A vibration signal analysis and diagnosis method for ball screw pair reverser failure

Abstract

The application provides a vibration signal analysis and diagnosis method for ball screw pair reverser failure, which comprises the following steps: 1, collecting screw vibration signals, using wavelet packet transform to perform modal decomposition on original signals, and realizing vibration signal frequency band screening; 2, performing frequency domain analysis and time domain analysis on the reserved vibration signals: decomposing the signals reserved in step 1 through an EMD method to obtain a limited number of intrinsic mode functions of different frequency bands, and selecting ball pass frequency and amplitude; using time domain statistics to calculate the variance, root mean square, kurtosis, pulse factor and margin factor in the domain signals of the signals reserved in step 1; 3, using the frequency domain analysis method and the time domain analysis method to formulate reverser performance detection standards, and taking the ratio of real-time signals to initial signals in the screw running process as the basis for judging whether the reverser performance is lost. The judgment result of the application is more accurate.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of fault diagnosis and analysis of ball screw pairs, and particularly relates to a vibration signal analysis and diagnosis method for faults of a ball screw pair reverser. BACKGROUND

[0002] Ball screw pairs are widely used in various industrial equipment and precision instruments due to their excellent friction characteristics. When the screw rotates at high speed, the high-frequency impact of the ball on the reverser causes fatigue damage to the reverser, which makes the reverser a weak link in the performance of the ball screw pair. During the circulation of the ball through the reverser, the stress state of the ball changes between bearing and non-bearing, and the motion state and stress condition are constantly changing. The instantaneous collision force generated thereby directly acts on the raceway of the reverser, greatly affecting the performance of the reverser. Therefore, research on the reverser is an important way to improve the performance of the ball screw pair.

[0003] At present, the research on fault diagnosis of ball screw pairs mainly focuses on the analysis of vibration signals. Most methods use the fault features of vibration signals obtained by wavelet packet, EMD or kernel principal component decomposition as samples, input neural networks or support vector machines, and then output the corresponding fault types such as screw bending and ball raceway wear. A large number of fault sample data are used for training. Finally, the trained program is used for real-time diagnosis of the type of fault.

[0004] (Wu Xi-xi et al. Ball Screw Fault Diagnosis Technology Based on Hyper-spherical Support Vector Machine uses wavelet packet decomposition to extract some fault features of vibration signals of ball screw pairs with faults, inputs a hyper-spherical support vector machine for learning and training, and outputs specific fault types. During real-time diagnosis, the collected vibration signals are directly input into the hyper-spherical support vector machine after decomposition, and the output diagnosis result will show the type of fault. Wen Guoqiang et al. Ball Screw Fault Diagnosis Based on EMD and Neural Network uses EMD to decompose some vibration signals of ball screw pairs with faults to obtain intrinsic mode function components, inputs a designed BP neural network for learning and training, and outputs specific fault types. During real-time diagnosis, the collected vibration signals are directly input into the trained neural network after decomposition, and the output diagnosis result will show the type of fault.)

[0005] The prior art has the following limitations in diagnosing the failure of the screw rod: 1. Each method does not consider the influence of different structural types, different lubrication states, different installations and the like on the failure diagnosis by using the vibration signal; 2. Since the methods collect the vibration signal after the ball screw pair has failed during the test, the signal is processed and then learned and trained, the process from normal to gradual wear to final failure of the ball screw pair cannot be represented, and the ball screw pair at the edge of failure cannot be diagnosed; 3. The neural network or support vector machine needs to be learned and trained with fault samples, in order to improve the judgment accuracy, the number of samples is bound to be large, a large amount of pre-test data needs to be accumulated, and the time and fund consumption is huge; 4. The above diagnosis methods are used to judge the three failures of the screw rod bending, ball wear failure and raceway wear failure, and the failure of the reverser is not specially studied, and the methods are not targeted.

[0006] At present, the research on the ball screw pair in China is more focused on the overall research, and the research on the reverser is not thorough, especially the research on the performance of the reverser is still very insufficient. SUMMARY

[0007] The purpose of the present application is to provide a vibration signal analysis and diagnosis method for the failure of the reverser of the ball screw pair, so as to diagnose the failure condition of the reverser of the ball screw pair.

