AT527989A1 - PROCEDURE FOR DETECTING AT LEAST ONE DAMAGE - Google Patents

Abstract

The invention relates to a method for detecting at least one defect in a mechanical system on a test bench, wherein vibrations of the system are measured and high-frequency vibration signals are recorded and fed into a fault analysis system. Simple and reliable early detection of damage can be achieved by performing the following steps: A. Processing the vibration signals, preferably at least by high-pass filtering; B. Transferring the processed vibration signals from a time domain to a frequency domain; C. Dividing the vibration signals into several frequency zones (Z0, Z1, Z2, Z3, Z4); D. Forming amplitude sums (Sum_Zone_X) by summing the amplitudes of the vibration signals of each frequency zone (Z0, Z1, Z2, Z3, Z4); E. Classifying the amplitude sums (Sum_Zone_X); F. Forming extended error indicators (AFI_x) by relating the classified amplitude sums (Sum_Zone_X) to an unfiltered base sum (Sum_Zone_0); G. Forming temporal histories of the error indicators (AFI_x); H. Detecting and assessing at least one anomaly based on the temporal histories of the error indicators (AFI_x).

Description

The invention relates to a method for detecting at least one damage of a mechanical system on a test bench, wherein vibrations of the system are measured and high-frequency vibration signals are recorded and

be subjected to error evaluation.

In order to detect damage occurring on test benches, suitable sensors are installed on test benches.

Examples of condition-monitoring sensors include accelerometers, torque and speed sensors, and temperature sensors. Typically, the test bench system to be monitored contains classic machine elements such as bearings, gears, and shafts, including their connecting elements, which are subject to natural aging.

and can be damaged prematurely through improper use.

The measurement data acquired by the sensors and recorded by a suitable system are centrally filtered, processed, stored, and compared with predefined limit values, for example, according to ISO standard 10816. If such a limit is exceeded, a warning or stop signal is transmitted to the test bench's automation system, which reacts by shutting down or entering a safe operating state. The statically parameterized limit values are defined in the system, for example, based on ISO standard 2954. In the event of a fault, stored data is a key indicator for determining the cause of damage or other unpredictable events.

By collecting the described data, it may seem logical to apply physical models for the remaining component lifetime, such as linear damage accumulation, to predict the lifetime. However, these methods fail in general, sometimes due to high application-related disturbances, as well as due to the fact that the reaction forces required for these methods

and moments cannot be measured directly.

Thus, the currently used calculation methods for condition assessment (e.g., ISO 2954) severely limit the informative value of changes in system variables, in the case of ISO 2954, vibrations or oscillations. The ISO standard requires filtering of signals at 10–1000 Hz, which fundamentally precludes further calculation of high-frequency early damage indicators. As a result, damage to components can only be detected very late or not at all. This results in downtimes of several weeks and/or consequential damage to other components. The need for early damage detection can therefore be directly derived from this.

become.

The object of the invention is therefore to provide a simple and reliable early detection of damage.

The object is achieved according to the invention with a method of the type mentioned at the outset by carrying out the following steps:

A. Processing the vibration signals, preferably at least by high-pass filtering;

B. Transferring the processed vibration signals from a time domain

into a frequency range;

C. Division of the vibration signals into several frequency zones;

D. Forming amplitude sums by summing the amplitudes of the

Vibration signals of each frequency zone;

E. Classifying the vibration signals;

F. Forming extended error indicators by referring to the

classified amplitude sums to an unfiltered basic sum;

G. Creating time series of error indicators;

H. Detecting and assessing at least one anomaly based on the fault indicator trends.

Step B provides the data to be classified. In step C, the classes

The actual classification takes place in or after step D.

The recorded signals are processed in a suitable form and processed using an algorithm to detect any damage due to atypical deviations in the time data. The algorithm is fundamentally based on frequency transformation, amplitude summation, and classification of the recorded, high-frequency time signal of a measured variable (e.g., acceleration). These classified amplitude sums are related to an unfiltered base sum, whereby unknown disturbances are detected or filtered by definition, eliminating unknown, irrelevant disturbances - particularly through autocorrelation. The relative amplitude sums (AFI, Advanced Failure Indicator) calculated in this way serve as the basis for assessing an occurring

Anomaly.

Within the scope of the invention, the assessment of an occurring anomaly can be carried out with or without the use of operating data, such as torque

and/or speed:

1. Existing operating data is included in the classification: By identifying load zones and a steady-state operating condition, the calculated amplitude sums can be assigned to an operating state and allow for classic limit value monitoring within these load zones. In one embodiment of the invention, it is thus provided that an anomaly is detected if at least one error indicator curve exceeds a defined

Exceeds or falls below the limit value for the error indicator.

