A New Method for Sensorless Estimation of the Speed and Position in Brushed DC Motors Using Support Vector Machines

Abstract

Abstrac/-Currently, for many applications, it is necessary to know the speecl and position oF motors. This can be achievecl nsing mechanical sensors coupled to the motor shaFt or using sensorless technic¡ues. The sensorless technic¡ues in brushed de motors can be classifiecl into two types: 1) technic¡ues basecl 011 the clynamic brushed de motor moclel ancl 2) technic¡ues basecl 011 the ripple componen! oF the curren!. This paper presents a new methocl, based on the ripple component, for speed ancl position estimation in brushed de motors, using support vector machines. The ¡>roposecl methocl only measures the curren! and detects the pulses in this signal. The motor speed is estimated by using the inverse clistance between the cletected pulses, and the position is estimat. ed by counting ali detected pulses. The ability to cletect ghost pulses and to discard false pulses is the main aclvantage oF this methocl over other sensorless methods. The perFormed tests on two fractional horsepower brushed de motors inclicate that lile methocl works correctly in a wicle range oF speeds ami situations, in which the speed is constan! or varíes clynamically. lndex Terms-Brushed de motor, curren! ripple, de motor, pattern recognition, position, sensorless, speed, support vector machines (SVMs). l. INTRODUCTIONS ENSORLESS techniques estímate the speed and position of motors without mechanical sensors coupled to the motor shaft, measuring only the current and/or the voltage of the mo tors. Sensorless techniques are not a recent idea. as is evidenced by the work of Allured and Strzelewigz [ 1 ]. Nevertheless, due to the complexity of these methods, they have not yet re placed conventional sensors such as encoders, potentiometers, lachometers, Hall effect sensors, or other mechanical sensors coupled to the motor shaft. The main advantages of these, compared to conventional sensors, are as follows: 1) decreased maintenance, number of connections. and cost of the final

Spatial-Aware Object-Level Saliency Prediction by Learning Graphlet Hierarchies

“…For example, in [6], Borji et al modeled top-down visual guidance under the influence of different tasks. Multiple factors, such as scene context, physical actions, and bottom-up saliency, are integrated either through popular statistical tools, such as K-NN and support vector machine (SVM) [14], [39], or through a conventional Bayesian network (BN). Borji et al [5] have further designed a new BN to describe task-driven visual attention, where global scene context, previously attended locations and motor actions, are integrated over time to predict the next attended location.…”

Endpoint Detection in Low Open Area TSV Fabrication Using Optical Emission Spectroscopy

“…, N, where x i denotes the components of the N-dimensional column vector that are the classifier inputs, y i denotes the class labels that are the classifier outputs (and can be 1, −1 for class 1, class −1, respectively), and N is the size of the training set. The linear function [22] that separates both classes is…”

Disturbance suppression method for position-sensorless motion control of DC brushed motor

“…Sensorless vector control has been applied to brushless DC motors, induction motors, and synchronous motors [2]- [4]. On the other hand, sensorless control methods of DC motors estimating rotor position or velocity from current ripple have been proposed [5]- [9]. However, these methods cannot estimate rotor velocity continuously since they use current ripple for estimation.…”