141 |
MACHINE CONDITION MONITORING USING NEURAL NETWORKS: FEATURE SELECTION USING GENETIC ALGORITHMHippolyte, Djonon Tsague 26 February 2007 (has links)
Student Number : 9800233A -
MSc dissertation -
School of Electrical and Information Engineering -
Faculty of Engineering and the Built Environment / Condition monitoring of machinery has increased in importance as more engineering processes
are automated and the manpower required to operate and supervise plants is reduced. The
monitoring of the condition of machinery can significantly reduce the cost of maintenance.
Firstly, it can allow an early detection of potential catastrophic fault, which could be extremely
expensive to repair. Secondly, it allows the implementation of conditions based maintenance
rather than periodic or failure based maintenance [1]. In these cases, significant savings can be
made by delaying schedule maintenance until convenient or necessary.
Although there are numerous efficient methods for modeling of mechanical systems, they all
suffer the disadvantage that they are only valid for a particular machine. Changes within the
design or the operational mode of the machine normally require a manual adaptation. Using
Neural Networks to model technical systems eliminates this major disadvantage. The basis for a
successful model is an adequate knowledge base on which the network is "trained". Without
prior knowledge of the machines systematic behavior or its history, training of a neural Network
is not possible. Therefore, it is a pre-requisite that the knowledge base contains a complete
behavior of the machine covering the respective operational modes whereby, not all rather the
most important modes are required. Neural networks have a proven ability in the area of
nonlinear pattern classification. After being trained, they contain expert knowledge and can
correctly identify the different causes of bearing vibration. The capacity of artificial neural
networks to mimic and automate human expertise is what makes them ideally suited for handling
nonlinear systems. Neural networks are able to learn expert knowledge by being trained using a
representative set of data [2]-[6]. At the beginning of a neural network’s training session, the
neural network fault detector’s diagnosis of the motor’s condition will not be accurate. An error
quantity is measured and used to adjust the neural network’s internal parameters in order to
produce a more accurate output. This process is repeated until a suitable error is achieved. Once
the network is sufficiently trained and the parameters have been saved, the neural network
contains all the necessary knowledge to perform the fault detection.
One of the most important aspects of achieving good neural network performance has proven to
be the proper selection of training features. The curse of dimensionality states that, as a rule of
thumb, the required cardinality of the training set for accurate training increases exponentially
with the input dimension [7]. Thus feature selection which is a process of identifying those
features that contribute most to the discrimination ability of the neural network is required.
Proposed methods for selecting an appropriate subset of features are numerous [8]-[11]. Methods
based on generating a single solution, such as the popular forward step wise approach, can fail to
select features which do poorly alone but offer valuable information together. Approaches that
maintain a population of solutions, such as genetic algorithms (GA) are more likely to speedily
perform efficient searches in high dimensional spaces, with strong interdependencies among the
features. The emphasis in using the genetic algorithm for feature selection is to reduce the
computational load on the training system while still allowing near optimal results to be found
relatively quickly.
To obtain accurate measure of the condition of machinery, a wide range of approaches can be
employed to select features indicative of condition. By comparing these features with features for
known normal and probable fault conditions, the machine’s condition can be estimated. The
most common approach is that of analysis in the frequency domain by applying a Fast Fourier
Transform (FFT) to the time domain history data. The idea is simply to measure the energy
(mean square value) of the vibrations. As the machine condition deteriorates, this measure is
expected to increase. The method is able to reveal the harmonics around the fundamental
frequency of the machine and other predominant frequency component (such as the cage
frequency) [12]. Frequency analysis is well established and may be used to detect, diagnose and
discriminate a variety of induction motor faults such as broken rotor bars, cage faults, phase
imbalance, inner and outer race faults. However, as common in the monitoring of any industrial
machine, background noise in recorded data can make spectra difficult to interpret. In addition,
the accuracy of a spectrum is limited due to energy leakage [12- 14].
Like many of the new techniques now finding application in machinery condition monitoring,
Higher Order Statistics was originally confined to the realms of non-linear structural dynamics. It
has of recent however found successful application to the identification of abnormal operation of diesel engines and helicopter gearboxes [5, 7]. Higher Order Statistics provide convenient basis
for comparison of data between different measurement instances and are sufficiently robust for
on-line use. They are fast in computation compared with frequency or time-domain analysis.
Furthermore, they give a more robust assessment than lower orders and can be used to calculate
higher order spectra. This dissertation reports work which attempts to extend this capability to
induction motors.
The aim of this project is therefore to examine the use of Genetic Algorithms to select the most
significant input features from a large set of possible features in machine condition monitoring
contexts. The results show the effectiveness of the selected features from the acquired raw and
preprocessed signals in diagnosis of machine condition. This project consists of the following
tasks:
#1; Using Fast Fourier transform and higher order signals techniques to preprocess data
samples.
#1; Create an intelligent engine using computational intelligence methods. The aim of
this engine will be to recognize faulty bearings and assess the fault severity from
sensor data.
#1; Train the neural network using a back propagation algorithm.