2. Operating data is missing or is not included in the classification: In this case, a dynamic limit for alarming can be determined by calculating a moving average of historical data (“Simple Moving Average”) and adding a standard deviation shifted by a defined factor. As an alternative or in addition to the classic limit value monitoring, an embodiment according to the invention therefore provides for the formation of a moving average of at least one error indicator curve, with an anomaly being detected if the moving average of the

Error indicator curve a defined target range for the mean value of the

error indicator curve exceeds or falls below.

If an anomaly is detected, an alarm can be triggered within the scope of the invention. The alarm is preferably triggered after a defined alarm delay to avoid false alarms. During the alarm delay, several calculation runs are performed. The alarm is only triggered when these consecutive calculations confirm the error.

The described method enables early detection of anomalies that cause spectrum changes, especially high-frequency damage phenomena, on the test bench. This detection was not possible with a conventional

Vibration monitoring not yet possible.

The invention is explained in more detail below with reference to the figures.

schematic

Fig. 1 shows the method according to the invention in a block diagram,

Fig. 2 shows a flow diagram of the method according to the invention in a first

Version,

Fig. 3 shows a flow diagram of the method according to the invention in a second

Version and

Fig. 4 shows an exemplary representation of the zones over a frequency spectrum.

The recorded signals are processed in a suitable form and processed using a sequence of mathematical transformation steps to detect damage due to atypical deviations in the time data. The method according to the invention is fundamentally based on frequency transformation, amplitude summation, and classification of the recorded, high-frequency time signal of a measured variable (e.g., acceleration). These classified amplitude sums Sum_Zone_X are related to an unfiltered base sum Sum_Zone_0, whereby unknown

Disturbances are detected and filtered by definition. The thus calculated

relative amplitude sums AFI_x (AFI=Advanced Failure Indicator) serve as

Basis for assessing an occurring anomaly.

Fig. 1 shows a block diagram with the essential steps of the method according to the invention. The method accordingly comprises the following main steps:

A. Processing of the vibration signals at least by high-pass filtering;

B. Transferring the processed vibration signals from one time domain to

a frequency range;

C. Division of the vibration signals into several frequency zones;

D. Forming amplitude sums by summing the amplitudes of the

Vibration signals of each frequency zone;

E. Classifying the vibration signals;

F. Forming error indicators by referring to the classified

Amplitude sums to an unfiltered basic sum;

G. Creating time series of error indicators;

H. Detecting and assessing at least one anomaly based on the

Error indicator trends.

Step B provides the data to be classified. In step C, the

Classes are defined. Classification occurs in or after step D.

The prepared measurement signals are divided into defined classes. Step C defines these ranges. For example, as explained below with reference to Figs. 2 and 3, zone Z2 in step C can represent a frequency range from 1 kHz to 3 kHz, zone Z3 for 3 kHz to 5 kHz, and zone Z0 for the

unfiltered class over the entire range.

Within the scope of the invention, the assessment of an occurring anomaly can be carried out with or without the use of operating data, such as torque and/or speed, as described below with reference to Figs. 2 and 3

is explained.

Fig. 2 shows a flowchart of the method according to the invention in a first embodiment using the example of fault detection in a bearing. Existing operating data is incorporated into the classification: By identifying load zones and a steady-state operating condition, the calculated classified amplitude sums Sum_Zone_X can be assigned to an operating state and allow for conventional limit value monitoring within these load zones. Thus, in the first embodiment of the invention, an anomaly is detected if at least one curve of an extended fault indicator AFI_x exceeds or falls below a defined limit value for the fault indicator.

The following positions are shown in Fig. 2:

1: Sampling the raw signal with a sampling rate fs;

2: High-pass filtering of the raw signal: Filter frequency: low frequency (e.g. 10 Hz) Order: 6 Method: Butterworth filter

3: Antialiasing of the signal by a suitable anti-aliasing filtering

4: Suitable image domain transformation, preferably a Fast Fourier Transformation (FFT)

5: Amplitude sum calculation for 5 frequency ranges ZO, Z1, Z2, Z3, Z4:

The amplitude sum Sum_Zone_X of zone X is:

Sum_Zone_X = J/X zone lines (line amplitude2)/2)

(The summation and averaging (RMS calculation) are explained in more detail below using Fig. 4.)

