Radial Basis Function Model Research Articles

Daily Runoff Forecasting Using Ensemble Empirical Mode Decomposition and Long Short-Term Memory

Hydrological series data are non-stationary and nonlinear. However, certain data-driven forecasting methods assume that streamflow series are stable, which contradicts reality and causes the simulated value to deviate from the observed one. Ensemble empirical mode decomposition (EEMD) was employed in this study to decompose runoff series into several stationary components and a trend. The long short-term memory (LSTM) model was used to build the prediction model for each sub-series. The model input set contained the historical flow series of the simulation station, its upstream hydrological station, and the historical meteorological element series. The final input of the LSTM model was selected by the MI method. To verify the effect of EEMD, this study used the Radial Basis Function (RBF) model to predict the sub-series, which was decomposed by EEMD. In addition, to study the simulation characteristics of the EEMD-LSTM model for different months of runoff, the GM(group by month)-EEMD-LSTM was set up for comparison. The key difference between the GM-EEMD-LSTM model and the EEMD-LSTM model is that the GM model must divide the runoff sequence on a monthly basis, followed by decomposition with EEMD and prediction with the LSTM model. The prediction results of the sub-series obtained by the LSTM and RBF exhibited better statistical performance than those of the original series, especially for the EEMD-LSTM. The overall GM-EEMD-LSTM model performance in low-water months was superior to that of the EEMD-LSTM model, but the simulation effect in the flood season was slightly lower than that of the EEMD-LSTM model. The simulation results of both models are significantly improved compared to those of the LSTM model.

A random search for discrete robust design optimization of linear-elastic steel frames under interval parametric uncertainty

This study presents a new random search method for solving discrete robust design optimization (RDO) problem of planar linear-elastic steel frames. The optimization problem is formulated with an explicit objective function of discrete design variables, unknown-but-bounded uncertainty in material properties and external loads, and black-box constraint functions of the structural responses. Radial basis function (RBF) models serve as approximations of structural responses. The anti-optimization problem is approximated by an RBF-based optimization problem that is solved using an adaptive strategy coupled with a difference-of-convex functions algorithm. The adaptive strategy is embedded in an iterative process for solving the upper-level optimization problem. This process starts with a set of candidate solutions and iterates through selecting the best candidate among the available candidates and generating new promising candidates by performing a small random perturbation around the best solution found so far to refine the RBF approximations. It terminates when the number of iterations reaches an upper bound, and outputs the optimal solution that is the best solution obtained through the optimization process. Two test problems and two design examples demonstrate that the exact optimal or a good approximate solution can be found by the proposed method with a few trials of the algorithm.

Comparison GIS-Based interpolation methods for mapping groundwater quality in the state of Qatar

Groundwater (GW) quality deterioration is one of the main water security issues in Qatar. GW in Qatar is over-exploited for agriculture activities that caused saltwater intrusion in the coastal areas and some inland GW. The selection of an appropriate interpolation method is crucial for producing a reliable map of spatial variability in environmental research and the decision-making process. In this study, 41 GW samples were used to produce interpolated maps and reveal the spatial variability of GW parameters namely: pH, electrical conductivity (EC), total dissolved solids (TDS), sodium adsorption ratio (SAR), major cations, and anions, and some trace elements, using spatial interpolated methods in geostatistical tool (ArcGIS Software). Multiple mapping methods were used to compare the current results with the national and international guidelines to give a relevant decision. Three different interpolated methods, namely inverse distance weighted interpolation (IDW), radial basis functions (RBFs), and simple Kriging (SK) were used to produce interpolated maps. Cross-validation was applied to assess the best-fit interpolation method. The result shows that the SK model was the optimum interpolated method for the most of the variable data. While the RBFs model was the best interpolation method for selenium and boron data. The IDW model was the best-interpolated method for sodium, TDS, SAR, and potassium data. The IDW and RBFs methods produced interpolation maps that help visualizing the location of GW within and above the guidelines. The analysis of spatial variability in the study area revealed high levels of TDS, EC, SAR, and some cations and anions in the coastal and south Qatar. While the spatial variability for nitrate and boron suggested the influence of human activities. A thorough understanding of the status of GW quality in Qatar is a significant tool for the GW management and decision-makers.

