1. A method for recognizing weak faults of a joint RV reducer is characterized by comprising the following specific steps:

(1) collecting servo motor current signal I (t) in the ascending or descending motion process of robot joint₁) And vibration signal S (t) of joint RV reducer operation state₂) Wherein t is₁Is a current signal I (t)₁) And t is the sampling time of₂Is a vibration signal S (t)₂) The sampling time of (a);

(2) using wavelet decomposition low-frequency reconstruction method to process current signal I (t)₁) Filtering to remove micro-fluctuation in the starting stage and the deceleration stagnation stage to obtain a reconstructed current signal I₁(t₁)；

(3) By the formulaWhere P is the Cauchy principal component, the slave current signal I₁(t₁) Obtaining a Hilbert transform result Y (t)₁) By the formulaAndcalculating the current signal I₁(t₁) Instantaneous frequency f (t)₁) To obtainCurrent signal I to the course of the rising or falling movement of the joint₁The time-frequency diagram of (2);

(4) obtaining a reconstructed current signal I by Fourier transform₁(t₁) The frequency spectrum graph takes the frequency corresponding to the highest amplitude as the instantaneous frequency f of the servo motor in the stable peak stage in the time frequency graph, and the rotating speed n of the servo motor in the stable peak stage in the time frequency graph is calculated through a formula n which is 60f/p, wherein p is the magnetic pole pair number of the servo motor; by the formula f₁Calculating to obtain the revolution frequency f of the sun gear of the joint RV reducer at the peak stationary stage as n/60₁；

(5) By the formulaAnd f_1c＝Z₁f₁＝Z₂f₂Calculating the planetary gear rotation frequency f of the joint RV reducer₂And a first order transmission mesh frequency f_1cI.e. sun and planet gear mesh frequency, wherein Z₁And Z₂The number of teeth of the sun gear of the RV reducer and the number of teeth of the planet gear are respectively;

(6) for the vibration signal S (t) collected in the step (1)₂) Denoising by adopting a morphological filtering method to obtain a filtered vibration signal S₁(t₂)；

(7) Using a current signal I of the joint ascending or descending motion process in the step (3)₁The time point corresponding to the maximum rotation frequency in the time-frequency diagram is taken as a time starting point, and the vibration signal S after filtering is carried out₁(t₂) Finding out the stable peak stage and intercepting the vibration signal S of the stable peak stage₂(t)；

(8) For vibration signal S₂(t) performing cepstrum analysis to obtain a cepstrum, judging that the joint RV reducer has a fault when an amplitude surge spectral line appears in the cepstrum, and obtaining the reciprocal of the reciprocal frequency at the highest position of the amplitude surge spectral line to obtain a fault point frequency f_p；

(9) Frequency f of fault point_pCarrying out comparative analysis with the characteristic frequency of the joint RV reducer, wherein the characteristic frequency of the joint RV reducerThe ratio being the sun gear revolution frequency f₁Planet wheel rotation frequency f₂First order transmission mesh frequency f_1c；

Frequency f at fault point_pWhen the characteristic frequency of the RV reducer is approximate to one of the characteristic frequencies, diagnosing that the fault is located at a characteristic position corresponding to the characteristic frequency;

(10) for vibration signal S₂(t) carrying out variation modal decomposition to obtain a vibration signal S₂And (t) taking the information entropy as a selection index, taking the component signal with the information entropy closest to that of the original signal as a reconstruction signal, carrying out Fourier transform on the reconstruction signal to obtain a fault signal spectrogram, and extracting fault features from the spectrogram.

2. The method of joint RV reducer weak fault identification of claim 1, characterized by: obtaining current signal I (t) of servo motor through current transformer₁)。

3. The method of joint RV reducer weak fault identification of claim 1, characterized by: obtaining vibration signal S (t) of joint RV reducer operation state through acceleration sensor₂)。

