English  |  正體中文  |  简体中文  |  全文筆數/總筆數 : 6507/11669
造訪人次 : 29712442      線上人數 : 335
RC Version 3.2 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
搜尋範圍 進階搜尋

請使用永久網址來引用或連結此文件: http://ir.ncue.edu.tw/ir/handle/987654321/18397

題名: 以振動訊號及適應模糊類神經理論為基礎之故障辨識系統
作者: 吳建達
貢獻者: 車輛科技研究所
關鍵詞: 振動訊號;齒輪故障診斷系統;離散小波轉換;適應性類神經模糊系統;能量頻譜
Vibration signal;Gear fault diagnosis;Discrete wavelet transform;Adaptive neuro-fuzzy interference system;Energy spectrum
日期: 2009
上傳時間: 2014-04-29T07:29:20Z
出版者: 行政院國家科學委員會
摘要: 本研究計畫主要是打算發展一套以振動訊號為基礎,並搭配離散小波轉換與適應性類神經模糊系統所建構成的智慧型故障診斷系統。此系統發展的主要目的在於利用定位性的察覺與智慧型的檢測方式,來預先得知故障齒輪的位置與狀態,以有效監控車輛中的轉動機械運轉之狀態,並利於維修工程師之維護保養,以減少不必要的維護時間,並且避免更大損壞情況的發生。在本研究計畫中,將以三年時間來深入進行相關性的研究與實現,包括:齒輪測試平台的建立、故障模擬與訊號特性之探討、離散小波分析理論之實現、智慧型分類技巧的實際運用、小波函數的適用性評估比較。在實驗方面,將預先建構3D 實驗平台以進行雛型評估,當平台建立完成之後,再利用實驗平台進行各種故障之模擬,例如:磨損損壞、單局部缺齒損壞、雙局部缺齒損壞等問題,並且透過加速規及數位訊號處理器來即時擷取各種運轉情況下之故障訊號,以作為診斷分析之基礎。在特徵擷取方面,將以離散小波分析技術作為特徵擷取之基礎,以多尺度的分析特性擷取出不同頻帶之特徵,然而,多層次的分解訊號雖然能夠擷取訊號之特徵,相對地將可能會造成資料量過於龐大,若以視覺性的評估檢測將可能困難去察覺。為了改善上述等缺失,本研究計畫將搭配能量頻譜擷取方法,來縮減經離散小波分析後的特徵值,以避免特徵資料庫過於龐大,此外將以不同的Daubechies 小波函數進行辨識率探討,並利用適應性類神經模糊系統將故障狀況進行定位性察覺與分類,以減少視覺性評估的錯誤率,增加系統診斷與分類的準確度,並建立相關硬體與技術,最後將此系統應用於實際行駛中車輛之機械故障監控與輔助診斷。
In this project, an intelligent diagnosis system of fault gear identification and classification based on vibration signal with discrete wavelet transform and adaptive neuro-fuzzy inference (ANFIS) system is proposed. These vibration signals will be measured using accelerometer and data acquisition system from an experimental platform. The discrete wavelet transform technique is one of important roles for pre-processing of signal feature extraction. Because of the multi-scale characteristic can be shown fault feature in the difference between sub-bands. Relatively, the decomposition signals of multi-level will cause the database quantity oversized reason, which are too hard to classify accurately by human vision. In the present project, the principle of feature extraction based on discrete wavelet transform with energy spectrum is proposed. The quantity of database can be reduced by energy spectrum which is used to as input pattern of training and identification process. The different wavelets are also considered in order to identification fault condition clearly. Furthermore, the ANFIS is proposed to identify and classify the fault condition in the present fault diagnosis system. The proposed ANFIS includes both the adaptive neural network capability and the fuzzy logic qualitative approach. The ANFIS training algorithm consisted of a combination of gradient descent algorithm and least squares algorithm. The experimental results verified that the proposed ANFIS has more possibilities in the present fault diagnosis system. The ANFIS achieved accuracy identification rate which was more satisfied than traditional vision inspection in the proposed system.
關聯: 國科會計畫, 計畫編號: NSC98-2221-E018-004; 研究期間: 98/08-99/07
顯示於類別:[車輛科技研究所] 國科會計畫

文件中的檔案:

檔案 大小格式瀏覽次數
index.html0KbHTML414檢視/開啟


在NCUEIR中所有的資料項目都受到原著作權保護.

 


DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - 回饋