Speed reducer early-stage composite fault detection method

Abstract

The invention relates to a speed reducer early-stage composite fault detection method, which comprises the following steps: collecting a vibration signal of a speed reducer and carrying out equiphase synchronous average processing to obtain a first processing signal; extracting a health index of the speed reducer shaft based on the first processing signal, and judging whether a speed reducer shaft fault exists or not according to a comparison result of the index and a corresponding threshold value; performing feature enhancement processing on the first processing signal to obtain a second processing signal; extracting a health index of the speed reducer gear based on the second processing signal, and judging whether a speed reducer gear fault exists or not according to a comparison result of the index and a corresponding threshold value; performing deperiodic coherent processing on the second processing signal to obtain a third processing signal; and extracting a health index of the reducer bearing based on the third processing signal, and judging whether a reducer bearing fault exists or not according to a comparison result of the index and a corresponding threshold value. According to the invention, real-time monitoring of the early health state of the speed reducer can be rapidly and accurately realized.

Description

Technical Field

[0001] The present invention relates to the technical field of speed reducer fault diagnosis and health management, and in particular to a speed reducer early composite fault detection method. Background Art

[0002] In helicopter transmission systems, the reducer is a typical component. It is the power output end of the drive shaft for accessories such as the tail rotor and rotor, the power input end of the engine, and the central load-bearing component of the helicopter. It must have high reliability, high safety, and high maintainability. The reducer mainly consists of three types of parts: shafts, bearings, and gears. Currently, most vibration signal analysis methods only detect faults in one of them, and cannot detect multiple complex faults that are more dangerous during actual operation. In particular, the early fault signals of bearings are particularly weak. In the process of being transmitted to the sensor, they need to be modulated by the transmission path and also contain strong interference from the rotor and gears. Existing fault detection methods cannot meet actual usage requirements. Therefore, there is a need for a method that can detect various types of faults in various objects of the reducer at an early stage to effectively improve the fault detection capabilities of the reducer and enhance the safety and reliability of the helicopter. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method for detecting early composite faults of a reducer, which can quickly and accurately realize real-time detection of the health status of the reducer, especially early composite faults of the reducer.

[0004] The technical solution adopted by the present invention to solve the technical problem is to provide a method for detecting early composite faults of a reducer, comprising the following steps:

[0005] Collecting the vibration signal of the reducer and performing equal-phase synchronous averaging processing to obtain a first processed signal;

[0006] extracting a health index of the reducer shaft based on the first processed signal, and determining whether there is a reducer shaft fault based on a comparison result of the index with a corresponding threshold;

[0007] performing feature enhancement processing on the first processed signal to obtain a second processed signal;

[0008] extracting a health index of the reducer gear based on the second processed signal, and determining whether there is a reducer gear fault based on a comparison result of the index with a corresponding threshold;

[0009] performing de-periodic coherence processing on the second processed signal to obtain a third processed signal;

[0010] A health index of the reducer bearing is extracted based on the third processed signal, and whether there is a reducer bearing fault is determined based on a comparison result between the index and its corresponding threshold.

[0011] Furthermore, performing equal phase synchronous averaging processing on the vibration signal includes:

[0012] Calculate the real-time phase of the reducer shaft according to the speed, and then deduce the sampling time points with constant phase intervals;

[0013] resampling the vibration signal based on the calculated sampling time point to obtain a fourth processed signal;

[0014] The fourth processed signal is subjected to synchronous averaging processing, and the first processed signal having the same length as the vibration signal is reconstructed using the signal subjected to the synchronous averaging processing.

[0015] Furthermore, when performing synchronous averaging processing on the fourth processed signal, the length of the sample per revolution is used as the segment period.

[0016] Furthermore, performing de-periodic coherence processing on the second processed signal to obtain a third processed signal includes:

[0017] initializing the coherent processed signal as a second processed signal;

[0018] Calculate the real-time phase of the reducer shaft according to the speed, and then deduce the sampling time points with constant phase intervals for this round;

[0019] resampling the current coherently processed signal based on the sampling time point of this round to obtain a fifth processed signal;

[0020] performing synchronous averaging processing on the fifth processed signal, and reconstructing a sixth processed signal having the same length as the coherent processed signal using the signal after the synchronous averaging processing;

[0021] Calculate the signal difference between the fifth processed signal and the sixth processed signal. If the signal difference remains unchanged within the set round, use the sixth processed signal of this round as the third processed signal. Otherwise, update the coherent processed signal to the signal difference of this round, and return to the step of calculating the real-time phase of the reducer shaft based on the speed.

