Theoretical Research On Tool Wear Condition Mornitoring And Wear Loss Prediction Based On Cloud Theary

Along with the development of equipment manufacturing industry,tool wear condition monitoring technology is a key technology for restricting the modern automation machine tool,the technology has not been solved effectively at present.Monitoring tool condition in real-time,it can improve the quality of machining parts and machine tool processing efficiency,reduce machine accidents,maximally reduce the operation of the machine tool,implement the intelligent of the machine tool and unmanned,ensure system running under the optimal parameters.Therefore,tool wear condition monitoring technology research is very urgent and important.This paper aimed at the research topic of tool wear condition monitoring and wear loss prediction under different cutting conditions,arranged cutting test in the way of orthogonal test.On the basis of the acquisition of the acoustic emission signal,the modern signal processing method of the wavelet packet analysis combined with the optimal entropy theory was applied to realize signal filter processing,and the uncertainty of cloud model theory was introduced into the feature extraction for different tool wear stage,and put forward method for tool wear state recognition based on cloud model theory and least squares support vector machine(LS-SVM).Finally,the cloud of uncertainty reasoning method was applied to implement wear prediction uncertainty.The main research content consists of the following several parts:1 ? In the past time,tool wear monitoring signal filtering used time domain analysis(empirical mode decomposition),analysis of frequency domain(power spectrum analysis)and other traditional signal pretreatment method,because of the non-stationary and non-linear characteristics of the acoustic emission signals,this paper introduced the wavelet packet analysis method,which is suitable for processing non-stationary signal,into the signal preprocessing to realize signal's filtering.Firstly,the distribution of the frequency band range under different stages of wear acoustic emission signal was obtained by frequency spectrum analysis,and took the result as the qualitative reference of wavelet packet decomposition levels.Second,Shannon entropy in the information entropy theory was applied to represent the size of the noise,to determine the best wavelet packet decomposition tree.At last,the optimal branch of best tree of wavelet packet decomposition was determined through statistical analysis,and the signal was reconstructed after threshold processed,and signal-to-noise ratio showed the denoising effect,the signal to noise ratio was more than 35 dB.2?The uncertain acoustic emission signal feature extraction method was proposed based on the cloud theory.First,the improved reverse cloud algorithm was applied to extract thecharacteristic parameters of different wear loss acoustic emssion signal,expectations,entropy,and hyper-entropy.Secondly,quantitative analysis of the three types of cloud characteristics parameters of cutting tool in different cutting conditions with the increase of wear is presented by change rule.Finally,the effectiveness of characteristic of three kinds of parameters on tool wear acoustic emission signal was analyzed during feature extraction.3? Cloud characteristics parameters and the combination of least squares support vector machine(LS-SVM)method for tool wear state recognition was presented.Aiming at the uncertainty of the tool wear acoustic emission signal and neural network learning algorithm's slow convergence speed,easy to fall into local minimum value,and higher feature requirements,least squares support vector machine method for tool wear state recognition was put forward based on cloud theory.Example analysis shows that under the condition of optimized parameters selection of support vector machine,the cloud-support vector machine(SVM)with the method of recognition rate is higher than the traditional neural network recognition method.4?A cloud of uncertainty reasoning model was applied to the prediction of cutting tool wear loss.Firstly,mined the relationship between different wear stages wear trends and the different stages of wear cloud characteristic parameter data.Second,built cloud prediction rules based on the condition of cloud generator.Last,set up many conditions and single rules wear loss prediction model based on this.Results shows that the cloud reasoning tool wear prediction model conformed to the law of tool wear.For non-deterministic model prediction,cloud reasoning is more in line with the actual situation than fuzzy reasoning.In addition,this method can reflect the real-time condition of the tool wear,has strong practicability.