A soft-computing approach to estimate soil electrical conductivity

Soil apparent electrical conductivity (EC a ) is an indirect and rapid measurement for soil salinity, but because of its dependency on some physical and chemical properties of soil in addition to salinity, consideration of the soil extract EC is preferred for monitoring soil salinity, especially in semi-arid areas, though its measurement needs laboratory processes. This study, therefore, sought to develop a multivariable model to estimate the soil EC e from soil EC a , temperature, moisture content, bulk density, and clay percentage, using radial basis function (RBF) artificial neural network (ANN). In the first step, a set of tests was performed in laboratory in Box-Behnken design (BBD) to train the RBF-ANN. The developed RBF estimated the soil EC e with R 2 = 0.99 and RMSE = 0.005 dS.m −1 . Moreover, a quadratic response surface model (RSM) was also developed to compare with the RBF model. The sensitivity analysis revealed that EC a , moisture, bulk density, and temperature had the maximum to minimum effect on the estimation of soil EC e , respectively. In the second step, the RBF and RSM models were validated by another dataset obtained from three sites located in a semi-arid area. They were applied in-field with a multi-sensor portable device. The R 2 and RMSE of the estimation of EC e by the RBF were equal to 0.801 and 0.350 dS.m −1 , respectively. While, R 2 and RMSE of the RSM model were 0.735 and 0.439 dS.m −1 , respectively. The results of the study indicated excellent ability of the RBF-ANN in the rapid and precise estimation of soil EC e . • Modelling soil ECe based on some soil physical and electrical properties. • Developing a portable multi-sensor soil surveying device to estimate soil ECe. • Investigating the effects of soil physical properties on the models. • Validating the model and the device in-field and comparing with the lab results.

Estimation of Sea Level Change in the South China Sea from Satellite Altimetry Data

The South China Sea is China’s largest marginal sea area, and it is rich in oil and gas mineral resources; thus, estimating its sea level changes is of practical significance. Based on linear and nonlinear sea level change characteristics, this paper decomposes 1992–2019 monthly mean sea level anomaly time series in the South China Sea into trend, seasonal, and random terms. This paper compares the seasonal autoregressive integrated moving average (SARIMA) and Prophet models for estimating the trend and seasonal terms and the long short-term memory (LSTM) and radial basis function (RBF) models for estimating random terms, and the more suitable models were selected. A Prophet-LSTM combined model was developed based on the accuracy results. This paper uses the combined model to study the effect of known data length on the experimental results and determines the best prediction duration. The results show that the combined model is suitable for short-term and medium-term estimations of 12–36 months. The accuracy at 36 months is 0.962 cm, which proves that the combined model has high application value for estimating sea level changes in the South China Sea.

Hyperspectral prediction of sugarbeet seed germination based on gauss kernel SVM

How to quickly and accurately select sugarbeet seeds with reliable germination is very important to sugarbeet planting. In this study, the hyperspectral images of 3072 sugarbeet seeds of the same variety were collected, and were successively processed by binarization, morphology, contour extraction and so on. The average spectrum of the single seed image was obtained by image segmentation. Comprehensive analysis of the evaluation parameters of the five spectral preprocessing methods revealed that the second derivative (2D) processing was optimal. Successive projections algorithm (SPA) was used to extract 16 characteristic wavelengths. Support vector machine radial basis function (SVM-RBF), k-nearest neighbor (KNN) and random forest (RF) models were established at the full wavelength and characteristic wavelength respectively to predict the germination of sugarbeet seeds. By analyzing the prediction accuracy of the three models, it was found that the SVM-RBF model provided the highest prediction accuracy in the test set (the prediction accuracy of the full wavelength was 95.5%, and the prediction accuracy of the characteristic wavelength was 92.32%). The research results showed that the hyperspectral image processing technology could accurately predict the germination rate of sugarbeet seeds, and realize the rapid and non-destructive prediction of the germination status of sugarbeet seeds.