Background

Industrial robot research is moving towards high precision, high speed, multi-axis and lightweight. The speed reducer is a core component of the industrial robot, and the health condition of the speed reducer determines the execution efficiency and precision of the industrial robot. The high-precision RV reducer (gear reducer) is the most commonly used reducer for industrial robots at present due to its characteristics of small size, large transmission ratio, high efficiency, etc. In the long-time operation process, the joint RV reducer has weak faults due to natural wear and fatigue life of parts and even sudden braking, and the weak faults do not affect the operation of the joint and have no obvious fault judgment, so that the feedback information displayed before the joint RV reducer has serious faults can be responded in advance, serious equipment faults can be avoided, and the intelligent fault diagnosis and health assessment of the industrial robot are particularly important. When the robot works, the monitoring of the attitude motion conventionally uses noise, vibration, current and the like as monitoring data sources. The RV reducer is sealed in a joint as a high-precision core component, and the joint RV reducer fault diagnosis is used as an early fault identification technology, and is the characteristic extraction and identification which needs to research and analyze weak signals. Therefore, the fault analysis by combining the current signal and the vibration signal is a fused fault processing method. The current signal is a signal which is finally fed back by the motion representation of the overall motion of the robot, the current signal acquisition mode is relatively simple, the external interference is small, and the vibration and noise are not required to be in a specific environment, so that more external interference is eliminated. The vibration signal is acquired by placing an acceleration sensor on the joint arm, the acquisition mode is relatively convenient, and the fault signal is more accurate and sensitive.

The stress of the joint RV reducer is changed in the operation process of the robot joint, so that the robot joint is unstable in vibration and does not complete periodic circular motion. The conventional method for acquiring the rotating speed of the servo motor is to analyze a key phase pulse signal of an encoder, but for a robot, the key phase pulse is not easy to acquire and the real running state of a joint cannot be accurately fed back. The current signal through the servo motor is more convenient and stable in speed extraction, and the current signal is less in interference. Because the current signal is different with vibration signal's collection mode, synchronous collection can lead to follow-up signal processing to become complicated because vibration signal's sampling frequency is far higher than current signal's sampling frequency, and asynchronous collection is favorable to the convenience of experiment, tests with different sampling frequencies.

Disclosure of Invention

Aiming at the problems in the prior art, the invention provides a method for identifying weak faults of a joint RV reducer based on cepstrum analysis and variational modal decomposition, which is characterized in that current signals and vibration signals are collected without synchronous collection and pulse signals are not required to extract rotating speed.

For the joint RV reducer, when weak faults occur, due to the fact that vibration signals can show amplitude modulation characteristics, amplitude is increased suddenly due to the fact that characteristic frequency is affected by weak fault amplitude modulation, the vibration signals of the running state of the joint RV reducer are analyzed based on current signals of a servo motor of the joint RV reducer, and the weak faults in the joint RV reducer can be accurately identified and diagnosed.

The invention relates to a method for identifying weak faults of a joint RV reducer based on cepstrum analysis and variational modal decomposition, which comprises the following steps:

(1) collecting servo motor current signal I (t) in the ascending or descending motion process of robot joint₁) And vibration signal S (t) of joint RV reducer operation state₂) Wherein t is₁Is a current signal I (t)₁) And t is the sampling time of₂Is a vibration signal S (t)₂) The sampling time of (a);

(2) using wavelet decomposition low-frequency reconstruction method to process current signal I (t)₁) Filtering to remove micro-fluctuation in the starting stage and the deceleration stagnation stage to obtain a reconstructed current signal I₁(t₁)；

(3) By the formulaWhere P is the Cauchy principal component, the slave current signal I₁(t₁) Obtaining a Hilbert transform result Y (t)₁) By the formulaAndcalculating the current signal I₁(t₁) Instantaneous frequency f (t)₁) Obtaining a current signal I of the ascending or descending movement process of the joint₁The time-frequency diagram of (2);

(4) obtaining a reconstructed current signal I by Fourier transform₁(t₁) The frequency spectrum graph takes the frequency corresponding to the highest amplitude as the instantaneous frequency f of the servo motor in the stable peak stage in the time frequency graph, and the rotating speed n of the servo motor in the stable peak stage in the time frequency graph is calculated through a formula n which is 60f/p, wherein p is the magnetic pole pair number of the servo motor; by the formula f₁The revolution frequency f of the sun gear of the joint RV reducer in the peak stationary stage is calculated and obtained as n/60₁；

