A periodic pulse feature extraction method based on vibration signal analysis

Abstract

The application discloses a periodic pulse feature extraction method based on vibration signal analysis and relates to the technical field of mechanical pulse extraction, and comprises the following steps: S1, collecting the vibration acceleration impact signal of a gear or a bearing by using an acceleration sensor and performing RMS envelope processing; S2, finding the position of each pulse center and the original signal amplitude of the corresponding position based on the envelope signal in the step S1, and then determining the pulse width through the pulse center position; the application is different from the traditional feature extraction method which calculates and statistically processes the overall vibration signal, the method focuses on the more interesting part of the signal, describes the running state information contained in the signal from the morphology, has stronger interpretability, and the extracted periodic pulse feature is a relative feature based on the signal itself, and the interference of the overall signal level is greatly weakened.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mechanical pulse extraction, and in particular to a periodic pulse feature extraction method based on vibration signal analysis. BACKGROUND

[0002] Bearing and gear as the key components in rotating machinery are widely used in aerospace, power, transportation, petrochemical and defense industry, etc. Its working environment is harsh and working conditions are complex. In the long-term operation process, the mechanical properties are continuously degraded, and faults occur frequently. Local defects of bearing and gear will produce exciting force, which is manifested as a series of periodic fault pulse impacts in the vibration signal. The periodic fault pulse signal is the key information for bearing and gear health monitoring. Therefore, effectively extracting the periodic fault pulse signal features from the actual observation signal is a difficult point in the field of mechanical fault diagnosis in recent years.

[0003] The features extracted by the existing vibration impact signal analysis technology are mostly studied for the whole vibration impact signal. The indexes such as kurtosis are extracted to characterize the impact features, or the original vibration signal is decomposed to separate multiple signal components, and then the features are extracted, which has achieved good results. However, these methods do not consider the characteristics of the fault pulse impact signal itself. The signal is described from the intuitive shape of the time domain signal impact. A periodic pulse feature extraction method based on vibration signal analysis is provided, which can accurately identify and quantitatively evaluate the periodic pulse based on the essence of the original fault periodic pulse from the morphology. SUMMARY

[0004] The present application relates to the technical field of mechanical pulse extraction, and in particular to a periodic pulse feature extraction method based on vibration signal analysis.

[0005] A periodic pulse feature extraction method based on vibration signal analysis, comprising the following steps:

[0006] S1, using an acceleration sensor to collect the vibration acceleration impact signal of the gear or bearing, and performing RMS envelope processing;

[0007] S2, based on the envelope signal in step S1, finding the position of each pulse center and the original signal amplitude corresponding to the position, and then determining the pulse width through the pulse center position;

[0008] S3, according to the determined multiple pulses, performing pulse saliency discrimination calculation to calculate the pulse average deviation level;

[0009] S4, for multiple pulses in the signal, according to the pulse interval, quantitatively discriminating and calculating the periodicity;

[0010] S5, calculate pulse sharpness index combined with pulse amplitude;

[0011] S6, obtain pulse quantitative index of self-defined vibration acceleration monitoring signal by weighted fusion of pulse significance, pulse periodicity and pulse sharpness.

[0012] In the above method for extracting periodic pulse characteristics based on vibration signal analysis, in step S1, let x(t) be defined as the collected original vibration impact signal, and the length of the original signal is:

[0013] L=length(x(t))=f s t

[0014] In the formula, f s represents the sampling rate, t represents the sampling time, and L represents the total length of the original signal. Take the RMS envelope x rms of the original signal x(t) as x rms , and the envelope data vector contains the upper envelope signal x rms_upper and x rms_lower , and the length is equal to the length of the original signal x(t).

[0015] In the above method for extracting periodic pulse characteristics based on vibration signal analysis, in step S2, it is assumed that there is one pulse in every N points in the envelope signal, and the maximum value x max in each segment of data is the pulse center. Exclude the data with a length of N / 2 on the left and right of each pulse center position in the envelope signal to obtain the pulse center position set:

[0016]

[0017] The corresponding amplitude set is:

[0018] val_set=[val1,val2,…,val k ]

[0019] The pulse width is defined as the number of points between the point on the left side of the pulse center first reaching the overall average of the envelope line and the point on the right side of the pulse first reaching the overall average of the envelope line, and the pulse width interval set is obtained:

[0020]

[0021] Wherein, the value of k is the number of pulses, which is related to the length of the original signal x(t).