22

Z1: Zone 1 -x < 1000 Hz

Z2: Zone 2 - 1000 < x < 3000 Hz

Z3: Zone 3 - 3000 < x < 7500 Hz

Z4: Zone 4 - 7500 < x

SO: Amplitude sum zone 0 => Sum_Zone_0

S1: Amplitude sum zone 1 => Sum_Zone_1

S2: Amplitude sum zone 2 => Sum_Zone_2

S3: Amplitude sum zone 3 => Sum_Zone_3

S4: Amplitude sum zone 4 => Sum_Zone_4

6: Calculating an advanced failure indicator AFI_x by forming relative amplitude sums (AFI=Advanced Failure Indicator), where classified amplitude sums Sum_Zone_X are related to an unfiltered basic sum Sum_Zone_0:

AFI_x = Sum_Zone_X /Sum_Zone_0

AFI_1: extended fault indicator for Zone 1

AFI_2: extended fault indicator for Zone 2

AFI_3: extended fault indicator for Zone 3

AFI_4: extended fault indicator for Zone 4

7: Output of the extended error indicators AFI_x over time;

10: Check whether the extended error indicators AFI_x are under a

limit value;

11: Terminate the routine if at least one extended error indicator

AFI_x is below the limit;

12: Implementing an alarm delay: To avoid a possible false alarm, several consecutive calculation runs must confirm the error. Otherwise, incorrect measurements or other influences would result in a false positive evaluation. The alarm delay

thus prevents false triggering;

13: Issue a warning if at least one extended error indicator

AFI_x is above the limit;

14: Starting a new calculation cycle.

Fig. 3 shows a further flow chart of the method according to the invention in a second embodiment using the example of fault detection in a bearing. The method differs from the first embodiment in that historical data is taken on a moving average basis in positions 8 and 9. This method is particularly recommended when operating data is missing or not included in the classification. In this case, a dynamic limit for alarming can be found by moving averaging historical data (“simple moving average”) and adding a standard deviation shifted by a defined factor. Alternatively or in addition to classic limit value monitoring, this embodiment provides for a moving average AFI_x_middle of at least one curve of the extended fault indicator AFI_x to be formed, with an anomaly being detected if the moving average AFI_x_middle of the curve of the extended fault indicator AFI_x exceeds or falls below a defined target range for the mean AFI_x_middle of the curve of the extended fault indicator AFI_x.

The following positions are shown in Fig. 3:

1: Sampling the raw signal with a sampling rate fs

2: High-pass filtering of the raw signal: Filter frequency: low frequency (e.g. 10 Hz) Order: 6 Method: Butterworth filter

3: Antialiasing of the signal by a suitable anti-aliasing filtering

4: Suitable image domain transformation, preferably a Fast Fourier Transformation (FFT)

5: Calculation of amplitude sums Sum_Zone_X for 5 frequency ranges ZO, Z1, Z2, Z3, Z4: The amplitude sum Sum_Zone_X of zone X is:

Sum_Zone_X = J/X zone lines (line amplitude2)/2)

(The summation and the averaging (RMS calculation) will be further

explained in more detail below using Fig. 4.) ZO: Zone 0 - unfiltered Z1: Zone 1 - x < 1000 Hz Z2: Zone 2 - 1000 Sum_Zone_0 S1: Amplitude sum Zone 1 => Sum_Zone_1 S2: Amplitude sum Zone 2 => Sum_Zone_2 S3: Amplitude sum Zone 3 => Sum_Zone_3 S4: Amplitude sum Zone 4 => Sum_Zone_4

S4: Amplitude sum zone 4 => Sum_Zone_4

6: Calculating an advanced failure indicator AFI_x by forming relative amplitude sums (AFI=Advanced Failure Indicator), where classified amplitude sums Sum_Zone_X are related to an unfiltered basic sum Sum_Zone_0:

AFI_x = Sum_Zone_X /Sum_Zone_0

AFI_1: extended fault indicator for Zone 1

AFI_2: extended fault indicator for Zone 2

AFI_3: extended fault indicator for Zone 3

AFI_4: extended fault indicator for Zone 4

7: Output of the extended error indicators AFI_x over time;

8: Calculation of a moving average AFI_x_middle from the number n

of extended error indicators AFI_x, where

AFLx(x) AFI_x_middle = 2m 1=0

9: Calculating an upper and lower limit for a permissible band range:

AFI_x_lower/upper = AFI_x_middle *k*5$,

with the adjustment factor k and the standard deviation $:

$ = EC —%)

i=1

10: Check whether at least one extended error indicator AFI_x is outside the

permissible band range.

11: End the routine if all extended error indicators AFI_x in the

permissible band range;

12: Implementing an alarm delay: To avoid a possible false alarm, several consecutive calculation runs must confirm the error. Otherwise, incorrect measurements or other influences would result in a false positive evaluation.

