RBF Model Research Articles

Development of Neuroxellar Algorithms for Adaptive Pi Adjustable Speed Control Parameters in Mining

This article discusses the adaptive Pi regulator based on the basic function of a neural network (RBF) and it is used to control the velocity of a vector directional asynchronous motor. The structure of the control circuit consists of the RBF identifier of the standard Pi controller. The RBF Identifier is used to identify the Jacobian meaning of the asynchronous motor online. The "online" change of neural network parameters is performed by the gradient access method without prior training. Pi controller training is performed by using the "online" identification RBF model. It is advisable to test this controller under different conditions to finally make sure the reliability of the management technology we offer. Research has shown that the proposed controller provides good steadiness and stability of the steering system compared to what a conventional Pi controller provides.

Modelling rate of penetration in drilling operations using RBF, MLP, LSSVM, and DT models

One of the most important problems that the drilling industry faces is drilling cost. Many factors affect the cost of drilling. Increasing drilling time has a significant role in increasing drilling costs. One of the solutions to reduce drilling time is to optimize the drilling rate. Drilling wells at the optimum time will reduce the time and thus reduce the cost of drilling. The drilling rate depends on different factors, some of which are controllable and some are uncontrollable. In this study, several smart models and a correlation were proposed to predict the rate of penetration (ROP) which is very important for planning a drilling operation. 5040 real data points from a field in the South of Iran have been used. The ROP was modelled using Radial Basis Function, Decision Tree (DT), Least Square Vector Machine (LSSVM), and Multilayer Perceptron (MLP). Bayesian Regularization Algorithm (BRA), Scaled Conjugate Gradient Algorithm and Levenberg–Marquardt Algorithm were employed to train MLP and Gradient Boosting (GB) was used for DT. To evaluate the accuracy of the developed models, both graphical and statistical techniques were used. The results showed that DT-GB model with an R2 of 0.977, has the best performance, followed by LSSVM and MLP-BRA with R2 of 0.971 and 0.969, respectively. Aside from that, the proposed empirical correlation has an acceptable accuracy in spite of simplicity. Moreover, sensitivity analysis illustrated that depth and pump pressure have the highest effects on ROP. In addition, the leverage approach approved that the developed DT-GB model is valid statistically and about 1% of the data are suspected or out of the applicability domain of the model.

Nonlinear system identification for speech model using linear predictive coefficients based radial basis function

In this paper, the authors have proposed a Linear Predictive coefficient-based Radial basis function neural network model (LPC-RBF) to solve the system identification problem. A recorded speech signal is considered to develop the model. The objective is to develop a machine learning model with the LPC values of the speech signal. The parameters of a speech signal are directly estimated using linear predictive coefficients. The proposed model is considered as the ARX model and trained with the LPC value of the speech signal. The comparison results are shown in the table.The performance is evaluated through mean square error (MSE). From the result analysis, the superiority of the proposed model is analyzed. The proposed model archives the best final MSE value 0.0027, as compared to the basic RBF model. From the comparison table, it can be seen that LPC-RBF has better performance as compared with the RBF model.

Heavy-load train model based on RBF neural network refers to the adaptive brake control strategy

Heavy-load train braking control is an important automatic driving function. The braking mode mainly used in China is air braking, which has great nonlinear characteristics. In order to reduce the influence of system uncertainty on control, improve the control accuracy and ensure the safety and efficiency of train transportation, an adaptive brake control strategy is designed.Firstly, establish the state space model of braking system according to the dynamic process of overloaded train braking, use the adaptive method and design the RBF model to track the train speed. Finally, the actual data from CK1466.1 to CK1431.3 in the line is used as the reference model. Experimental results The RBF neural network model designed by the reference adaptive controller realizes the tracking control of a given speed curve with high accuracy, verifying the effectiveness of the strategy.