[0022] In the periodic pulse feature extraction method based on vibration signal analysis, after the pulse center and the pulse width of the first pulse are determined in step S2, the envelope data in the position interval corresponding to the current pulse width is set to zero, and the updated upper and lower envelope signals are obtained by processing, and the pulse center and the pulse width of the next pulse are determined based on the updated envelope signals, and the above process is repeated.

[0023] In the periodic pulse feature extraction method based on vibration signal analysis, in step S3, the original data corresponding to the pulse width region where the k pulses are located is cut off to obtain the remaining length data The signal envelope average level of the remaining length data is calculated and normalized, and the calculation formula is as follows:

[0024]

[0025]

[0026] wherein, and are the upper envelope average value and the lower envelope average value of the remaining length data respectively, and Q1 is a pulse prominence index.

[0027] In the periodic pulse feature extraction method based on vibration signal analysis, in step S4, the k pulses obtained are arranged in ascending order of position index:

[0028]

[0029] The adjacent pulse interval is calculated and normalized:

[0030] P j =p j+1 -p j , j = 1, 2, …, k-1

[0031]

[0032]

[0033] wherein, is the average value of all adjacent pulse intervals, and Q2 is a pulse periodicity index.

[0034] In the periodic pulse feature extraction method based on vibration signal analysis, in step S5, the sharpness of the pulse as a whole is calculated based on the waveform shape and the pulse amplitude of the pulse for the vibration acceleration original signal.

[0035]

[0036] The area of each pulse surrounded by the upper and lower envelope curves is calculated and summed, and then the pulse sharpness index is calculated and normalized:

[0037]

[0038]

[0039]

[0040] Wherein, V pulse The sum of the pulse maximum values is represented by Q3, and the pulse sharpness index is represented by Q3.

[0041] In the above-mentioned periodic pulse feature extraction method based on vibration signal analysis, in step S6, the pulse significance, pulse periodicity and pulse sharpness are weighted and fused to calculate:

[0042] Q=r1Q1+r2Q2+r3Q3

[0043] Wherein, r1, r2 and r3 are adjustable coefficients, and the sum of the three coefficients is 1.

[0044] Compared with the prior art, the present application has the following advantages:

[0045] 1. Unlike the conventional feature extraction method which calculates and statistically analyzes the overall vibration signal, the present application focuses on the more interesting part of the local signal, and describes the running state information contained in the signal from the morphological aspect, and has better interpretability. The periodic pulse feature extracted by the present application is a relative feature based on the signal itself, which greatly weakens the interference of the overall signal level.

[0046] 2. The present application can quickly distinguish the periodic pulse impact from the time domain signal itself, without the need for complex signal transformation and frequency domain and time-frequency domain analysis, and has good monitoring effect on local defects, which is beneficial to subsequent fault diagnosis research work.

[0047] 3. The periodic pulse feature extraction method based on vibration signal analysis can quickly understand the fault problems of gears or bearings, and distinguish the fault types, improve the diagnosis efficiency, and can judge the component fault problems without disassembly for some gears or bearings which are difficult to disassemble. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 The present application is a periodic pulse feature extraction method based on vibration signal analysis.

[0049] Figure 2The schematic diagram for determining parameters of periodic pulse of vibration acceleration data in the application.

[0050] Figure 3 The schematic diagram for determining the maximum value of pulse and the pulse width in the periodic pulse parameters of the collected data in the embodiment of the application.

[0051] Figure 4 The different periodic pulse signals analyzed in the application.

[0052] Figure 5 The schematic diagram for comparing effects of the extracted pulse quantitative indexes of the periodic pulse signals of the gear components in different failure modes in the embodiment of the application. DETAILED DESCRIPTION

[0053] In order to make the technical means, creative features, purposes and effects achieved by the application easy to understand, the application is further described below in combination with specific embodiments.

[0054] REFERENCE Figures 1-5 As shown in the figure, a periodic pulse feature extraction method based on vibration signal analysis comprises the following steps:

[0055] S1, using an acceleration sensor to collect multi-channel vibration acceleration data of a gear or bearing monitoring, selecting a signal with strong representation for subsequent impact feature extraction, and performing RMS envelope processing;

[0056] S2, based on the envelope signal in step S1, finding the position of each pulse center and the original signal amplitude at the corresponding position, and then determining the pulse width through the pulse center position;

[0057] S3, according to the determined multiple pulses, performing pulse saliency discrimination calculation, and calculating the pulse average deviation level;

[0058] S4, for multiple pulses in the signal, quantitatively discriminating and calculating the periodicity according to the pulse interval;

[0059] S5, combining the pulse amplitude, calculating the pulse sharpness index;

[0060] S6, by weighting and fusing the pulse saliency, pulse periodicity and pulse sharpness three indexes, obtaining the pulse quantitative index of the self-defined vibration acceleration monitoring signal.

