A data-driven method for extracting and analyzing single impact signals of piston pumps

Abstract

The present invention discloses a method for extracting and analyzing a single impact signal of a reciprocating plunger pump based on data-driven: according to the maximum spectral kurtosis principle, the original signal is subjected to adaptive bandpass filtering to obtain the impact signal in the original signal; according to the envelope demodulation spectrum, the speed data of the plunger pump is automatically searched by combining the secondary envelope spectrum peak with the optimal harmonic energy and algorithm; the single cycle of the plunger pump is segmented according to the speed to extract the single impact signal of each sealing valve in each large cycle; the original signal is subjected to morphological filtering, and the start and end time of the single impact of each sealing valve is determined according to the number of sealing valves in each large cycle; according to the start and end time, a periodic time-varying filter group is designed to filter the original signal, calculate the characteristic index of the single sealing valve impact signal, and determine the health status of the sealing valve according to the combination of characteristic indexes. This method can be applied to the mechanical fault judgment of the sealing ring or connecting rod of various common multi-cylinder plunger pumps.

Description

Technical Field

[0001] The invention relates to the technical field of digital signal processing, and in particular to a data-driven method for extracting and analyzing a single impact signal of a plunger pump. Background Art

[0002] Reciprocating plunger pumps rely on the reciprocating motion of the plunger in the cylinder to change the volume of the sealed working chamber to achieve the suction and discharge of liquid media. They are widely used in high pressure, large flow and flow adjustment occasions. When this plunger pump is running, the camshaft drives the piston to reciprocate in the cylinder, and each pressure relief will cause a collision in the sealing valve assembly. This cyclic reciprocating motion causes the sealing rubber ring in the plunger pump to be easily damaged by alternating loads. This sealing valve assembly has been developed in the 1980s and has been used to this day. At present, with the update of my country's oilfield operation technology, higher requirements are placed on the pressure, displacement and stability of the cement truck in the initial oil well construction process. There are many problems with the existing plunger pump sealing valve assembly in China, resulting in unstable liquid flow, high maintenance rate, easy loss of sealing rings, low pump efficiency, and even sealing rubber ring failures that lead to construction interruptions.

[0003] In response to this type of problem, many manufacturers have proposed installing sensors on the pump body to monitor the overall operating status of the sealing valve. Among the many current monitoring solutions, most use vibration sensors to collect vibration signals during the operation of the plunger pump, and judge the health status of the equipment by refining the corresponding indicators. However, among the various components of the plunger pump, the rubber ring of the sealing valve is the most susceptible to damage, and the vibration signal that can most directly reflect its status is the impact signal of the sealing valve when each single cylinder reciprocates. Therefore, the key to effectively monitoring this type of plunger pump lies in separating and extracting the impact signals of the sealing valves of different cylinders, and calculating the impact index of a single sealing valve after reorganization, so as to effectively monitor each valve and achieve the effect of early warning.

[0004] In the past 20 years, many experts and scholars have proposed various solutions for the decomposition of vibration signals and the separation and extraction of impacts. Professor Yang Guoan of Southeast University proposed the use of wavelet packet decomposition to realize the time domain feature extraction of mechanical fault signals. By decomposing the signal in different frequency bands at a certain decomposition scale, features are extracted from the reconstructed time series. However, this decomposition method is not targeted and may lead to the reconstruction of effective impact signals and noise, and is not very practical (Yang Guoan, Research on Time Domain Feature Extraction Method of Wavelet Packet Decomposition of Mechanical Fault Signals, Southeast University); Guo Yuanjing et al. used S transform to represent the signal in time and frequency, and then used SVD to denoise the fault vibration signal containing impact features. Finally, the denoised data matrix was inversely transformed to reconstruct the time domain impact features. This method uses SVD for one-dimensional signal processing. , the method can effectively enhance the signal background noise, but the extracted time domain shock is a unified waveform, and it is impossible to classify and extract each shock separately (Guo Yuanjing, S transform time spectrum SVD denoising shock feature extraction method, Zhejiang University); Zhu Yajun proposed a mechanical shock feature extraction method based on short-time Fourier transform. This method uses STFT to identify the time of the shock in the two-dimensional time-frequency surface. When using the time-frequency spectrum diagram for time and frequency positioning, it will be affected by the mutual constraints of time and frequency resolution, and it is impossible to achieve high-precision phase recognition of the shock signal and separation of different vibration source signals (Zhu Yajun, Short-time Fourier transform mechanical shock feature extraction method based on frequency domain window function, patent publication number: CN 112101245A); Invention (Zhang Jinjie, A piston engine impact feature extraction method based on EMD, Patent Publication No.: CN 110220711B) proposed a scheme for decomposing and reconstructing the engine vibration signal using the EMD method. The monitoring scheme for each cylinder of the engine is similar to the present invention, but the separation effect of the impact signal of each cylinder is not obvious when using EMD decomposition, and the scheme of installing a sensor separately for each cylinder is not applicable in many scenarios involved in the present invention.