[0022] Furthermore, when performing synchronous averaging processing on the fifth processed signal, the processing is performed in sequence with the reciprocal of each interference frequency as a segment period.

[0023] Furthermore, the interference frequency includes the rotation frequency and its multiples, the meshing frequency and its sidebands, and the multiples of the meshing frequency and its sidebands.

[0024] Furthermore, the performing feature enhancement processing on the first processed signal includes:

[0025] obtaining a frequency spectrum of the first processed signal;

[0026] The gear meshing frequency and its sidebands are filtered out from the obtained spectrum and then converted into the time domain.

[0027] Furthermore, extracting the health indicator of the reducer shaft based on the first processed signal includes:

[0028] Acquire the spectrum of the first processed signal and perform normalization processing on the spectrum amplitude;

[0029] Calculate the health index C of the reducer shaft r for

[0030]

[0031] Among them, f is the rotation frequency, A S (f) is the rotation frequency amplitude after normalization, and a and b are the index adjustment coefficients.

[0032] Furthermore, extracting the health index of the reducer gear based on the second processed signal includes:

[0033] Calculate the kurtosis Ku of the first processed signal respectively S and effective value R S , and the kurtosis Ku of the second processed signal G and effective value R G ;

[0034] Calculate the health index C of the reducer gear G for

[0035]

[0036] Among them, c and d are indicator adjustment coefficients.

[0037] Furthermore, extracting the health indicator of the reducer bearing based on the third processed signal includes:

[0038] Calculating the fault characteristic frequencies of the reducer bearing, wherein the fault characteristic frequencies include the outer ring fault characteristic frequency, the inner ring fault characteristic frequency, the rolling element fault characteristic frequency, and the cage fault characteristic frequency;

[0039] Perform square envelope spectrum analysis on the third processed signal to calculate the health index C of the reducer gear B for

[0040]

[0041] Among them, f B is the fault characteristic frequency appearing in the square envelope spectrum, A(f B ) is the amplitude of the fault characteristic frequency.

[0042] Beneficial effects

[0043] Due to the adoption of the above-mentioned technical solution, the present invention has the following advantages and positive effects compared with the prior art: the present invention targets the characteristics of fault signals of various monitoring objects, uses vibration signal analysis as a means, and constructs the health factor of each monitoring object through vibration signal analysis, thereby realizing the separation and detection of early complex faults of the reducer. In addition, the health indicators have clear physical meanings and fast calculation speeds, which can effectively realize real-time monitoring of the health status of the reducer and provide a basis for maintenance and usage decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of an embodiment of the present invention;

[0045] Figure 2 1 is a diagram showing the results of equal-phase synchronous averaging processing according to an embodiment of the present invention;

[0046] Figure 3 is a graph showing the relationship between the signal-to-noise ratio and the average number of segments according to an embodiment of the present invention;

[0047] Figure 4 is a diagram of shaft fault detection results according to an embodiment of the present invention;

[0048] Figure 5 This is a feature enhancement processing result diagram according to an embodiment of the present invention;

[0049] Figure 6 is a diagram of gear fault detection results according to an embodiment of the present invention;

[0050] Figure 7 1 is a diagram showing the result of de-periodic coherence processing according to an embodiment of the present invention;

[0051] Figure 8 4 is a diagram showing the bearing fault detection results according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] Below in conjunction with specific embodiment, further set forth the present invention.Should be understood that these embodiments are only used to illustrate the present invention and are not used in limiting the scope of the present invention.In addition, should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall equally within the scope limited by the appended claims of the application.

[0053] An embodiment of the present invention relates to a method for detecting early composite faults of a reducer based on vibration analysis.