A Virtual Combustion Sensor Based on Ion Current for Lean-Burn Natural Gas Engine

In this study, an innovative sensor was designed to detect the key combustion parameters of the marine natural gas engine. Based on the ion current, any engine structurally modified was avoided and the real-time monitoring for the combustion process was realized. For the general applicability of the proposed sensor, the ion current generated by a high-energy ignition system was acquired in a wide operating range of the engine. It was found that engine load, excess air coefficient (λ) and ignition timing all generated great influence on both the chemical and thermal phases, which indicated that the ion current was highly correlated with the combustion process in the cylinder. Furthermore, the correlations between the 5 ion current-related parameters and the 10 combustion-related parameters were analyzed in detail. The results showed that most correlation coefficients were relatively high. Based on the aforementioned high correlation, the novel sensor used an on-line algorithm at the basis of neural network models. The models took the characteristic values extracted from the ion current as the inputs and the key combustion parameters as the outputs to realize the online combustion sensing. Four neural network models were established according to the existence of the thermal phase peak of the ion current and two different network structures (BP and RBF). Finally, the predicted values of the four models were compared with the experimental values. The results showed that the BP (with thermal) model had the highest prediction accuracy of phase parameters and amplitude parameters of combustion. Meanwhile, RBF (with thermal) model had the highest prediction accuracy of emission parameters. The mean absolute percentage errors (MAPE) were mostly lower than 0.25, which proved a high accuracy of the proposed ion current-based virtual sensor for detecting the key combustion parameters.

Forecasting of groundwater level fluctuations using a hybrid of multi-discrete wavelet transforms with artificial intelligence models

Abstract Groundwater is often one of the significant natural sources of freshwater supply, especially in arid and semi-arid regions, and is of paramount importance. This study provides a new and high accurate technique for forecasting groundwater level (GWL). The artificial intelligence (AI) models include the artificial neural network (ANN) of multi-layer perceptron (MLP) and radial basis function network (RBF), and adaptive neural-fuzzy inference system (ANFIS) models. Input data to the models is the monthly average GWL of 17 piezometers. In this study, a preprocessing of data including the discrete wavelet transform (DWT) and multi-discrete wavelet transform (M-DWT) simultaneously was utilized. The results showed that the hybrid M-DWT-ANN, M-DWT-RBF, and M-DWT-ANFIS models compared to the DWT-ANN, DWT-RBF, and DWT-ANFIS models as well as than regular ANN, RBF, and ANFIS models, had the highest accuracy in forecasting GWL for the 1-, 2-, 3-, and 6-months ahead. Also, the M-DWT-ANN model had the best performance. Overall, the results illustrated that using the M-DWT method as preprocessing of input data can be a valuable tool to increase the predictive model's accuracy and efficiency. The results of this study indicate the potential of M-DWT-AI hybrid models to improve GWL forecasting.

Design of Aging Smart Home Products Based on Radial Basis Function Speech Emotion Recognition.

The rapid development of computer technology and artificial intelligence is affecting people’s daily lives, where language is the most common way of communication in people’s daily life. To apply the emotion information contained in voice signals to artificial intelligence products after analysis, this article proposes a design based on voice emotion recognition for aging intelligent home products with RBF. The authors first aimed at a smart home design, and based on the problem of weak adaptability and learning ability of the aging population, a speech emotion recognition method based on a hybrid model of Hidden Markov/Radial Basis Function Neural Network (HMM/RBF) is proposed. This method combines the strong dynamic timing modeling capabilities of the HMM model and the strong classification decision-making ability of the RBF model, and by combining the two models, the speech emotion recognition rate is greatly improved. Furthermore, by introducing the concept of the dynamic optimal learning rate, the convergence speed of the network is reduced to 40.25s and the operation efficiency is optimized. Matlab’s simulation tests show that the recognition speed of the HMM/RBF hybrid model is 9.82–12.28% higher than that of the HMM model and the RBF model alone, confirming the accuracy and superiority of the algorithm and model.

