Digital hardware implementation of a radial basis function neural network

Thanh, Nguyen Phan; Kung, Ying‐Shieh; Chen, Seng-Chi; Chou, Hsin-Hung

doi:10.1016/j.compeleceng.2015.11.017

Cited by 20 publications

(16 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Such function is used, for example, in the realizations of the Radial Basis Function Neural Networks (RBF-NNs). NNs of this type are mainly realized using the FPGA platforms [42][43][44][45]. Such realizations, despite some advantages (time and cost) suffer from several limitations.…”

Section: Neighborhood Functionmentioning

confidence: 99%

An Energy-Efficient, Parallel Neighborhood and Adaptation Functions for Hardware Implemented Self-Organizing Maps Applied in Smart Grid

Kolasa

2020

Energies

View full text Add to dashboard Cite

Smart Grids (SGs) can be successfully supported by Wireless Sensor Networks (WSNs), especially through these consisting of intelligent sensors, which are able to efficiently process the still growing amount of data. We propose a contribution to the development of such intelligent sensors, which in an advanced version can be equipped with embedded low-power artificial neural networks (ANNs), supporting the analysis and the classification of collected data. This approach allows to reduce the energy consumed by particular sensors during the communication with other nodes of a larger WSN. This in turn, facilitates the maintenance of a net of such sensors, which is a paramount feature in case of their application in SG devices distributed over a large area. In this work, we focus on a novel, energy-efficient neighborhood mechanism (NM) with the neighborhood function (NF). This mechanism belongs to main components of self learning ANNs. We propose a realization of this component as a specialized chip in the CMOS technology and its optimization in terms of the circuit complexity and the consumed energy. The circuit was realized as a prototype chip in the CMOS 130 nm technology, and verified by means of transistor level simulations and measurements.

show abstract

Section: Neighborhood Functionmentioning

confidence: 99%

An Energy-Efficient, Parallel Neighborhood and Adaptation Functions for Hardware Implemented Self-Organizing Maps Applied in Smart Grid

Kolasa

2020

Energies

View full text Add to dashboard Cite

show abstract

“…And the error between the estimated and given speed is controlled to adjust the PI parameters online by the back‐propagation (BP) NN, so that the PI value can keep the optimal control rule in speed‐closed and current‐closed loops. In , the radial basis function neural network (RBF NN) is trained to optimize its parameters by the supervised learning mechanism based on the stochastic gradient descent (SGD) method and achieves sensorless control. A co‐simulation work validates the effectiveness of RNF NN, and the maximum position error is ±6 ∘ in .…”

Section: Mid‐high‐speed Sensorless Control Strategiesmentioning

confidence: 99%

Review of sensorless control techniques for PMSM drives

Wang

Chen

2019

IEEJ Transactions Elec Engng

View full text Add to dashboard Cite

Permanent magnet synchronous motor (PMSM) drive systems have been increasingly applied in all kinds of high-power AC drives because of their inherent advantages of good performance, rapid dynamic response, as well as high reliability and efficiency. Precise control of the system traditionally depends on the shaft encoder, but the presence of the encoder could result in increased cost and volume, reduced reliability, etc. Therefore, the development of sensorless control techniques is important to avoid the above issues by the elimination of the sensor. Based on the running process of the motor, sensorless control techniques can be classified into four categories: initial position detection methods (IPDMs), low-speed control strategies, mid-high-speed control strategies, and full-speed control strategies. The fundamental principles, their advantages and disadvantages, the control precision, and the application range of these strategies on surface and interior permanent magnet synchronous motors (SPMSMs, IPMSMs) are described in detail in this paper. Finally, the development trend of PMSM sensorless control technologies is discussed. Initial position detection methodsThe initial rotor position is very crucial to the motor starting process by FOC. If the initial position is not accurate, many problems will be caused, such as excessive starting current, small starting torque, reverse rotating, or even start-up failures. At present, according to the motion state of the rotor in the detection process, IPDMs are divided into initial position identification methods with the rotor rotating, and initial position identification methods with rotor at standstill. The former is generally applied to the starting process of applications demanding low steadiness, such as pumps and fans, while the latter is mainly applied to cranes, electric vehicles, and elevator traction machines, which demand stringent requirements for riding comfort. Initial position identification methods with rotor rotatingThe presetting method is the simplest initial position identification method and utilizes the electromagnetic torque caused by the given current to pull the rotor to the fixed location [9][10][11][12]. The process of identification of the rotor's initial position based on the presetting method is shown in Figure 1. In addition, the method is independent of saliency, so it is suitable for both of SPMSM and IPMSM.T e = 1.5p(ψ f i s * sin( θ) + (L d − L q )i 2 s * sin( θ) cos( θ)) (1)

show abstract

“…The method always converges at the same point when trained with the orthogonal least squares algorithm. The advantage of the classifier is that the network has no local minima problem [50].…”

Section: Radial Basis Function Neural Networkmentioning

confidence: 99%

Development of Application Specific Electronic Nose for Monitoring the Atmospheric Hazards in Confined Space

Bakar¹,

Abdullah²,

Saad³

2019

Adv. sci. technol. eng. syst. j.

View full text Add to dashboard Cite

The presence of atmospheric hazards in confined space can contribute towards atmospheric hazards accidents that threaten the worker safety and industry progress. To avoid this, the environment needs to be observed. The air sample can be monitored using the integration of electronic nose (e-nose) and mobile robot. Current technology to monitor the atmospheric hazards is applied before entering confined spaces called pre-entry by using a gas detector. This work aims to develop an instrument to assist workers during pre-entry for atmosphere testing. The developed instrument using specific sensor arrays which were identified based on main hazardous gasses effective value. The instrument utilizes multivariate statistical analysis that is Principal Component Analysis (PCA) for discriminate the different concentrations of gases. The Support Vector Machine (SVM) and Artificial Neural Network (ANN) that is Radial Basis Function Neural Network (RBFNN) are used to classify the acquired data from the air sample. This will increase the instrument capability while the portability will minimize the size and operational complexity as well as increase user friendliness. The instrument was successfully developed, tested and calibrated using fixed concentrations of gases samples. The results proved that the developed instrument is able to discriminate an air sample using PCA with total variation for 99.42%, while the classifier success rate for SVM and RBFNN indicates at 99.28% for train performance and 98.33% for test performance. This will contribute significantly to acquiring a new and alternative method of using the instrument for monitoring the atmospheric hazards in confined space.

show abstract

Digital hardware implementation of a radial basis function neural network

Cited by 20 publications

References 7 publications

An Energy-Efficient, Parallel Neighborhood and Adaptation Functions for Hardware Implemented Self-Organizing Maps Applied in Smart Grid

An Energy-Efficient, Parallel Neighborhood and Adaptation Functions for Hardware Implemented Self-Organizing Maps Applied in Smart Grid

Review of sensorless control techniques for PMSM drives

Development of Application Specific Electronic Nose for Monitoring the Atmospheric Hazards in Confined Space

Contact Info

Product

Resources

About