人工智慧應用於雷射切割矽鋼片之品質預測與閉迴路控制;Artificial Intelligent Based Approaches for Quality Prediction and Closed Loop Control in Laser Cutting of Electrical Steel Sheet

NCU Institutional Repository > 工學院 > 機械工程研究所 > 博碩士論文 > Item 987654321/90150

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/90150

题名:	人工智慧應用於雷射切割矽鋼片之品質預測與閉迴路控制;Artificial Intelligent Based Approaches for Quality Prediction and Closed Loop Control in Laser Cutting of Electrical Steel Sheet
作者:	阮庭四;Nguyen, Dinh Tu
贡献者:	機械工程學系
关键词:	熱影響區;矽鋼片;品質預測;閉迴路控制;人工智慧方法;Heat-Affected Zone;Electrical Steel;Quality Prediction;Closed Loop Control;Artificial Intelligent
日期:	2022-07-28
上传时间:	2022-10-04 12:12:29 (UTC+8)
出版者:	國立中央大學
摘要:	在金屬薄板的雷射切割過程中，雷射功率、脈衝頻率與切割速度等幾個因素會影響切口品質。由雷射切割過程中的熱效應產生的熱影響區也受切割速度和雷射功率的影響。為了提高切口品質並降低熱影響區，透過即時調整雷射功率及切割速度有助於控制工件工作區域的溫度，因此需要開發合適的控制方法。本研究的目的是開發用於雷射切割矽鋼片品質預測與閉迴路控制的人工智慧方法，首先，發展一卷積神經網路(CNN)模型，用於分析和預測薄矽鋼片雷射切割的切口寬度，考慮三個加工參數做為模型的輸入，即雷射功率、切割速度及脈衝頻率，同時評估一個模型輸出品質參數，即切口寬度。與深度神經網路(DNN)模型與極限機器學習(ELM)模型相比，所開發的CNN模型在預測切口寬度的最終測試數據集上具有最低的平均絕對百分比誤差(MAPE)為4.76%。其次，本研究開發了一種基於模糊增益調整(FGS)控制的改進方法，用於在矽鋼片雷射切割過程中對PID參數進行調整，並即時精確控制工作溫度。因此，藉由此改進的FGS控制，即時調整控制參數，產生了更精細和更均勻的切口寬度，及具有較小HAZ和熔渣附著的平行切邊。此外，平均切口寬度與FGS控制中的溫度設定值有很好的線性關連性。第三，本研究提出了一種帶有新的模糊調節器的PID型模糊邏輯控制器(FLC)來控制具有可變參數的二階系統。模擬結果顯示，與相對速率觀測器及模糊參數調節器整合(RRO-FPR)方法相比較，所使用的最佳PID型FLC與所提出的新模糊調節器產生更優的系統性能與更短的穩定時間及上升時間。此外，亦提出了一種只需調節一個單一輸出比例因子之新的線上模糊調節器結構，以克服使用RRO-FPR方法的主要缺點。所提出帶有模糊調節器的最佳 PID型FLC，係藉由混合PSO-GWO方法優化，其結果證明優於本研究中所比較的其他方法，於實際應用中表現出最短的穩定時間、上升時間及最低的過衝。最後，將所開發具有線上模糊調節器的PID型FLC應用於雷射切割薄矽鋼片過程中，即時穩定控制工件切口前緣溫度，並與固定切割速度的開迴路控制比較。此新型閉迴路控制器透過控制參數的即時調節產生了具有更佳真圓度與較小HAZ的更好切割品質；此外，隨著切割速度的優化調整，切割工作時間亦減少。所提出的閉迴路控制亦可以減少高雷射脈衝頻率下薄鋼板熱量積累的影響。總之，本研究為使用人工智慧方法進行雷射切割薄矽鋼片的品質預測和閉迴路控制提供了極具應用可能的潛力。 ;In the laser cutting processes of thin metallic sheets, several factors affect the kerf quality such as laser power, pulse frequency, and cutting speed. Heat affected zone (HAZ) which results from the thermal effect during laser cutting is also affected by cutting speed and laser power. To improve kerf quality and reduce HAZ, it is beneficial to adjust laser power and cutting speed in situ to control the temperature of the working area such that a suitable controller needs to be developed. Thus, the aim of this study was to develop artificial intelligent approaches for quality prediction and closed loop control in laser cutting of non-oriented electrical steel. Firstly, a convolutional neural network (CNN) model was developed for analysis and prediction of kerf width in laser cutting of thin non-oriented electrical steel sheet. Three input process parameters were considered, namely laser power, cutting speed, and pulse frequency, while one output parameter, kerf width, was evaluated. Compared with a deep neural network (DNN) model and an extreme learning machine (ELM) model, the developed CNN model had the lowest mean absolute percentage error (MAPE) of 4.76% for the final test dataset in predicting kerf width. Secondly, an improved method based on fuzzy gain scheduling (FGS) control was developed in this study to tune the PID parameters and precisely control the working temperature in situ during laser cutting of non-oriented electrical steel sheets. A better quality of finer and uniform kerf width with parallel kerf edges of a smaller HAZ and dross attachment was thus produced by the improved FGS control through real-time tuning of control parameters. In addition, the average kerf width was well correlated with the set-point temperature in FGS control by a linear equation. Thirdly, a PID-type fuzzy logic controller (FLC) with a new fuzzy tuner was proposed to control a second-order system with varying parameters. Simulation results showed that all the given optimal PID-type FLCs with the proposed new fuzzy tuner produced a better system performance and exhibited a shorter settling time/rise time than a relative rate observer and fuzzy parameter regulator (RRO–FPR) approach. Moreover, a new online fuzzy tuner structure was proposed by tuning a single output scaling factor to overcome the major disadvantages of the approach using RRO–FPR. The proposed optimal PID-type FLC with a fuzzy tuner, which was optimized by a hybrid PSO–GWO method, proved to be superior to others given in this study by exhibiting the shortest settling time/rise time and the lowest overshoot in a practical application. Finally, the developed PID-type FLC with online fuzzy tuner was applied to stably control the cutting front temperature in situ during laser cutting of thin non-oriented electrical steel sheet. The obtained optimal parameters were applied to the experiments, in comparison with an open loop control of constant cutting speed. A much better quality of cut kerf with a smaller roundness and HAZ was produced by the PID-type FLC with online fuzzy tuner via in-situ tuning of control parameters. In addition, the cutting time was reduced as the cutting speed was optimally adjusted under PID-type FLC with online fuzzy tuner. The proposed closed loop control could lessen the effect of heat accumulation in the steel sheet under high laser pulse frequency. In summary, this study provide significant potentials in using artificial intelligent approaches for quality prediction and closed loop control in laser cutting of thin non-oriented electrical steel sheet.
显示于类别:	[機械工程研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	37	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....