智能化多模型表面粗糙度預測系統之研究

Research on Intelligent Multi-model Surface Roughness Prediction System

范晉維(Chin-Wei Fan)

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(2020年)

https://doi.org/10.6840/cycu202000525

摘要

CNC加工品質的預測在製程管理上被視為相當重要的一個環節，利用預測技術可以提前避免不良品的產生並節省量測的步驟，然而現今多數的預測模型或是監控系統大多受限於學者提出的各項限制且為單一的預測模型，或許所得出之預測結果較為主觀但研究者卻無從得知，因此本研究目的為透過有效的感測技術配合適當的預測系統，讓工具機可以透過人工訓練、學習的方式來具備判斷能力，達到使工具機智能化的目標，並利用多種的預測模型互相輔助佐證，以合適的方法進行比較，藉此判斷預測模型在預測中是否出現顯著差異，若模型間的預測值相近，本研究也將套用最佳公式結合相近的預測值，進一步提升整個系統的預測準確率以及可靠性。
本研究建構三種預測模型，分別為線性迴歸預測模型、支持向量迴歸預測模型以及類神經網路預測模型，最終透過多預測模型演算法結合各個單一預測模型的預測值，並以MAPE計算預測準確率，可得到有顯著提升的研究成果，得以證實本研究提出之多預測模型預測系統的可行性。

關鍵字

CNC銑削加工；線性迴歸；支持向量迴歸；類神經網路；表面粗糙度；多預測模型

並列摘要

The prediction of CNC machining quality is regarded as a very important link in process management. The use of predictive technology can prevent defective products in advance and save the measurement steps. However, most of the current predictive models or monitoring systems are restricted by scholars. The proposed restrictions are a single decision model. Perhaps the predictions results obtained are subjective but the purpose of this researcher have no way of knowing.In order to achieve the goal of intelligence, using effective sensing technology with an appropriate prediction system to allow machine have the ability to make decision through training and learning and use a variety of forecasting models to assist each other and to compare with appropriate methods to determine whether there are significant differences in the forecasting of the forecasting models. If the predicted values between the models are similar, this study will also apply the best formula to combine the similar predicted values to further improve the prediction accuracy and reliability of the entire system.
This research constructs three prediction models, linear regression prediction model, support vector regression prediction model, and neural network prediction model. Finally, multiple decision model algorithms are used to combine the prediction values of each single prediction model, and the prediction accuracy rate is calculated through MAPE. Significantly improved research results can be obtained to confirm the feasibility of the multi-decision model prediction system proposed in this research.

並列關鍵字

CNC Milling；Linear regression；Support Vector Regression；Artificial Neural Networks；Surface Roughness；Multi-model

參考文獻

參考文獻

Ahmad, M., Alam, K., Tariq, S., Anwar, S., Nasir, J., Mansha, M. (2019). Estimating fine particulate concentration using a combined approach of linear regression and artificial neural network. Atmospheric Environment, 219, 117050.

Asiltürk, I., Neşeli, S., Ince, M. A. (2016). Optimisation of parameters affecting surface roughness of Co28Cr6Mo medical material during CNC lathe machining by using the Taguchi and RSM methods. Measurement, 78, 120-128.

Aydemir, A., Terzioglu, Y., Akin, T. (2016). A new design and a fabrication approach to realize a high performance three axes capacitive MEMS accelerometer. Sensors and Actuators A: Physical, 244, 324-333.

Bhuiyan, M. S. H., Choudhury, I. A., Dahari, M., Nukman, Y., Dawal, S. Z. (2016). Application of acoustic emission sensor to investigate the frequency of tool wear and plastic deformation in tool condition monitoring. Measurement, 92, 208-217.