|
基于AdaBoost算法實(shí)現(xiàn)的機(jī)床刀具磨損節(jié)能集成研究
馬瑞 2024/12/15 18:48:09
淮安生物工程高等職業(yè)學(xué)校，機(jī)電工程系，江蘇淮安 223002
摘要：為了提高刀具磨損狀態(tài)識別能力，開發(fā)出了刀具磨損階段建立模型方法。把AdaBoost集成算法加入模型，降低磨損過程中模型預(yù)測誤差。研究結(jié)果表明：平穩(wěn)磨損階段所需時間最短，最長為急劇磨損階段。進(jìn)行刀具磨損識別期間，集成學(xué)習(xí)算法可以獲得比單獨(dú)算法更優(yōu)性能。磨損期間誤差受到各階段磨損變化率的較大影響。采用集成方法AdaBoost得到了較小MAE，只有36.3%，可以有效促進(jìn)非集成算法模型的性能提升，實(shí)現(xiàn)集成學(xué)習(xí)算法模型的改善效果。
關(guān)鍵詞：刀具磨損；模型；集成算法；狀態(tài)識別
中圖分類號：TG156
Research on tool wear energy-conservation integration of machine tools based on AdaBoost algorithm
Ma Rui
Department of Mechanical and Electrical Engineering, Huaian Higher Vocational School of Biological Engineering, Huaian 223002, China
Abstract: In order to improve the ability of tool wear state recognition, a regression model of tool wear stage was developed. AdaBoost integrated algorithm is added to the regression model to reduce the prediction error of the regression model in the process of wear. The results show that the time required for smooth wear stage is the shortest, while the time required for sharp wear stage is the longest. During tool wear identification, the integrated learning algorithm can obtain better performance than the single algorithm. The error during wear is greatly affected by the rate of wear change at each stage. The integration method AdaBoost obtained a small MAE, only 36.3%, which can effectively promote the performance improvement of the non-integrated algorithm model and achieve the improvement effect of the integrated learning algorithm model.
Key words: Tool wear; Regression model; Integration algorithm; State recognition
刀具是數(shù)控機(jī)床上的關(guān)鍵設(shè)備，去受刀具磨損因素影響而引起機(jī)床運(yùn)行停止的情況時有發(fā)生，已經(jīng)成為機(jī)床故障的主要方式之一[1-2]�？梢酝ㄟ^一些檢測手段對刀具磨損狀態(tài)進(jìn)行檢測，進(jìn)而采用智能控制技術(shù)進(jìn)行磨損狀態(tài)識別。那么如果整理并獲取刀具磨損狀態(tài)數(shù)據(jù)是一個熱點(diǎn)方向[3]。
目前，有許多學(xué)者針對刀具磨損機(jī)制開展研究，并取得了一定的研究成果。例如，楊莉等[4]提出基于堆疊稀疏自動編碼器和多傳感器特征融合的刀具磨損預(yù)測方法，應(yīng)用磨損數(shù)據(jù)集來驗(yàn)證預(yù)測性能,證明學(xué)習(xí)方法可以提高預(yù)測性能。高鳴等[5]通過采集主軸振動信號來實(shí)現(xiàn)銑刀磨損狀態(tài)監(jiān)測，使用小波閾值降噪消除信號干擾，引入雙向長短周期記憶網(wǎng)絡(luò)對特征信號編碼，通過全連接層預(yù)測刀具磨損狀態(tài)。周粵等[6]提出將希爾伯特黃變換與卷積神經(jīng)網(wǎng)絡(luò)相結(jié)合的監(jiān)測方法，運(yùn)用卷積神經(jīng)網(wǎng)絡(luò)生成刀具磨損監(jiān)測遷移模型，改進(jìn)后希爾伯特黃變換能避免虛假固有模態(tài)分量問題，監(jiān)測效果得到提升。
本文選擇銑削加工刀具作為測試對象，按照具體階段構(gòu)建刀具磨損模型后，把（未完，下一頁）
附件下載：基于AdaBoost算法實(shí)現(xiàn)的機(jī)床刀具磨損節(jié)能集成研究
|