Research On Monitoring Technology Of Turning Tool Wear Condition Based On Multi-feature Fusion

Tool wear status monitoring technology is an important part of advanced manufacturing technology.Real-time monitoring of tool wear plays an important role in improving product quality,reducing manufacturing costs,and improving production efficiency.However,this technology has not been able to be employed in actual machining since its development,and it has not been able to solve the problem of accurately identifying tool wear status under variable processing conditions.For this reason,this paper establishes a turning tool wear status recognition model based on multi-feature fusion and a majority voting method based on the research of turning tool wear status monitoring technology.The main research contents and results are as follows:First,the research background and significance of tool wear state monitoring are summarized.According to the different stages of tool wear,the tool flank wear VB value is selected as the tool wear monitoring index.By analyzing and comparing the advantages and disadvantages of various monitoring methods and monitoring signals,vibration and acoustic emission signals are selected as the monitoring signals for the research;according to the performance parameters of the machine tool,the test plan is designed and the full factor test is carried out.The data acquisition system built by LABVIEW 2018 a software collects the vibration and acoustic emission signals generated in the turning test.Analyze the collected signals in the time domain,frequency domain,and time-frequency domain respectively,and obtain the feature vector with the strong correlation with the tool wear status as the original feature.Then,the Relief-F algorithm is used to screen the original features twice,and the final feature parameters that are most relevant to the tool wear state in the vibration signal and acoustic emission signal are obtained.Then,the PCA(Principal Component Analysis)method is used to reduce the original features,and the principal metadata features of the same dimensions as those filtered by the Relief-F algorithm are obtained,and the corresponding feature vectors in the two cases are obtained as The respective final feature samples.Finally,2/3 of the final feature samples obtained by the Relief-F and PCA algorithm are used as the training set to input the established GA-BP neural network model,ELM model,and SVM model for training,and input 1/3 of the final feature sample respectively.3 as a test set,test the above three sub-models respectively.The correct recognition rates of the prior three sub-models were 88.889%,92.592%,96.256%,and the correct recognition rates of the latter three sub-models were 81.48%,77.7778%,and 77.7778%,respectively.Afterward,the majority voting method was utilized to integrate the output results of the three sub-models in the case of Relief-F and PCA algorithms.The results showed that the established turning tool wear status recognition model based on multi-feature fusion and majority voting method is in Relief-F.The correct recognition rates in the case of Relief-F and PCA are 96.296%and 85.185%,respectively.This demonstrates that the performance of the model obtained after multi-feature fusion is better than that of a single sub-model,and the tool wear state recognition model based on the Relief-F algorithm and the majority voting method is also significantly better than the PCA algorithm.The model was established after dimensionality reduction.Therefore,the recognition model based on the multi-feature fusion and majority voting method established by the Relief-F algorithm has a good recognition and monitoring effect on the tool wear status.