[0054] During the operation of the reducer, whether or not a fault occurs, noise interference (including noise from the operating environment and periodic noise during the operation of the reducer) always exists. When the reducer fault is serious, the energy of the shaft-gear-bearing fault signal becomes larger, and it can also show obvious performance under noise interference. Therefore, the composite signal can be processed directly, and the shaft-gear-bearing fault monitoring can be performed at the same time. When the reducer is in the early stage of failure, the fault signal is weak and the collected vibration signal-to-noise ratio is low. It is impossible to extract the fault features by directly processing the composite signal. Therefore, it is necessary to extract features in sequence according to the features of each component of the reducer. The specific method is as follows. Figure 1 As shown, the following steps are included:

[0055] S1 performs equal-phase synchronous averaging on the collected vibration signal Y to obtain a noise reduction signal S (i.e., the first processed signal), thereby increasing the frequency and its multiples in the signal and reducing noise interference;

[0056] S2 extracts the health indicators of the axis and sets a threshold. If the threshold is exceeded, an axis fault alarm is issued;

[0057] S3 performs feature enhancement processing on the noise reduction signal S to obtain the gear signal G (i.e., the second processed signal), removes the influence of the rotation frequency and meshing frequency, and highlights the gear fault component;

[0058] S4 Extract Gear Health Indicator C r , and set a threshold value. If the threshold value is exceeded, a gear fault alarm will be issued;

[0059] S5 performs de-periodic coherence processing on the gear signal G to obtain signal B (i.e., the third processed signal), removes the influence of the periodic coherent signal, and highlights the bearing fault component;

[0060] S6 performs envelope demodulation analysis on signal B, extracts the bearing's health indicators, and sets a threshold. If the threshold is exceeded, a bearing fault alarm is issued.

[0061] The phase synchronization averaging process in step S1 can be performed using the following method:

[0062] S101 calculates the phase of the axis according to the speed;

[0063] S102 calculates a new sampling time for subsequent samples with a constant phase;

[0064] S103 resamples the signal Y according to the new sampling time to obtain an equal-phase signal P;

[0065] S104 performs synchronous averaging processing on the phase signals with the sample length per revolution as a period to obtain the signal SA, thereby removing random noise interference;

[0066] S105 connects the signal SA according to the length of the signal Y to obtain the signal S.

[0067] The processing results are as follows Figure 2 As shown, vibration signal Y = original fault signal A + noise N. This signal consists of several revolutions of measured signals. The first-revolution measured signal is a portion of vibration signal Y (the first revolution). Signal SA, obtained after equal-phase synchronous averaging of the signal, is the first-revolution synchronous average signal. By segmenting vibration signal Y by time and performing equal-phase synchronous averaging, the signal more closely matches the original fault signal A, effectively removing the influence of noise.

[0068] from Figure 3 It's easy to see that as the number of segments increases, the impact of noise gradually decreases until it stabilizes. When the number of segments is an integer multiple of the number of gear teeth, the interference from noisy gears is eliminated, and the signal-to-noise ratio is significantly improved. Beyond a certain number of segments, the signal-to-noise ratio begins to fluctuate, while the signal-to-noise ratio for segments that are integer multiples of the number of gear teeth decreases. Therefore, the number of segments can be determined based on the signal-to-noise ratio for integer multiples of the number of gear teeth to improve efficiency.

[0069] The shaft fault signal is most obvious in the reducer parts (shaft-gear-bearing). This is because during the actual operation of the reducer, the shaft must have a certain degree of imbalance, misalignment and bending. Therefore, the shaft rotation frequency and its multiples are always present in the collected vibration signal. Therefore, in the process of shaft health index extraction and fault detection, it is only necessary to highlight the shaft rotation frequency and its multiples through equal phase synchronous averaging processing. By comparing the amplitude of the rotation frequency and its multiples with the threshold set based on the normal signal, the shaft fault condition can be clearly identified. The steps for extracting the shaft health index in step S2 are:

[0070] S201 performs Fourier transform on the noise reduction signal S to obtain a spectrum, and performs normalization on the spectrum amplitude A:

[0071]

[0072] S202 records the rotation frequency as f, the rotation frequency amplitude as A(f), and the shaft health index C r for

[0073]

[0074] Among them, a and b are index adjustment coefficients, a∈[0,1], b∈[0,1], and a=0.5 can be generally taken.

[0075] The health indicators are extracted for 100 normal samples and 100 fault samples. The results are as follows: Figure 4 shown.

