DE102011121789B4 - Method for determining a defective part in a transmission and device set up for carrying out the method - Google Patents

Abstract

Method for determining a defective part in a gearbox, whereby a sensor for detecting acoustic vibrations is brought into contact with the gearbox, whereby the gearbox is operated in neutral, with the input shaft of the gearbox being operated at a speed, with the sensor signal being sent to an evaluation unit is supplied, where (i) the kurtosis of the time profile of the sensor signal is determined, where (ii) the kurtosis of an envelope of the sensor signal is determined, where (iii) a kurtogram, i.e. the kurtosis as a function of the frequency, is determined, with one of a respective point in time from the beginning of the time segment of the sensor signal is Fourier transformed, with the point in time being shifted after each Fourier transformation and the Fourier transformation then being carried out again in turn, so that a value profile dependent on time, i.e. on the respective points in time, is present for each frequency value, whose kurtosis value is determined, where the j the kurtosis value determined in each case is compared with a specified value and the envelope of the sensor signal is determined and if the kurtosis value exceeds the specified value, the frequency values associated with the amplitudes protruding from the background noise are determined, with these frequency values being compared with the determined speeds of the toothed parts of the Transmission and its harmonics, ie its harmonics located in the detected spectrum, where the defective part is determined if they match, and where a bearing is reported and/or displayed as a defective part if they do not match.

Description

The invention relates to a method for determining a defective part in a transmission and a device set up for carrying out the method.

It is generally known to detect acoustic waves on a housing part with a structure-borne noise sensor.

From the publication Klein Ulrich: Vibration diagnostic assessment of machines and systems. In: 2000 Verlag Stahleisen GmbH, Düsseldorf, ISBN 3-514-00663-6, 2000, pages 19-21 and 43 and 76-112, a method for sound evaluation is known as the closest prior art.

From Randall R.B.: Applications of Spectral Kurtosis in Machine Diagnostics and Prognostics. In: Key Engineering Materials Vols. 293 - 294, 2005, 21 - 30 a kurtogram is known.

From the publication Immovilli F, Cocconcelli M, Bellini A, Rubini R: Detection of Generatized-Roughness Bearing Fault by Spectral-Kurtosis Energy of Vibration or Current Signals. In: IEEE Transactions on Industrial Electronics. Vol. 56 No. 11, November 2009, 4710-4717 a fault detection system is known that uses a kurtogram.

The invention is therefore based on the object of further developing a method for sound evaluation, with a defective part being able to be identified.

According to the invention, the object is achieved in the method for determining a defective part in a transmission according to the features specified in claim 1 and in the device according to the features specified in claim 5.

Important features of the invention in the method are that the method is intended to determine a defective part in a transmission,
a sensor for detecting acoustic vibrations being brought into contact with the transmission,
wherein the transmission is operated at least in neutral, with the input shaft of the transmission being operated at a speed,
the sensor signal being fed to an evaluation unit,
wherein at least one value for kurtosis is determined from the sensor signal.

The advantage here is that the defective part can be determined by determining a statistical variable of the discretized signal, that is to say in particular that defective part which causes the damage can be determined. It is thus possible to replace the part and therefore a cost-effective repair can be carried out.

According to the invention, the kurtosis of the sensor signal is determined, in particular the kurtosis of the time profile of the sensor signal. The advantage here is that the discretized sensor signal, i.e. a simply recorded set of values of the sensor signal, in particular acceleration values recorded at respective times, can be used in order to use a moment of the distribution of the values that can be determined with little computing effort to determine the defective one Part.

According to the invention, the kurtosis of an envelope of the sensor signal is determined
In particular, the sensor signal is pre-filtered, rectified, low-pass filtered and then Fourier-transformed to form the envelope. The advantage here is that the frequencies generated by the damage can be determined with a significant signal-to-noise ratio.