[0061] In the step S1, let x(t) be defined as the collected original vibration impact signal, and the length of the original signal is:

[0062] L=length(x(t))=f s ·t

[0063] In the formula, fs represents the sampling rate, t represents the sampling time, L represents the total length of the original signal, and x rms represents the RMS envelope of the original signal x(t) rms is the envelope data vector, which contains the upper envelope signal x rms_upper and the lower envelope signal x rms_lower , and the length thereof is equal to the length of the original signal x(t)

[0064]

[0065]

[0066] In the step S2, it is assumed that there is one pulse in every N points in the envelope signal, and the maximum value x max in each segment of data is the pulse center. The data with a length of N / 2 on the left and right sides of each pulse center position in the envelope signal is excluded to obtain a set of pulse center positions:

[0067]

[0068] The corresponding amplitude set is:

[0069] val_set=[val1,val2,…,val k ]

[0070] The pulse width is defined as the number of points between the point on the left side of the pulse center that first reaches the overall average of the envelope line and the point on the right side of the pulse that first reaches the overall average of the envelope line, and a set of pulse width intervals is obtained:

[0071]

[0072] wherein k is the number of pulses, and the value thereof is related to the length of the original signal x(t)

[0073] After the pulse center and the pulse width of the first pulse are determined, the envelope data in the position interval corresponding to the current pulse width is set to zero, and the updated upper and lower envelope signals are obtained by processing. The pulse center and the pulse width of the next pulse are determined based on the updated envelope signals, and the above is repeated. The determination of the first three pulses will be described in detail below.

[0074] For the first pulse:

[0075] In the iterative calculation, the envelope signal of the original signal is directly processed, and the average values of the upper envelope and the lower envelope are

[0076]

[0077]

[0078] The difference between the maximum value of the pulse and the minimum value of the pulse is:

[0079]

[0080] Considering the upper envelope and the lower envelope, a point with the largest amplitude is selected as the position pos1 of the pulse peak point, and the amplitude val1 is recorded. The pulse domain range is [pos1-N / 2+1, pos1+N / 2], and for the pulse domain with a width less than N / 2 on both sides, the actual number of points is measured.

[0081] In the determined pulse domain range, the upper and lower limits of the pulse actual pulse width are found. The upper limit of the pulse width is defined as the position of the point on the right side of the pulse peak point, which first reaches the overall average value of the envelope line. The lower limit of the pulse width is defined as the position of the point on the left side of the pulse peak point, which first reaches the overall average value of the envelope line. The determination of is related to the determination of the pulse peak point, and is one of and .

[0082] After determining the first pulse center and the upper and lower limits of the pulse width, the envelope data in the position interval corresponding to the current pulse width is set to zero.

[0083] Second pulse:

[0084] After removing the first pulse, the envelope signal is obtained. On this basis, in the same way, the maximum value of the remaining signal is found, and the amplitude val2 and the pulse domain range [pos2-N / 2+1, pos2+N / 2] are determined.

[0085] After determining the second pulse center and the upper and lower limits of the pulse width, the envelope data in the position interval corresponding to the current pulse width is set to zero.

[0086] Third pulse:

[0087] On the basis of the above-processed signal, the envelope signal is updated, and the position pos3, the amplitude val3, and the pulse width of the third impact pulse are determined by combining the upper and lower envelope signals. ​​

[0088] The step S3, the k pulse width region corresponding to the original data cutting off the remaining length data The remaining length data The signal envelope average level is calculated and normalized, and the formula is:

[0089]

[0090]

[0091] Wherein, And The upper envelope average value and the lower envelope average value of the remaining length data Q1 is the pulse significance index.

[0092] The step S4, the k pulse obtained according to the position subscript ascending arrangement:

[0093]

[0094] The adjacent pulse interval is calculated and normalized:

[0095] P j =p j+1 -p j J=1,2,…,k-1

[0096]

[0097]

[0098] Wherein, P is the average value of all adjacent pulse intervals, and Q2 is the pulse periodicity index.

[0099] The step S5, for the vibration acceleration original signal, the pulse is calculated from the waveform shape, combined with the pulse amplitude, the sharp degree of the whole pulse.

[0100]

[0101] The area of each pulse surrounded by the upper and lower envelope curves is calculated and summed, and then the pulse sharpness index is calculated and normalized:

[0102]

[0103]

[0104]

[0105] Wherein, V pulseThe sum of the pulse maximum values is represented, and Q3 is a pulse sharpness index.

