Heterogeneous Data Fusion Based On Ensemble Empirical Mode Decomposition And D-S Evidence Theory

In machining of parts,the surface quality of parts,such as flatness,is not only related to processing methods and technological parameters,but also closely related to tools' wear.Maintaining a good tool condition and changing a worn tool in time are more than important in modern manufacturing.Starting from indirect method of tool condition monitoring,this paper focuses on heterogeneous signals from sensors installed on milling machines,studies multi-scale signal decomposition algorithms and heterogeneous data fusion methods,and then establishes the relationship between the sensor signals and tool wear degree.The main research results of the dissertation are as follows:Firstly,this paper proposed an improved EEMD algorithm that can adaptively selects the best white noise amplitude coefficient and the ensemble number for the original signal by taking extreme points distribution into consideration.Taking sound signal,vibration signal and acoustic emission signal as research objects,the paper compared the improved EEMD algorithm with traditional EEMD algorithm,and found out the advantage of the improved EEMD algorithm in effectiveness and accuracy in decomposition.Secondly,the paper puts forward an interval value transformation method for D-S evidence theory in order to fuse heterogeneous data at feature level.Features of actual milling signals are extracted through the improved EEMD,transformed into mass function that D-S evidence theory can identify through an interval value table,and then fused through D-S evidence theory.Thus,tool condition can be recognized.Results show that the heterogeneous data fusion method is greatly better than a single signal character method in tool condition recognition.Finally,the above method is verified by a milling experiment.Moreover,flank wear of the tools and flatness of the parts are measured in this experiment.The paper studies their relationship and confirms that the greater the flank wear is,the worse the surface flatness is.This reflects the practical significance of the paper.