| Summary: | 碩士 === 中原大學 === 工業與系統工程研究所 === 103 === As the changes of the market trend over the years, production orientation of machining process has evolved from mass production to high-quality, small-volume production; therefore, quality control and production prediction is becoming increasingly important. Some prediction methods such as statistical forecasting, simulation modeling, regression analysis and soft computing, have been proposed in recent years to construct prediction systems, which demonstrate significant increasing in accuracy. However, there is a defect to these prediction models. That is they require large amount of data to perform modeling, which is not suitable to small-volume productions. Not only does it take too much time for these systems to collect data, but remodeling or readjustment is required every time once the settings are changed. Due to the inconvenience, these prediction systems are rarely used in the industry. While the Grey theory contrary coincided with the above theory, Grey theory can utilize "small data", "small information" to find the information trend and optimize it. Therefore, in this research, a tool life prediction system for milling process has been constructed, which is able to predict a tool condition and tool life based on only a small amount of data and without the influence of the changed process settings.
The Grey carbide inserts of tool life prediction system for milling operations is developed in this research through grey generating, a method which emphasizes the input data trend and uses its small sample feature as the fundamental structure for building an in-process prediction system. Grey prediction is a small-volume prediction model built upon accumulated generating sequence and the forecasting characteristics among data. It is used on CNC milling process in this research in which a tool life is predicted based on a small amount of surface roughness data. The grey prediction system can quickly predict a tool life under specific processing settings without considering the machining parameters and environment, also does not need any sensors.
To prove the proposed method is both accurate and reliable, three different sets of machining parameters are used to perform the milling operations. A small amount of the surface roughness data is placed in the grey prediction system for the prediction of surface roughness. Based on the information obtained, we can determine a tool life is determined and its accuracy is investigated to verify the feasibility of the prediction system.
|
|---|
