A multi-sensor tool breakage detection system is introduced that characterizes the state of measurements during normal (no-fault) condition and at tool breakage by the two columns of a multi-valued influence matrix (MVIM). In this system the measurements are monitored on-line and flagged upon the detection of abnormalities. Tool breakage detection is performed by matching this vector of flagged measurements against the two columns of MVIM, which are estimated during a training session so as to minimize the error in detection. The detection system is implemented in turning. Experimental results indicate that this system provides excellent detection when the full range of tool breakage effect on the measurements is included in training, and that its performance is less dependent upon the training set than a multilayer neural net.

This content is only available via PDF.
You do not currently have access to this content.