Analysis model and classification of the geometry of gas turbine engine blades

Аuthors

Pechenin V. A.^*, Bolotov M. A.^**

Samara National Research University named after Academician S.P. Korolev, 34, Moskovskoye shosse, Samara, 443086, Russia

*e-mail: vadim.pechenin2011@yandex.ru
**e-mail: maikl.bol@gmail.com

Abstract

The quality of aircraft engines (specific consumption, gas-dynamic stability, reliability, service life) is embedded in the course of the design process, fixed outat the stage of finalization, providedin the course of production, and implemented during operation. Account for actual geometry of the partsduring design process, and forming the best for current manufacturing environment conditions geometry of the products during manufacturing and assembling is one of the reserves allowing aircraft engine quality increase. Realization of the above mentioned reserve becomes possible with implementation of intellectual system of aircraft engines quality provision. The article presents the technique for the intellectual system of aircraft engines quality provision forming. It shows the structure of the intellectual system of aircraft engines quality provision. Mathematical model of the system with regard to analysis and clustering of complex shapes and surfaces is built. The analysis allows calculating parameters of a part location deviation from nominal geometry, and calculation deviation of parts shapes. Analysis tools are as follows: an iterative algorithm of nearest points, piecewise splines approximation of surface points, Fourier transform. We used an algorithm of k-means and self-organizing Kohonens neural networks as clustering methods. For adequate profile parameters clastering (shape and position) these parameters were normalized to the range of [0,1]. Complex shape geometry intellectual analysis model was implemented in MATLAB^®. Within the framework of the developed system using obtained models, clustering of series of GTE backet compressor blades, obtained as a result of measuring with CMMs DEA Global Performance 07.10.07. Implementation of neural networks seems to be more promising for the tasks of clustering solving compared to static methods, due to the possibility of learning and clustering of new objects on the basis of accumulated know-how. For further model realization in manufacturing building of new program modules is necessary

Keywords:

analysis, Fourier transformation, clustering, neural network

References

1. Besl P.J., Mckay N.D. A method for registration of 3-D shapes. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1992, vol. 14, no. 2, pp. 239-256.

2. Zalmanzon L.A. Preobrazovaniya Fure, Uolsha, Khaara i ikh primenenie v upravlenii, svyazi i drugikh oblastyakh (Fourier, Walsh, Haar transform and their application in management, communication and other areas), Мosсow, Nauka, 1989, 496 p.

3. Jayanti S., Kalyanaraman Y., Ramani K. Shape-based clustering for 3D CAD objects: A comparative study of effectiveness. Computer-Aided Design, 2009, no. 41, pp. 999-1007.

4. Demuth H., Beale M. Neural Network Toolbox For Use with MATLAB. Neural Network Toolbox Users Guide, 2003, 840 p.

5. Niu Y., Wong Y., Hong G. Intelligent sensor system approach for reliable toolflank wear recognition. International Journal of Advanced Manufacturing Technology, 1998, no. 14(2), pp. 77-84.

6. Emel E. Tool wear detection by neural network based acoustic emission sensing, American Society of Mechanical Engineers. Dynamic Systems and Control Division (Publication) DSC, 1991, no. 28, pp.79-85.

7. Osovskii S. Neironnye seti dlya obrabotki informatsii (Neural network for information processing), Мoscow, Finansy i statistika, 2002, 344 p.

8. Ketkov Yu.L., Ketkov A.Yu., Shtults M.M. MATLAB 7: programmirovanie, chislennye metody (MATLAB 7: programming, numerical methods), St. Petersburg, BKhI-Peterburg, 2005, 752 p.