(5) By the formulaAnd f_1c＝Z₁f₁＝Z₂f₂Calculating the planetary gear rotation frequency f of the joint RV reducer₂And a first order transmission mesh frequency f_1cI.e. sun and planet gear mesh frequency, wherein Z₁And Z₂The number of teeth of the sun gear of the RV reducer and the number of teeth of the planet gear are respectively;

(6) for the vibration signal S (t) collected in the step (1)₂) Denoising by adopting a morphological filtering method to obtain a filtered vibration signal S₁(t₂)；

(7) Using the current signal I in the ascending or descending motion stage of the joint in the step (3)₁The time point corresponding to the maximum rotation frequency in the time-frequency diagram is taken as the time starting point, and the vibration signal S after filtering is carried out₁(t₂) Finding out the stable peak stage and intercepting the vibration signal S of the stable peak stage₂(t)；

(8) By applying a vibration signal S₂(t) performing cepstrum analysis to obtain a cepstrum, judging that the joint RV reducer has a fault when an amplitude surge spectral line appears in the cepstrum, and obtaining the reciprocal of the reciprocal frequency at the highest position of the amplitude surge spectral line to obtain a fault point frequency f_p；

(9) Frequency f of fault point_pCarrying out comparative analysis on the characteristic frequency of the joint RV reducer, wherein the characteristic frequency of the joint RV reducer is the revolution frequency f of the sun gear₁Planet wheel rotation frequency f₂First order transmission mesh frequency f_1c；

Frequency f at fault point_pWhen the characteristic frequency of the RV reducer is approximate to one of the characteristic frequencies, diagnosing that the fault is located at a characteristic position corresponding to the characteristic frequency;

(10) for vibration signal S₂(t) carrying out variation modal decomposition to obtain a vibration signal S₂And (t) taking the information entropy as a selection index, taking the component signal with the information entropy closest to that of the original signal as a reconstruction signal, carrying out Fourier transform on the reconstruction signal to obtain a fault signal spectrogram, and extracting fault features from the spectrogram.

In the method, the current transformer is used for acquiring the current signal I (t) of the servo motor₁) (ii) a Obtaining vibration signal S (t) of joint RV reducer operation state through acceleration sensor₂)。

The invention has the beneficial effects that:

1. the acquisition way of acquiring the rotating speed of the servo motor from the current signal is superior to that of recalculating the rotating speed by the key phase pulse of the encoder;

2. the method can effectively and accurately extract the vibration signal of the RV reducer in the stable stage of the joint operation process, avoids mixing the acceleration or deceleration stage during blind extraction, and only needs to ensure that the robot joint has the stable stage reaching the peak rotating speed in the operation process;

3. the method has the advantages that the target for extracting the fault characteristics of the unstable vibration is clear, the fault point frequency obtained by cepstrum analysis corresponds to the characteristic frequency of the joint RV reducer, the characteristic frequency with the amplitude increased due to the influence of weak fault amplitude modulation is found near the fault frequency for analysis, the type of the weak fault is determined, and the identification application in the weak fault of the joint RV reducer is realized.

Drawings

FIG. 1 is a schematic view of an experimental bench for simulating joint movement RV reducers;

FIG. 2 shows a fault current signal I (t) of a servo motor of the RV reducer for joint elevation₁) A schematic diagram of an original waveform;

FIG. 3 shows a fault vibration signal S (t) of the joint rising RV reducer operation state₂) A schematic diagram of an original waveform;

FIG. 4 shows a reconstructed fault current signal I of a servo motor of the joint elevation RV reducer₁(t₁) Schematic diagram of time domain waveform of (a);

FIG. 5 is a schematic diagram of a frequency spectrum of a fault current signal of a servo motor;

FIG. 6 is a schematic frequency spectrum diagram of a fault current signal of a servo motor

FIG. 7 is a filtered fault vibration signal S of joint rising RV reducer operation state₁(t₂) Time domain waveform of

FIG. 8 is a diagram illustrating the extraction of the fault vibration signal S at the peak stationary phase₂(t) a time domain waveform schematic;

FIG. 9 shows a fault vibration signal S₂(t) a cepstrum schematic of a cepstrum analysis;

FIG. 10 shows a fault vibration signal S₂(t) a schematic diagram of the envelope spectrum;