Control Method for Improving Voltage Response and Ripple of Isolated DC-DC Converter

Affected by the abrupt changes of load and unregulated first-stage rectifiers, it is difficult for the output voltage of the isolated DC-DC converter with traditional proportional-integral (PI) control to meet the needs of industrial applications such as automotive electrophoretic painting. A voltage response improving the method based on a radial basis function neural network (RBFNN) is proposed in this study for the output voltage mutation of DC converter caused by load mutation. By detecting the output voltage deviation and load current value, the RBF model gives the compensation amount of the duty ratio to enhance the PI controller to speed up the voltage regulation and improve voltage response. For the six-pulse voltage ripple introduced by the unregulated rectifier in the previous stage, a delay link is adopted, where the compensation amount of the duty ratio and its delay time are reasonably designed to reduce the ripple value. The experimental test platform is built, and the proposed method is verified. The results show that the proposed method can effectively improve the output voltage response and ripple as well as improve the output voltage quality of the DC converter, which is of certain engineering application value.

Mobile Performance Intelligent Evaluation of IoT Networks Based on DNN

The rapid development of the sensor equipment has promoted the rapid growth of the Internet of Things (IoT). The IoT has been widely employed in the multidimensional signal processing and gradually formed the IoT networks. Mobile communication promotes the wide application of the IoT networks. In this study, the transmit antenna selection (TAS) scheme is employed to investigate the average symbol error probability (ASEP) performance of mobile IoT networks over the 2-Rayleigh channels. We first employ moment-generating function (MGF) approach to derive the exact ASEP expressions. We also investigate the outage probability (OP) performance and derive OP expressions. Employing the deep neural network (DNN), an OP intelligent prediction algorithm is proposed. Then, the numerical simulations are conducted to confirm the ASEP and OP performance analysis. The effect of different channel parameters is also analyzed. Compared with Nakagami and Rayleigh channel models, the 2-Rayleigh model has 83.6% and 59.1% increase in ASEP values, respectively. Compared with ELM and RBF models, the DNN model has 31.7% and 22.5% increase in OP prediction accuracy, respectively.

Machine learning models for prediction the Higher Heating Value (HHV) of Municipal Solid Waste (MSW) for waste-to-energy evaluation

Time-resolved knowledge of physico-chemical properties of Municipal Solid Waste (MSW) materials and their thermal energy content is one of the important subjects needed to build waste incineration power plants around the world. For this purpose, machine learning models were used to predict the Higher Heating Value (HHV) of MSW based on the initial materials. Four types of machine learning methods: Radial Bias Function Artificial Neural Network (RBF-ANN), Multilayer Perceptron Artificial Neural Network (MLP-ANN), Support Vector Machine (SVM) and Adaptive Nero-Fuzzy Inference System (ANFIS) were used for modeling the HHV with six different inputs (carbon, water, hydrogen, oxygen, nitrogen, sulfur, and ash). The results showed that RBF-ANN can predict the HHV of MSW with higher accuracy than other models. The overall Mean Absolute Percentage Error (MAPE) for MLP-ANN, SVM and ANFIS models were 7.3, 11.77 and 23.76%, respectively. The MAPE of the best topology for RBF model (6-17-1) with spread factor of 0.8 reach to 0.45%. Finally, the results of this study proved that ANN's can be used as a practical tool with high accuracy and reliability for design and management of waste incineration plants.

