With the modern metrology we can measure almost all variables in the phenomenon field of a working machine, and many of measuring quantities can be symptoms of machine condition. On this basis we can form the symptom observation matrix (SOM) intended for condition monitoring. On the other hand we know, that contemporary complex machines may have many modes of failure, so called faults. The paper presents a method for the extraction of fault information from the symptom observation matrix by means of singular value decomposition (SVD) in the form of generalized fault symptoms. As the readings of the symptoms can be unstable, the moving average of the SOM was applied with success. The attempt to assess the diagnostic contribution of primary symptom was undertaken, and also some approach to connect SVD methodology with neural nets is considered. These possibilities are illustrated in the paper by processing data taken directly from the vibration condition monitoring of the machine.