請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/78804
標題: | 工具機主軸之熱特徵系統辨識 Thermal-Feature System Identification for a Machine Tool Spindle |
作者: | Ping-Jung Chen 陳品蓉 |
指導教授: | 張培仁(Pei-Zen Chang) |
關鍵字: | 機器學習,工具機主軸,系統識別,溫度感測器,熱特徵模型, Machine learning,Machine tool spindle,System identification,Temperature sensor,Thermal feature model, |
出版年 : | 2019 |
學位: | 碩士 |
摘要: | 主軸的內部溫度是預防保養的重要指標,但是其內部並無足夠空間可置入溫度感測器,而將主軸鑽孔則會降低其機械剛度,故難以直接量測主軸內部的溫度,因此建立主軸的熱特徵模型並據以預測其內部溫度成為重要的研究。本文提出了兩種系統識別(System Identification,SID)方法,分別為灰箱系統識別(Grey-box System Identification,GSID)和黑箱系統識別(Black-box System Identification,BSID),用來建立外部驅動式主軸的熱特徵模型。本研究包含溫度感測及無線傳輸模組(Temperature Sensing and Wireless Transmission Module,TSWTM)的自製硬體和用於系統識別(SID)的軟體; TSWTM用來蒐集SID所需的溫度訓練數據,以識別主軸熱特徵模型的參數。
在GSID中,主軸的熱特性被描述為以熱阻(thermal resistance),熱容(thermal capacity)和熱源(heat source)所組成的熱網絡模型(Thermal Network Model,TNM),該TNM的參數則由MATLAB的System ID Toolbox提供的命令“greyest”識別出來。TNM可以預測出各特徵點在不同轉速下的溫升曲線,在主軸的工作模式和自然冷卻模式下,其預測溫度和測量數據的一致性分別達到71.83%和93.14%。在BSID中,主軸的熱特性由線性非時變狀態空間模型建模,BSID整合由MATLAB的System ID Toolbox提供的命令“n4sid”和機器學習中的K-fold交叉驗證方法,將模型的參數識別出來,所產生的熱特徵模型能夠在操作期間從其表面溫度準確地預測出主軸的內部溫度或者以轉速預測主軸的內外部溫度,其驗證準確率高於98.5%。BSID可以有效地最佳化模型的複雜度,在模型的偏差和變異之間取得平衡,避免擬合不足或是過度擬合的情況。本文比較了兩種用於建立主軸熱特徵模型的系統辨識方法,並說明了利用BSID算法從其表面溫度(可直接檢測)精確計算出主軸內部溫度(無法直接檢測)的可行性。 The internal temperature is an important index for the prevention and maintenance of a spindle. However, the temperature inside the spindle is undetectable directly because there is no space to embed temperature sensor and drilling holes will reduce its mechanical stiffness. Therefore, it is worthwhile to understand the thermal-feature of a spindle and predict the internal temperature variation of a spindle from its thermal-feature model. This article presents two method of system identification (SID), namely grey-box system identification (GSID) and black-cox system identification (BSID) to identify the thermal-feature model of an externally driven spindle. The works contains a self-made hardware of the temperature sensing and wireless transmission module (TSWTM) and a software for the SID; TSWTM is to acquire the temperature training data while SID to identify the parameters of the thermal-feature model of spindle. For GSID, the thermal-feature model of a spindle is modeled as a thermal network model (TNM) composed of equivalent thermal resistances, thermal capacity and heat source, and the parameters of the TNM are identified by the command “greyest” provided by the System ID Toolbox of MATLAB. The resulting TNM is used to figure out the temperature variation inside and outside the spindle via rotational speeds and the consistence with measurement data reaches to 71.83% and 93.14% during the operation and natural-cooling modes of the spindle respectively. For BSID, the thermal-feature of a spindle is modeled by a linearly time-invariant state-space model whose parameters are identified by BSID which integrates the command “n4sid” provided by the System ID Toolbox of MATLAB and the k-fold cross-validation that is common in machine learning. The present BSID can effectively strike a balance between the bias and variance of model, such that both under-fitting and over-fitting can be avoided. The resulting thermal-feature model can not only predict the temperature of spindle rotating at various speeds but also figure out the internal temperature of spindle from its surface temperature. Its validation accuracy is higher than 98.5%. This article compares two kind of system identification in thermal-feature modeling, and illustrates the feasibility of accurately figuring out the internal temperature (undetectable directly) of spindle from its surface temperature (detectable directly) with BSID. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/78804 |
DOI: | 10.6342/NTU201800972 |
全文授權: | 有償授權 |
顯示於系所單位: | 應用力學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-108-R05543089-1.pdf 目前未授權公開取用 | 10.26 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。