The dressing process creates the grinding wheel shape and topography required for the respective machining task. Fluctuations of the actual depth of dressing cut, temperature drift of the machine axes relative to each other and further sources of disturbance cause dressing errors and thus fluctuations in workpiece quality as well. In order to guarantee a consistent grinding tool quality therefore, the actual achieved depth of dressing cut must be monitored during the process. By means of monitoring systems, the dressing process can be automated and visualised [AVER82, BAUS03, INAS77, KOEN82a, KOEN86]. Present-day strategies for monitoring the dressing process primarily include static thresholds, the lower deviation of which by the URMS-value, which is proportional to the actual depth of dressing cut, signals a faulty depth of cut during the dressing stroke. The prerequisite of this however is a constant AE-signal throughout the dressing stroke (Fig. 10-9). When dressing profiled grinding wheels however, the continuous AE-signal splits into a group of individual signal sections, the number of which corresponds to the geometric elements of the wheel. The level of each signal section is not constant depending on the motion of the dresser. Furthermore, different dressing errors, like dressing tool overheating or dressing tool wear, can occur but nonetheless not manifest themselves in a static
For monitoring complex grinding wheel shapes, the procedure illustrated in Fig. 10-10 is capable. During the dressing cycle, every geometric element is first identified, the gliding average calculated and the static and dynamic thresholds accompanying the process are formed. In this way, we can monitor whether the signal of every geometric element remains within a tolerance range throughout the reference progression. Any deviations that occur can be diagnosed and indicated online.
In addition, the mean signal increase, the average value as well as the standard deviation are calculated from the AE-signals for every recognized geometric ele
ment. Since these parameters are linked significantly with typical dressing errors, a parameter comparison in sections suffices to identify potential process errors. Because of the vectorisation and reduction of classification to a few parameters, this strategy is realisable with minimal need for evaluating computer storage space, and the dressing errors can be diagnosed online.