SVM法在齒輪轉子系統異常檢測研究

Damage Detect for a Rotor-Gear System Based on Support Vector Machine

蔡旻修(Min-Hsiu Tsai)

指導教授 ： 范憶華

中原大學/工學院/機械工程研究所/碩士(2020年)

https://doi.org/10.6840/cycu202000436

摘要

齒輪的傳動系統被廣泛的使用在各個種類的機械設備中，目前普遍是靠著技術人員來診斷減速機等齒輪機構的健康狀態，在實務中若齒輪損毀沒有及時進行更換，不僅可能會影響生產物品的良率以及產量，更甚還可能會因此損壞其他的零件，增加維修的成本以及難度。隨著科技的進步無人工廠或智慧型工廠的概念的展開，一個能自動診斷齒輪健康狀態的系統是目前的趨勢。
　　本文實驗主要分為兩個階段，第一階段使用加速規分別對六種不同的運轉中齒輪箱的三軸X、Y以及Z軸進行震動訊號的抓取，而齒輪箱一共分為六種類別，分別為正常、齒輪偏心、鬆脫、轉軸不對心、斷齒和磨耗，將資料經過快速傅立葉變換(Fast Fourier Transform, FFT)，將時域頻域整合資料輸入至分類器有K個最近鄰居法(K-nearest neighbor, K-NN)以及支持向量機器(Support Vector Machine, SVM)中進行，支持向量機器的分類對於單個故障齒輪箱可達到100％的準確度，此分類結果對齒輪異常檢測有著極大的幫助。
　　第二階段延續使用第一階段的資料擷取方式以及支持向量分類器的過程，利用決策函數給每個樣本中的每類一個評分並進行排序，透過排序可以使得在診斷齒輪異常時可以優先進行排查，可以給有複合異常齒輪的判定提供一個檢測一個有效的方式。

關鍵字

齒輪轉子系統 ； 快速傅立葉轉換 ； 支持向量機 ； 決策函數 ； 多類別分類

並列摘要

The transmission system of gears is widely used in various types of mechanical equipment. Currently, technicians diagnose the health of gear mechanisms such as reducers. In practice, if the gear is damaged and not replaced in time, it may not only affect the yield and output of the production items, but also may damage other parts, increasing the cost and difficulty of maintenance. With technological development, the concept of an unmanned factory or a smart factory is a system that can automatically diagnose the health status of gears.

The experiment in this article is mainly divided into two stages. In the first stage, the acceleration gauge is used to capture the vibration signals of the three axes X, Y and Z of the six different running gearboxes. The gearboxes are divided into six categories, Good, BC, Loose, Shaft, Tooth break and Wear. The time-domain data undergoes Fast Fourier Transform (FFT), and the frequency-domain data is input to the classifier by K nearest neighbors (K-nearest neighbor, K-NN), and support vector machine (SVM). Support vector machine classification can achieve 100% accuracy for a single faulty gearbox. This classification result is of great help to the detection of abnormal gear.

The second stage continues the process of using the data extraction method of the first stage. The support vector classifier is used the decision function to score and sort each category in each sample. Through sorting, priority can be given to the diagnosis of gear abnormalities, which can provide an effective method for the determination of compound abnormal gears.

並列關鍵字

Gear rotor system ； fast Fourier transform ； support vector machine ； decision function ； Multiclass Classification

參考文獻

[1] 高金吉，機器故障診治與自愈化，中國北京：高等教育出版社，2012。

[2] R. B. Randall, A new method of modeling gear faults, Journal of Mechanical Design. vol.104, pp.259-267, 1982.

[3] R. B. Randall and J. Hee, Cepstrum analysis, Wireless World. vol. 88, pp. 77-80, 1982.

[4] F. Filippetti, G. Franceschini, C. Tassoni, and P. Vas, Recent Developments of Induction Motor Drives Fault Diagnosis Using AI Technique. IEEE Transactions on Industrial Electronics, Vol.47, pp. 994-1004, 2000.

[5] R. A. Fisher, The Use of Multiple Measurements in Taxonomic Problems. Annals of Eugenics. 7 (2): 179–188, 1936.