According to the invention, a kurtogram, i.e. the kurtosis as a function of the frequency, is determined, with a time segment of the sensor signal beginning at a particular point in time being Fourier transformed, with the point in time being shifted after each Fourier transformation and the Fourier transformation then being carried out in turn becomes,
so that for each frequency value there is a value curve which is dependent on time, that is to say on the respective points in time, and whose kurtosis value can be determined. The advantage here is that a very good significant signal-to-noise ratio can be achieved and thus a low susceptibility to errors.

In an advantageous embodiment, several values for kurtosis are determined from the sensor signal. The advantage here is that an even more reliable determination is made possible.

According to the invention, the determined kurtosis value is compared with a predetermined value and, if the predetermined value is exceeded, a defective part is concluded. The advantage here is that a simple detection of the defective part is made possible.

In an advantageous embodiment, the determined kurtosis value is compared with a predefined value and if the predefined value is exceeded, the frequency value assigned to the kurtosis value is used to determine the defective part, in particular where the frequency value is referred to as the damage frequency value. The advantage here is that a simple, rapid determination of the defective part is made possible.

In an advantageous embodiment, the kurtosis value determined in each case is compared with a predetermined value and the envelope of the sensor signal is determined and, in particular if the kurtosis value exceeds the predetermined value, the frequency values associated with the amplitudes protruding from the background noise are determined,
whereby these frequency values are compared with the determined speeds of the toothed parts of the gear and their harmonics, i.e. in particular their harmonics in the recorded spectrum,
where the defective part is determined if they match
and where a bearing is reported and/or displayed as a defective part in the event of non-compliance. The advantage here is that it is possible to determine the defective part, ie the toothed part or the bearing.

In an advantageous embodiment, the speed of a shaft of the transmission is determined
in particular by means of a speed sensor connected to the shaft, by means of an angle sensor connected to the shaft or by means of a speed-controlled electric motor driving the transmission, the setpoint speed provided for the speed control or the actual speed determined during the speed control being used as the speed of the shaft of the transmission. The advantage here is that a speed value can be determined without any effort.

In an advantageous embodiment, the speed values of the other rotatably mounted toothed parts of the gearbox are determined from the gearing data of the gearbox from the determined speed of the shaft of the gearbox
and associated harmonics, in particular harmonics, of the speed values of all rotating toothed parts of the transmission are determined. The advantage here is that the possible frequencies can be determined in a simple manner and the damage can therefore be clearly assigned if its frequency corresponds to one of the possible frequencies, in particular essentially, ie within the scope of the measurement accuracy and/or accuracy of the determination.

In an advantageous embodiment, the frequency value assigned to the kurtosis value is compared with the determined speed values of the shafts or toothed parts of the transmission and with the determined harmonics
and if there is a match, in particular a substantial match, the toothed part is determined as the defective part
and if there is disagreement, a bearing is determined as a defective part. The advantage here is that a simple method can be used.

Important features of the device for carrying out an aforementioned method are that a structure-borne noise sensor is used to detect the sensor signal, which is positioned touching a housing part of a device, in particular a gear or a device comprising a gear. The advantage here is that only one structure-borne sound sensor with an evaluation unit is required, with the evaluation unit having an analog/digital converter and having a computer for evaluating the digital data.

• In der 1 ist ein Ablauf zur Auswertung des erfassten Sensorsignals gezeigt.

The invention will now be explained in more detail using an illustration:

• In the 1 a sequence for evaluating the detected sensor signal is shown.

