Development of a method for optimizing the structure of static neural networks intended for categorizing technical state of gas-turbine engines

The presence on modern aviation gas-turbine engines of dozens and even hundreds of sensors for continuous registration of various parameters of their operation makes it possible to collect and process large amounts of information. This stimulates the development of monitoring and diagnostic systems. At the same time the presence of great volumes of information is not always a sufficient condition for making adequate managerial decisions, especially in the case of evaluation of the technical condition of aviation engines. Thus it is necessary to consider, that aviation engines it is objects which concern to individualized, i.e. to such which are in the sort unique. Therefore, the theory of creating systems to assess the technical state of aircraft engines is formed on the background of the development of modern neural network technology and requires the formation of specific methodological apparatus. From these positions in the article the methods which are used at carrying out clustering of the initial information received at work of modern systems of an estimation and forecasting of a technical condition of aviation gas-turbine engines are considered. This task is particularly relevant for creating neural network multimode models of aircraft engines used in technical state estimation systems for identification of possible failures and damages. Metric, optimization and recurrent methods of input data clustering are considered in the article. The main attention is given to comparison of clustering methods in order to choose the most effective of them for the aircraft engine condition evaluation systems and suitable for implementation of systems with meta-learning. The implementation of clustering methods of initial data allows us to breakdown diagnostic images of objects not by one parameter, but by a whole set of features. In addition, cluster analysis, unlike most mathematical-statistical methods do not impose any restrictions on the type of objects under consideration, and allows us to consider a set of raw data of almost arbitrary nature, which is very important when assessing the technical condition of aircraft engines. At the same time cluster analysis allows one to consider a sufficiently large volume of information and sharply reduce, compress large arrays of parametrical information, make them compact and visual.

Download Full-text

Fault detection and isolation of gas turbine engines using a bank of neural networks

Journal of Process Control ◽

10.1016/j.jprocont.2015.08.007 ◽

2015 ◽

Vol 36 ◽

pp. 22-41 ◽

Cited By ~ 57

Author(s):

S. Sina Tayarani-Bathaie ◽

K. Khorasani

Keyword(s):

Neural Networks ◽

Fault Detection ◽

Gas Turbine ◽

Fault Detection And Isolation ◽

Turbine Engines ◽

Gas Turbine Engines

Download Full-text

Creep Life Prediction for Aero Gas Turbine Hot Section Component Using Artificial Neural Networks

Volume 3: Cycle Innovations; Education; Electric Power; Fans and Blowers; Industrial and Cogeneration ◽

10.1115/gt2012-68856 ◽

2012 ◽

Author(s):

M. F. Abdul Ghafir ◽

Y. G. Li ◽

L. Wang

Keyword(s):

Neural Network ◽

Neural Networks ◽

Gas Turbine ◽

Life Prediction ◽

Turbine Engines ◽

Gas Turbine Engines ◽

Creep Life ◽

Model Based ◽

Creep Life Prediction ◽

Prediction Approach

Accurate and reliable component life prediction is crucial to ensure safety and economics of gas turbine operations. In pursuit of such improved accuracy and reliability, model-based creep life prediction methods have become more and more complicated and therefore demand more computational time although they are more flexible in applications, in particular for new gas turbine engines. Therefore, there is a need to find an alternative approach that is able to provide a quick solution to creep life prediction for production engines while at the same time maintain the same accuracy and reliability as that of the model-based methods. In this paper a novel creep life prediction approach using Artificial Neural Networks is introduced as an alternative to the model based creep life prediction approach to provide a quick and accurate estimation of gas turbine creep life. Multilayer feed forward back propagation neural networks have been utilised to form three neural network-based creep life prediction architectures known as the Range Based, Functional Based and Sensor Based architectures. The new neural network creep life prediction approach has been tested with a model single spool turboshaft gas turbine engine. The results show that good generalisation can be achieved in all three neural network architectures. It was also found that the Sensor-Based architecture is better than the other two in terms of accuracy, with 98% of the post-test samples possessing prediction errors within ± 0.4%. Overall, it can be concluded that the proposed neural network approach in creep life prediction is able to provide a good alternative to the more complicated model-based creep life prediction algorithms and can be applied to different types of gas turbine engines.

Download Full-text

Analysis and directions of technological and organization problem updating in gas turbine aircraft engine repair

Science intensive technologies in mechanical engineering ◽

10.30987/2223-4608-2020-8-42-48 ◽

2020 ◽

Vol 2020 (8) ◽

pp. 42-48

Author(s):

Vyacheslav Bezyazychnyy ◽

Andrey Smirnov

Keyword(s):

Gas Turbine ◽

New Technologies ◽

Aircraft Engine ◽

Technical State ◽

Turbine Engines ◽

Gas Turbine Engines ◽

Turbine Engine ◽

Technology Application ◽

Organization Problem ◽

Repair Techniques

There are presented technological (requirements in new technologies for repair of gas turbine engines (GTE) repaired according to a technical state, growing requirements on reliability, high cost of repair, a limited access to new technologies) and organization (absence of repair localization, high competition etc.) problems of aircraft gas turbine engine repair. The direction for updating: development of new repair technologies at the transition to the concept of repair on a technical state; module technology application; repair production localization; creation of flexible repair techniques etc. is considered.

Download Full-text

Neural Networks for the Diagnostics of Gas Turbine Engines

ASME 1996 Turbo Asia Conference ◽

10.1115/96-ta-039 ◽

1996 ◽

Cited By ~ 6

Author(s):

G. Torella ◽

G. Lombardo

Keyword(s):

Neural Networks ◽

Gas Turbine ◽

Back Propagation ◽

Learning Rate ◽

Turbine Engines ◽

Gas Turbine Engines ◽

Turbine Engine ◽

Classical Architecture ◽

Set Up ◽

The One

The paper describes the activities carried out for developing and testing Back Propagation Neural Networks (BPNN) for the gas turbine engine diagnostics. One of the aims of this study was to analyze the problems encountered during training using large number of patterns. Each pattern contains information about the engine thermodynamic behaviour when there is a fault in progress. Moreover the research studied different architectures of BPNN for testing their capability to recognize patterns even when information is noised. The results showed that it is possible to set-up and optimize suitable and robust Neural Networks useful for gas turbine diagnostics. The methods of Gas Path Analysis furnish the necessary data and information about engine behaviour. The best architecture, among the ones studied, is formed by 13, 26 and 47 neurons in the input, hidden and output layer respectively. The investigated Nets have shown that the best encoding of faults is the one using a unitary diagonal matrix. Moreover the calculation have identified suitable laws of learning rate factor (LRF) for improving the learning rate. Finally the authors used two different computers. The first one has a classical architecture (sequential, vectorial and parallel). The second one is the Neural Computer, SYNAPSE-1, developed by Siemens.

Download Full-text