[0008] The technical solution for achieving the purpose of the present application is as follows:

[0009] A vibration signal analysis and diagnosis method for the failure of the reverser of the ball screw pair, comprising the following steps:

[0010] Step 1, collecting the vibration signal of the screw rod, using wavelet packet transform to modal decomposition of the original signal, realizing the selection of the vibration signal frequency band;

[0011] Step 2, frequency domain analysis and time domain analysis are performed on the retained vibration signal:

[0012] The signal retained in step 1 is decomposed by EMD method to obtain a finite number of intrinsic mode functions of different frequency bands, and the ball pass frequency and amplitude are selected;

[0013] The variance, root mean square, kurtosis, pulse factor and margin factor in the domain signal are obtained by using time domain statistics to calculate the signal retained in step 1;

[0014] Step 3, the frequency domain analysis method and the time domain analysis method are used to develop the performance detection standard of the reverser, and the ratio of the real-time signal to the initial signal during the running-in process of the screw rod is used as the basis for judging whether the performance of the reverser is lost.

[0015] Compared with the prior art, the present application has the following advantages:

[0016] (1) The present application compares the amplitude of ball pass frequency before and after running-in, which avoids the large difference between the ball pass frequency and amplitude of vibration signals of different screws caused by processing errors, and makes the judgment result more accurate.

[0017] (2) The present application fully considers the influence of sensor installation position, screw lubrication condition, screw installation condition, and disassembly of reverser on vibration signal collection in the preliminary test, and excludes the influence of external factors on multiple control test data as much as possible, so that the signal diagnosis method of reverser failure is more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a flow chart of the present application.

[0019] Figure 2 is a ball pass frequency amplitude change curve.

[0020] Figure 3 is a variance change curve with life.

[0021] Figure 4 is a pulse factor change curve with life.

[0022] Figure 5 is a margin factor change curve with life. DETAILED DESCRIPTION

[0023] The present application will be further described below in combination with the drawings and specific embodiments.

[0024] In combination with Figure 1 , the vibration signal analysis and diagnosis method for reverser failure of the ball screw pair of the present embodiment includes the following steps:

[0025] Step 1, when the screw is running at high speed, the main failure mode of the reverser is pitting of the reverser raceway and fatigue failure of the reverser inlet and outlet. The vibration sensor is installed on the outer end surface of the reverser on the nut side to collect vibration signals. The influence factors of reverser vibration test are studied through comparison test, so as to reduce the influence of external factors on vibration signal collection.

[0026] The influence of sensor installation position on vibration signal collection: in the test, the sensor is installed on the front end surface, upper end surface, side end surface and outer end surface of the reverser on the nut flange respectively. From the test results, it can be seen that the amplitudes of the ball pass frequency collected by different sensor installation positions are different. When the screw speed is low, the sensor installed on the flange will cause a lot of signal pollution, and the effective ball pass frequency amplitude cannot be extracted. When the screw speed is high, the influence of sensor installation position on amplitude is relatively small.

[0027] The influence of lubrication condition on vibration signal acquisition: The lubricating oil film generated by the screw under different lubrication conditions may be different, which may affect the collision force generated by the ball impacting the reverse device, and then affect the acquisition result of the vibration signal. Data acquisition was carried out on the screw pair under the conditions of no lubrication, oil lubrication and grease lubrication. It was found that using oil lubrication and grease lubrication can reduce the amplitude value of the ball pass frequency of the reverse device in the signal by about 50%, indicating that reasonable use of lubricant can reduce the vibration intensity inside the reverse device.

[0028] The influence of screw installation on vibration signal acquisition: When collecting vibration signals of the same screw at different workstations, it was found that the installation workstation of the screw had a huge difference on the test results, which may be due to the inherent frequency of the workstation. Once the workstation is selected, the signal collection of the screw cannot be changed; when repeatedly disassembling the vibration signal of the same screw at the same workstation, it was found that the repeated disassembly of the screw had little effect on the vibration amplitude at the ball pass frequency, and the error after disassembly was within 10%; when collecting vibration signals of different screws at the same workstation, it was found that there were great differences in the vibration characteristics between different screws, and the difference in the ball pass frequency could be controlled within 10%, but the difference in the vibration amplitude could reach about 50%, so it was difficult to develop a unified judgment standard for different types of screws, and the initial vibration data of the screw itself should be used as the judgment benchmark for subsequent vibration signal tests.

