Research On On-Line Monitoring Technology For FSJ Of SiCp/Al Composites

As a new type of solid-state connection technology,Friction Stir Joining(FSJ)has the unparalleled advantages of traditional welding and has drawn wide attention both at home and abroad.Due to the wear of the stirring head,the current FSJ is mainly used in the connection of light metal alloys such as aluminum and magnesium,while the connection and processing of titanium alloy and composite materials still belongs to the current research hotspot.Among them process monitoring research is one of the most important research directions.The characteristic signals of temperature,vibration and acoustic emission during FSJ process can reflect the machining status and the wear status of the stirring head.Therefore,it is very important and valuable to research the FSJ on-line monitoring technology of SiCp / Al composite.The main work carried out in the thesis is as follows:(1)Design and develop a friction stir welding online monitoring system based on Labview virtual instrument technology.Through the infrared temperature sensor,acceleration sensor,acoustic emission sensor and corresponding signal conditioner,the system can achieve signal acquisition,filtering,real-time display,storage and other functions.The system also uses a modular design,so there is less interference when collecting signals,which is convenient for the expansion of other function modules.(2)Design an experiment to explore the optimum welding parameters range of SiCp/Al composites;and study the characteristics of various sensing signals through the self-developed online monitoring system.Then use time domain analysis and wavelet analysis to handle the signals and obtaining the change rule of root mean square and wavelet energy distribution and other indicators.(3)Design an experiment to study the wear rules of the stirring head in SiCp / Al composites during FSJ process.Use online monitoring system and Matlab signal processing system to collect and process vibration and acoustic emission signals under different wear conditions.Then extract the characteristic parameters that is related to the wear states and construct a wear state mapping model of the stirring head,which lay the foundation for subsequent online monitoring of the stirring head's wearing states.(4)Research the BP neural network and decision-making neural network.According to the relationship in the mapping model,it is successfully applied to the predict the wear state of the stirring head.After the training and testing of the sample data,a good recognition effect has been achieved.And it is further proved that the combination of multi-signal and multi-feature quantities can reflect the wear state of the stirrer more accurately and comprehensively than single-signal single feature.