Research On Monitoring The Tool Wear On Line Based On The Milling Forces

With the rapid development of the society, a fierce competition on intelligent manufacturing has emerged around the world. The development of CNC machines fundamentally changed the manufacturing industry in that they increased manufacturing flexibility while ensuring consistency of manufacturing output, however automatically tool changing is still a tough problem which we are facing. For a long time traditional manufacturing required workers to change almost all the tools and to identify faults pertaining to tool failure depending on worker's experience including the knowledge of sound, the color of the chips or the feeling of their hands between the normal condition and tool failure condition. Nowadays, automatic tool changing in most machining processes must take place at some time and is dependent on the intelligent manufacturing system available during the process of machining, which can be achieved through some useful physical signal collected in real time. Therefore it is of great significance to solve this global problem to improve the manufacturing level. To resolve these problems, this main research of this dissertation is focusing on monitoring the tool wear during the process of milling titanium(TC4) based on the milling forces as summarized some advanced tool wear monitoring methods, detailed as follows:Firstly, different advanced tool wear monitoring methods were summarized, and the characteristics of tool wear when milling titanium were analyzed, milling forces and work piece surface roughness were determined to monitor tool wear conditions. Afterwards a single tool milling titanium alloy(TC4) experiment platform was built. Then multiple groups of experimental measurements were made and 6 different kinds of tool wear data in different cutting parameters were eventually obtained.Secondly, the research studied the time domain and frequency domain related to the cutting forces, the relationship between cutting forces & surface roughness and the cutting power was analyzed. Furthermore, the study analyzed the cutting forces & surface roughness changes under various cutting conditions. Some useful signal characteristics in both the frequency domain and the time domain were also extracted. The research also analyzed the energy values & standard deviation in different frequency bands and finally extracted the characteristic parameters by the correlations between tool wear and different characteristic signals.Thirdly, the research analyzed neural network technology used to predict the tool flank wear(VB) value and tool wear state identification after network training. The test results showed that it is accurate to use the neural network to predict tool wear. Meanwhile the Mahalanobis distance was used to identify the tool wear.