Cempel C., Tabaszewski M.

Paper concerns the case of multidimensional vibroacoustical condition monitoring of unstable running machines, working for example with stochastic environment, where the load or some control parameters can not be simple accessed. In the last papers of the first author it was shown, that the use of SVD (Singular Value Decomposition) of transformed symptom observation matrix can be useful in the extraction of fault information. From this decomposition it is possible to obtain some generalized fault symptom, which can be used for example for calculation of symptom limit value, needed for diagnostic decision. As the method of reduction of symptom unstability, the moving average procedure has been taken. The paper contains some comparison of estimation of symptom limit value obtained with, and without, moving average procedure. This was illustrated by some example of real data obtained from condition monitoring subsystem of fans, and also taken form rolling bearing test stand, has been used in the paper. It was shown, that averaging operation applied to the symptom observation matrix can be valuable help, enabling to smoothen the life course of generalized fault, and much better calculate the symptom limit value needed for the final diagnostic decision.