Rotating machine abnormality detection device and method, and rotating machine

A rotating machine abnormality detection device to detect abnormality in a rotating machine having at least a first rotating body and a second rotating body which rotate around respective predetermined shafts while engaging with each other with a predetermined gap provided therebetween in a normal state, the device comprising:

• a central processing unit (CPU);
• a sensor to measure vibration caused by at least one of the first and second rotating bodies;
• a spectrum processing unit, to be executed by the CPU, to obtain a frequency spectrum of measurement data measured by the sensor;
• a feature amount processing unit, to be executed by the CPU, to obtain a predetermined feature amount related to a predetermined frequency component, based on the frequency spectrum obtained by the spectrum processing unit;
• an abnormality determination unit, to be executed by the CPU, to determine presence/absence of an abnormality in the rotating machine, based on the predetermined feature amount obtained by the feature amount processing unit;
• and
• a noise determination unit, to be executed by the CPU, to determine, when determination is made that abnormality is present by the abnormality determination unit and based on the frequency spectrum obtained by the spectrum process unit, whether the determination of abnormality is caused by noise or not,
• wherein when the noise determination unit determines that the determination of abnormality is not caused by the noise, the abnormality determination unit is to confirm the determined abnormality as a confirmed abnormality,
• when a first number of teeth and a first rotation frequency of the first rotating body are represented as MA [unit] and VA [Hz], respectively, a second number of teeth of the second rotating body is represented as MB [unit], a least common multiple of the first number of teeth MA and the second number of teeth MB is represented as XAB, a basic wave mesh frequency VA/MA obtained by dividing the first rotation frequency VA by the first number of teeth MA is represented as f1 [Hz], and n-th harmonic mesh frequencies (n is an integer of 2 or more) with respect to the basic wave mesh frequency f1 are represented as fn [Hz],
• the feature amount processing unit obtains, from the frequency spectrum, a basic wave component F1 of the basic wave mesh frequency f1 and n-th harmonic components Fn of the n-th harmonic mesh frequencies fn (n=2 to XAB) up to the least common multiple XAB, and obtains the predetermined feature amount based on the obtained basic wave component F1 and n-th harmonic components Fn.

A rotating machine abnormality detection method for detecting abnormality in a rotating machine having a first rotating body and a second rotating body which rotate around respective predetermined shaft while engaging with each other with a predetermined gap provided therebetween in a normal state, the method comprising:

• a vibration measuring step for measuring vibration caused by at least one of the first and second rotating bodies;
• a spectrum process step for obtaining a frequency spectrum of measurement data measured in the vibration measuring step;
• a feature amount process step for obtaining a predetermined feature amount related to a predetermined frequency component, based on the frequency spectrum obtained in the spectrum process step;
• an abnormality determination step for determining presence/absence of abnormality in the rotating machine, based on the predetermined feature amount obtained in the feature amount process step;
• a storage step for storing in a memory, as a normal frequency spectrum, the frequency spectrum of measurement data measured by a sensor that indicates that abnormality is absent, and
• a noise determination step for determining presence/absence of noise superimposing on the measurement data measured in the vibration measuring step, when determination is made in the abnormality determination step that abnormality is present and based on differences between the frequency spectrum obtained in the spectrum process step and the normal frequency spectrum stored in the memory,
• wherein
• when determination is made in the noise determination step that the noise is absent, the determined abnormality is confirmed as a confirmed abnormality in the abnormality determination step, and
• determination as to whether the determination of abnormality is caused by noise or not is made in the noise determination step, based on a subtraction result obtained by subtracting one of an average value and a median value of the differences between the frequency spectrum obtained by the spectrum processing unit and the normal frequency spectrum stored in the memory from a maximum value of the differences.