Methods of diagnostics of electromechanical equipment removal based on technologies of artificial intelligence.
Journal Title: Наукові праці Донецького Національного Технічного Університету серія: Електротехніка і Енергетика - Year 2017, Vol 1, Issue 1
Abstract
In this paper, the latest developments and applications of artificial intelligence for monitoring state and diagnostics of faults are considered, according to the categories of malfunctions’ diagnostics. A comparative analysis of the characteristics of computational methods of diagnosis, which can be considered when choosing the appropriate method in a concrete situation for future research, is carried out. It has been determined that artificial neural networks are one of the most commonly used classifiers of methods of intelligent diagnostics of faults, which has the potential of high level of training and general characteristics. The accuracy of artificial neural networks (АNN) is highly dependent on the training sample. In the case of a limited number of sample volumes, АNNs often show weak generalization capabilities, therefore, of course, the АNN is used in the case of a sufficient size of the training sample. The method of support vectors machine (SVM) is introduced into the fault’s diagnosis and prediction of machine failure for a small sample size, taking into account its high accuracy and good generalization. But SVM must be specially combined for a multi -class classification. Training of this method also takes a lot of time to work with large -scale data. The deep neural network (DNN) can adaptively choose the necessary information from the source without the need for prior knowledge through the deep structure, so it can be used for intelligent fault diagnosis, but only when it is difficult to identify faults, however, the DNN needs more time to study than ANN through depth structure. The basis of fuzzy rules is a key point and a bottleneck in the development of fuzzy logic (FL) based on expert knowledge and experience. In the absence of self -learning and self-realization, FL is often combined with other algorithms such as neural network, tree failure and expert system for to achieve error detection and predictions. Evolutionary algorithms are the most commonly used diagnostic failure. EAs have been applied to malfunctioning, along with other signal processing techniques, such as wavelet transforms, stochastic resonances, and others. Merging different methods can form a new hy brid algorithm which combines the benefits of different methods.
Authors and Affiliations
Т. ALTUKHOVA
Keywords
Related Articles
- EP ID EP660755
- DOI -
- Views 168
- Downloads 0
The modernized system of the automatic control of metal level in the mold.
Stabilization of the metal in the mold during continuous casting of steel is needed to ensure a high quality ingot, increasing the metal yield, reduce the time and improve the safety of the casting process. Maintainin...
Analysis of the methods of the normalization of the specific consumption of heat and electric power in the institu-tions of the public sector.
The article describes the real situation of normalization of specific consumption of fuel and energy resources by the institutions of the public sector of the Kharkiv region. Identified shortcomings from use of the ca...
Influence on the durability of work of high-temperature rotors of long-term static strength and low cycle fatigue
A significant part of heat and mechanical equipment at Ukrainian thermal power plants has reached the end of normative and extended operating life. The power plants were commissioned 30-40 years ago and now appear to be...
Determination of the pre-defective state of element of metal constructions at fatigue failure by magnetometric methods.
The paper has proved the relevance and substantiation of the use of magnetometric methods for diagnostics of the defect state of metal structures. The causal relationship between the occurrence of the defect state wit...
Engineering implementation of detection radar based on the algorithm of probability distribution function.
Creation of modern radio and radio physical devices is impossible without taking into account the random processes occurring in them. Sufficiently detailed description of random processes in relation to...