Basis Function Neural Network Technique Research Articles

Prescribed performance adaptive fault‐tolerant control for nonlinear systems with actuator faults and input dead‐zone

AbstractThis paper investigates the prescribed performance adaptive fault‐tolerant control problem for nonlinear systems with actuator faults and input dead‐zone (IDZ). The partial loss of effectiveness fault and bias fault are considered. First, the asymmetric nonlinear IDZ is transformed into an affine function in the backstepping process. Second, the radial basis function neural network technique is applied to deal with the unknown uncertainty of the studied system. The state observer is designed to estimate the unmeasurable states in the nonlinear systems. Third, a fault‐tolerant controller is developed such that all signals are semi‐globally uniformly ultimately bounded for the closed‐loop system, and the tracking error remains within the specified performance requirements even though the actuator faults and IDZ occur. Finally, an electromechanical system example is provided to exhibit the utilizability of the obtained control technique.

Damage detection in cracked structure rotating under the fluid medium through radial basis function neural network technique

Damage detection in cracked structure rotating under the fluid medium through radial basis function neural network technique

Noncontrast MRI in assessing venous reflux of legs using QFlow analysis and radial basis function neural network technique

Since venous reflux is difficult to quantify, triggered angiography non-contrast-enhanced (TRANCE)-magnetic resonance imaging (MRI) is a novel tool for objectively evaluating venous diseases in the lower extremities without using contrast media. This study included 26 pre-intervention patients with superficial venous reflux in the lower extremities and 15 healthy volunteers. The quantitative flow (QFlow) analyzed the phase shift information from the pixels within the region of interest from MRI. The fast and simple radial basis function neural network (RBFNN) learning model is constructed by determining the parameters of the radial basis function and the weights of the neural network. The input parameters were the variables generated through QFlow, while the output variables were morbid limbs with venous reflux and normal limb classification. The stroke volume, forward flow volume, absolute stroke volume, mean flux, stroke distance, and mean velocity of greater saphenous veins from QFlow analysis could be used to discriminate the morbid limbs of pre-intervention patients and normal limbs of healthy controls. The neural network successfully classified the morbid and normal limbs with an accuracy of 90.24% in the training stage. The classification of venous reflux using the RBFNN model may assist physicians in clinical settings.

L1 neural network adaptive fault-tolerant controller for unmanned aerial vehicle attitude control system

In this article, a L1 neural network adaptive fault-tolerant controller is exploited for an unmanned aerial vehicle attitude control system in presence of nonlinear uncertainties, such as system uncertainties, external disturbances, and actuator faults. A nonlinear dynamic inversion controller with sliding mode control law is designed as the outer-loop controller to track the attitude angles quickly and accurately which reduces dependence on model accuracy. A L1 neural network adaptive controller of the inner loop is introduced to compensate the nonlinear uncertainties and have a good attitude tracking. The radial basis function neural network technique is introduced to approximate a lumped nonlinear uncertainty and guarantee the stability and transient performance of the closed-loop system, instead of converting it to a half-time linear system by the parametric linearization method. Simulation results demonstrate the effectiveness of the proposed controller.

Hybrid technique for optimal location and cost sizing of thyristor controlled series compensator to upgrade voltage stability

In this study, a hybrid technique is employed for ameliorating the voltage stability of test system with thyristor controlled series compensator (TCSC) installation. Improved gravitational search algorithm is used to first set the optimal location of TCSC to generate optimal power flow solution under normal and contingency condition while firefly algorithm is utilised to improve voltage stability and TCSC cost sizing, respectively. The proposed technique is implemented in MATLAB working platform and the power flow security and voltage stability is evaluated with IEEE 6, 14 and 30 bus transmission systems. The optimal TCSC placement is governed by new proposed index of apparent power flow index. The performance of the proposed method is deduced to be superior when results are compared with traditional gravitational search algorithm (GSA) and other hybrid methods of fuzzy-GSA and radial basis function neural network techniques.

A hybrid control strategy to control active filter and improve the PQ of non-linear load

Nowadays, the power quality (PQ)-related issues are more absorbed in power electronic devices. For compensating PQ disturbances, various types of power devices are used. The shunt active power filter (SAPF) is one of the power electronics devices that are used for enhancing the PQ. To reduce the PQ disturbances, the performance of SAPF must be improved. In this paper, a hybrid controller is proposed for improving the performance of SAPF and compensating the PQ. The hybrid technique is the combination of cuckoo search algorithm and radial basis function neural network technique which is used for extracting the reference current for compensating the SAPF and reducing the PQ problem. The SAPF is used for reducing the harmonic disturbances. The DC-link voltage is regulated by utilizing the proposed hybrid controller technique. Subsequently, the Fast Fourier Transform analysis is used for measuring total harmonic distortion values of the proposed system. The proposed method is implemented in MATLAB/Simulink platform and its performance is evaluated. The performance of the proposed method is evaluated and compared with the proportional–integral controller and fuzzy controller.

On‐line identification of biomass fuels based on flame radical imaging and application of radical basis function neural network techniques

In biomass fired power plants a range of biomass fuels are used to generate electric power. It is desirable to identify the type of biomass fuels on-line continuously in order to achieve an improved combustion efficiency, and reduced pollutant emissions. This paper presents the recent investigations into the on-line identification of biomass fuels based on the combination of flame radical imaging and radical basis function (RBF) neural network (NN) techniques. The characteristic values of flame radicals (OH*, CN*, CH* and C2*), including the intensity ratio, intensity contour, mean intensity, area and eccentricity, are computed to reconstruct two types of RBF NN, that is, accurate and probabilistic RBF networks. Experimental results obtained for three types of biomass fuels (flour, willow sawdust and palm kernel shell) firing on a laboratory-scale combustion test rig are presented to demonstrate the effectiveness of the proposed method.

Modeling and prediction of wear at revolute clearance joints in flexible multibody systems

This article proposes a numerical approach for the modeling and prediction of wear at revolute clearance joints in flexible multibody systems by integrating the procedures of wear prediction with multibody dynamics. In the approach, the flexible component is modeled based on the absolute nodal coordinate formulation. The contact force in the clearance joint is applied using the continuous contact force model proposed by Lankanrani and Nikravesh and the friction effect is considered using the LuGre friction model. The simulation of wear is performed by an iterative wear prediction procedure based on Archard’s wear model. The radial basis function neural network technique is employed to deal with the pin-on-disc experimental data for obtaining the wear coefficient used in the wear prediction procedure at different contact conditions. The comparison of the wear predicted at the clearance joint in the rigid and flexible planar slider-crank mechanisms demonstrates that the proposed approach can be used to model and predict wear at revolute clearance joints in flexible multibody systems, and the wear result predicted is slightly reduced after taking the flexibility of components into account.