Wyniki 1-4 spośród 4 dla zapytania: authorDesc:"Jacek Wojciechowski"

A graph clustering-based method of the assessment of rough sets efficiency in the diagnostics of analog systems

Czytaj za darmo! »

A modern diagnostics of analog systems is strongly supported by the machine learning and artificial intelligence (AI) methods. They are widely used to design computer diagnostic modules, run on both general purpose computers and specialized hardware. Their main advantages are high autonomy (resulting in automatic decision making about the state of the examined system), ability to emulate human behavior (allowing it to work as the expert system), or ability to generalize, i.e. correct reaction to the patterns that the module did not encounter in the past. Multiple approaches, such as neural networks [1], rules induction algorithms (decision trees [2]), or fuzzy logic [3] proved their efficiency. The main aim of the diagnostic methods is to determine that the system under test ([...]

Tree-based access network design under requirements for an aggregation network

Czytaj za darmo! »

Nowadays network operators invest in upgrading access networks - the most expensive part of the whole telecommunication system - to improve delivery of high-bandwidth services for the maximum number of subscribers. It implies that new access nodes are installed to extend the service area. Many operators build completely new access networks, e.g. Passive Optical Networks (PON). In this situation they face the problem of selecting the most suitable access network architecture. Tree structures in telecommunications networks topology are taking on large importance because of their simplicity and relatively low cost. Most network operators build access networks in this topology, taking into account such criteria like: cost, robustness to failures, and traffic delay. Desirable value[...]

Overview of optimization methods in diagnostics of analog systems DOI:10.15199/59.2015.6.5


  1. Introduction. applications of analog systems despite their unification with digital parts still remains important in various technical domains, such as military, acoustic or radio frequency (RF). The continuously growing the number of elements makes their testability (i.e. the ability to distinguish between particular fault sources) difficult to obtain. In case of high frequency and data acquisition systems, conducting the diagnostics of analog and digital parts separately is crucial in achieving satisfying work evaluation results. The testability of digital circuits have already been well defined, there are no such procedures prepared for the analog and mixed systems [1]. Identification of such systems is a key to decrease the production costs of modern electronics [2]. The analog systems, diagnostics is complicated by the tolerances of elements or noise (which must be treated during the signal processing operations prior to the fault detection or identification). The testability of mixed systems is included in the IEEE 1149.4 norm [3]. For analog systems, no such standards exist. The Artificial Intelligence (AI) approaches have been extensively used to monitor the state of analog systems during the last twenty years. Currently, sophisticated computer systems are able to perform fault detection, location and identification in the real time. The advanced concepts of the AI include the classifier fusion [7] or the combination of supervised and unsupervised learning systems [23]. Contemporary approaches are heuristic and their parameters must be optimized to maximize the accuracy. Therefore one of the most pressing issues in the fault detection and location domain is the investigation and applications of optimization methods. The paper presents the overview of optimization algorithms and their applications to particular problems. The most widely used approaches are briefly described and their diagnostic implementations consi[...]

 Strona 1