建構田口殘差修正GM(1,1)灰模型之微鑽孔刀具壽命預測系統

Construction of Taguchi Residual Modification GM(1,1) Grey Model of Tool Life Prediction System for Micro Drilling

林明慶(Ming-Qing Lin)

指導教授 ： 黃博滄

中原大學/電機資訊學院/工業與系統工程研究所/碩士(2016年)

https://doi.org/10.6840/cycu201600736

摘要

隨著科技的發展、資訊快速的變化，預測技術已成了各國重視的領域之一，因此各種不同的預測模型應運而生。其中灰色預測模型能夠利用少量數據便能進行預測的特性已被廣泛的應用於各研究領域中。而現今的市場趨勢受到消費者的行為改變，產品推陳出新的速度隨之增快，造就了需求變化快速的市場，進而使產品的製造方式由大量生產轉變到少量多樣的型態。故本研究為符合現今少樣多量的加工生產型態，利用灰色理論，針對微型鑽孔建構刀具壽命預測系統，並配合田口方法加以改良，發展田口殘差修正GM(1,1)(Taguchi Residual Modification GM(1,1), TRGM(1,1))預測模型，使TRGM(1,1)預測模型能夠提升其預測準確度。
　　本研究所開發之TRGM(1,1)刀具壽命預測系統，係利用電腦數值控制(Computer Numerical Control, CNC)工具機進行微鑽孔加工，透過少量的孔徑數據建構原始GM(1,1)預測模型，並利用預測產生的殘差值配合田口方法進行修正、改良，延續灰色理論其小樣本預測特性作為基礎架構，能夠快速建構TRGM(1,1)刀具壽命預測系統。
　　為了驗證本研究所發展的TRGM(1,1)其預測準確性與可靠性，本研究將先以一組刀具與工件及兩組不同的加工參數進行微型鑽孔加工，透過光學顯微鏡測量加工孔徑，並將少量孔徑值投入原始GM(1,1)預測系統，進行孔徑大小的趨勢預測，藉此預測該加工參數下的刀具壽命。接著將同樣的孔徑值投入TRGM(1,1)預測系統進行刀具壽命預測，並與原始GM(1,1)預測系統及學者提出的RGM(1,1)進行比較，透過兩次的比較，驗證本研究所提出的預測系統其可行性。原始GM(1,1)預測模型與TRGM(1,1)預測模型的比較部分，原始GM(1,1)的 為12.58%，屬優良的預測、TRGM(1,1)的 為6.11%，屬高準確的預測；TRGM(1,1)預測模型與RGM(1,1)預測模型的比較部分，TRGM(1,1)的準確度分別為95.15%與96.97%，皆高於RGM(1,1)的89.14%與92.42%。故可驗證本研究所發展的TRGM(1,1)具有一定的預測準確性與可靠性。

關鍵字

灰色預測 ； 田口方法 ； 殘差修正

並列摘要

Research on various prediction models has drawn national attention with the technological development. Among one of them, Grey prediction Model (GM) has successfully use fewer data to achieve better results, thus been widely used in many different research topics. Today, consumers’ behaviors have changed along with the emerging new products, turning the original massive production mode into current rapid but diverse marketing strategies. Under the above observation, this thesis focuses on the Grey prediction Model, GM(1,1), using practical life prediction system of micro drilling tools.
The original tool life prediction system is based on few CNC micro drilling data to build a GM(1,1) baseline model. We proposed Taguchi Residual Modification GM(1,1), or TRGM(1,1), which can efficiently improve the model prediction accuracy without additional new data.
To verify the prediction accuracy and reliability of TRGM(1,1) tool life prediction system, we collected the data under the two different parameter settings, but with the same CNC micro drilling tool. For each setting, we used optical microscope to collect the hole diameters into sequential data. The GM(1,1) prediction system can then utilize the data to predict the tool life. We compare the original model performance with RGM(1,1) and TRGM(1,1). The experimental results show that the of TRGM(1,1) is 6.11%, better than that of original GM(1,1) 12.58%. And the accuracy of TRGM(1,1) over our two datasets are 95.15% and 96.97%, better than that of RGM(1,1) 89.14% and 92.42%. We can conclude that TRGM(1,1) has apparently strengths over GM(1,1) and RGM(1,1).

並列關鍵字

Grey prediction ； Taguchi method ； Residual Modification

參考文獻

郭家豪. (2014) 微型鑽孔刀具壽命灰色預測系統之開發(碩士)　中原大學　桃園市

張黃傑. (2013) 平面銑削之灰色即時可調式學習表面粗糙度預測系統開發(碩士)　中原大學　桃園市

羅傑瀛, 林彥宏, 王正賢. (2002) 應用灰色理論於時間序列轉折點之分析與預測, Da-Yeh University, Vol. 11, No.2, 115-127

Cavalcante, R. C., Brasileiro, R. C., Souza, V. L., Nobrega, J. P., & Oliveira, A. L. (2016). Computational Intelligence and Financial Markets: A Survey and Future Directions. Expert Systems with Applications, 55, 194-211.

Chung, C. C., Chen, H. H., & Ting, C. H. (2010). Grey prediction fuzzy control for pH processes in the food industry. Journal of food engineering, 96(4), 575-582.