Online feature trend monitoring and signal pattern clustering in milling process

Abstract

Data mining is a powerful technology used in the manufacturing industries to discovery useful information. Data mining technology could be integrated with computer integrated manufacturing system in order to analyze the data of the real situation of the manufacturing process. Clustering is one form of unsupervised learning used in data mining. There are a number of applications in a wide variety of area such as pattern recognition, image processing, automatic control, communication, and bio-information.

Downtime and malfunction of industrial equipments represents a significant cost burden and productivity loss. Fault diagnosis of such industrial equipments is carried out to pinpoint the location of these fault(s) and their cause(s). Statistical methods and clustering methods can be used to build and recognize archived error patterns to predict errors expected to occur in the manufacturing system. Each method has advantages and disadvantages. There are limitations of statistical clustering, namely the lack of basic ability to learn patterns such as K-means clustering, that may affect the accurate of the clustering result. Clustering differs from classification because no pre-specified category defines the observation; there is no clearly defined outcome variable. Therefore a different set of techniques is needed to analyze the data. Because there are no outcome categories, there are no right answers. Clustering focuses on determining whether the groupings found are meaningful. In this project, research will be carried out on K-means clustering as well as online statistical clustering, which arm to develop methodologies to help for milling tool optimized cutting parameter determination and tool degradation detection in tool condition monitoring.