English  |  正體中文  |  简体中文  |  全文筆數/總筆數 : 78728/78728 (100%)
造訪人次 : 33543361      線上人數 : 520
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
搜尋範圍 查詢小技巧:
  • 您可在西文檢索詞彙前後加上"雙引號",以獲取較精準的檢索結果
  • 若欲以作者姓名搜尋,建議至進階搜尋限定作者欄位,可獲得較完整資料
  • 進階搜尋


    請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/87025


    題名: 脈衝雷射切割無定向矽鋼片及人工智能質量預測的實驗研究;Experimental Study on Pulsed Laser Cutting of Thin Non-Oriented Silicon Steel and Quality Predication Using Artificial Intelligence
    作者: 阮覃懷;Tan, Nguyen Hoai
    貢獻者: 機械工程學系
    關鍵詞: ??激光切割;Pulsed Laser Cutting
    日期: 2021-10-18
    上傳時間: 2021-12-07 13:46:31 (UTC+8)
    出版者: 國立中央大學
    摘要: 中 文 摘 要

    本研究旨在研究雷射參數對無定向矽鋼薄片切割質量的影響,然後通過應用人工智能 (AI) 模型來預測和優化切割質量。首先,本研究提出了四個實驗,實驗一和實驗二以較慢的切割速度研究;實驗三和實驗四則以較快的切割速度研究。在四個實驗中都檢查了熱影響區(HAZ),因為它對矽鋼材料的磁性影響最大,而熱影響區通過在空氣及溶劑的環境條件下以快慢兩種切割速度進行研究。此外,還評估了幾何形狀、表面粗糙度、切割時間、尺寸精度等質量。
    實驗一考察了三種雷射參數對切口波紋度和熱影響區兩種切割質量的影響,即雷射功率、切割速度和脈衝重複率。實驗二表明,在水和氯化鈉的環境條件中切割,擁有比在空氣中進行切割更小的熱影響區。實驗二顯示了HAZ和磁通密度之間的負相關。實驗三研究了雷射參數對切割時間(TC)、切口寬度(KT)、切口錐角()、HAZ、表面粗糙度(Sa)等多種質量的影響。實驗四評估了環境條件(空氣、去離子水、酒精、潤滑油、氯化鈉和 5%、10%、15% 硝酸酒精溶液)對重鑄層、切割邊緣的浮渣/碎屑的影響。
    由於在較短的切割時間內切口的質量良好,因此選擇了初始 10% 的條件來研究三個雷射參數 (P、v、f) 對電機鐵芯疊片、HAZ、尺寸精度(內徑誤差ED1、外徑誤差ED2、齒寬誤差EL、內徑圓度C1、外徑圓度C2)、切削時間(TC)。然後,應用極限學習機(ELM)、多元回歸分析(MRA)、人工神經網絡(ANN)和隨機森林(RF)四種人工智能模型來預測切割質量。結果表明,在所使用的四種模型中,ELM 模型的預測最為準確。此外,隨機森林法還用於確定與響應相關的輸入參數的相對重要性。在全部的四個實驗中都表明,雷射功率對 HAZ 的影響最為顯著。
    最後,由於ELM在四個 AI 測試模型中預測響應準確度最高,因此選擇 ELM 模型來優化切割質量。通過實施實驗二預測的最佳加工參數,得到的極限學習機-遺傳算法(ELM-GA)進行優化和實驗確認,得到的MFD和HAZ為1.639 T和30.40 ?m。在比較三種模型(ELM、MRA、ANN)的預測精度後,開發了一種多目標優化模型,即基於 ELM - GA 算法的偏好選擇指數(PSI)方法,以優化初始 10% 條件下七種疊片鐵芯的品質。進行實驗驗證以評估最佳預測的準確性,HAZ 為 33.4 ?m,ED1 為 14 ?m,ED2 為 19 ?m,C1 為 21 ?m,C2 為 26 ?m,EL 為 13 ?m,TC 為 34.14 s,其中對於 HAZ、ED1、ED2、C1、C2、EL 和 TC,最佳和實驗確認之間的誤差分別為 4.04%、6.52%、4.02%、0.48%、2.14%、5.09% 和 1.25%。因此,無需任何後處理即可直接形成可組裝鐵芯疊片的優點,使所提出的雷射切割方案成為一種經濟而有效的方法。
    ;This study aims to investigate the effects of laser process parameters on the cutting qualities of thin non-oriented silicon steel; then, the cutting qualities are predicted and optimized by the application of the artificial intelligence (AI) models. Firstly, four experiments are presented in the research: experiment I and experiment II study with the slow cutting speed, experiment III and experiment IV research in higher cutting speed. The heat-affected zone (HAZ) is examined in both four experiments because it has the most influence on the magnetic property of silicon steel material, considered by laser cutting in air and solvent conditions with slow and high cutting speed. Besides, the qualities such as geometry, surface roughness, cutting time, dimensional precision are also investigated. Experiment I examines the influences of the three laser process parameters on two cutting qualities of the kerf waviness and HAZ, which are the laser power, cutting speed and pulse repletion rate. Experiment II shows that cutting in water and sodium chloride achieve a smaller HAZ than cutting in air condition. Experiment II also indicates a negative correlation between HAZ and magnetic