[0005] The above-mentioned existing methods for extracting or decomposing vibration signals basically adopt strategies such as filtering noise reduction, EMD and SVD decomposition, time-frequency analysis, and impact feature calculation. All of them have one or more of the following defects and deficiencies:

[0006] (1) Most existing methods focus on signal filtering, denoising, feature extraction, etc. for a single fault body or a single vibration source, and cannot effectively separate signals from multiple impact sources.

[0007] (2) In the absence of prior knowledge such as rotation speed and phase, the fully data-driven approach poses a challenge to existing analysis methods. The lack of rotation speed input leads to deviations, losses, or misidentifications in the positioning of each impact.

[0008] (3) Directly using relevant eigenvalue calculation and separation algorithms, such as wavelet decomposition and SVD decomposition, will result in the inability to close the loop of impacts at each moment. It is easy to achieve separation in the entire frequency domain dimension, but extraction in the time dimension cannot be completed.

[0009] (4) Using the distribution characteristics of the time-frequency domain to calculate the number of shocks and obtain the start and end positions of the shocks will be limited by the resolution of the time-frequency analysis. The resolutions of the two dimensions restrict each other, affecting the accuracy and effect of the analysis.

[0010] (5) Some of the methods currently proposed require the installation of a separate sensor at the equivalent position of each cylinder. After data collection, the data at each measuring point is analyzed independently and monitored in a single manner. This is difficult to achieve in the scenario designed by the present invention, and the hardware cost is high. The impact data of a single cylinder will also fluctuate with changes in working conditions and media, and a longitudinal comparison with historical data will be more effective than a horizontal comparison. Summary of the invention

[0011] In view of the problems and shortcomings of the current methods, the present invention provides a data-driven method for extracting and analyzing a single impact signal of a plunger pump.

[0012] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0013] A data-driven method for extracting and analyzing a single impact signal of a plunger pump comprises the following steps:

[0014] S1. Collect the original signal generated by the impact of the sealing valve when the plunger pump reciprocates, design an adaptive envelope demodulation band search algorithm based on the maximum spectrum kurtosis principle, and perform adaptive bandpass filtering on the original signal to obtain the impact signal in the original signal;

[0015] S2. According to the envelope demodulation spectrum, the current real-time speed data of the plunger pump is automatically searched by combining the secondary envelope spectrum peak with the optimal harmonic energy and the adaptive envelope demodulation frequency band search algorithm;

[0016] S3, dividing the single cycle of the plunger pump according to the speed data, and calculating the noise floor of the impact signal to extract the single impact signal of each sealing valve in each large cycle;

[0017] S4, performing morphological filtering on the original signal, and determining the start and end time of a single impact of each sealing valve according to the number of sealing valves in each large cycle;

[0018] S5. According to the start and end times, a periodic time-varying filter group is designed to filter the original signal to obtain a single sealing valve impact signal, calculate the characteristic index of the single sealing valve impact signal, and judge the health status of the sealing valve according to the combination of the characteristic indexes.

[0019] A further improvement is that the specific operation of step S1 is:

[0020] S1-1, s(t)(t∈[0,T])) represents the original input signal, and firstly designs the adaptive envelope demodulation band search algorithm of the signal according to the maximum spectral kurtosis principle: construct a tree search structure, and the frequency band corresponding to each node in the structure is calculated according to the predetermined sampling frequency and the number of nodes in each layer;

[0021] S1-2, then perform bandpass filtering on s(t) according to the filtering parameters of each node to obtain the filtered signal s corresponding to each node f (t); for s f (t) performing envelope demodulation to obtain the envelope signal corresponding to each node and calculating its kurtosis;

[0022] S1-3. Select the envelope frequency band corresponding to all nodes with the largest kurtosis as the optimal filter, and perform bandpass filtering on the original signal to obtain the impact signal in the original signal.

