小型直流無刷風扇線上檢測系統開發

風扇運轉都會伴隨著振動，當振動過大時，則表示可能發生故障現象，對於風扇而言，在其發生故障時，若使用單一分析方法，有時未能正確診斷出故障原因，透過多種振動分析方法加以診斷，而每一種分析方式，皆有一定鑑別的特性。為了能提高故障鑑別率，傳統典型的診斷是使用感測器量取對象的時域振動信號，再透過頻域信號作頻譜分析，然後交由專人或有經驗的工程師以時域或頻域兩組信號判斷發生何種故障及其可能性的大小。近幾年來風扇或機械故障判斷技術逐漸由傳統方式轉換成在不需專人判讀的狀況下完成，讓產業達到自動化並節省成本。本文旨在利用類神經網路中的「倒傳遞演算法」，來使得辨別風扇故障的方法能有所準則，且達到80%以上的辨識率。並藉由此提出的判斷法則，能夠使得風扇不論在內部零件部分故障，或者是外部外殼損壞或不全，皆能有所依偱規則與依據，並藉由此方法使得風扇的生產過程能夠增加產能，提高生產良率。

關鍵字

頻譜分析；類神經網路；倒傳遞演算法

並列摘要

Mechanical vibrations occur in the normal DC brushless fan working. The large magnitude vibrations may indicate the defects of the fan rolling parts. In this study, an on-line diagnosis system was developed to monitor the fan vibration and judge the failure conditions. Due to the complex mechanical interferences, spectrum analysis results can’t give enough information for the experimental judge of the fan failures. The artificial neural networks analysis was employed for the on-line diagnosis system to realize the automatic fan defects. The networks weighting was performed by the back propagation algorithm. After well system trainingprocesses, the on-line system can judge the DC brushless fan failures according to the vibrations spectra and the 80% hit rate can be achieved.

並列關鍵字

Spectrum Analysis ； Artificial Neural Networks (ANN) ； Back-Propagation Algorithm

參考文獻

[1] W. S. McCulloch, W. Pitt, “A Logical Calculus of the Ideas Immanent in Nervous Activity,” Bulletin of Mathematical Biophysics, Vol.5, 1943, pp.115-133.

[2] L. A. Zadeh, “Fuzzy Sets,” Information Contr, Vol.8, 1965, pp. 338-353.

[3] D. E. Rumelhart, “Learning Representations by Back-Propagating Errors,” Nature, Vol. 323, 1986, pp. 533-536.

[4] A. Fanni, A. Giua, E. Sandoli, “Neural Newtworks for Multiple Fault in Analog Circuits,” The IEEE International Workshop on Defect and Fault Tolerance in VLSI, 1993, pp.303-310.

[5] P. V. Goode, M. Y. Chow, “Neural/Fuzzy Systems for Incipient Fault Detection in Induction Motors,” Proceedings of IECON’93, 1993.

國際替代計量

小型直流無刷風扇線上檢測系統開發

未授權

主題瀏覽