JP2550066B2 - Method for diagnosing machine abnormality using acoustic signals - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a method for diagnosing a machine abnormality by means of an acoustic signal, and is particularly suitable for performing a failure diagnosis of, for example, a computer peripheral device using an acoustic sensor. The present invention relates to a method for diagnosing a machine abnormality using an acoustic signal.

[Conventional technology]

2. Description of the Related Art Conventionally, development of failure prediction technology for rotating machines and the like has been advanced, and as a method for diagnosing rotating machines in large plants and equipment, for example, a technology described in JP-A-59-63528 is known.

The technique is to detect the acceleration of a rotating machine, obtain a vibration spectrum in a predetermined frequency range based on the detection result, compare the spectrum with a reference spectrum, or obtain a spectrum effective value and compare it with a reference value. Therefore, the method of diagnosing abnormality was used.

In this method, a vibration sensor such as an acceleration detector is attached at an appropriate position on the rotating machine, and the same sensor is used for one machine.

[Problems to be solved by the invention]

The above-mentioned conventional technique has a problem that assembly and dimensional accuracy vary when diagnosing a large number of machines of the same model, or vibration when an unspecified number of people try to adopt the diagnosis method. No consideration has been given to the fluctuation of the acoustic spectrum caused by the variation of the sensor mounting position, and there is a problem that the reliability of the diagnostic ability is lowered.

Further, with respect to a machine that is not suitable for measurement of detecting the acceleration of a rotating machine, such as a computer peripheral device, abnormality diagnosis cannot be performed by the above rotating machine diagnosis method.

The present invention has been made in order to solve the above-mentioned problems of the prior art, and it is possible to correct the fluctuation of the acoustic spectrum for each frequency of the spectrum and use the same judgment criterion for a large number of machines. Can increase the ability to diagnose abnormalities,
It is an object of the present invention to provide a method for diagnosing a machine abnormality using an acoustic signal.

[Means for solving problems]

The above-mentioned purpose is to detect the sound pressure of a machine, perform a plurality of measurements for obtaining an acoustic spectrum within a predetermined frequency range based on the detection result, and perform acoustic processing by statistically processing the acoustic spectrum obtained by the plurality of measurements. This is achieved by obtaining a variation amount indicating the magnitude of the variation of the spectrum, correcting the acoustic spectrum by this variation amount, and comparing the corrected acoustic spectrum with a preset reference spectrum to make a machine abnormality determination. .

It should be noted that, for the purpose of the diagnosis target machine, the acoustic spectrum is measured by a large number of machines in a steady state by a large number of people, and for each frequency of the acoustic spectrum, statistical processing is performed on the variation of the magnitude, For example, the standard deviation is obtained, and it is added to the acoustic spectrum that is measured for each period as a varying acoustic spectrum that represents variation.
Subtract or divide to correct, convert to the acoustic spectrum originally sounded by the machine to be diagnosed, and determine the abnormality of the machine based on whether the ratio of this to the preset reference spectrum exceeds a predetermined value. Is achieved by

[Action]

The sound pressure measured is emitted from the machine, but generally the sound of the machine is slightly different for each individual. Further, when a large number of measurers attach the sensor to a large number of machines at a certain position, an error is likely to occur in the attaching position or the attaching direction.

Therefore, assuming that the n-th acoustic spectrum in a predetermined frequency region is given a subscript of n, the n-th acoustic spectrum is represented by P (n), and the acoustic spectrum measured many times is represented by P _i (n), The average value of them is (n)
This average acoustic spectrum (n) can be expressed by the following equation.

(n) = {P ₁ (n) + P ₂ (n) + ... + Pm (n)} / m When the fluctuating acoustic spectrum is represented by Pσ (n), Pσ ² (n) = [{P ₁ ( n)-(n)} ² + {P ₂ (n)-(n)}
² + ... + {Pm (n)-(n)} ² ] / m.

Since the fluctuating acoustic spectrum Pσ (n) includes the above error factors, the measured value for the diagnostic target machine is corrected. When the corrected spectrum is expressed as Pa (n), for example, Pa (n) is expressed as Pa (n) = P (n) · (n) / Pσ (n), and this Pa (n) is expressed as By comparing with (n), Pa (n) becomes small in the frequency region where the fluctuation is large, that is, Pσ (n) is large, and its effect becomes small when compared with (n), so the diagnostic ability is improved. .