[0029] The influence of reverse device disassembly on vibration signal acquisition: During the subsequent test process of developing standards for reverse device performance detection, the topography of the reverse device raceway needs to be observed, and the reverse device will be frequently disassembled. Since manual installation will affect the positioning accuracy of the reverse device and the nut raceway, it is necessary to collect vibration signals before and after disassembly of the reverse device, and to judge whether disassembly of the reverse device will have a great impact on the amplitude of the ball pass frequency. The reverse device was repeatedly disassembled and tested, and vibration data was collected after each disassembly. It was found that repeated disassembly of the reverse device would have some effect on the amplitude of the ball pass frequency, but this effect could be ignored.

[0030] In summary, the sensor should be installed on the side outer end surface of the reverse device, the screw should be kept in the same workstation and have the same lubrication condition during each signal collection, and since the vibration characteristics of different screws are different, the vibration signal of the screw before running-in must be used as the reference data for judging whether the reverse device has failed.

[0031] Because the relative sliding between the ball and the screw raceway, nut raceway may exist, which leads to the actual ball pass frequency may appear in the range of 0~2f. The essence of wavelet packet transform is to obtain information of different frequency bands by using a set of low-frequency filters and band-pass filters. The energy sum of the signal in each frequency band is the same as the original signal. Through wavelet packet transform, the frequency band of the vibration signal can be optimized. Therefore, the collected vibration signal can be denoised by wavelet packet transform to retain the frequency band within 0~2f for subsequent research. The specific calculation formula of the ball pass frequency is:

[0032]

[0033]

[0034]

[0035] Wherein, f is the theoretical ball pass frequency of the ball impacting the reverser; m is the number of rollers contained in each raceway of the ball screw pair; ω M is the angular velocity of the ball; D M is the nominal diameter of the screw; D B is the ball diameter; L is the lead of the screw; ω is the axial speed of the screw; R B is the radius of the ball; R M is the pitch radius of the screw; and α is the contact angle of the ball and the raceway.

[0036] Step 2, frequency domain analysis (EMD decomposition of the filtered signal): the signal retained in step one is decomposed by EMD method to obtain a finite number of intrinsic mode functions of different frequency bands. Each intrinsic mode function component contains the local characteristic signal of the signal retained in step one. Through experiments, it is found that the frequency corresponding to the intrinsic mode function component has a very small error with the theoretical ball pass frequency. Therefore, the EMD analysis method can be used to screen the signal in step one to select the ball pass frequency and amplitude with clear characteristics, effectively extract the vibration characteristics at the theoretical ball pass frequency of the reverser, and effectively exclude the interference of other factors.

[0037] Time domain analysis: Because the vibration signal of the screw is a nonlinear signal, the processed vibration signal amplitude of the reverser collected in step 1 can be used to judge its performance by statistical method.

[0038] (1) Root mean square:

[0039] (2) Variance: The larger the value is, the greater the energy contained in the vibration signal is, and the greater the possibility of failure of the reverser is.

[0040] (3) Kurtosis: The standard value of kurtosis is 3, and the distribution curve greater than 3 indicates that the signal is more obviously impacted, and the possibility of failure of the reverser is greater.

[0041] (4) Pulse factor: The value represents the strength of the impact on the signal, and the greater the value, the stronger the impact.

[0042] (5) Margin factor: It can be used to detect the wear condition of the equipment, and the more serious the wear condition, the greater the value of the margin factor.

[0043] wherein, is the root mean square; x i is the amplitude of the i-th signal; i is the i-th signal; n is the number of signals; σ is the variance; k u is the kurtosis; X max is the peak value of the signal.