Laser Beam Welded Aluminum-Titanium Dissimilar Sheet Metals: Neural Network Based Strength and Hardness Prediction Model

Abstract ‘Laser Beam Welding (LBW) is a welding technique used to join pieces of metal or thermoplastics with the aid of laser’. The beam offers a concerted heat source, which enabled higher, deeper welds and narrower welding rates. The procedure is commonly exploited in higher volume appliances using mechanization. It is dependent on penetration or keyhole mode welding. This paper intends to design a novel prediction model on LBW using the Optimized Neural Network (NN) framework. The input to the optimized NN is the welding properties like ‘Laser power, welding speed, offset, shielding gas, flow/pressure, focal distance and frequency (where power, speed and offset gets varied)’ that directly predict the hardness and tensile strength of welds since the NN is already trained with the provided data. In order to make the prediction model more accurate, this paper aims to train the NN using a new improved Trial Integer-based Whale Optimization Algorithm (TI-WOA) via updating the weight. Finally, the betterment of the suggested scheme is validated with respect to error analysis. Accordingly, from the analysis, it is observed that the proposed methods are 50%, 13.33%, 6.67% and 4% better than ANN-BP, RBF, ANN-GA and NN-WOA models, respectively, at 70th learning percentage.

Properties of water-based fly ash-copper hybrid nanofluid for solar energy applications: Application of RBF model

The hybrid nanofluids were used as absorber fluids in solar energy applications, which could further increase the efficiency of solar devices. The use of nanofluids in solar devices with the laminar and turbulent flow has received much attention. Presently, the effect of temperature and concentration on thermal conductivity and viscosity of fly ash-copper (80:20% by volume) hybrid nanofluid is investigated. The thermal conductivity and viscosity measurements were carried in the temperature range of 30–60 °C for a concentration range of 0–4.0 vol%. The nanoparticles and nanofluids were characterized by XRF, XRD, SEM, TEM, zeta potential, and DLS techniques. The maximum augmentation in the hybrid nanofluid's dynamic viscosity and thermal conductivity at a concentration of 4 vol% is 45.18% and 49.8%, respectively, at 30 and 60 °C. Correlations to estimate the hybrid nanofluid's dynamic viscosity and thermal conductivity have been proposed considering the results obtained from the present study. A radial basis function-based neural network is used to model nanofluids' effective thermal conductivity and relative viscosity. The outcomes of the experiments were used to calculate the Mouromtseff number and heat transfer efficiency for solar energy applications. • HyNF shows Newtonian behavior in the studied range. • The maximum augmentation in viscosity is 45.2%. • The highest amplification in thermal conductivity is 49.8%. • Radial basis function based neural network was used.

Execution Time Prediction for Cypher Queries in the Neo4j Database Using a Learning Approach

With database management systems becoming complex, predicting the execution time of graph queries before they are executed is one of the challenges for query scheduling, workload management, resource allocation, and progress monitoring. Through the comparison of query performance prediction methods, existing research works have solved such problems in traditional SQL queries, but they cannot be directly applied in Cypher queries on the Neo4j database. Additionally, most query performance prediction methods focus on measuring the relationship between correlation coefficients and retrieval performance. Inspired by machine-learning methods and graph query optimization technologies, we used the RBF neural network as a prediction model to train and predict the execution time of Cypher queries. Meanwhile, the corresponding query pattern features, graph data features, and query plan features were fused together and then used to train our prediction models. Furthermore, we also deployed a monitor node and designed a Cypher query benchmark for the database clusters to obtain the query plan information and native data store. The experimental results of four benchmarks showed that the average mean relative error of the RBF model reached 16.5% in the Northwind dataset, 12% in the FIFA2021 dataset, and 16.25% in the CORD-19 dataset. This experiment proves the effectiveness of our proposed approach on three real-world datasets.

A New Neural Network Based Sliding Mode Adaptive Controller for Tracking Control of Robot Manipulator

In this paper a New RBF Neural Network based Sliding Mode Adaptive Controller (NNNSMAC) for Robot Manipulator trajectory tracking in the presence of uncertainties and disturbances is introduced. The research offers a learning with minimal parameter (LMP) technique for robotic manipulator trajectory tracking. The technique decreases the online adaptive parameters number in the RBF Neural Network to only one, lowering computational costs and boosting real-time performance. The RBFNN analyses the system's hidden non-linearities, and its weight value parameters are updated online using adaptive laws to control the nonlinear system's output to track a specific trajectory. The RBF model is used to create a Lyapunov function-based adaptive control law. The effectiveness of the designed NNNSMAC is demonstrated by simulation results of trajectory tracking control of a 2 dof Robotic Manipulator. The chattering effect has been significantly reduced.

