以類神經法修正轉子系統之分析誤差

Error Improvement of rotordynamic coefficient by using neural networks

徐文明(Wenming Xu)

指導教授 ： 楊大中

元智大學/工程學院/機械工程學系/碩士(2004年)

摘要

摘要
由於轉動機械的轉子動力系統，廣泛的被應用在各式各樣的工業，因此，過去數十年來發表的有關轉子動力系統的研究論文及實驗報告不勝枚舉，研究主題囊括了剛性轉子及撓性轉子的振動、顫動、旋動，軸承中轉子的不平衡反應....等等，這些結果不外乎要使轉子動力系統在高速旋轉下能有較高的穩定性，或期使系統的臨界旋轉速度提高。本文使用類神經網路法來鑑別影響轉子系統的軸承參數，使用其中的倒傳遞網路法來預測轉子系統中的參數進而對理論模型進行調整，使理論的模擬系統更逼近於實際的轉子系統。 為達成上述的目的，本文使用了兩個不同之模型，分別討論類神經網路對於參數的預測分析是否具有可行性，以及預測實際轉子系統影響係數並調整,以期理論模型對於實際轉子系統能有效的進行模擬。

關鍵字

類神經網路 ； 軸承 ； 倒傳遞網路法 ； 轉動機械

並列摘要

Abstract
Rotating machinery has been utilized in various industrial applications. Research on rotor dynamics is extensive, including unbalance response, resonance, stability, oil whirling, balancing, etc., in aiming at stable high speed operation of the rotors. However, the estimation of the dynamic coefficients of hydrodynamic bearings is a rather difficult and critical part in the analysis of rotating machines. In this study, back-propagation method of Neural Networks is used to estimate the bearing coefficients, without resorting to conventional hydrodynamic analysis using Reynolds’ equation. The neural networks of bearing coefficient prediction are trained by using the simulated natural frequencies, then the real natural frequencies are fed into the trained neural networks to obtain a set of predicted bearing coefficients which makes the results of analysis close to the real behavior of the machines. Two rotor models are used to demonstrate the feasibility and effectiveness of this estimation approach of bearing coefficients.

並列關鍵字

Neural Network ； Back-propagation ； Rotating Machinery ； Bearing

參考文獻

[3] 彭善謙，”綜合振動信號於馬達故障診斷”， 中原大學機械工程研究所，2004.

[4] 趙安民，”馬達故障診斷之模糊類神經網路”， 中原大學機械工程研究所，2004.

[1] K. Brian. and H. Ian.,” Vibration fault detection of large turbogenerators using neural networks, ” IEEE International Conference,Vo1, P P.121-126,1995.

[2] Zhang. S. and Ganesan. R.,” Self-organizing neural networks for automated machinery monitoring systems, ” Mechanical Systems & Signal Processing, V10,PP.517-532,1996.

[5] Luxhoj. J. T.,” Artificial neural network for nonlinear estimation of the turbine flow-meter coefficient, “Engineering Applications of Artificial Intelligence,V11,PP.723-734,1998.