[0023] A further improvement is that the specific operation of step S2 is:

[0024] S2-1. Calculate the filtered signal s using Hilbert transform f (t) envelope signal:

[0025]

[0026] S2-2, perform Fourier transform to obtain the spectrum distribution of the envelope signal The envelope spectrum is further subjected to Fourier transform to obtain the inverse envelope spectrum P s (t);

[0027] S2-3, Automatic search P s (t) of the (2n-1)th order harmonic energy and maximum value, where n is the number of cylinders of the plunger pump; the maximum harmonic energy and the corresponding fundamental period T r That is the real-time speed data at the pump output end. Theoretically, there will be 2n impacts in each cycle.

[0028] A further improvement is that the specific operation of step S3 is:

[0029] S3-1. According to the speed data and the number of piston pump cylinders, the structural element expression is g(k)(k∈[0,T r / 20n]), where T r / 20n is the scale of the structural element;

[0030] S3-2. The mathematical morphological expansion operation result of the original signal is calculated using the structural element to extract the single impact signal of each sealing valve in each large cycle.

[0031] A further improvement is that in step S3-2, a straight line structural element is used and the expansion operation formula is: Where i represents the result of digital discretization of continuous signal time; based on morphological filtering, twice the signal amplitude corresponding to the maximum number of crossings is calculated as the optimal noise level N Level ; The signal sequence (s⊕g)(i) after morphological filtering, all less than N Level Set the point of zero to get S g (i)((s⊕g)(i)>N Level ).

[0032] A further improvement is that the specific operation of step S4 is:

[0033] S4-1. Calculate S g (i) The largest front T / T r (Round up) impact peak value, according to T r And the principle of optimal main impact position (interval is T r , assuming that each shock is the main shock, the number of main shocks searched is the largest), and the peak position belonging to the main shock is obtained;

[0034] S4-2. Determine the main shock starting time t of all time ranges T from the main shock peak value obtained. peak1 (j) and the end time t peak2 (j);

[0035] S4-3, in each base period T r Inside, with T r / 2n is the search cycle, and S is searched in each search cycle. g (i) Maximum impact S g (i') as the effective impact in the search period, and obtain the corresponding time as the starting time t of the impact peak1 (i') and the end time t peak2 (i');

[0036] S4-4, thus each base period T r The internal search obtains 2n-1 impact start and end times, and all main impact times are combined to obtain the start and end times t of all single impacts of the sealing valves in the original signal. peak1 and t peak2 ;

[0037] Where T is the signal period, T r is the base period.

[0038] A further improvement is that the specific operation of step S5 is: according to the start and end time information t peak1 and t peak2 , a time domain signal processing filter for a single sealing valve is designed. The filter is configured as follows: when the impact signal of one of the sealing valves is obtained, the filter is used to weaken the impact strength of other valves. The signal obtained by processing retains the impact characteristics of the current sealing valve. For the impact signal of each sealing valve, the characteristic index of the single sealing valve impact signal is calculated, and the health status of the sealing valve is judged according to the combination of characteristic indicators.

[0039] A further improvement is that the characteristic indicators include kurtosis, high-frequency energy and high-frequency frequency center of gravity.

[0040] The beneficial effects of the present invention are:

[0041] (1) The present invention proposes an effective separation method for multi-source impact signals, which can realize the separation and extraction of impacts from different vibration sources in a unified collected signal. The processed signal type is not limited, and can be a vibration sound signal of a device, or other digital signals such as sound collected by a microphone;

[0042] (2) The invention solves the problem that most existing methods can only achieve filtering, denoising, feature extraction, etc. for a single fault body or a single vibration source, and is more universally applicable to scenarios such as multi-stroke engines and plunger pumps;

[0043] (3) The invention proposes a method for adaptively extracting the rotation speed of impact equipment based on the calculation of the inverse envelope spectrum with the optimal spectral kurtosis. This method is proposed for the first time and can achieve the acquisition of the rotation speed information contained in the signal completely based on data drive without the rotation speed information.

[0044] (4) A two-step extraction strategy for shocks is proposed. First, the main shock component is obtained once in each large cycle, and then the sub-shock is searched in each large cycle, which greatly improves the phase accuracy of shock acquisition.

[0045] (5) The start and end times of the impact are determined by using morphological filtering. The parameters of the morphological filter are set adaptively according to the signal without the need for prior knowledge, which ensures the accuracy of the calculation of the start and end times of a single impact and avoids the time-frequency resolution limitation commonly encountered when using time-frequency distribution analysis.

[0046] (6) The invention proposes a method for designing a time domain filter group for different impact sources. The method designs an amplitude compression filter based on the information of the amplitude level of the non-impact signal and the envelope contour of the original signal. The designed filter group is a new type of full-sequence time series filter without singular point divergence.