The Relationship between Soil Electrical Parameters and Compaction of Sandy Clay Loam Soil

Soil spatial variability mapping allows the delimitation of the number of soil samples investigated to describe agricultural areas; it is crucial in precision agriculture. Electrical soil parameters are promising factors for the delimitation of management zones. One of the soil parameters that affects yield is soil compaction. The objective of this work was to indicate electrical parameters useful for the delimitation of management zones connected with soil compaction. For this purpose, the measurement of apparent soil electrical conductivity and magnetic susceptibility was conducted at two depths: 0.5 and 1 m. Soil compaction was measured for a soil layer at 0–0.5 m. Relationships between electrical soil parameters and soil compaction were modelled with the use of two types of neural networks—multilayer perceptron (MLP) and radial basis function (RBF). Better prediction quality was observed for RBF models. It can be stated that in the mathematical model, the apparent soil electrical conductivity affects soil compaction significantly more than magnetic susceptibility. However, magnetic susceptibility gives additional information about soil properties, and therefore, both electrical parameters should be used simultaneously for the delimitation of management zones.

Prediction of hypericin content in Hypericum perforatum L. in different ecological habitat using artificial neural networks

BackgroundHypericum is an important genus in the family Hypericaceae, which includes 484 species. This genus has been grown in temperate regions and used for treating wounds, eczema and burns. The aim of this study was to predict the content of hypericin in Hypericum perforatum in varied ecological and phenological conditions of habitat using artificial neural network techniques [MLP (Multi-Layer Perceptron), RBF (Radial Basis Function) and SVM (Support Vector Machine)].ResultsAccording to the results, the MLP model (R2 = 0.87) had an advantage over RBF (R2 = 0.8) and SVM (R2 = 0.54) models and it was relatively accurate in predicting hypericin content in H. perforatum based on the ecological conditions of site including soil types, its characteristics and plant phenological stages of habitat. The results of sensitivity analysis revealed that phenological stages, hill aspects, total nitrogen, altitude and organic carbon are the most influential factors that have an integral effect on the content of hypericin.ConclusionsThe designed graphical user interface will help pharmacognosist, manufacturers and producers of medicinal plants and so on to run the MLP model on new data to easily discover the content of hypericin in H. perforatum by entering ecological conditions of site, soil characteristics and plant phenological stages.

A Thermal–Elastic–Plastic Constitutive Model using the Radial Basis Function Neural Network and Application for an Energy Efficient Warm Forming Process

This work presents a thermal–elastic–plastic constitutive equation based on the radial basis function (RBF) artificial neural network and application with the finite element (FE) analysis. In order to capture the stress data in the coupled temperature-strain doamin, a constitutive equation was defined based on the RBF model, and the trained model was validated by test data that were not used in the training. The RBF based constitutive model was then combined with the stress integration and tangent modulus formulation of FE analysis to apply the new model to a warm V-bending process that includes elastic–plastic deformation and elastic recovery under non-isothermal conditions. The heating method was the infrared (IR) local heating method. The results show that the RBF constitutive model can provide good agreement with the experimental data for V-bending in the non-isothermal conditions. The effects of the parameters of the basis function are also discussed in this work.