Der Mittelwert einer Verteilung von statistisch schwankenden Werten beträgt $$\overline{x}=\frac{1}{N}\cdot {\displaystyle \sum _{i=1}^N{x}_{\text{i}}}$$

und die Standardabweichung
${\sigma }^2=\frac{1}{N-1}\cdot {{\displaystyle \sum _{i=1}^N\left(x_i-\overline{x}\right)}}^2$

ist, wobei N die Anzahl der Werte ist.For understanding the background of the invention, reference is made to the fact that a normal distribution of statistically fluctuating values is determined by $p\left(x\right)=\frac{1}{\sqrt{2\cdot \pi \cdot {\sigma }^2}}\cdot e^{\frac{-{\left(x-\overline{x}\right)}^2}{2{\sigma }^2}}.$

The mean of a distribution of statistically varying values is $$\overline{x}=\frac{1}{N}\cdot {\displaystyle \sum _{i=1}^N{x}_{\text{i}}}$$

and the standard deviation
${\sigma }^2=\frac{1}{N-1}\cdot {{\displaystyle \sum _{i=1}^N\left(x_i-\overline{x}\right)}}^2$

where N is the number of values.

wobei N die Anzahl der Werte ist.The associated kurtosis is a higher moment for a distribution of N statically varying values and is determined according to $$g_3\left[\frac{1}{N}{\displaystyle \sum _{i=0}^{N-1}{\left(\frac{x_i-\overline{x}}{\sigma }\right)}^4}\right]-\mathrm{3,}$$

where N is the number of values.

In the case of the invention, a sensor is brought into contact with the housing of a transmission so that acoustic waves can be detected. A structure-borne noise sensor is suitable as a sensor, which can record frequencies up to 5 kHz or up to 50 kHz with a good signal-to-noise ratio. The sensor preferably works on the basis of piezo technology.

The sensor signal thus corresponds to the acceleration profile of the surface of the transmission housing at the point of contact between the sensor and the transmission.

The sensor signal is fed to an evaluation unit, which makes it possible to evaluate the time-dependent acceleration profile and enables local damage to be detected.

Damage to a tooth of a toothed part of the transmission or to a bearing of the transmission is possible as damage, for example.

The toothed part is non-rotatably connected to one of the shafts of the transmission, ie to the input shaft, the output shaft or an intermediate shaft.

The input shaft is driven at a speed by a motor. The method according to the invention can also be used when the geared motor is idling, ie without a load to be driven. When idling, the impulsiveness of the sensor signal caused by damage to a rotating part is low.

In a first evaluation variant, the evaluation unit determines the kurtosis of the time profile of the sensor signal. The sensor signal is shown as discrete in time and thus has a sequence of values xi detected at a particular point in time, with i counting through the N points in time, with N being an integer.

The kurtosis is determined as the mean value of the fourth power of the deviation from the mean value. If the kurtosis exceeds a critical value, a warning signal is generated, since exceeding it indicates the presence of damage.

In a second evaluation variant, a time section of the sensor signal is Fourier-transformed by the evaluation unit, after which the time section is shifted by a small amount in the time direction and is then Fourier-transformed again. This is repeatedly carried out, resulting in a Fourier spectrum as a function of time. An FFT (Fast Fourier Transformation) is used as the Fourier transformation.

Thus, for each frequency value there is a time profile of the amplitude, ie a set of amplitude values. In the next step, the kurtosis is calculated for each frequency value. When there is no damage, the kurtosis has an essentially vanishing value. If a critical value, which in turn can be specified, is exceeded, a warning signal is also generated, since exceeding it indicates the presence of damage.

In a third evaluation variant, the evaluation unit determines the kurtosis of the spectrum of the sensor signal or its envelope.

To form the kurtosis in the frequency spectrum, the sensor signal is Fourier transformed, in particular by means of FFT, and then the kurtosis of the amplitude values, which were determined for each frequency, is formed. If a critical value, which in turn can be specified, is exceeded, a warning signal is also generated, since exceeding it indicates the presence of damage.

To form the kurtosis of the envelope, the sensor signal is first bandpass filtered, rectified, lowpass filtered and then Fourier transformed, in particular by means of FFT, and then the kurtosis of the amplitude values, which were determined for each frequency, is formed. If a critical value, which in turn can be specified, is exceeded, a warning signal is also generated, since exceeding it indicates the presence of damage. By means of the envelope formation mentioned, in particular by means of rectification with subsequent low-pass filtering, the impulse surges introduced by damage, i.e. amounts of energy, are more strongly emphasized from the background noise, so that an improved signal-to-noise ratio can be achieved for recognizing the presence of impulses.