[0076] Gear failure is relatively obvious in the parts of the reducer (shaft-gear-bearing). Compared with shaft failure, gear failure signals are easily interfered by the meshing frequency and its side frequencies (these frequencies also appear in normal signals and have a greater impact on the entire signal). Therefore, in addition to highlighting the rotational frequency and its multiples of the shaft through equal-phase synchronous averaging, it is also necessary to remove the interference of the meshing frequency and its side frequencies (these are also multiples of the rotational frequency). Then, the gear failure condition is determined based on the effective value and kurtosis index before and after removing the interference of the meshing frequency and its side frequencies. The feature enhancement processing in step S3 includes the following steps, and the processing results are as follows: Figure 5 As shown:

[0077] S301 performs Fourier transform on the noise reduction signal S to obtain a spectrum;

[0078] S302 spectrum filter gear meshing frequency f G and its sidebands (f G +nf,f G -nf) and then perform inverse Fourier transform to obtain signal G. G =m*f, m is the number of gear teeth, n is the number of sidebands, n is a positive integer, generally n=2.

[0079] The gear health index extraction step in step S4 is as follows: Figure 6 As shown:

[0080] S401 calculates the kurtosis Ku of the signal S S and effective value R S ;

[0081] S402 Calculate the kurtosis Ku of signal G G and effective value R G ;

[0082] S403 calculates the health index C of the gear G for,

[0083]

[0084] Among them, c and d are empirical parameters, generally c = 0.5, d = 0.5.

[0085] The de-periodic coherent signal in step S5 includes the following steps, and the processing result is as follows: Figure 7 As shown:

[0086] S500 initializes the signal to be processed Q to signal G;

[0087] S501 calculates the phase of the axis according to the speed;

[0088] S502 calculates a new sampling time for subsequent samples with a constant phase;

[0089] S503 resamples the signal Q according to the new sampling time to obtain an equal-phase signal P;

[0090] S504 performs synchronous averaging processing on the equal-phase signals in sequence with each interference frequency as a period to obtain signal S1, thereby removing random noise interference;

[0091] S505 connects signal S1 according to the length of signal Q to obtain signal S2;

[0092] S506 uses the equal-phase signal P to subtract the signal S2 to obtain the signal B1, and Q=B1 is recorded. Steps S501 to S505 are repeated until the signal B1 remains unchanged and is output as the bearing signal B2.

[0093] Interference frequencies include the rotation frequency and its multiples, the meshing frequency and its sidebands, and the multiples of the meshing frequency and its sidebands. Because there are multiple interference frequencies, they need to be removed one by one. The result B1 after each removal is used as the input Q for the de-periodic coherence process. The input for the final de-periodic coherence process is the penultimate result Q = B1. The final de-periodic coherence process result is B1. When B1 remains unchanged, the interference is completely removed. At this time, the extracted bearing fault signal B2 is obtained.

[0094] The bearing fault signal is the weakest among the reducer parts (shaft-gear-bearing). If a composite fault occurs, the bearing fault signal can only be extracted after removing the shaft and gear fault signals and various interferences. Therefore, the bearing fault detection is performed last. The bearing health index extraction step in step S6 is as follows: Figure 8 As shown:

[0095] S601 calculates the outer ring fault characteristic frequency f according to the bearing fault characteristic frequency calculation formula o , inner race fault characteristic frequency f i , rolling element fault characteristic frequency f b , cage fault characteristic frequency f c ;

[0096] S602 calculates the square envelope spectrum of the bearing signal B2, assuming that the fault characteristic frequency is f B , the amplitude is A(f B ), bearing health index C B for

[0097]

[0098] The health index is calculated for the fault characteristic frequencies of the inner ring, outer ring, rolling element and cage respectively.

[0099] The threshold setting method in step S2, step S4 and step S6 is:

[0100] Calculate the health index C based on the healthy samples;

[0101] By calculating the multi-level threshold of the health indicator C

[0102] T v =M+n*σ

[0103] T v —Health indicator threshold; M—Health indicator C mean; σ—Health indicator C variance; n—Variance coefficient. Different thresholds can be obtained by setting the variance coefficient. Generally, the degradation alarm threshold n can be 3, and the severe fault alarm threshold n can be 6.

