Health Condition Prognostics of Complex Equipment Based on Discrete Input Process Neural Networks

Abstract:

Considering the problem of health condition prognostics of complex equipment, a discrete input process neural networks (DPNN) model based on process neural networks (PNN) is proposed in this paper. DPNN utilizes vector inputs together with convolution operator to gain the capability of time and spatial aggregation operation, which is implemented with continuous function inputs and integral operator by PNN. Different from PNN, DPNN can use discrete samples as inputs directly, thus can avoid precision loss during procedures of data fitting and function expanding required by PNN. The application of DPNN to health condition prognostics of complex equipment is described through the prediction of the future health state of the civil aircraft engines, where the short-term and long-term predictions of the health condition represented by the exhausted gas temperature time series are conducted. Moreover, the performance of DPNN is compared with common artificial neural networks (NN) and PNN. The results show that DPNN has satisfied performance for health condition prognostics of civil aircraft engines, and DPNN performs better than both NN and PNN, which prove that DPNN is suitable for health condition prognostics of complex equipment.

• Periodical: Applied Mechanics and Materials (Volumes 423-426)
• Pages: 2347-2354
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.423-426.2347
• Online since: September 2013
• Authors: Gang Ding*, Da Lei, Wei Yao
• Keywords: Complex Equipment, Discrete Inputs, Health Condition Prognostics, Process Neural Networks

References

[1] A Hess, G Calvello, T Dabney, in: Proceedings of 2004 IEEE Aerospace Conference, IEEE(2004).

[2] O.E. Dragomir, R. Gouriveau, F. Dragomir, et al, in: Proceedings of European Control Conference, Budapest, Hungary( 2009).

[3] J.W. Sheppard, M.A. Kaufman and T.J. Wilmer, Aerospace and Electronic Systems Magazine, IEEE, Vol. 24(2009), pp.34-41.

[4] Z.K. Peng, F.L. Chu, Mechanical Systems and Signal Processing, Vol. 18(2004), pp.199-221.

[5] M Pecht, R Jaai, Microelectronics Reliability, Vol. 50(2010), pp.317-323.

[6] C.S. Byington, M Watson and D Edwards, in: Proceedings of 2004 IEEE Aerospace Conference, IEEE(2004).

[7] X.G. He, J.Z. Liang, Engineering Science, Vol. 2(2000), pp.40-44 (in Chinese).

[8] K Ouyang, R Zhou and W.F. Liu, Beijing Biomedical Engineering, (1997), pp.93-101 (in Chinese).

[9] S S Zhong, Y Li, G Ding, et al, Neural network world, Vol. 16(2007), pp.483-495.

[10] S S Zhong, G Ding and D Z Su, in: Proceedings of the Second international conference on Advances in Neural Networks, IEEE(2005).

[11] G Ding, L Lin, in: Proceedings of the 8th International Symposium on Neural Networks, IEEE(2011).

[12] S H Xu, X G He, Chinese Journal of Computers, Vol. 27(2004), pp.645-650 (in Chinese).

[13] A V Oppenheim, A S Willsky, and S H Nawab. Signals and systems. (Prentice Hall, New Jersey 1997).

[14] M T Hagan, M B Menhaj, IEEE Transactions on Neural Networks, Vol. 5(1994), pp.989-993.

[15] Y A Cheng, L Zhang, Y B Liu, et al, Computer Integrated Manufacturing Systems, Vol. 16 (2010), pp.2026-2037(in Chinese).