The evaluation unit thus carries out the three evaluation variants mentioned and generates a damage warning in each case when the respective critical value is exceeded.

In order to determine the defective part as clearly and automatically as possible, the speed of the input shaft is first determined. For this purpose, either the known, predetermined speed value of the input shaft caused by the motor, in particular by a converter feeding the motor, is transmitted to the evaluation unit or, alternatively, the speed of the input shaft is determined by the evaluation unit. To determine the speed using the evaluation unit, the meshing speed of the input gear stage is preferably determined in the frequency spectrum of the sensor signal, since the gear meshing frequency of the first gear stage is clearly recognizable in the spectrum. The rotational speed of the input shaft and the rotational speed of the output shaft as well as the intermediate shaft or intermediate shafts can then be determined from the known gearing data determine. The corresponding meshing frequency is assigned to each toothed part.

In addition to the speed values determined in this way, the harmonics are also determined, i.e. those harmonics that can occur in the spectral range. In a subsequent step, the envelope curve spectrum is checked to determine whether the damage frequency determined essentially corresponds to one of the rotational speed values of one of the shafts or one of the meshing frequency values or one of the determined harmonics. If a match is found, a warning message is displayed by the evaluation unit, identifying the defective part. If there is no match, it indicates that there is bearing damage in one of the bearings on one of the shafts.

In this way, an automatic detection of a damage is possible without an experience database.

In the 1 a sequence according to the invention of the sensor signal evaluation carried out in the evaluation unit is shown.

The detected sensor signal is initially subjected to two evaluations, with the first evaluation being illustrated in the left half and the other evaluation in the right half.

In the first evaluation, the spectral acceleration density, i.e. the (averaged) discrete FFT spectrum, of the sensor signal is first determined.

The meshing frequency of the first gear stage is estimated from the gearing data and the specified, at least approximate, speed of the input shaft, in particular the synchronous speed. The exact associated frequency value is then searched for and determined in the FFT spectrum. The actually existing or detected tooth meshing frequency value is thus then known.

From this frequency value, taking into account the known gearing data, the frequency values, in particular rotational speeds, of the geared parts or shafts of the transmission, in particular the input shaft, the output shaft and at least one intermediate shaft are determined.

In addition, the harmonics of the specified shaft speeds are determined.

The shaft speed values determined in this way and associated harmonic frequency values are also referred to below as damage frequencies.

alternatively, in 1 marked with OR, the speed of rotation of the input or output shaft can also be transmitted directly to the evaluation unit, so that the associated frequency values of the other shafts and the associated harmonics can be determined. These values then form the damage frequencies.

In the other evaluation, i.e. the one in the right half of the 1 shown evaluation, the sensor signal is examined for impulsiveness.

For this purpose, the sensor signal is first pre-filtered and/or subjected to noise suppression. A bandpass filter can preferably be used here. In addition or as an alternative, however, a noise suppression method such as prewhitening or an adaptive noise suppression method such as adaptive noise cancellation can also be used.

• - die Kurtosis des zeitabhängigen, insbesondere unbehandelten, Sensorsignals,
• - die Kurtosis der Hüllkurve und
• - das Kurtogramm.

Then the three kurtosis values are determined, i.e

• - the kurtosis of the time-dependent, in particular untreated, sensor signal,
• - the kurtosis of the envelope and
• - the kurogram.

To determine the kurtogram, a frequency spectrum belonging to a sliding time segment of the sensor signal is determined, in particular by means of short-time FFT, i.e. an FFT transformation, and then for each frequency value the kurtosis of the amplitude value belonging to this respective frequency value.