The alarm delay thus prevents false triggering;

13: Issue a warning if at least one extended error indicator AFI_x is outside the permissible range.

The described method according to the invention makes it possible to detect anomalies, in particular high-frequency damage phenomena, which cause a spectrum change

cause, to be detected early on the test bench.

Fig. 4 shows an exemplary representation of the zones ZO, Z1, Z2, Z3 from step C, plotted against a frequency spectrum f. For each of these zones ZO, Z1, Z2, Z3, the vibration acceleration as in the frequency spectrum of this zone is summed as the amplitude of each line. For example, the vibration acceleration as formed by the amplitude is 0.5 m/s at a frequency of 10 Hz, 0.7 m/s at a frequency of 11 Hz, and 0.3 m/s at a frequency of 12 Hz; this results in a root mean square value for these three frequencies.

Amplitude sum Sum_Zone_X of 0.91 m/s2.,

PATENT CLAIMS

Method for detecting at least one defect of a mechanical system on a test bench, wherein vibrations of the system are measured and high-frequency vibration signals are recorded and fed to a fault evaluation, characterized in that

the following steps must be carried out:

A. Processing the vibration signals, preferably at least by

high-pass filtering;

B. Transferring the processed vibration signals from a time domain

into a frequency range;

C. Division of the vibration signals into several frequency zones (ZO, Z1, Z2, Z3, Z4);

D. Forming amplitude sums (Sum_Zone_X) by summing the amplitudes of the vibration signals of each frequency zone (ZOO, Z1, Z2, Z3, Z4);

E. Classify the amplitude sums (Sum_Zone_X);

F. Formation of extended error indicators (AFI_x) by relating the classified amplitude sums (Sum_Zone_X) to an unfiltered

Basic sum (Sum_Zone_0);

G. Formation of time courses of the error indicators (AFI_x);

H. Detecting and assessing at least one anomaly based on the

temporal courses of the error indicators (AFI_x).

Method according to claim 1, characterized in that an anomaly is detected if at least one course of an error indicator (AFI_x) exceeds or falls below a defined limit value for the error indicator (AFI_x).

falls below.

3. Method according to claim 1 or 2, characterized in that a moving average (AFI_x_middle) of at least one course of an extended error indicator (AFI_x) is formed, wherein an anomaly is detected if the moving average (AFI_x_middle) of the course of the extended error indicator (AFI_x) exceeds or falls below a defined target range for the moving average (AFI_x_middle) of the course of the error indicator (AFI_x).

4. Method according to one of claims 1 to 3, characterized in that an alarm is triggered if at least one anomaly is detected

becomes,

5. Method according to one of the claims 4, characterized in that the

Alarm is only triggered after a defined alarm delay.

31.01.2024 FÜ

Claims (1)

(new) patent claims 1. A method for detecting at least one defect in a mechanical system on a test bench, wherein vibrations of the system are measured and high-frequency vibration signals are recorded and fed to a fault evaluation, wherein the vibration signals are preferably processed at least by high-pass filtering and transferred from a time domain to a frequency domain, thereby marked that the following steps are carried out: e Division of the vibration signals into several frequency zones (ZOO, Z1, Z2, Z3, Z4); e Forming classified amplitude sums (Sum_Zone_X) by summing the amplitudes of the vibration signals of each frequency zone (ZO, Z1, Z2, Z3, Z4); e Formation of error indicators (AFI_x) by relating the classified amplitude sums (Sum_Zone_X) to an unfiltered basic sum (Sum_Zone_0); e Formation of time courses of the error indicators (AFI_x); e Detecting and assessing at least one anomaly on the basis of the temporal progression of the error indicators (AFI_x), wherein a moving average (AFI_x_middle) of at least one progression of an error indicator (AFI_x) is formed, and wherein an anomaly is detected if the moving average (AFI_x_middle) of the progression of the error indicator (AFI_x) exceeds a defined target range for the moving average (AFI_x_middle) of the progression of the error indicator (AFI_x). or falls below. 2. Method according to claim 1, characterized in that an anomaly is detected if at least one course of an error indicator (AFI_x) 21/22 MOST RECENT CLAIMS exceeds or exceeds a defined limit value for the error indicator (AFI_x) falls below. 3. Method according to claim 1 or 2, characterized in that a Alarm is triggered if at least one anomaly is detected. 4. Method according to claim 3, characterized in that the alarm is only is triggered after a defined alarm delay. 25.03.2025 FÜ 22/22 MOST RECENT CLAIMS