#1; Implement a feature selection algorithm using genetic algorithms to minimize the
number of selected features and to maximize the performance of the neural network.
#1; Retrain the neural network with the reduced set of features from genetic algorithm
and compare the two approaches.
#1; Investigate the effect of increasing the number of hidden nodes in the performance of
the computational intelligence engine.
#1; Evaluate the performance of the system using confusion matrices.
The output of the design is the estimate of fault type and its severity, quantified on a scale
between 0-3. Where, 0 corresponds to the absence of the specific fault and 3 the presence of a
severe machine bearing fault. This research should make contribution to many sectors of industry
such as electricity supply companies, and the railroad industry due to their need of techniques
that are capable of accurately recognizing the development of a fault condition within a machine system component. Quality control of electric motors is an essential part of the manufacturing
process as competition increases, the need for reliable and economical quality control becomes
even more pressing. To this effect, this research project will contribute in the area of faults
detection in the production line of electric motor.
|
142 |
The performance and compatibility of thin client computing with fleet operationsLandry, Kenneth J. 06 1900 (has links)
This research will explorethe feasibility of replacing traditional networked desktop personal computers (PC) with a thin client/server-based computing (TCSBC) architecture. After becoming nearly extinct in the early 1990s, thin clients are emerging on the forefront of technology with numerous bandwidth improvements and cost reduction benefits. The results show that TCSBC could provide a practical and financially sound solution in meeting the Navy's need to reduce costs and propagate the latest technology to all personnel. This solution may not meet the requirements of all naval commands. A thorough performance analysis should be conducted of the applications employed and the overall expenditures prior to implementation. / US Navy (USN) author.
|
143 |
Some aspects of a code division multiple access local area networkPearce, Richard Sargon January 1987 (has links)
No description available.
|
144 |
An Interactive Tool to Investigate the Inference Performance of Network Dynamics From DataVeenadhar, Katragadda 08 1900 (has links)
Network structure plays a significant role in determining the performance of network inference tasks. An interactive tool to study the dependence of network topology on estimation performance was developed. The tool allows end-users to easily create and modify network structures and observe the performance of pole estimation measured by Cramer-Rao bounds. The tool also automatically suggests the best measurement locations to maximize estimation performance, and thus finds its broad applications on the optimal design of data collection experiments. Finally, a series of theoretical results that explicitly connect subsets of network structures with inference performance are obtained.
|
145 |
Monitoring and management of OSI networksModiri, Nasser January 1989 (has links)
No description available.
|
146 |
Optimization of large systemsHamam, Y. January 1972 (has links)
No description available.
|
147 |
Developing a methodology model and writing a documentation template for network analysisSkagerlind, Mikael January 2016 (has links)
This report focuses on finding best practices and a better methodology when performing computer network analysis and troubleshooting. When network analysis is performed, computer network data packets are captured using data capturing software. The data packets can then be analysed through a user interface to reveal potential faults in the network. Network troubleshooting is focusing more on methodology when finding a fault in a network. The thesis work was performed at Cygate where they have recently identified needs for an updated network analysis methodology and a documentation template when documenting the network analysis results. Thus, the goal of this thesis has been to develop an elaborated methodology and discover best practices for network analysis and to write a documentation template for documenting network analysis work. As a part of discovering best practices and a methodology for network analysis, two laboratory tests were performed to gather results and analyse them. To avoid getting too many results but to still keep the tests within the scope of this thesis, the laboratory tests were limited to four network analysis tools and two test cases that are explained below. In the first laboratory test during three different test sequences, voice traffic (used in IP-phones and Skype etc.) is sent in the network using a computer program. In two of the test sequences other traffic is also congesting the network to disturb the sensitive voice traffic. The program used to send the voice traffic then outputs values; packet delay, jitter (variation in delay) and packet loss. Looking at these values, one can decide if the network is fit for carrying the sensitive voice traffic. In two of the test cases, satisfying results were gathered, but in one of them the results were very bad due to high packet loss. The second laboratory test focused more on methodology than gathering and analysing results. The goal of the laboratory test was to find and prove what was wrong with a slow network, which is a common fault in today’s networks due to several reasons. In this case, the network was slow due to large amounts of malicious traffic congesting the network; this was proven using different commands in the network devices and using different network analysis tools to find out what type of traffic was flowing in the network. The documentation template that was written as part of this thesis contains appealing visuals and explains some integral parts for presenting results when network analysis has been performed. The goal of the documentation template was an easy-to-use template that could be filled in with the necessary text under each section to simplify the documentation writing. The template contains five sections (headlines) that contain an explanation under it with what information is useful to have under that section. Cygate’s network consultants will use the documentation template when they are performing network analysis. For future work, the laboratory test cases could be expanded to include Quality of Service (QoS) as well. QoS is a widely deployed technology used in networks to prioritise different types of traffic. It could be used in the test cases to prioritise the voice traffic, in which case the results would be completely different and more favourable.
