Model and experimental verification of grinding surface roughness based on acoustic emission

Abstract

To predict the surface roughness in the grinding process, an acoustic emission device (AE) was incorporated into the grinding process to monitor the grinding state using AE signals. The variations in AE signal characteristic parameters and frequency spectrum with respect to grinding parameters, such as grinding depth ap, grinding wheel speed vs and feed speed vw were analyzed. The results show that as ap and vw increase, the effective values and ringing count values of the AE signal's characteristic parameters both increase. The main energy concentration spectrum of the AE signal is between 90 and 140 kHz, and the corresponding spectrum amplitude shows a gradual increasing trend. With the gradual increase of vs, the effective value of AE signal characteristic parameters gradually decreases, the ringing count value gradually increases, and the spectral amplitude corresponding to the frequency band shows a gradual decreasing trend. Further data analysis reveals the corresponding relationship between AE signal characteristic parameters and machining surface roughness, providing a sample for establishing a surface roughness prediction model. The multi-information fusion algorithm, based on a BP neural network, is used to reasonably fuse various characteristic parameters of the AE signal. And the multi-information fusion prediction model for grinding surface roughness based on AE signal was established. After experimental verification, this model can predict the roughness of the ground surface in actual production.

Keywords

• grinding process
• acoustic emission（ae）
• surface roughness
• multi-information fusion