FIG. 11 is a schematic diagram of a reconstructed fault vibration signal envelope spectrum;

FIG. 12 is a schematic of a reconstructed fault vibration signal spectrum;

FIG. 13 shows a fault-free current signal I (t) of a servo motor of the RV reducer for joint elevation₁) A schematic diagram of an original waveform;

FIG. 14 shows a failure-free vibration signal S (t) of the joint-lift RV reducer in the operating state₂) A schematic diagram of an original waveform;

FIG. 15 is a reconstructed fault-free current signal I of the joint rising RV reducer servo motor₁(t₁) Schematic diagram of time domain waveform of (a);

FIG. 16 is a schematic diagram of a frequency spectrum of a servo motor without a fault current signal;

FIG. 17 is a schematic frequency spectrum diagram of a fault-free current signal of a servo motor

FIG. 18 is a filtered fault-free vibration signal S of joint-up RV reducer operating condition₁(t₂) Time domain waveform of

FIG. 19 is a diagram of extracting a fault-free vibration signal S at the peak plateau₂(t) a time domain waveform schematic;

FIG. 20 shows a fault-free vibration signal S₂(t) a cepstrum schematic of a cepstrum analysis;

in fig. 1: the system comprises a 1-joint arm, a 2-servo motor, a 3-accelerator sensor, a 4-RV reducer, a 5-experiment table base, a 6-current transformer and a 7-electric cabinet.

Detailed Description

The present invention is further illustrated by the following examples, but the scope of the invention is not limited to the above-described examples.

Example 1: fault analysis of planet gear tooth root cracks of joint RV reducer

An RV-40E reducer is adopted, and FIG. 1 is a schematic diagram of an RV reducer experiment table for simulating joint movement, and the RV reducer experiment table comprises a 1-swing arm, a 2-servo motor, a 3-accelerator sensor, a 4-RV reducer, a 5-experiment table base, a 6-current transformer and a 7-electric cabinet; presetting an experiment: the motion range of the swing arm is 100-180 degrees (single 80 degrees, rising and falling are 160 degrees in a single cycle), the running speed of the swing arm is 32 degrees/s, and the specific operation flow is as follows:

1. in the ascending process of the swing arm, a current transformer 6 is adopted to obtain a current signal I (t) of the servo motor 2₁) As shown in fig. 2; the acceleration sensor 3 is adopted to obtain the vibration signal S (t) of the running state of the joint RV reducer₂) As shown in fig. 3; t is t₁Is a current signal I (t)₁) And t is the sampling time of₂Is a vibration signal S (t)₂) The sampling time of (a); the current signal is collected as a U-phase current signal at the output end of the driver, and the sampling frequency is 25.6 KHz; the vibration signal is collected to be placed at the top end of the servo motor, and the sampling frequency is 1 MHz;

2. the current signal I (t) is subjected to a wavelet 6-layer decomposition low-frequency reconstruction method by using MATLAB analysis software₁) Filtering to remove micro-fluctuation in the starting stage and the deceleration stagnation stage to obtain a filtered current signal I₁(t₁) The reconstructed current signal is shown in fig. 4;

the following calculation processes and output results are all completed in analysis software;

3. by the formulaWhere P is the Cauchy principal component, the slave current signal I₁(t₁) Obtaining a Hilbert transform result Y (t)₁) By the formulaAndcalculating the current signal I₁(t₁) Instantaneous frequency f (t)₁) Obtaining a current signal I of the joint in the ascending motion process₁The time-frequency diagram of (1) is shown in figure 5;

4. obtaining a reconstructed current signal I by Fourier transform₁(t₁) As shown in fig. 6; taking the frequency corresponding to the highest amplitude position in the spectrogram as the instantaneous frequency f of the servo motor in the stable peak value stage in the time-frequency diagram, wherein f is 62.7Hz, and calculating to obtain the peak value of the servo motor through a formula n which is 60f/pThe rotating speed n in the stationary stage is 752.4r/min, wherein p is the number of pole pairs of the servo motor, and p is 5; by the formula f₁Calculating to obtain the rotation frequency f of the sun gear of the joint RV reducer in the stable stage as n/60₁，f₁＝12.54Hz；

