On the issue of automatic form accuracy during processing on CNC machines

Abstract

This work aims to develop technical solutions that allow providing the specified parameters of the accuracy of the shape of parts in the cross-section during processing on a CNC machine. Experimental studies were performed on a screw-cutting lathe. An acoustic signal in the range from 6 to 12 kHz was used as a diagnostic sign to assess the wear of the cutting tool, since during preliminary studies, it was found that this range is most sensitive to changes in processing modes. Studies were performed at different values of wear of the cutting tool (estimated by the width of the wear chamfer). For estimating the life of a cutting tool, a neuro-fuzzy model has been developed. Using models of this class allows adjusting to specific conditions (machine, tool), and correctly evaluating the tool life. The model error for the test sample does not exceed 10%. The test results showed that using the proposed solutions makes it possible to increase the accuracy of the manufacturing of shut-off valve parts by 20-30%.

Keywords

• vibration signal
• correlogram analysis
• neuro-fuzzy models
• correlation entropy
• neural network database