神經網路於齒輪故障診斷之應用

Neural Network Application in the Gear Fault Diagnosis

許景榮(Jing-Rong Syu)

指導教授 ： 張信良

國立虎尾科技大學/工程學院/機械與機電工程研究所/碩士(2012年)

https://doi.org/10.6827/NFU.2012.00226

摘要

齒輪故障的種類相當多，本研究主要建立一套基本的診斷模式雛型，首先利用快速傅立葉、功率譜、經驗模態分解法和小波分析進行訊號處理，接著提取故障特徵參數並正規化，最後結合倒傳遞類神經網路和機率神經網路。本研究分為模擬和實例驗證兩部分，第一部分故障類型由虛擬訊號產生，測試此診斷系統之可行性，首先判斷訊號為正常或異常，如果屬於異常訊號則判斷訊號為單一故障或複合故障並且判斷各類故障類型，在單一故障類型中裂縫和磨損故障更細分診斷故障程度為何。第二部分為實例驗證，以實例配合此診斷系統進行故障診斷，測試此診斷系統之準確性。診斷後發現結果有不一致的情況則進行投票機制，針對不同訊號處理判斷的故障類型進行多數決，以判斷故障種類與程度。

關鍵字

齒輪故障 ； 快速傅立葉 ； 功率譜 ； 經驗模態分解法 ； 小波分析 ； 倒傳遞類神經網路 ； 機率神經網路

並列摘要

Due to the variable types of gear faults, the main purpose of this study is to establish a model prototype of diagnostic. Firstly, apply FFT (Fast Fourier Transform), power spectrum, EMD (Empirical Mode Decomposition), and wavelet analysis for signal processing. Then extract the fault characteristic parameters and regularize. BPN (Back-propagation Network) and PNN (Probabilistic Neural Network) are then combined to develope the diagnosis system. This study consists of simulations and examples validation. The first part of the fault types are generated by the virtual signal for the feasibility test of this diagnostic system. The signals are judged to be normal or abnormal at first step. If the signal is abnormal, the signal is divided to be single fault or double faults, and to determine the type of fault. A single fault type of crack, wear and tear fault are proceeded the level of fault diagnosis. The second part is the case of the verification. The verification can be used to prove the diagnosis, accuracy of this diagnostic system. If the diagnosis found that the results are inconsistent, the different signal processing are voted to determine the fault type and the level.

並列關鍵字

Gear fault ； FFT ； Power spectrum ； EMD ； Wavelet ； BPN ； PNN

參考文獻

[22]張智傑，「齒輪故障診斷之模糊類神經網路」，碩士論文，中原大學機械工程學系，2005。

[1]Welch, P. D. “The Use of Fast Fourier Transform for the Estimation of Power Spectra :A Method Based on Time Averaging Over Short, Modified Periodograms.” IEEE Trans. Audio Electroacoust. Vol. AU-15 , pp. 70-73, 1970.

[3]Huang, N. E., Shen, Z., Long, S. R., Wu, M. C., Shih, H. H., Zheng, Q., Yen, N. C., Tung, C.C., and Liu, H. H., “The Empirical Mode Decomposition and the Hilbert Spectrum for Nonlinear and Non-stationary Time Series Analysis”, Proc. Soc. Lond. A, Vol. 454, pp. 903-995, 1998.

[4]Liu, B., Riemenschneider, S., Xu, Y., “Gearbox Fault Diagnosis using Empirical Mode Decomposition and Hilbert spectrum”, Mechanical Systems and Signal Processing, Vol. 20, pp. 718-734, 2006.

[6]Zheng, H., Li, Z., and Chen, X., “Gear Fault Diagonsis Based On Continuous Wavelet Transform”, Mechanical Systems and Signal Processing, Vol. 16, pp. 447-457, 2002.