Wyniki 1-1 spośród 1 dla zapytania: authorDesc:"Dariusz MAJCHRZAK"

Faults detection in PMSM drive using Artificial Neural Network DOI:10.15199/48.2017.06.06

Czytaj za darmo! »

Introduction The permanent magnet synchronous motors (PMSM) are becoming increasingly popular in industry due to their high power density, low inertia and high efficiency. Thanks to their excellent dynamic performance, they are widely used in robots, machine tool, winders and similar systems that require precise speed and torque control. Nowadays, electrical drives often work in human life-critical systems, where high reliability is required [1]. In these applications the traditional control algorithms do not provide a sufficient safety, so fault tolerant control (FTC) is commonly used. FTC algorithms require information about type and location of fault [2], therefore the fault detection and diagnosis systems are necessary. There are many methods of fault detection and identification. They can be divided into signal processing based and model-based categories. First of them uses measured signals analysis methods such as spectral analysis [3] or wavelet transform [4]. In general, they only uses output signals of drive, but no input signals, so influence of input on output may be ignored [5]. Modelbased methods use information about structure and parameters of dynamic model of plant. These include state estimation methods, for example observers or Extended Kalman Filter [6]. Moreover, model parameters estimation methods like recursive last square algorithm can be used [7]. Model-based methods generate residuals, by estimating output signals (or parameters of the plant) and computing estimation error vector [8]. Next the residual evaluation system generates diagnosis. Fig. 1 presents the block diagram of model-based method of fault detection. Symbols shown in Fig. 1 are u - plant inputs, y - plant outputs, z - disturbance, f - fault, and r - generated residuals. The main disadvantage of mentioned methods is the need for a reliable model [5]. In this paper, fault detection method based on model is connected with computational [...]

 Strona 1