| 摘要 |
<p>PURPOSE:To prevent the deterioration of the learning efficiency caused by the noises and at the same time to improve the recognizing factor by using the cross spectrums of the time fluctuation value set at different time points as the input signals of a neural. network. CONSTITUTION:The cross spectrum Wxy is obtained by a cross spectrum computing element 5 between a vibration signal. (x) and a sound signal (y) and then inputted to a neural network 6. The network 6 produces the output signals showing the degree of abnormality of a motor 1, the factor of the abnormality, etc. Thus more plentiful information are available for decision with use of both signals (x) and (y) than a case where just one of both signals is used. In addition, the spectrum Wxy of both signals (x) and (y) is used as the input signal of the network 6. As a result, the noise components included in both signals (x) and (y) independently of each other are inputted with erasion or attenuation.</p> |