Application of machine learning for solar radiation modeling

Solar radiation is an important parameter that affects the atmosphere-earth thermal balance and many water and soil processes such as evapotranspiration and plant growth. The modeling of the daily and monthly solar radiation by Gaussian process regression (GPR) with K-fold cross-validation model has been discussed recently. This study evaluated different neural models such as artificial neural network (ANN), support vector machine (SVM), adaptive network-based fuzzy inference system (ANFIS), and multiple linear regression (MLR) for estimating the global solar radiation (daily and monthly) with K-fold cross-validation method. For the appropriate comparison of the models, the randomized complete block (RCB) design applied in the training and test phases. Also, different data sets were evaluated by K-fold cross-validation in each model. The results showed that radial basis function (RBF) model has the lowest error for estimating the monthly and daily solar radiation. In this study, the result of RBF was compared with the GPR models. The conclusion indicated that RBF methodology can predict solar radiation with higher accuracy relative to the GPR model. The results of yearly solar radiation estimation (2009–2014) showed that the RBF model can estimate solar radiation with the MAPE and RMSE of 5.1% and 0.29, respectively. Also, the coefficient of correlation (R2) between actual and estimated values throughout the year is 98% and can be used by the engineers and other researchers for solar and thermal applications.

A fast and non-destructive approach to identify the heavy mineral oil trace evidence based on spectral fusion treatment and chemometrics

As common trace evidence, heavy mineral oil is particularly essential in the field of forensic science. An approach based on spectral fusion analysis and chemometrics was proposed to realize the fast, accurate and nondestructive identification of heavy mineral oil. The spectral data of one hundred and sixty heavy mineral oil samples from different categories and manufacturers were collected by a Fourier transform Raman spectroscopy. These heavy mineral oil samples were analyzed by chemometric methods, including principal component analysis (PCA), a radial basis function (RBF), a multilayer perceptron (MLP), a k-nearest neighbor (KNN) algorithm and Fisher discriminant analysis (FDA), and the results were discussed. The results showed that the models based on the fused spectra had a higher identification capacity than the individual data. Among them, the fusion models based on the first and second derivative spectra were superior to the other models. Compared with the KNN model, the RBF, MLP and FDA models reached 100.0% accuracy for the 5 categories of the 160 samples. For the different brands in the same category, the RBF, MLP, and FDA model further realized 98.1%, 100.0%, and 100.0% accuracy, respectively, which was ideal. As an emerging analysis technology, fusion spectroscopy is worthy of promotion and application in forensic science and other fields. The identification ability of models could be improved and strengthened by derivative processing of the spectra. The FDA model based on the first and second derivative spectral data treated by PCA was a more useful and practical approach than the other models for identifying the 160 heavy mineral oil samples.

Designing a committee of machines for modeling viscosity of water-based nanofluids

Viscosity is a crucial thermophysical feature of a substance that must be accurately determined before designing a system with nanofluid as the working fluid. In this study, the modern technique of committee machine intelligent system (CMIS) is used for establishing a predictive model for the relative viscosity of the water-based nanofluids. The model was developed by considering 1440 experimental data points of different types of water-based nanofluids containing Al2O3, SiC, SiO2, TiO2, CuO, nanodiamond, and Fe3O4 nanoparticles. The CMIS model combines three intelligent models including a multilayer perceptron (MLP) model trained with Levenberg-Marquardt (LM), an MLP model trained by Bayesian Regularization (BR) and a radial basis function (RBF) approach to estimate the relative viscosity of different water-based nanofluids. Statistical and graphical error criteria revealed that the CMIS technique successfully estimates the relative viscosity of all data points over the whole ranges of operational conditions with a mean absolute relative error of approximately 1.25%. According to their precision and performance, the established CMIS system provides the best performance, followed by the BR-MLP, LM-MLP, and RBF models. Moreover, the performance and estimation capability of the CMIS model was verified against 13 theoretical and empirical models.