Claims

1. A method for detecting early composite faults of a reducer, characterized in that: The following steps are involved: Collecting the vibration signal of the reducer and performing equal-phase synchronous averaging processing to obtain a first processed signal; extracting a health index of the reducer shaft based on the first processed signal, and determining whether there is a reducer shaft fault based on a comparison result of the index with a corresponding threshold; performing feature enhancement processing on the first processed signal to obtain a second processed signal; extracting a health index of the reducer gear based on the second processed signal, and determining whether there is a reducer gear fault based on a comparison result of the index with a corresponding threshold; performing de-periodic coherence processing on the second processed signal to obtain a third processed signal; A health index of the reducer bearing is extracted based on the third processed signal, and whether there is a reducer bearing fault is determined based on a comparison result between the index and its corresponding threshold.

2. The method according to claim 1, characterized in that The equal phase synchronous averaging processing of the vibration signal includes: calculating the real-time phase of the reducer shaft according to the rotation speed, and then calculating the sampling time points with a constant phase interval; resampling the vibration signal based on the calculated sampling time point to obtain a fourth processed signal; The fourth processed signal is subjected to synchronous averaging processing, and the first processed signal having the same length as the vibration signal is reconstructed using the signal subjected to the synchronous averaging processing.

3. The method according to claim 2, characterized in that When performing synchronous averaging processing on the fourth processed signal, the length of the sample per revolution is used as the segment period.

4. The method according to claim 2, characterized in that The performing de-periodic coherence processing on the second processed signal to obtain a third processed signal includes: initializing the coherent processed signal as a second processed signal; Calculate the real-time phase of the reducer shaft according to the speed, and then deduce the sampling time points with constant phase intervals for this round; resampling the current coherently processed signal based on the sampling time point of the current round to obtain a fifth processed signal; performing synchronous averaging processing on the fifth processed signal, and reconstructing a sixth processed signal of the same length as the coherently processed signal using the signal after the synchronous averaging processing; Calculate the signal difference between the fifth processed signal and the sixth processed signal. If the signal difference remains unchanged within the set round, use the sixth processed signal of this round as the third processed signal. Otherwise, update the coherent processed signal to the signal difference of this round, and return to the step of calculating the real-time phase of the reducer shaft based on the speed.

5. The method according to claim 4, characterized in that When performing synchronous averaging processing on the fifth processed signal, the processing is performed in sequence with the reciprocal of each interference frequency as a segment period.

6. The method according to claim 5, characterized in that The interference frequency includes the rotation frequency and its multiples, the meshing frequency and its sidebands, and the multiples of the meshing frequency and its sidebands.

7. The method according to claim 1, characterized in that The performing feature enhancement processing on the first processed signal includes: obtaining a frequency spectrum of the first processed signal; The gear meshing frequency and its sidebands are filtered out from the obtained spectrum and then converted into the time domain.

8. The method according to claim 1, characterized in that The extracting the health indicator of the reducer shaft based on the first processed signal includes: Acquire the spectrum of the first processed signal and perform normalization processing on the spectrum amplitude; Calculate the health index C of the reducer shaft r for Among them, f is the rotation frequency, A S (f) is the rotation frequency amplitude after normalization, and a and b are the index adjustment coefficients.

9. The method according to claim 1, characterized in that The extracting the health indicator of the reducer gear based on the second processed signal includes: Calculate the kurtosis Ku of the first processed signal respectively S and effective value R S , and the kurtosis Ku of the second processed signal G and effective value R G ; Calculate the health index C of the reducer gear G for Among them, c and d are indicator adjustment coefficients.

10. The method according to claim 1, characterized in that Extracting the health indicator of the reducer bearing based on the third processed signal includes: Calculating the fault characteristic frequencies of the reducer bearing, wherein the fault characteristic frequencies include the outer ring fault characteristic frequency, the inner ring fault characteristic frequency, the rolling element fault characteristic frequency, and the cage fault characteristic frequency; Perform square envelope spectrum analysis on the third processed signal to calculate the health index C of the reducer gear B for Among them, f B is the fault characteristic frequency appearing in the square envelope spectrum, A(f B ) is the amplitude of the fault characteristic frequency.

Images