Application of a neural network to predict the FAC rate of NPP equipment
NPP piping elements manufactured from carbon steel are subjected to Flow-Accelerated Corrosion (FAC) process. Consequence of this process is broken piping element until the leak appears. So prediction of the FAC rate is very important for NPP safety and economy. The rate of FAC process is defined by the large number of parameters influencing each to other by different manner. Such factors are about 12. There are some procedures to predict FAC rate but the problem of mathematical model parameters estimation is very complicated. Using of neural network (NNW) to predict FAC rate allows estimating of factors mutual influence, defining important of data and improving of prediction accuracy without definition of functions between factors. The paper deals with development and education of optimal NNW to predict the FAC rate at NPP piping. From review of some different NNW, methods of network education and analysis of control data (control of thickness of piping) there is realizing the NNW to predict the FAC rate at NPP straight pipe with single-phase flow for PWR by means of MatLab. Some configurations of NNW are researching. To arise NNW extension capacity there is solved to build decremented network. To estimate the number of latent layers the Kolmogorov theorem is applied. For example NNW with four factors is considering. For some areas of factors values correct behavior of NNW is shown. As result there is suggesting the conceptual scheme of intellectual system as complex of three types of NNW: replicate NNW, self-adjusting card of Cokhonnen and inverse propagation NNW.
- Baranenko V.I., Yanchenko Y.A., Gulina O.M., Dokukin D.A. O raschyote skorosti erozionno-korrozionnogo iznosa i ostatochnogo resursa truboprovodov AES (Estimation of FAC Rate and Service Life for NPP Piping). Izvestiya vuzov. Yadernaya energetika. 2010, no. 2, pp. 55-63.
- Chexal V.K. (Bind), Horowitz J.S. Chexal-Horowitz Flow-Accelerated Corrosion Model-Parameter and Influences. Current perspective of Inter. Pressure vessels and Piping: Codes and Standard. Book No. 409768, 1995, pp. 231-243.
- Gulina O.M., Baranenko V.I., Dokukin D.A. Metodologicheskaya osnova prognozirovaniya erozionno-korrozionnogo iznosa oborudovaniya AES metodom neirosetevogo modelirovaniya (Methodological base to predict the FAC rate by Neural Network Modeling). Izvestiya vuzov. Yadernaya energetika. 2008, no. 1, pp. 3-8.
- Wassermann F. Neirokomp’yuternaya tehnika: teoriya i praktika (Neural computer technology: Theory and practice). 1992. Avaiable at http://www.codenet.ru/progr/alg/nks/ (in Russian).
- Swingler K. Primenenie Neural Networks. Prakticheskoe rukovodstvo (Neural Networks Application. Guidance). Avaiable at http://matlab.exponenta.ru/neuralnetwork/book4/3_3.php/ (in Russian).