Using Machine Learning Radial Basis Function (RBF) Method for Predicting Lubricated Friction on Textured and Porous Surfaces

The coefficient of friction (CoF) obtained from tribological tests conducted on textured and porous surfaces was analysed using the machine learning Radial Basis Function (RBF) method. Non-textured and non-porous samples were taken as reference surfaces. Test parameters, such as entrainment velocity and slide-roll ratio (SRR), along with geometric characteristics of surface features (e.g. texture width and depth, coverage area, circularity, spatial distribution and directionality, among others), were selected as training dataset for the machine learning RBF model. The surface features were divided into designed patterns (dimples and grooves) manufactured by laser texturing, and randomised cavities (surface pores) resulted from the sintering process. The principal outcomes of this study are the effective use of the machine learning RBF method for tribological applications, as well as a critical discussion on its feasibility for the experimental dataset selected and the preliminary results obtained. Main results show that the Hardy multiquadric radial basis function provided an overall correlation coefficient of 0.934 for 35 poles. The application of the suggested machine learning technique and methodology can be extended to other experimental results available in the literature to train more robust models for predicting tribological performances of textured and structured surfaces.

Comparative study on different neural networks for network security situation prediction

AbstractThis article mainly studied the performance of different neural networks in the processing network security situation prediction (NSSP). Radial basis function (RBF) and back propagation neural network (BPNN) models were optimized by particle swarm optimization (PSO) algorithm and seeker optimization algorithm (SOA), respectively. Then the PSO‐RBF model and SOA‐BPNN model were obtained, and comparative experiments were carried out on CNCERT/CC data set. The results suggested that the improved models were more accurate in predicting the situation value compared with RBF and BPNN models; the PSO‐RBF mode had three prediction errors, with 0.05 mean square error (MSE) and 0.05 mean absolute error (MAE), and the SOA‐BPNN model had six prediction errors, with 0.2 MSE and 0.13 MAE, which showed that the PSO‐RBF model had better performance. The experimental results show that the PSO‐RBF model has an excellent performance in processing NSSP and can be promoted and applied in practice.

Classification of retinal fundus image using MS-DRLBP features and CNN-RBF classifier

The most common retinal diseases that are to be diagnosed are Diabetic Retinopathy (DR), Age-related Macular Degeneration (AMD) and Choroidal Neovascularization (CNV). For the people above 60 years of age, detection of these retinal diseases is an important task for treatment that reduces the risk of vision loss. Retinal fundus images play a significant role in the detection of DR, AMD and CNV disease diagnosis and treatment. The existing techniques for the detection of DR, AMD and CNV have not fulfilled with the classification accuracy of the retinal diseases effectively. This research work proposes an efficient classification framework for retinal fundus image recognition to overcome these drawbacks. Initially, the input image from the publicly available STARE database is preprocessed with the following three steps (a) Specular reflection removal and smoothing, (b) contrast enhancement and (c) retinal region expansion. With the preprocessed image, the features are extracted using Multi-Scale Discriminative Robust Local Binary Pattern (MS-DRLBP), based on RGB component selection, Gradient operation, and LBP descriptor. Finally, classification was done using hybrid Convolution Neural Network (CNN) and Radial Basis Function (RBF) model (CNN-RBF) which classifies the retinal fundus images into four classes such as DR, AMD, CNV and Normal (NR). Experimental results of the proposed method gives an accuracy of 97.22% compared with the existing other methodologies.

A novel surrogate-assisted evolutionary algorithm with an uncertainty grouping based infill criterion

For tackling expensive optimization problems (EOPs), surrogate-assisted evolutionary algorithms (SAEAs) will run the evolutionary search and then select some promising solutions to be evaluated as predicted by the surrogate models. Different model management criteria for the surrogate models, such as improvement probability, expected improvement, and lower confidence bound, have shown their effectiveness when solving EOPs. In this paper, a novel SAEA with an uncertainty grouping based infill criterion, called SAEA-UGC, is proposed, in which the uncertainty is treated as an indicator to select solutions for training the models. The selected solutions are adopted to train an ensemble surrogate model and a radial basis function model, which run the global search and local search respectively. After obtaining the predicted value and uncertainty for all solutions in global search, they are evenly grouped according to uncertainty and a best predicted solution from each group is selected to form a new population. The global search and local search are cooperative to find the optimal value for the target EOP, i.e., global search or local search will continually run if an improved value for the target EOP can be found in each iteration; otherwise, the switch between global search and local search will happen. The performance of SAEA-UGC is validated when tacking 20 widely used test problems with various properties. The experimental results confirm the superiority of SAEA-UGC over four representative SAEAs in solving a majority of test EOPs.