|
148 |
Correlations of Higher Order in Networks of Spiking NeuronsJovanovic, Stojan January 2016 (has links)
The topic of this dissertation is the study of the emergence of higher-order correlations in recurrentlyconnected populations of brain cells.Neurons have been experimentally shown to form vast networks in the brain. In these networks, eachbrain cell communicates with tens of thousands of its neighbors by sending out and receiving electricalsignals, known as action potentials or spikes. The effect of a single action potential can propagate throughthe network and cause additional spikes to be generated. Thus, the connectivity of the neuronal networkgreatly influences the network's spiking dynamics. However, while the methods of action potentialgeneration are very well studied, many dynamical features of neuronal networks are still only vaguelyunderstood.The reasons for this mostly have to do with the difficulties of keeping track of the collective, non-linearbehavior of hundreds of millions of brain cells. Even when one focuses on small groups of neurons, all butthe most trivial questions about coordinated activity remain unanswered, due to the combinatorialexplosion that arises in all questions of this sort. In theoretical neuroscience one often needs to resort tomathematical models that try to explain the most important dynamical phenomena while abstractingaway many of the morphological features of real neurons.On the other hand, advances in experimental methods are making simultaneous recording of largeneuronal populations possible. Datasets consisting of collective spike trains of thousands of neurons arebecoming available. With these new developments comes the possibility of finally understanding the wayin which connectivity gives rise to the many interesting dynamical aspects of spiking networks.The main research question, addressed in this thesis, is how connectivity between neurons influences thedegree of synchrony between their respective spike trains. Using a linear model of spiking neurondynamics, we show that there is a mathematical relationship between the network's connectivity and theso-called higher-order cumulants, which quantify beyond-chance-level coordinated activity of groups ofneurons. Our equations describe the specific connectivity patterns that give rise to higher-ordercorrelations. In addition, we explore the special case of correlations of third-order and find that, in large,regular networks, it is the presence of a single subtree that is responsible for third-order synchrony.In summary, the results presented in this dissertation advance our understanding of how higher-ordercorrelations between spike trains of neurons are affected by certain patterns in synaptic connectivity.Our hope is that a better understanding of such complicated neuronal dynamics can lead to a consistenttheory of the network's functional properties. / <p>QC 20161003</p>
|
149 |
Essays in Network Economics and Game TheoryTan, Hi-Lin January 2009 (has links)
Thesis advisor: Richard J. Arnott / This dissertation comprises three papers that are concerned with the implications of strategic interactions between a finite set of agents in private goods economies. One form of strategic behavior I consider arises in a social network when the consumption decisions of agents are influenced by those around them. The other form of strategic behavior I consider arises when agents bargain with one another. The first paper focuses on undirected networks in which consumers care about the average of their neighbors' consumption. The main contribution is to show how social networks affect equilibrium prices. I show that if every consumer has the same number of neighbors, then each consumer's influence on the market is independent of the number of neighbors. Due to the tradeoff between more neighbors responding and less sensitive responses, greater network intensity may not result in greater average influence of all consumers. In addition, I show that a consumer who is central in the network may not have the highest influence on the market because of the need to consider not only the number of neighbors that he has or his distances to other consumers, but also the number of neighbors that his neighbors have. The second paper examines strategic consumption in a directed network. The main contribution is to show how directed networks affect equilibrium outcomes. I show how the critical and promising links, and the key players in a social network can be identified. In doing so, I introduce the impact centrality and reaction centrality measures, and show how these measures are used to determine the effects on aggregate centrality of removing any agent from the network, and of removing or adding any directed link. The third paper considers bargaining under two-sided incomplete information in a market with multiple buyers and sellers, each with either high or low independent private values. I show that there exists a mechanism that guarantees efficient trading outcomes even when gains from trade are uncertain. The main contribution of this paper to show that a large number of traders is not necessary to guarantee efficient trading if there are at least as many sellers as there are buyers, and there is at least one low valuation buyer. / Thesis (PhD) — Boston College, 2009. / Submitted to: Boston College. Graduate School of Arts and Sciences. / Discipline: Economics.
|
150 |
A Primal-Dual Approximation Algorithm for the Concurrent Flow ProblemNahabedian, Aaron Joseph 29 April 2010 (has links)
The multicommodity flow problem involves shipping multiple commodities simultaneously through a network so that the total flow over each edge does not exceed the capacity of that edge. The concurrent flow problem also associates with each commodity a demand, and involves finding the maximum fraction z, such that z of each commodity's demand can be feasibly shipped through the network. This problem has applications in message routing, transportation, and scheduling problems. It can be formulated as a linear programming problem, and the best known solutions take advantage of decomposition techniques for linear programming. Often, quickly finding an approximate solution is more important than finding an optimal solution. A solution is epsilon-optimal if it lies within a factor of (1+epsilon) of the optimal solution. We present a combinatorial approximation algorithm for the concurrent flow problem. This algorithm consists of finding an initial flow, and gradually rerouting this flow from more to less congested paths, until an epsilon-optimal flow is achieved. This algorithm theoretically runs much faster than linear programming based algorithms.
|
Page generated in 0.0509 seconds