5. By the formulaAnd f_1c＝z₁f₁＝z₂f₂Wherein Z is₁12 and Z₂42, calculating the planetary wheel rotation frequency f of the joint RV reducer₂And primary transmission engagement frequency (i.e. sun and planet gear engagement) f_1cPlanetary gear rotation frequency f₂At 3.58Hz, a primary drive engagement frequency f_1c150.48 Hz;

6. the collected vibration signal S (t) is filtered by a morphological filtering method₂) De-noising to obtain filtered signal S₁(t₂) The results are shown in FIG. 7;

7. current signal I during step 3 joint up-movement₁The time point corresponding to the maximum rotation frequency obtained in the time-frequency diagram (fig. 5) is 0.3S, and in order to increase the accuracy and reduce the selection range, it can be seen that 0.5S to 2.5S are peak stationary phases, so that the vibration signal S after filtering of fig. 7 at the same time is the same₁(t₂) In the figure, the vibration signal S can be accurately judged to be 0.5S to 2.5S₁(t₂) At the peak stationary phase of the filtered vibration signal S₁(t₂) Intercepting any 1S signal from 0.5S to 2.5S as vibration signal S₂(t) avoiding the signals in the acceleration and deceleration stages from being mixed, i.e. the signals from 1s to 2s selected in the embodiment, as shown in fig. 8;

8. for vibration signal S₂(t) carrying out cepstrum analysis to obtain a cepstrum, wherein as shown in fig. 9, if spectral lines with amplitude surge appear in the cepstrum, the joint RV reducer is judged to have faults; taking the reciprocal of the reciprocal frequency at the highest position of the amplitude surge spectral line to obtain the frequency f of the fault point_p，f_p＝1000/6.68＝149.7Hz；

9. Frequency of fault pointf_pThe characteristic frequency of the joint RV reducer is compared and analyzed, and the frequency f of the fault point can be known_pFrequency of engagement with primary drive f_1c(150.48Hz) is close, the fault is diagnosed to be positioned at the primary transmission meshing part, namely the meshing part of the sun wheel and the planet wheel; meanwhile, envelope spectrum analysis is adopted to verify the effectiveness of the method in the step, and the vibration signal S is subjected to₂(t) envelope analysis to obtain an envelope spectrum, as shown in FIG. 10, with the result showing the primary drive engagement f_1cAnd the frequency multiplication has side frequency spectral lines, which also indicates that the fault is positioned at the primary transmission meshing;

10. for vibration signal S₂(t) carrying out variable mode decomposition analysis, setting the decomposition scale K in the variable mode decomposition to be 3, and setting other parameters to be conventional default values to obtain a vibration signal S₂(t) the component signal; calculating the number of complete cycles of planetary wheel operation in unit time to be 3 according to the tooth numbers of the sun wheel and the planetary wheel, dividing each component signal into 3 sections for information entropy calculation, then taking the information entropy as a selection index, taking the component signal with the information entropy of the component signal closest to the information entropy of the original signal as a reconstruction signal, see table 1, and taking the information entropy of IMF1 in table 1 as the closest to the information entropy of the original signal, therefore taking IMF1 as the reconstruction signal, performing Fourier transform on the reconstruction signal to obtain a fault signal spectrogram, as shown in fig. 12, observing a side band f of meshing frequency in the spectrogram_1c±f₁And the amplitude of its multiple frequency is abruptly increased by root crack failure, beyond its primary drive mesh frequency and its multiple frequency, and Δ f₁The interval of the planet gear is in accordance with the rule of the crack fault of the gear, so that the crack fault of the planet gear is inferred;

TABLE 1 information entropy of original signal and each component signal

Meanwhile, the effectiveness of the method in the step is verified by adopting envelope spectrum analysis, the reconstructed signal IMF1 is subjected to envelope analysis to obtain an envelope spectrum, as shown in figure 11, the prominent characteristic frequency of the envelope spectrum can be seen from the figure, and the reconstruction is carried outFiltering out other irrelevant frequencies by envelope spectrum of vibration signal, and obviously observing first-order meshing frequency f_1cAnd the frequency multiplication thereof prove the accuracy of the information entropy selection index.

