Study On Sick Tool Recognition Technology Of Horizontal Drilling Machine Based On Sound Frequency Characteristics

Sick tool is a kind of fault tool caused by damage or serious wear in long-term machining.Tool condition monitoring is one of the key technologies to ensure smooth machining process and realize intelligent machining.It is of great significance to ensure product processing quality,improve production efficiency and realize automatic production line.Therefore,the development of accurate,reliable and low-cost tool condition monitoring system has become a research hotspot.In many tool condition monitoring methods,the sound monitoring signal directly comes from the work area,has the advantages of low installation conditions,high signal sensitivity and fast response speed,is conducive to accurate and rapid diagnosis of the tool condition,is a relatively new monitoring means.Since the sound frequency of tool drilling is low,it is easily disturbed by various noise factors in the complex working environment.How to extract the characteristic quantity related to the tool state under the interference noise and build a novel and practical tool state recognition system is the key of tool state recognition.In this paper,the following research works have been completed:(1)Based on the analysis of the geometric structure of the tool,the forms of breakage and the causes of its formation,firstly,the sick tool is divided into two types:the broken tool and the serious wear.Then,taking the horizontal drilling machine of production line as an example,the data acquisition platform was built,and the sound signals of different cutting tools in different states were collected by classification.(2)The related monitoring technology of tool state feature extraction is studied.Based on the statistical feature extraction in time domain and frequency domain of preprocessed monitoring signals,the wavelet packet decomposition technology was proposed to collect the energy value of each frequency band wavelet packet sensitive data,and the feature sample data set for tool state recognition was constructed.(3)An ELM based model for pathological tool pattern recognition is presented,and a comparative study is made with BP model and RBF model as the reference.The research shows that after 600 sets of sample data were trained to establish the model,the recognition accuracy of the model was tested by another 300 sets of data.Comparing the research results,it is verified that the ELM model has higher accuracy and better real-time performance for tool state recognition.(4)With the help of MATLAB software,a tool breakdown state identification system is realized.The preliminary test and application practice show that the tool condition monitoring system based on ELM is reasonable and feasible.