Runoff modeling in Kolar river basin using hybrid approach of wavelet with artificial neural network

AbstractIn this paper, the Kolar River watershed, Madhya Pradesh is taken as the study area. This study area is located in Narmada River in Central India. The data set consists of monthly rainfall of three meteorological stations, Ichhawar, Brijesh Nagar, and Birpur rainfall stations from 2000 to 2018, runoff data at Birpur and temperature data of Sehore district. In this paper, radial basis function neural network models have been studied for generation of rainfall–runoff modeling along with wavelet input and without wavelet input to the RBF neural network. A total of 15 models were developed in this experiment based on various combinations of inputs and spread constant of RBF model. The evaluation criteria for the best models selected are based on R2, AARE, and MSE. The best predicting model among the networks is model 8, which has input of R(t-1), R(t-2), R(t-3), R(t-4), and Q(t-1). For the RBFNN model, the maximum value of R2 is 0.9567 and the lowest values of AARE and MSE are observed. Similarly, for the WRBFNN model, the maximum value of R2 is 0.9889 and the lowest values of AARE and MSE are observed. WRBF performs better than RBF with any data processing techniques which shows the proposed model possesses better predictive capability.

A COMPARATIVE STUDY FOR THE PREDICTION OF HEART ATTACK RISK AND ASSOCIATED FACTORS USING MLP AND RBF NEURAL NETWORKS

Abstract— Aim: The aim of this study is to develop a predictive classification model that can identify risk factors for heart attack disease.
 Materials and Methods: In the study, patients with low and high probability of having a heart attack were examined. Variable importance was calculated to identify risk factors. The radial basis function and multilayer perception neural networks were used to compare the classification prediction results.
 Results: MLP model criteria; Accuracy 0.911, F1 score 0.918, Specificity 0.92, Sensitivity 0.903, while RBF model criteria were obtained as accuracy 0.797, F1 score 0.812, Specificity 0.84, Sensitivity 0.765. The first three most important factors that may be associated with having a heart attack were obtained as trestbps, oldpeak, and chol. 
 Conclusion: According to the prediction results of the heart attack, it can be said that the model created with the MLP neural network has more successful predictions than the model created with the RBF neural network. In addition, estimating the importance values of the factors most associated with heart attack (obtaining the most important biomarkers that may cause heart attack) is a promising result for the diagnosis, treatment and prognosis of the disease.
 
 Keywords— Heart Attack, machine learning, neural networks, classification, variable importance.

Using UAV and Multispectral Images to Estimate Peanut Maturity Variability on Irrigated and Rainfed Fields Applying Linear Models and Artificial Neural Networks

Using UAV and multispectral images has contributed to identifying field variability and improving crop management through different data modeling methods. However, knowledge on application of these tools to manage peanut maturity variability is still lacking. Therefore, the objective of this study was to compare and validate linear and multiple linear regression with models using artificial neural networks (ANN) for estimating peanut maturity under irrigated and rainfed conditions. The models were trained (80% dataset) and tested (20% dataset) using results from the 2018 and 2019 growing seasons from irrigated and rainfed fields. In each field, plant reflectance was collected weekly from 90 days after planting using a UAV-mounted multispectral camera. Images were used to develop vegetation indices (VIs). Peanut pods were collected on the same dates as the UAV flights for maturity assessment using the peanut maturity index (PMI). The precision and accuracy of the linear models to estimate PMI using VIs were, in general, greater in irrigated fields with R2 > 0.40 than in rainfed areas, which had a maximum R2 value of 0.21. Multiple linear regressions combining adjusted growing degree days (aGDD) and VIs resulted in decreased RMSE for both irrigated and rainfed conditions and increased R2 in irrigated areas. However, these models did not perform successfully in the test process. On the other hand, ANN models that included VIs and aGDD showed accuracy of R2 = 0.91 in irrigated areas, regardless of using Multilayer Perceptron (MLP; RMSE = 0.062) or Radial Basis Function (RBF; RMSE = 0.065), as well as low tendency (1:1 line). These results indicated that, regardless of the ANN architecture used to predict complex and non-linear variables, peanut maturity can be estimated accurately through models with multiple inputs using VIs and aGDD. Although the accuracy of the MLP or RBF models for irrigated and rainfed areas separately was high, the overall ANN models using both irrigated and rainfed areas can be used to predict peanut maturity with the same precision.