[0106] In the step S6, the three indexes of pulse saliency, pulse periodicity, and pulse sharpness are fused and calculated by weighting:

[0107] Q = r1Q1 + r2Q2 + r3Q3

[0108] wherein, r1, r2, r3 are adjustable coefficients, the weights of the three indexes of the weighted fusion can be set by the user, and the sum of the three weight coefficients is 1, so that the pulse quantitative index meets the requirements of fault diagnosis.

[0109] Embodiment

[0110] In this embodiment, the data is collected by a rotating machinery fault test bench built in the laboratory to show the effectiveness of the algorithm. The test bench includes a servo drive motor, a gear box, a magnetic powder brake, and a set of control circuit. The data is collected by an acceleration sensor (sensitivity 100 mv / g, range 50 g). The sensor is installed on a magnetic base, and the sensor is installed in different positions and directions of the gear box. The digital signal is obtained by discrete sampling, and the sampling rate is 20 KHz. The motor speed is unknown, and the fault gear is obtained by simulation processing. The local fault is processed under laboratory conditions.

[0111] The specific steps of this embodiment are as follows:

[0112] Step one, vibration acceleration original impact vibration signal collection and envelope signal processing.

[0113] Select the acceleration sensor data with close distance to the fault gear excitation signal transmission path and large basic stiffness. The data length is 20000, and the RMS envelope processing is performed to obtain the upper and lower envelope signals x rms_upper , x rms_lower .

[0114] Step two, periodic pulse discrimination stage.

[0115] Through preliminary judgment of multiple sampling original signal waveforms, the average impact pulse number k = 3, and it is assumed that there is 1 pulse in every 6000 points, that is, N = 6000. As shown in Figure 3 , taking a sampling signal as an example, the average value of the original signal upper envelope x envlope_upper,mean = 0.4408, and the average value of the lower envelope x envlope_lower,mean = -0.4490. The three pulse subscript position sets are:

[0116] pos_set = [pos1, pos2, …, pos k ] = [1678, 8369, 14893], k = 3

[0117] Corresponding amplitude set:

[0118] val_set = [val1, val2, …, val k ] = [-3.084, -2.0328, -2.6110]

[0119] Pulse width interval set:

[0120]

[0121] Step three, pulse saliency calculation.

[0122] Remove the original data corresponding to the pulse width region where the three pulses are located, calculate the average level of the signal envelope except for the pulse impact, and normalize it.

[0123]

[0124]

[0125] Step four, pulse periodicity calculation.

[0126] Pulse position index ascending order P_set = [p1, p2, p3] = [1678, 8369, 14893], calculate the interval between adjacent impact pulses, P1 = p2-p1 = 6691, P1 = p3-p2 = 6524, and thus calculate the pulse periodicity index,

[0127]

[0128]

[0129] Step five, pulse sharpness calculation.

[0130] Sum of pulse amplitudes:

[0131]

[0132] Sum of areas of the regions where the pulses are located surrounded by the upper and lower envelope curves:

[0133]

[0134]

[0135] Step six, pulse saliency, pulse periodicity, and pulse sharpness three indicators are fused to obtain the pulse quantitative indicators of the self-defined vibration acceleration monitoring signal.

[0136] Here, the coefficients r1, r2, and r3 are selected as 0.6, 0.2, and 0.2, respectively.

[0137] Q = r1Q1 + r2Q2 + r3Q3 = 0.6Q1 + 0.2Q2 + 0.2Q3 = 0.7849

[0138] In addition, in order to verify the effect of the pulse quantitative index of the application, under laboratory conditions, gears at the same position are selected to simulate machining faults, the fault types include single-sided wear of gear teeth, double-sided wear of gear teeth and gear tooth breakage, the speed of the servo motor is kept constant, other operating conditions and environmental conditions are kept unchanged, only the faulty parts are replaced, and results are shown in Figure 5 The pulse quantitative index is extracted by collecting a plurality of groups of fault part monitoring vibration acceleration signals.

[0139] Due to the simulation machining, the gear tooth breakage machining fault position is at the corner and is relatively small, and the characteristic effect is not strong, but for the single-sided wear and double-sided wear two fault types, the fault position participates in meshing and generates a large exciting force, the trend of the two types of fault characteristic indexes is distinguished, and further support can be provided for fault diagnosis.