[0044] The variance and the root mean square value can represent the energy contained in the vibration signal, and the greater the energy, the greater the value; the kurtosis, whose value represents the flatness of the signal, has a standard value of 3, and the distribution curve greater than 3 indicates that the signal is more obviously impacted; the pulse factor, whose value represents the strength of the impact on the signal, the greater the value, the stronger the impact; the margin factor, which can be used to detect the wear condition of the equipment, the more serious the wear condition, the greater the value of the margin factor.

[0045] Step 3, respectively, using frequency domain analysis method and time domain analysis method to develop reverser performance detection standard, with the ratio of real-time signal to initial signal in the process of screw running as the basis for judging whether the performance of the reverser is lost. The reverser performance detection standard includes the following four indicators: amplitude at ball pass frequency, variance, pulse factor, margin factor.

[0046] Frequency domain: when the actual amplitude at ball pass frequency during the screw running process is lower than 200% of the initial amplitude, the performance of the reverser is good; when the amplitude reaches 200-250% of the initial amplitude, it indicates that the reverser has entered a severe wear stage, and the reverser will fail at this time.

[0047] Time domain: when any of the following occurs during the use of the screw, the variance reaches 300% of the initial value of the screw; the pulse factor reaches 130% of the initial value; the margin factor reaches 180% of the initial value, it can be considered that the performance of the reverser may have degraded, and the amplitude change of the ball pass frequency needs to be combined for comprehensive judgment, and the reverser is disassembled to observe the raceway morphology to determine whether the reverser has failed.

[0048] Through the test, it is known that the frequency domain index, i.e. the amplitude at ball pass frequency, is used as the standard when judging whether the reverser has failed.

[0049] The present application collects the vibration signal of the screw, utilizes the wavelet packet transform to carry out the modal decomposition to the original signal, then carries out the frequency domain analysis through the EMD analysis method, carries out the time domain analysis through the kurtosis, variance and the like standards, finally utilizes the ratio of the real-time signal in the running-in process of the screw and the initial signal as the basis whether the performance of the reverser loses.

[0050] The present application improves the existing vibration signal analysis method, proposes the wavelet packet transform-EMD method to detect the reverser fault, judges whether the performance of the reverser changes through the change of the time domain signal index and the frequency domain signal index relative to the initial value as the standard, eliminates the influence of the machining error and the like, and makes the judgment result more accurate.

[0051] Example 1

[0052] In the present embodiment, two JF2003 type screws of NG company are tested and analyzed, and the two screws are named as No. 1 screw and No. 2 screw, and the time domain and frequency domain indexes are used to characterize the performance of the reverser.

[0053] 1. Test parameters

[0054]

[0055] 2. Test data

[0056] The screw is installed in different stations of the test table, after the initial data is collected, the running-in test is carried out, and the test data is recorded as follows, and the corresponding curve is generated as Figures 2-5 .

[0057] Time-frequency characteristic change table of two screws

[0058]

[0059] 3. Test result analysis

[0060] The amplitude change at the ball pass frequency is the most important index to judge whether the reverser appears fault. When the service life of the screw reaches 5.4 million revolutions (i.e. 30h), the amplitude change of the two screws at the ball pass frequency relative to the initial value is 242% and 264% respectively, which exceeds the critical value 200% representing the performance loss of the reverser, indicating that the reverser may have appeared fault.

[0061] The variance change is inaccurate. When the screw reaches 5.4 million revolutions (i.e. 30h), the reverser has failed, and the variance of the two screws relative to the initial value is 199% and 514%, respectively. For the No. 1 screw, the critical value 300% is not reached. For the No. 2 screw, when the screw reaches 3.6 million revolutions (i.e. 20h), the variance relative to the initial value is 318%. After disassembling the reverser to observe the raceway profile, it is found that the reverser raceway is still in the running-in wear stage. Therefore, the variance value change can only indicate the wear trend of the reverser, and cannot accurately determine whether the reverser has failed.

[0062] When the screw reaches 5.4 million revolutions, i.e. the time when the reverser fails, the pulse factor of the two screws relative to the initial value is 133.6% and 132.8%, respectively, both of which are greater than the critical value 130% of the reverser performance loss; the margin factor of the two screws relative to the initial value is 201% and 211%, respectively, both of which are greater than the critical value 180% of the reverser performance loss.