RBF Model for the Mass Loss of a Brass in Cavitation Field

This article aims at presenting the model of the mass loss of a brass sample in ultrasonic cavitation field in saline water. The experiments done for data collecting was performed in three scenarios. In the first one, the high frequency generator worked at three power levels - 80 W, at the second one - at 120 W, and in the third one - at 180 W. The Model has been built using the series of the mass loss on surface.

A Study of the Performances of Widely Used External Magnetic Field Models in the Outer Zone of the Earth's Radiation Belts by Comparing the Field Observations From Van Allen Probe‐A and the Model Estimations

AbstractUsing magnetic field observations from Van Allen Probe‐A (VAP‐A), we evaluated the performances of 13 widely used external magnetic field models in the Earth's outer radiation belt region in detail. It is shown that later models tend to match observations better than earlier models except for T03s, which has a top performance similar to the latest TA16_RBF model. Additionally, in its applicable domain from quiet to moderately disturbed geomagnetic conditions, T02 has comparable performance to T03s. In the dusk, dawn and night sectors, highly stretched magnetic fields are seen in both the observations and model output between LmIGRF ∼5.7 and 6.1, which suggests the existence of steep radial gradients in the plasma pressures inside the outer belt region. The contributions from the currents located in the dusk sector during a disturbed period, such as the partial ring current and the substorm current wedge, are not adequately accounted for by the models, and the contribution of field‐aligned currents may also be underestimated. The results of this study will help select reliable external field models for future radiation belt studies and help assess uncertainties caused by the selected models. In addition, detailed analysis of the field components depicted by the empirical models could shed light on understanding the contributions from different current systems in this region.

Calibration uncertainty of MEMS thermopile imagers for quantitative temperature measurement

This article presents the calibration procedure of three MEMS thermopile imagers from the same manufacturer. We did not find in the literature a study that put the accuracy of these sensors to the test. We divided the methodology into the following parts: data collection, mathematical modeling (radiation heat transfer, non-uniformity correction, and regression models), estimation of uncertainty sources, and expanded uncertainty. The Guide to the Expression of Uncertainty in Measurement was the base of our uncertainty estimations. We considered the flat plate radiator, the radiation heat transfer model, the correction model, and the IR sensors as uncertainty sources. The coefficients ABC (Sakuma-Hattori model) and RBF are considered dependent variables in the regression models (we did not calculate ABC or RBF using the half-width and the operating wavelength range). Results led to a maximum deviation of 0.46 °C in the RBF model and 0.49 °C in the Sakuma-Hattori model. Uncertainties related to the flat plate radiator, mainly uniformity and temperature, were dominant in the uncertainty budget, followed by the mathematical routines and temporal noise. Finally, we compared our expanded uncertainties with the scientific literature and the manufacturer’s estimations. The proposed methodology resulted in a minor expanded uncertainty, compared to the uncertainty provided by the manufacturer. One of the three MEMS thermopile imagers presented pixels with relatively high expanded uncertainty, which illustrates the importance of a pixel-by-pixel analysis. Readers from this journal can benefit from this article as an alternative to check the reliability and accuracy of infrared thermometers..