[0140] In summary, through the periodic pulse feature extraction method based on vibration signal analysis, the fault problem of the gear or bearing can be quickly understood, the fault type can be distinguished, the diagnosis efficiency is improved, and in application occasions, for some gear or bearing installation positions which are difficult to disassemble, if the gear or bearing is disassembled for fault inspection, it is not only troublesome, but also affects the equipment work and production efficiency, and through the method, the component fault problem can be judged without disassembling the gear or bearing.

[0141] From the technical common sense, the application can be realized through other embodiments without departing from the spirit or essential characteristics thereof. Therefore, the above disclosed embodiments are only examples, and are not the only ones. All changes within the scope of the application or within the scope equivalent to the application are included in the application.

Claims

1. A method for extracting periodic pulse features based on vibration signal analysis, characterized in that: Includes the following steps: S1. Use an accelerometer to collect vibration acceleration impact signals of gears or bearings and perform RMS envelope processing. S2. Based on the envelope signal in step S1, find the location of each pulse center and the original signal amplitude at the corresponding location, and then determine the pulse width through the pulse center location. S3. Based on the determined multiple pulses, perform pulse significance discrimination calculation and calculate the average pulse deviation level; S4. For multiple pulses in the signal, perform periodicity quantization and discrimination calculation based on the pulse interval; S5. Calculate the pulse sharpness index by combining the pulse amplitude; S6. By weighted and fused three indicators—pulse significance, pulse periodicity, and pulse sharpness—a quantitative pulse index for the custom vibration acceleration monitoring signal is obtained.

2. The method for extracting periodic pulse features based on vibration signal analysis according to claim 1, characterized in that: In step S1, let x(t) be defined as the acquired original vibration and impact signal, and the length of the original signal is: L=length(x(t))=f s ·t In the formula, f s Let represent the sampling rate, t represent the sampling time, and L represent the total length of the original signal. The RMS envelope x(t) is taken for the original signal. rms x rms The envelope data vector contains the upper envelope signal x rms_upper and x rms_lower Its length is equal to the length of the original signal x(t).

3. The method for extracting periodic pulse features based on vibration signal analysis according to claim 2, characterized in that: In step S2, it is assumed that there is one pulse within every N points in the envelope signal, and the maximum value x in each data segment is... max To find the pulse center, exclude the data of length N / 2 to the left and right of each pulse center position in the envelope signal, resulting in the set of pulse center positions: The corresponding amplitude set: val_set=[val1,val2,…,val k ] Pulse width is defined as the number of points between the point on the left side of the pulse center that first reaches the overall average value of the envelope and the point on the right side of the pulse that first reaches the overall average value of the envelope, thus obtaining the set of pulse width intervals: Here, k is the number of pulses, and its value is related to the length of the original signal x(t).

4. The method for extracting periodic pulse features based on vibration signal analysis according to claim 3, characterized in that: In step S2, after determining the pulse center and pulse width of the first pulse, the envelope data in the position interval corresponding to the current pulse width is set to zero. The updated upper and lower envelope signals are obtained through processing. The pulse center and pulse width of the next pulse will be determined under the updated envelope signal, and so on.

5. The method for extracting periodic pulse features based on vibration signal analysis according to claim 1, characterized in that: In step S3, the original data corresponding to the pulse width region of the k pulses is cut and removed to obtain the remaining length data. Calculate the remaining length data The average level of the signal envelope is calculated and normalized, using the following formula: in, and These are the remaining length data. The upper envelope mean and lower envelope mean, Q1 is the impulse significance index.

6. The method for extracting periodic pulse features based on vibration signal analysis according to claim 5, characterized in that: In step S4, the obtained k pulses are arranged in ascending order according to their position indices: Calculate the interval between adjacent pulses and normalize it: P j =p j+1 -p j ,j=1,2,…,k-1 in, Q1 is the average value of all adjacent pulse intervals, and Q2 is the pulse periodicity index.

7. The method for extracting periodic pulse features based on vibration signal analysis according to claim 6, characterized in that: In step S5, for the original vibration acceleration signal, the overall sharpness of the pulse is calculated based on its waveform shape and amplitude. Calculate and sum the areas of each pulse enclosed by the upper and lower envelope curves, then calculate the pulse sharpness index and normalize it. Among them, V pulse Q3 represents the sum of the maximum pulse values, and Q3 is the pulse sharpness index.

8. The method for extracting periodic pulse features based on vibration signal analysis according to claim 7, characterized in that: In step S6, the three indicators of pulse significance, pulse periodicity, and pulse sharpness are weighted and fused together: Q = r1Q1 + r2Q2 + r3Q3 Among them, r1, r2, and r3 are adjustable coefficients, and the sum of the three coefficients is 1.

Images