Mechanical properties of UN-5100 envelope material for stratospheric airship

ABSTRACTStratospheric airships are promising aircraft, usually designed as a non-rigid airship. As an essential part of the non-rigid airship, the envelope plays a significant role in maintaining its shape and bearing the external force load. Generally, the envelope material of a flexible airship consists of plain-weave fabric, composed of warp and weft fibre yarn. At present, biaxial tensile experiments are the primary method used to study the stress–strain characteristics of such flexible airship materials. In this work, biaxial tensile testing of UN-5100 material was carried out. The strain on the material under unusual stress and the stress ratio were obtained using Digital Image Correlation (DIC) technology. Also, the stress–strain curve was corrected by polynomial fitting. The slope of the stress–strain curve at different points, the Membrane Structures Association of Japan (MSAJ) standard and the Radial Basis Function (RBF) model were compared to identify the stress–strain characteristics of the materials. Some conclusions on the mechanical properties of the flexible airship material can be drawn and will play a significant role in the design of such envelopes.

A Novel Method for Output Characteristics Calculation of Electromagnetic Devices using Multi-kernel RBF Neural Network

The action performance and reliability of electromagnetic devices is critical to the entire working system. In this paper, a new method for calculating the output characteristics of electromagnetic devices is proposed. This method uses the multi-kernel radial basis function neural network (MK-RBFNN) approximation modeling by the finite element calculation results at the key nodes. It obtains the output response of the electromagnetic device under different coil voltages and air gaps. The key of establishing a MK-RBFNN is to obtain the weight coefficients of each single-kernel radial basis function (RBF) model by using a heuristic weighting strategy. When the electromagnetic output characteristics is calculated in the optimization design of the electromagnetic device, this method solves the problem that the traditional method is difficult to balance the calculation accuracy and speed. The effectiveness of the method is verified by the calculation results of the electromagnetic torque of a typical electromagnetic relay.

Monitoring and predicting Fusarium wilt disease in cucumbers based on quantitative analysis of kinetic imaging of chlorophyll fluorescence.

Cucumber (Cucumis sativus L.) is a widely cultivated and economically profitable crop. However, Fusarium wilt disease can seriously affect cucumber yields, as it is difficult to prevent and eliminate. Therefore, a reliable method is needed for the rapid and early detection of Fusarium infection in cucumbers, which could be provided via the kinetic imaging of chlorophyll fluorescence (ChlF). In this study, ChlF imaging and kinetic parameters were utilized with gray and radial basis function models to monitor cucumber Fusarium wilt disease. The results indicate that the disease can be detected and predicted using this imaging technique before symptoms become visible.

Deep learning based reference model for operational risk evaluation of screw chillers for energy efficiency

Operational risk evaluation for chillers is beneficial for reducing building energy wastage. An accurate reference model can significantly improve the evaluation performance. This paper presents a novel framework for developing and applying the reference model by integrating density clustering and deep learning. Unsupervised density-based spatial clustering of applications with noise (DBSCAN) is adopted to construct the library of operating conditions and recognize operating pattern of chillers. Deep learning approach, deep belief network (DBN), is presented to learn all process data in each operating pattern. Subsequently, multiple DBN models are developed for matching various operating patterns in the condition library. A simple strategy for tuning the hyperparameters of DBN is further presented to obtain a better performance. The prediction and generalization abilities of the proposed approach are validated and compared with multivariate linear regression (MLR), support vector regression (SVR) and radial basis function (RBF) models based on the experimental data obtained from a real screw chiller. Results reveal that the proposed method yields a significant performance advantage than MLR, SVR and RBF models, especially for the extended conditions in actual applications, the mean relative errors of MLR, SVR, RBF and the proposed method are 5.8%, 11.41%, 13.73%, and 2.11%, respectively.