灰模糊微鑽孔孔徑預測之研究

A Study of Grey-fuzzy Dimension Predection in a Micro-Drilling Operation

顏經哲(Jing-Zhe Yan)

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(2017年)

https://doi.org/10.6840/cycu201700779

摘要

伴隨著科技的進步，生產品質的控管與預測在生產加工過程中越來越受到矚目，產品生產方式也從大量生產轉變為高精密品質的加工方式。許多學者針對加工品質預測提出許多方法進行研究，例如：模擬建模、統計預測、迴歸分析及各種軟性計算的方式，與導入感測器進行建構預測系統，皆有不錯的成效；但若加工設置改變時，在建構過程中仍然需要大量數據與時間進行系統建制，相對於製程流暢度而言是一項困難要素。故本研究將透過灰色理論、感測器訊號與模糊理論，透過少數據建制模糊規則庫補正灰色預測值，針對微型鑽孔建構一個不受影響並能快速預測孔徑的即時預測系統。
本研究所開發之預測系統，利用灰色預測能依靠少量數據建模的特性為基礎，建構即時預測系統、改良模糊理論的模糊規則庫建制方式，減少建制複雜程度與所需之專家經驗，透過相對的感測器訊號與灰色預測值之殘差值的投入，建立補正殘差值的模糊規則庫，藉此達到快速建模、即時預測更準確的加工孔徑，並進行整體刀具壽命的判斷，探討其準確性。
　　為證實所提出的方法其可行性與準確性，本研究將設置兩組不同加工參數，進行微鑽孔加工，透過光學電子顯微鏡與非接觸式感測器，將少量資料之孔徑値投入灰預測系統，並利用感測器頻率值進行模糊規則庫之建制，補正孔徑殘差值，兩組補正後之孔徑預測值皆比原始灰預測精準，刀具壽命在補正後重新判別，其平均精準度高達98.58%，比原始預測89.54%提升將近10%的準確率，故得以證實本研究所提出之預測系統其可行性與準確性。

關鍵字

灰色預測 ； 感測器訊號 ； 田口方法 ； 模糊規則庫 ； 微型鑽孔

並列摘要

Through the improvement of technology, quality of process and forecast becomes more and more important and the way to process has change from mass production to high quality production. Many scholars propose a lot of method in process quality forecast, like simulation modeling, tradition statistical forecasting, regression analysis, soft computing and with the use of sensor modeling forecast system, these gives good affects. However, if the process setup changes, it needs a lot of data and time to build system, this make it difficult for the process to run smoothly. In this research, through the use of fuzzy theory, sensor signal and gray theory, hope to use less data in building Fuzzy rule bank for correcting grey prediction value and build a real-time forecast system which will not be effected by process parameters through Micro-drilling.
The forecast system we develop uses Grey prediction which base on using few data to build a system moreover, improve the way how Fuzzy rule bank built with Fuzzy theory, decrease the need of Expert experience and complex level. Through the use of sensor signal and Grey Prediction Residual value, build Fuzzy rule bank which can correct the Residual value in a short time for real-time modeling for more accurate process Drilling Diameters, moreover, determine the tool life with the accuracy Drilling Diameters forecast.
To verify the accuracy and feasible, we use two different process parameters with Micro-drilling, through the use of Optical electron microscopy and Non-contact sensor, use few Drilling Diameters value into grey prediction system to build Fuzzy rule bank and correct the Drilling Diameters Residual value. The two drilling forecast value are more accurate than the original forecast, tool life has been re-differentiate and the average accuracy raises from 89.54% to 98.58% which verify the accuracy and feasible that we proposed.

並列關鍵字

Grey prediction ； Sensor signal ； Taguchi method ； Fuzzy rule bank ； Micro-drilling

參考文獻

Abbasi, F., Mojtahedi, A., & Ettefagh, M. M. (2015). Fault diagnosis using noise modeling and a new artificial immune system based algorithm. Earthquake Engineering and Engineering Vibration, 14(4), 725-741.

Abu-Mahfouz, I. (2003). Drilling wear detection and classification using vibration signals and artificial neural network. International Journal of Machine Tools and Manufacture, 43(7), 707-720.

Ahmed, F., Capretz, L. F., & Samarabandu, J. (2008). Fuzzy inference system for software product family process evaluation. Information Sciences, 178(13), 2780-2793.

Ahmed, M. A., Saliu, M. O., & AlGhamdi, J. (2005). Adaptive fuzzy logic-based framework for software development effort prediction. Information and Software Technology, 47(1), 31-48.

Albarbar, A., Mekid, S., Starr, A., & Pietruszkiewicz, R. (2008). Suitability of MEMS accelerometers for condition monitoring: An experimental study. Sensors, 8(2), 784-799.