Example 2: the method for identifying the weak fault of the joint RV reducer comprises the following steps:

the RV-40E reducer is adopted, the schematic diagram of the experiment table of the joint motion RV reducer in the embodiment is the same as that in the embodiment 1, the difference is that the joint RV reducer has no fault, the running speeds are different, the sampling rates are different, and the running speed of a swing arm is 45 degrees/s;

1. in the ascending process of the swing arm, a current transformer 6 is adopted to obtain a current signal I (t) of the servo motor 2₁) As shown in fig. 13; the acceleration sensor 3 is adopted to obtain the vibration signal S (t) of the running state of the joint RV reducer₂) As shown in fig. 14; t is t₁Is a current signal I (t)₁) And t is the sampling time of₂Is a vibration signal S (t)₂) The sampling time of (a); the current signal is collected as a U-phase current signal at the output end of the driver, and the sampling frequency is 8192 Hz; the vibration signal is collected to be placed at the top end of the servo motor, and the sampling frequency is 50 kHz;

2. the current signal I (t) is subjected to a wavelet 6-layer decomposition low-frequency reconstruction method by using MATLAB analysis software₁) Filtering to remove micro-fluctuation in the starting stage and the deceleration stagnation stage to obtain a filtered current signal I₁(t₁) The reconstructed current signal is shown in fig. 15;

the following calculation processes and output results are all completed in analysis software;

3. by the formulaWhere P is the Cauchy principal component, the slave current signal I₁(t₁) Obtaining a Hilbert transform result Y (t)₁) By the formulaAndcalculating the current signal I₁(t₁) Instantaneous frequency f (t)₁) Obtaining a current signal I of the joint in the ascending motion process₁The time-frequency diagram of (1) is shown in fig. 16;

4. obtaining a reconstructed current signal I by Fourier transform₁(t₁) As shown in fig. 17; taking the frequency corresponding to the highest amplitude position in the frequency spectrogram as the instantaneous frequency f of the servo motor in the stable peak value stage in the time-frequency diagram, wherein f is 88Hz, and calculating to obtain the rotating speed n of the servo motor in the stable peak value stage through a formula n which is 60/p, wherein n which is 1056r/min, wherein p is the magnetic pole logarithm of the servo motor, and p which is 5; by the formula f₁Calculating to obtain the rotation frequency f of the sun gear of the joint RV reducer in the stable stage as n/60₁，f₁＝17.6Hz；

5. By the formulaAnd f_1c＝z₁f₁＝z₂f₂Wherein Z is₁12 and Z₂42, calculating the planetary wheel rotation frequency f of the joint RV reducer₂And primary transmission engagement frequency (i.e. sun and planet gear engagement) f_1cPlanetary gear rotation frequency f₂At 5.03Hz, a primary drive engagement frequency f_1cIs 211.2 Hz;

6. the collected vibration signal S (t) is filtered by a morphological filtering method₂) De-noising to obtain filtered signal S₁(t₂) The results are shown in FIG. 18;

7. current signal I during step 3 joint up-movement₁The time point corresponding to the maximum rotation frequency obtained in the time-frequency diagram (fig. 16) is 0.35S, and in order to increase the accuracy and reduce the selection range, it can be seen that 0.5S to 1.7S are peak stationary phases, so that the vibration signal S filtered in the same time point fig. 18 is the same₁(t₂) In the figure, the vibration signal S can be accurately judged to be 0.5S to 1.7S₁(t₂) At the peak stationary phase of the filtered vibration signal S₁(t₂) Intercepting any 1s signal from 0.5s to 1.7s as vibration signalNumber S₂(t) avoiding the mixing of signals in the acceleration and deceleration stages, which are selected from 0.6s to 1.6s in the embodiment, as shown in fig. 19;

8. for vibration signal S₂(t) carrying out cepstrum analysis to obtain a cepstrum, wherein as shown in fig. 20, if spectral lines with no obvious amplitude surge in the cepstrum can be seen from the graph, it is judged that the joint RV reducer has no fault;

in a word, the method can effectively identify and diagnose the weak fault of the joint RV reducer, can be used for early fault diagnosis of equipment, and prevents equipment damage caused by the fact that the weak fault cannot be processed in time.