[0047] (7) The method proposed in the present invention can be used in various application scenarios for separating single impact signals with multiple impact sources, which solves the limitation of the existing method that a sensor needs to be installed separately for each impact source, thereby reducing hardware costs; and the separated signal eliminates mutual interference from other vibration sources, thereby improving analysis accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 The present invention is a flow chart of the method involved;

[0049] Figure 2 is the original waveform of the impact signal in the example and its spectrum diagram;

[0050] Figure 3 It is a schematic diagram of the calculation results of the filtered kurtosis of different search nodes in the example;

[0051] Figure 4 Schematic diagram of envelope demodulation signal, envelope spectrum and inverse envelope spectrum under optimal filtering parameters;

[0052] Figure 5 This is a schematic diagram of the original signal morphology filtering and noise threshold calculation results;

[0053] Figure 6 Schematic diagram of the main impact position extraction result;

[0054] Figure 7 Design schematic diagram for all impact start and end times and timing filter for a single valve body;

[0055] Figure 8 This is the result of separating the impact signal of a single valve body. DETAILED DESCRIPTION

[0056] The present application is further described in detail below in conjunction with the accompanying drawings. It is necessary to point out here that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technical personnel in this field can make some non-essential improvements and adjustments to the present application based on the above application content.

[0057] like Figure 1 As shown, a data-driven method for extracting and analyzing a single impact signal of a plunger pump comprises the following steps:

[0058] S1. Collect the original signal generated by the impact of the sealing valve when the plunger pump reciprocates, design an adaptive envelope demodulation band search algorithm based on the maximum spectrum kurtosis principle, and perform adaptive bandpass filtering on the original signal to obtain the impact signal in the original signal;

[0059] S2. According to the envelope demodulation spectrum, the current real-time speed data of the plunger pump is automatically searched by combining the secondary envelope spectrum peak with the optimal harmonic energy and the adaptive envelope demodulation frequency band search algorithm;

[0060] S3, dividing the single cycle of the plunger pump according to the speed data, and calculating the noise floor of the impact signal to extract the single impact signal of each sealing valve in each large cycle;

[0061] S4, performing morphological filtering on the original signal, and determining the start and end time of a single impact of each sealing valve according to the number of sealing valves in each large cycle;

[0062] S5. According to the start and end times, a periodic time-varying filter group is designed to filter the original signal to obtain a single sealing valve impact signal, calculate the characteristic index of the single sealing valve impact signal, and judge the health status of the sealing valve according to the combination of the characteristic indexes.

[0063] The specific implementation method and algorithm process of this method are described in detail below in conjunction with actual processing routines.

[0064] (i) Install the acceleration sensor on the surface of the equipment to be monitored, sample at a certain sampling frequency, and obtain the vibration signal x(i). In this example, a three-cylinder pump is used, that is, n=3 in the above parameters, the sampling frequency Fs=12.8kHz, and the number of sampling points N=65536. Figure 2 The original signal shown in the figure is used as an example to further illustrate the implementation steps of the present invention. The above figure is the time domain waveform of the original signal. Figure 2 Obvious periodic impact can be seen in the figure, which is caused by the impact of the sealing valve when the plunger pump reciprocates; Figure 2 The lower part is its corresponding spectrum. It can be seen from the spectrum of the signal that both the speed cycle and the impact cycle in the original signal are difficult to be found in the point frequency band. This is because the frequency is modulated to the high frequency band and cannot be distinguished in the spectrum.

[0065] (ii) For the original signal x(i), a tree search structure is constructed to search the filtering results of different frequency bands to obtain the frequency band with the largest kurtosis. The number of child nodes in the kth layer of the tree search structure is designed to be:

[0066]

[0067] Where NLevel is the number of tree levels of the tree search structure. In this example, NLevel=12 is set. Thus, in the structure of the kth layer, the lth (0≤l≤N k The filtering frequency band of ) basis points is:

[0068]

[0069] In the above formula, Fs is the original signal sampling frequency, FB k,l is the filter frequency band used by the lth search node of the kth layer. At each node position, the original signal is filtered using the corresponding filter. The filter uses a 7th-order ButterWorth filter. The kurtosis of the filtered signal is calculated and the calculation results are plotted as shown in Figure 3 As shown in the figure, the darker the color, the larger the value, and the lighter the color, the smaller the value. It can be seen that the optimal filter band is [3200Hz, 4798.7Hz]. Using this filter as the filter parameter of the Hilbert transform, the envelope signal and envelope spectrum are obtained as follows: Figure 4 As shown in (1, 2), the envelope spectrum is further subjected to Fourier transform to obtain the inverse envelope spectrum P x (t) Figure 4 (3).