〔Example〕

 An embodiment of the present invention will be described below with reference to the drawings.

The drawing is a block diagram of a preferred apparatus for carrying out a method for diagnosing a machine abnormality by an acoustic signal according to an embodiment of the present invention.

In the drawings, 1 is a sound pressure sensor that detects the sound pressure of a machine to be diagnosed, 2 is an amplifier that amplifies the signal of the sound pressure sensor 1, and 3 is a filter that passes a signal of a certain frequency. , By the micro computer 7,
Switching is performed via the frequency switching unit 4. 5 is an effective value circuit, 6 is an A / D converter, and the output of the filter 3 is
It becomes an effective value by the effective value circuit 5, is digitized by the A / D converter 6, and then inputted to the microcomputer 7 and stored in the memory 9. The memory 9 stores a system for operating a microcomputer such as a calculation program and a display program.

In the present embodiment, according to the above, the microcomputer 7
The acoustic spectrum taken in is stored in the memory 9 at the sound pressure level SPL (n) as in the following equation.

SPL (n) = 20log ₁₀ P (n) / P ₀ Here, P ₀ : × 10 ⁻⁵ (N / m ² ) Hereinafter, a method for obtaining the corrected acoustic spectrum will be described.

Average sound pressure level measured over m times ▲
▼ (n) is ▲ ▼ (n) = {SPL ₁ (n) + SPL ₂ (n) + ... + SPLm (n)}
/ m, and its standard deviation SPLσ (n) is calculated by the following equation.

The corrected acoustic spectrum SPLa (n) is obtained by using SPLA (n) = SPL (n) −SPLσ (n) as a corrected acoustic spectrum.

In this way, by subtracting the corrected acoustic spectrum SPLσ (n) from the corrected acoustic spectrum SPLa (n), the influence of machine sound variations and measurement variations is reduced, and the diagnostic ability is improved.

The acoustic spectrum SPLa (n) after the correction is calculated by using a reference spectrum, which is empirically determined in advance and set, for example, an average spectrum ▲ ▼ (n), ▲ ▼ (n) +
When it exceeds the reference spectrum represented by 10 (db), the display circuit 8 displays an error.

According to this embodiment, the measurement error peculiar to the sound pressure sensor,
Since it is possible to correct the difference in the sound emission characteristics of a large number of machines, the diagnostic ability is improved. Further, since the same judgment standard can be used for a large number of machines, it is expected that the judgment can be speeded up and the diagnostic efficiency can be improved.

〔The invention's effect〕

As described above, according to the present invention, the fluctuation of the acoustic spectrum can be corrected for each frequency of the spectrum, and the same criterion can be used for a large number of machines, thereby increasing the abnormality diagnosis capability. It is possible to provide a method for diagnosing a machine abnormality using an acoustic signal.

[Brief description of drawings]

The drawing is a block diagram of a preferred apparatus for carrying out a method for diagnosing a machine abnormality by an acoustic signal according to an embodiment of the present invention. 1 ... Sound pressure sensor, 4 ... frequency switching unit, 5 ... effective value circuit, 7 ... microcomputer, 8 ... display circuit,
9 ... Memory.

Front page continued (72) Inventor Hidetetsu Yasuda 504-2 Shinanomachi, Totsuka-ku, Yokohama City Hitachi Electronics Service Co., Ltd. (72) Hajime Fujita 502 Kandachimachi, Tsuchiura City Hitachi Ltd. Mechanical Engineering Laboratory

Claims (1)

(57) [Claims] 1. A sound pressure of a machine is detected, an acoustic spectrum within a predetermined frequency range is measured based on the detection result, and the acoustic spectrum obtained by the plural times is statistically processed. It is characterized in that a variation amount representing the magnitude of variation of the acoustic spectrum is obtained, the acoustic spectrum is corrected by this variation amount, and the machine abnormality determination is performed by comparing the corrected acoustic spectrum with a preset reference spectrum. Method for diagnosing machine abnormality by sound signal.