Study On On-line Condition Monitoring For Grinding Wheel Based On Acoustic Emission

Usually, Grinding is very important in manufacturing. It is very important to judge whether the grinding wheel contact with the workpiece, and to know the condition of the in the grinding process. The additional methods is that workers judge the grinding state with their experience, then dress the wheel at regular time, which has great limitations. If dress grinding wheel earlier than necessary, it will trim the grinding efficiency, cause great waste; if later, it will affect the quality of the products. This seriously hindered to grinding technology automation and intelligent development.Acoustic Emission signals can be generated in the grinding process. Based on the technology status of tool condition monitoring, this paper has carried out research on the grinding wheel condition monitoring by acoustic emission signals, primarily deals with the following research efforts:(1) Construct the grinding wheel monitoring experiment system in grinding. Acoustic emission transducer and A/D data acquisition card are used to collect AE signals in the different working condition. The analyzing of the AE signals shows that the relevant data will change with the different working condition. These changes have monotony, repeatability, which proves the effectiveness of hardware data acquisition system and the feasibility that the tool condition could be monitored well by acoustic emission signals.(2) Analyse the AE signals based on wavelet transform, and extract the eigenvector of the signals. We filter the AE signal collected in the grinding processing, analyze it by wavelet analysis. Eigenvectors of grinding condition can be constructed by Wavelet transform coefficients using RMS method, and used to judge the different grinding conditions to achieve the state of the online grinding monitoring, also can be used as the input of neural network. The result shows that this method can effectively describe processing condition.(3) Construct neural network for pattern recognition. We establish three layers of BP neural network as grinding wheel state recognition classifier. The chose parameters are used to determine the optimized neural network structure. Network training and simulation test results shows that this technology can recognize the wheel states effectively and achieve the goal of recognize the different working condition in the grinding process, accuracy in exceeds 96%.