[0063] From the results, it can be seen that for the two screws, when the screw reaches 5.4 million revolutions (i.e. 30h), it is determined through the frequency domain and time domain vibration signal analysis that the reverser has failed. After disassembling the screw to observe the reverser raceway, it is found that the raceway surface is damaged and the base unquenched internal alloy material is exposed, at which time the reverser has failed.

Claims

1. A method for vibration signal analysis and diagnosis of ball screw pair reversing device faults, characterized in that, Includes the following steps: Step 1: Collect the vibration signal of the lead screw, and use wavelet packet transform to perform mode decomposition on the original signal to achieve frequency band selection of the vibration signal; Step 2: Perform frequency domain analysis and time domain analysis on the retained vibration signal: The signal retained in step 1 is decomposed using the EMD method to obtain a finite number of eigenmode functions in different frequency bands, and the ball pass frequency and amplitude are selected. The inverter's spool frequency is: In the formula, m is the number of rollers contained in each revolution of the ball screw pair; ω M ω is the angular velocity of the ball's revolution. Where D M D is the nominal diameter of the leadscrew. B L is the ball diameter; L is the lead of the leadscrew; ω is the axial speed of the leadscrew; R B R is the radius of the ball bearing; M α is the pitch circle radius of the leadscrew; α is the contact angle between the ball and the raceway. The variance, root mean square, kurtosis, impulse factor, and margin factor of the signal retained in step 1 are calculated using time-domain statistics. Step 3: Use frequency domain analysis and time domain analysis methods respectively to formulate inverter performance testing standards, and use the ratio of real-time signal to initial signal during the lead screw running-in process as the basis for judging whether the inverter performance has been lost. The frequency domain analysis method is as follows: when the amplitude at the actual ball-pass frequency during the lead screw running-in process is less than 200% of the initial amplitude, the inverter does not malfunction; when the amplitude reaches 200-250% of the initial amplitude, it indicates that the inverter has entered the severe wear stage, at which point the inverter will malfunction. The time-domain analysis method is as follows: when any of the following occurs during the use of the lead screw, the variance reaches 300% of the initial value of the lead screw itself; the impulse factor reaches 130% of the initial value; or the margin factor reaches 180% of the initial value, the performance of the inverter may degrade.

2. The vibration signal analysis and diagnosis method for ball screw pair reverser faults according to claim 1, characterized in that, The formulas for calculating variance, root mean square, kurtosis, impulse factor, and margin factor in step 2 are as follows: Root mean square: variance: kurtosis: Pulse factor: Margin factor: Where x i Let X be the amplitude of the i-th signal, n be the number of signals, and X be the amplitude of the i-th signal. max This represents the peak value of the signal.

3. The vibration signal analysis and diagnosis method for ball screw pair reverser faults according to claim 1, characterized in that, The magnitudes of variance and root mean square (RMS) represent the amount of energy contained in the vibration signal; kurtosis represents the smoothness of the signal; impulse factor represents the strength of the impact on the signal; and margin factor is used to detect the wear condition of the equipment.

4. The vibration signal analysis and diagnosis method for ball screw pair reverser faults according to claim 1, characterized in that, When collecting vibration signals from a lead screw, the influence of external factors must be considered: The effect of sensor installation position on vibration signal acquisition: By installing the sensor at different positions on the nut flange, the effect of different sensor installation positions on the amplitude of the acquired ball pass frequency was obtained; The influence of lead screw lubrication conditions on vibration signal acquisition: Data was collected on the lead screw pair under unlubricated, oil-lubricated, and grease-lubricated conditions to obtain the influence of lubrication conditions on the amplitude of the ball pass frequency. The impact of lead screw installation on vibration signal acquisition: data acquisition of vibration signals of the same lead screw at different work positions, and acquisition of vibration signals of different lead screws at the same work position, to obtain the influence of lead screw type and work position on the amplitude of the acquired ball-pass frequency. The impact of reversing on vibration signal acquisition: Vibration signals before and after reversing are acquired to determine the impact of reversing on the amplitude of the acquired ball-pass frequency.

Images