flux density. Experiment III studies the influences of laser process parameters on multi-qualities, cutting time (TC), kerf width (KT), kerf taper angle (), HAZ, and surface roughness (Sa). Experiment IV evaluates the influences of environmental conditions (air, deionized water, alcohol, lubricant oil, sodium chloride, and nital solution 5%, 10%, 15%) on the recast layer, dross/debris at the cutting edge. Due to the good quality at the cutting edge in the short cutting time, the nital 10% condition is selected for investigating the effects of three process parameters (P, v, f) on the seven cutting qualities of a motor core’s lamination, HAZ, dimensional precision (inner diameter error ED1, outer diameter error ED2, tooth width error EL, inner diameter roundness C1, outer diameter roundness C2), and cutting time (TC).
    Then, four AI models, which are an extreme learning machine (ELM), multiple regression analysis (MRA), artificial neural network (ANN), and random forest (RF), are applied to predict the cutting qualities. Results reveal that among the four models used, the most accurate predictions are found by the ELM model. Besides, the random forests method is also addressed to determine the relative importance of input parameters associated with the responses. All four experiments show that laser power is the most significant influence on HAZ.
    Finally, the ELM model is selected to optimize the cutting qualities due to the highest predictive response accuracy among the four AI tested models. By implementing the predicted optimal processing parameters for experiment II, the resulting Extreme learning machine - genetic algorithm (ELM-GA) for the optimization and experimental confirmation, the obtained MFD and HAZ is 1.639 T and 30.40 mm. A multi-objective optimization model, the ELM - GA algorithm based preference selection index (PSI) method, is developed to optimize the seven output qualities of laminated core in the nital 10% condition after comparing the predicted accuracy of three models applied (ELM, MRA, ANN). The experimental validation is performed to evaluate the accuracy of optimal prediction with HAZ of 33.4 ?m, ED1 of 14 ?m, ED2 of 19 ?m, C1 of 21 ?m, C2 of 26 ?m, EL of 13 ?m, and TC of 34.14 s, which errors between optimal and experimental confirmation were 4.04%, 6.52%, 4.02%, 0.48%, 2.14%, 5.09%, and 1.25% for HAZ, ED1, ED2, C1, C2, EL, and TC, respectively. Consequently, the merit of direct formation is ready-for-assembly core laminations without the need of any post-processing renders the proposed laser cutting scheme an economical and effective approach for manufacturing ready-for-assembly core laminations from the thin silicon steel sheets.
    顯示於類別:[機械工程研究所] 博碩士論文

    文件中的檔案:

    檔案 描述 大小格式瀏覽次數
    index.html0KbHTML140檢視/開啟


    在NCUIR中所有的資料項目都受到原著作權保護.

    社群 sharing

    ::: Copyright National Central University. | 國立中央大學圖書館版權所有 | 收藏本站 | 設為首頁 | 最佳瀏覽畫面: 1024*768 | 建站日期:8-24-2009 :::
    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - 隱私權政策聲明