| The selection of cutting tools and their processing state is important in the process of metal cut,because they will directly influence the machining accuracy and surface roughness. And in theproduction process, if the key equipment is not in normal operation due to the tool failure, theentire system will be affected and a huge economic loss even casualty will accompany, so studyingthe tool failure state has a great significance.The current signal has the characteristic of high integration information and can exert itssuperiority when the machine tool is not only one, so the total working current signal was selectedto be the tool failure monitoring signal and in the same time, the hall current sensor whichpossessing the noninvasive installation characteristic was chosen to acquire the current signal.After the possibilities of using the current signal to monitor the tool failure state were analyzed, theC2616-1B lathes were selected to be the experiment object. The single lathe tool failure and lathetool failure were studied on the basic of the real and effective turning current signal were acquired.The wavelet packet decomposition method was chosen to decompose the obtained currentsignal and the four order standard entropy of the reconstruction signal of the nodes which aresensitive to the tool failure were extracted, then some appropriate treatment method was chosen toprocess them and in the end realize the single machine tool failure detection. The collected currentwas processed by wavelet analysis and Fourier transform combining method, the variance of theamplitude of50HZ harmonic was extracted as the characteristic parameters, and the fault state wasdecided by the wave characteristics of them. And in the same time, a carrier cycle energy signalmethod was put forward for the e tool failure state detection.In the multi-machine tools failure diagnosis, the method of power spectrum eliminating andFourier transform were combined to determine the failure state of tools first, then support vectormachine (SVM) method was used to decide the state of the remaining current signal, and thesingle tool failure diagnosis in two sets of machine tools under the experimental conditions wererealized. The judgment of the failure tool was preliminary realized according frequency bandwhich sensitive to the tool failure of the Fourier transform spectrum graph of the total workingcurrent signal. Finally, after the contrast analysis the advantages of the empirical mode decomposition (EMD)method to wavelet packet decomposition technique, the EMD method was chosen to decomposethe total working current of three machine tool, and the energy and variance of four intrinsic modefunctions were extracted as training characteristic vector of the tool failure diagnosis for supportvector, and then single lathe tool failure in three lathes was preliminary realized. |