If one of the kurtosis values determined in the kurtogram or the kurtosis value of the time-dependent, in particular untreated, sensor signal or the kurtosis value of the envelope of the pre-filtered and/or noise-suppressed sensor signal exceeds an associated limit value, a warning is issued that there is local damage. In the other case, it indicates that there is no local damage.

If there is local damage, the part of the transmission that has the damage is determined. From the kurtosis-based monitoring described above, only the frequency value of the damage is known, but not the part showing the damage. The part is determined by first using the envelope of the sensor signal. As above , the envelope is generated by bandpass filtering, rectifying, lowpass filtering and Fourier transforming the sensor signal. Thus, the amplitudes of the shaft speeds, meshing frequencies and their harmonics clearly stand out from a background noise component. The amplitudes that clearly stand out from the background noise component are easy to determine and thus the associated frequency values can also be determined. By comparison with the frequency value of the damage, the corresponding shaft frequency, meshing frequency or at least one harmonic, ie overtone, of a meshing frequency can be determined and the damaged part can be inferred from this.

If the frequency value of the damage frequency is far away from one of the frequency values of one of the shafts, a tooth mesh or a harmonic thereof, another rotating part is indicated as the damaged part, i.e. bearing damage in particular.

Band-pass filtering and/or high-pass filtering is preferably carried out for noise suppression.

Reference List

1

left half

2

right half

Claims (5)

Procedure for determining a defective part in a gearbox, a sensor for detecting acoustic vibrations being brought into contact with the transmission, wherein the gearbox is operated in neutral, with the input shaft of the gearbox being operated at a speed, the sensor signal being fed to an evaluation unit, where (i) the kurtosis of the time course of the sensor signal is determined, where (ii) the kurtosis of an envelope of the sensor signal is determined, where (iii) a kurtogram, i.e. the kurtosis, is determined as a function of the frequency, with a time segment of the sensor signal beginning at a particular point in time being Fourier transformed, with the point in time being shifted after each Fourier transformation and the Fourier transformation then being carried out in turn is, so that for each frequency value there is a value curve dependent on time, i.e. on the respective points in time, whose kurtosis value is determined, wherein the respectively determined kurtosis value is compared with a predetermined value and the envelope of the sensor signal is determined and when the predetermined value is exceeded by the kurtosis value, the frequency values belonging to the amplitudes protruding from the background noise are determined, whereby these frequency values are compared with the determined speeds of the toothed parts of the gear and their harmonics, i.e. their harmonics in the recorded spectrum, where the defective part is determined if there is a match, and where a bearing is reported and/or displayed as a defective part in the event of non-compliance. procedure after claim 1 , characterized in that the respectively determined kurtosis value is compared with a predetermined value and the envelope of the sensor signal is determined and when the predetermined value is exceeded by the kurtosis value, the frequency values belonging to the amplitudes protruding from the background noise are determined, these frequency values being compared with the specific speeds of the toothed parts of the transmission and their harmonics, i.e. their harmonics located in the detected spectrum, the defective part being determined if they match and a bearing being reported and/or displayed as the defective part if they do not match. Method according to one of the preceding claims, characterized in that the speed of a shaft of the transmission is determined by means of a speed sensor connected to the shaft, by means of an angle sensor connected to the shaft or by means of a speed-controlled electric motor driving the transmission, the setpoint speed provided for the speed control or the actual speed determined during speed control is used as the speed of the shaft of the gearbox. Method according to one of the preceding claims, characterized in that the speed values of the other rotatably mounted toothed parts of the gearbox are determined from the gearing data of the gearbox and from the determined speed of the shaft of the gearbox and that associated harmonics, i.e. harmonics of the speed values of all rotating toothed parts of the gearbox to be determined. Device set up to carry out a method according to one of the preceding claims, characterized in that a structure-borne noise sensor is used to detect the sensor signal, which is positioned in contact with a housing part of a device.

Images