sediment transport modeling in sewers for conditions of non-deposition with a bed deposit

It is known that construction of large sewers based on consideration of flow with non-deposition without a bed deposit is not economical. Sewer design based on consideration of flow with non-deposition with a bed deposit reduces channel bed slope and construction cost in which the presence of a small depth of sediment deposition on the bed increases the sediment transport capacity of the flow. This paper suggests a new Pareto-optimal model developed by the multigene genetic programming (MGGP) technique to estimate particle Froude number (Frp) in large sewers with conditions of sediment deposition on the bed. To this end, four data sets including wide ranges of sediment size and concentration, deposit thickness, and pipe size are used. On the basis of different statistical performance indices, the efficiency of the proposed Pareto-optimal MGGP model is compared to those of the best MGGP model developed in the current study as well as the conventional regression models available in the literature. The results indicate the higher efficiency of the MGGP-based models for Frp estimation in the case of no additional deposition onto a bed with a sediment deposit. Inasmuch as the Pareto-optimal MGGP model utilizes a lower number of input parameters to yield comparatively higher performance than the conventional regression models, it can be used as a parsimonious model for self-cleansing design of large sewers in practice.

Uncertainty Analysis and Calibration of Swat Model for Estimating Impacts of Conservation Methods on Streamflow and Sediment Yield in Thika River Catchment, Kenya

Despite their imperative role in water resources management, distributed hydrological models like SWAT require calibration that can be challenging due to uncertainties of parameters involved. Prior to modelling of hydrological processes, these parameters and their uncertainty range need to be identified. The objective of this study was to conduct uncertainty analysis of hydrological processes and to calibrate the Soil and Water Assessment Tool (SWAT) for stream flow and sediment yield modelling in Thika River catchment. Sequential Uncertainty Fitting program (SUFI-2) was used to conduct sensitivity and uncertainty analysis. Stream flow was calibrated and validated between the years 1998 to 2013 for gauging stations 4CB05 and 4CB04. Manual sediments calibration was achieved by constraining the MUSLE parameters using the bathymetric survey data. Two uncertainty indices, p and r factor, were obtained as 0.72 and 0.65, 0.65 and 0.45 during calibration and validation, respectively. Statistical performance indicators showed a good match between the observed and simulated values which indicated that the model was well calibrated for simulation of stream flow and sediments yield in the catchment.

Intelligent computing to analyze the dynamics of Magnetohydrodynamic flow over stretchable rotating disk model

In this study a novel application of neurocomputing technique is presented for nonlinear fluid mechanics problem arising in the model of the flow over stretchable rotating disk in the presence of strong magnetic field. The scheme comprises of the power of effective modelling of neural networks supported with integrated optimization strength of genetic algorithm and interior-point method. The governing partial differential equation of the system is converted to nonlinear systems of simultaneous ordinary differential equations by incorporating the similarity variables. Neural network based approximate differential equation models are formulated for the transformed system that are used to construct the merit function in mean squared error sense. The networks are trained initially by genetic algorithm for the global search and rapid local refinements is attained through efficient interior point method. The given scheme is applied for dynamical analysis of the system model in terms of radial, tangential, axial velocities and heat effects by varying magnetic interaction parameters, unsteadiness factors, disk stretchable magnitudes, and Prandtl numbers. The statistical performance indices based on error from standard numerical solutions are used to validate the correctness, consistency, robustness and stability of the proposed stochastic solver.

Neuro-heuristics for nonlinear singular Thomas-Fermi systems

A neuro-heuristic scheme is design to solve nonlinear singular second order system based on Thomas-Fermi equation using the strength of universal approximation capabilities of feedforward artificial neural networks supported with optimization power of genetic algorithms and sequential quadratic programming. An error function is constructed by differential equations artificial neural networks and optimization of design parameters of networks is carried out initially with genetic algorithms for the global search while sequential quadratic programming algorithm is used for further rapid local refinements. The performance of the design is analyzed by solving variants of Thomas-Fermi equation. Comparison of the results with standard numerical as well as analytical solvers establish the significance of the method on the basis of accuracy and convergence through statistical performance indices.

Application of computational intelligence techniques to forecast daily PM10 exceedances in Brunei Darussalam

Particulate matter (PM10) is the pollutant causing exceedances of ambient air quality thresholds, and the key indicator of air quality index in Brunei Darussalam for haze related episodes caused by the recurrent biomass fires in Southeast Asia. The present study aims at providing suitable forecasts for PM10 exceedances to aid in health advisory during haze episodes at the four administrative districts of the country. A framework based on random forests (RFs), genetic algorithm (GA) and back propagation neural networks (BPNN) computational intelligence techniques has been proposed in which the final prediction is made by the BPNN model. A hybrid combination of GA and RFs is initially applied to determine optimal set of inputs from the initial data sets of largely available meteorological, persistency of high pollution levels, short and long term variations of emissions rates parameters. The inputs selection procedure does not depend on the back propagation training algorithm. The numerical results presented in this paper show that the proposed model not only produced satisfactory forecasts but also consistently performed better via several statistical performance indicators when compared with the standard BPNN and GA optimisation based on back propagation training algorithm. The model also showed satisfactory threshold exceedances forecasts achieving for instance best true predicted rate of 0.800, false positive rate of 0.014, false alarm rate of 0.333 and success index of 0.786 at Brunei-Muara district monitoring station. Overall, the current study has profound implications on future studies to develop a real-time air quality forecasting system to support haze management.