Study to Improve Drum Shape Error in Rough Cutting of Wire-EDM

線切割放電加工是模具製作及精密加工不可或缺的一種材料移除方法，因為可以加工具有複雜外型與輪廓的導體、高強度、耐高溫材料(HSTR)時提供最佳的選擇或者有時候甚至是唯一選擇，但在粗加工過之工件，往往在工件之Z軸向會遺留下鼓形的腹部，必須靠後續之數道次精加工逐步來予以修正，以使工件Z軸向任一點均能滿足公差之尺寸精度。鼓形量較大，則需要較多道次的精加工，如此增加了加工時間與成本，也增添了事先規劃工件預留量的困擾，若預留量不足，則粗加加工後之工件無法修復必須重作。本文目的在於探討粗加工鼓形量誤差之成因，並提出改進方案。研究中是以振動的影響與殘渣觀測方式研究影響鼓形的因素，發現加工槽中的沖水與排渣會影響鼓形的位置與分佈，用高速攝影機觀察實際放電加工與振動情形，加工並未產生明顯振動行為。第二部分是以田口式實驗計劃方法為基礎，嘗試尋求出影響加工特性的參數，結果發現，相對而言工件厚度與放電時間的影響較大，但統計上沒有單獨的顯著因子，鼓形量是由多重因子的交互作用造成。據此，使用類神經網路建立了一具預測鼓形量的模式，經驗證其誤差大小約在6%，利用此模式可以在粗加工階段精確的預估鼓形量，而應用於線切割放電加工之製程規劃。利用各個加工參數與鼓形量的關係作為加工參數調整的原則，配合建立出之模式，發現可以將鼓形量從一般依經驗法則加工得到的40 μm 降至10 μm 左右，除了可以降低工件的鼓形量外，可以得到較佳的真直度與高精度之放電加工品質，最後並結合類神經網路執行最佳化技術，可以降低工件的鼓形量至6 μm。Wire electrical discharge machining (Wire-EDM) is an essential material removal process in die and mold making and precision machining. This is because Wire-EDM provides the best alternative or sometimes the only alternative for machining conductive, exotic and high strength and temperature resistive (HSTR) materials with the scope of generating intricate shapes and profiles. However, it usually results in the drum shape along vertical axis (bowing) of the machined part. Several subsequent finishing cuts are needed to remove this inaccurate part so that the required dimensional tolerance is satisfied. A too large bowing will need more finishing cuts, which in turn increases machining time and cost. It also makes process planning of the allowance in advance more difficult. In the most serious case, re-do of the machined part all over again is needed. There have been very few studies on the bowing of the wire-cut machined part so far. The purpose is to find out the reason why the drum shape machined workpiece is formed, and the factors that would affect its magnitude. The effect of the vibration of wire on drum shape error is studied. The experimental results show that the drum shape machined workpiece is formed mainly due to the dielectric fluid flow and debris distribution in the machining gap. There is positive relationship of the location distribution of discharge with drum shape error. Also, the behaviour of the wire vibration observed by high speed camera. The objectives of this thesis are to investigate the relationship between the machining parameters and the characteristic of WEDM. The objectives of this dissertation are to investigate if there are significant parameters affecting barrel shape machined part (drum shape error), and to develop a mathematical model to predict drum shape error. Based on the Taguchi Quality design and the analysis of variance, it is found that comparatively workpiece thickness and ignition on time play more important roles. But statistically there is no significant factor affecting drum shape error. The size of the barrel is the result of cross effect of machining parameters. Based on this finding, the neural network with genetic algorithms to find the optimal operation machined was adopted and a model related the drum shape error and machining parameters was established. Experimental results show that the predicted error is less than 6%. By using this model the amount of drum shape error at rough machining stage can be accurately predicted, and applied for process planning of Wire-EDM. Incorporating the machining settings adjustment principles with the developed model, it is also found that the drum shape error can be reduced from 40μm to around 10μm. Thus, part with less amount of drum shape error and a better straightness 6 μm can be obtained.目錄 謝 一文摘要 二文摘要 三錄 五目錄 七目錄 九號說明 一?一章 緒 論 1-1 引言 1-2相關文獻回顧 3-3 研究目的與研究方法 7-4 本文結構 13二章 放電加工原理 15-1 放電加工過程 15-2 放電基本轉換過程及火花結構 17-3 放電電源及波形 20-3-1 放電電源 20-3-2 放電波形 26-4 線切割放電加工之特性 28-5 類神經網路理論 31-5-1 類神經網路概論 31-5-2 類神經網路理論 32-5-3 倒傳遞網路架構 34-5-4基本網路演算法 37-5-5網路參數之影響 40-5-6 遺傳基因演算法 42三章 實驗設備與實驗規劃 48四章 鼓形量成因問題研究 54-1 殘渣濃度分佈對於鼓形之影響 54-2線極振動的影響之研究 61-3應用高速攝影機之振動行為觀察 67五章 加工參數的影響研究 77-1 實驗設計原理 77-1-1實驗數據轉換與各水準別之總和的計算 82-1-2變異數分析(Variance Analysis) 85-2加工參數之水準對加工特性之影響 88-3類神經網路應用 91-3-1網路的建構與收斂情形評估 91-3-2倒傳遞網路的預測與實驗結果比較 93-3-3 基因遺傳演算法的應用 94-3-4 加工參數最佳化之求解 96六章 結論與建議 100考文獻 102者簡歷及發表之論文 10