[0070] Automatic search P s The fifth-order harmonic energy and maximum value of (t) (the plunger pump used in the example is a three-cylinder pump, that is, the above parameter n = 1), the maximum harmonic energy and the corresponding fundamental period T r =0.2662s, which is the speed period of the plunger pump, and the corresponding speed frequency is 3.7566Hz.

[0071] (III) Constructing structural elements g(k), T r / 20n is the scale of the structural element. The scale of the discrete signal is 57 according to the sampling rate. The original signal x(i) is expanded using the structural element. The result is as follows: Figure 5 (1) As shown; Based on morphological filtering, the signal amplitude corresponding to the maximum number of crossings is calculated, and the optimal noise level N is obtained. Level like Figure 5 (1) The gray horizontal line indicates that all the filtered values ​​less than N Level Set the point of zero to get S g (i)((s⊕g)(i)>N Level )like Figure 5 (2).

[0072] (IV) According to T r = 0.2662s, signal period T = 5.12s, calculate S g (i) The first 20 peaks, Figure 5 In each impact interval in (2), the rising edge time is extracted as the impact start time, and the peak sequence obtained is as follows: Figure 6 (1), the top 20 largest peaks are marked in gray in the figure. Each peak is taken as the main impact position, according to T rBy calculating the number of effective main shocks, it is found that the number of effective main shocks for 18 peaks is 17, and the number of effective main shocks for the other two peaks is 1. Therefore, 18 of the above 20 peaks are real main shock peaks.

[0073] Search the entire morphological filter after S g (i) sequence, supplement the missing main shocks, and finally obtain all the main shock peaks as follows Figure 6 (2) Solid line position, the black solid line is the supplementary main shock position, and the black dashed line is the ineffective main shock among the first 20 peaks.

[0074] (V) Determine the start and end time of the main impact within the time range T from the main impact peak value obtained; in each basic period T r Within (from the start time of one main impact to the start time of the next main impact), T r / 6 is the search cycle, and S is searched in each small cycle. g (i) Maximum impact S g (i') is taken as the effective shock in this small cycle, and S g (i) The rising and falling edges of the noise level are taken as the starting time t of the impact peak1 (i') and the end time t peak2 (i').

[0075] After completing all searches, each base cycle T r The five impact start and end times are searched and combined with all the main impact times to obtain the start and end times t of all the impact signals from the sealing valve in the original signal. peak1 and t peak2 like Figure 7 (1) As shown in the figure: the rising edge of the black square wave is the starting time of each effective impact in the base period, and the falling edge is the ending time; the rising edge of the gray square wave is the starting time of the main impact, and the falling edge is the ending time.

[0076] (VI) Using Hilbert transform to calculate the envelope signal of the original signal At the same time, calculate all the impact moments (t peak1 <t<t peak2 ) is set to zero, and a time domain filter array for each sealing valve impact signal is constructed. The construction method is as follows:

[0077] Construct an array of all 1s, for the impact signal of sealing valve 1, at all impact moments 2-6, so that the value of the array is The newly formed array is the filter array for sealing valve 1, and the other five filter groups are similar. Figure 7 (2, 3) show the filter array construction results for sealing valves 1 and 2.

[0078] The original signal is processed by six filter groups respectively to obtain the impact separation signal for each sealing valve as follows: Figure 8 As shown in (1-6), it can be seen that the impact signal of each sealing valve is well separated, which is convenient for targeted analysis of each object and calculation of the effective value, kurtosis, pulse and other indicator data of a single valve body, so as to realize independent monitoring and diagnostic analysis of each valve body.

[0079] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A data-driven method for extracting and analyzing a single impact signal of a plunger pump, characterized in that the steps include: S1. Collect the original signal generated by the impact of the sealing valve when the plunger pump reciprocates, design an adaptive envelope demodulation band search algorithm based on the maximum spectrum kurtosis principle, and perform adaptive bandpass filtering on the original signal to obtain the impact signal in the original signal; S2. According to the envelope demodulation spectrum, the current real-time speed data of the plunger pump is automatically searched by combining the secondary envelope spectrum peak with the optimal harmonic energy and the adaptive envelope demodulation frequency band search algorithm; S3, dividing the single cycle of the plunger pump according to the speed data, and calculating the noise floor of the impact signal to extract the single impact signal of each sealing valve in each large cycle; S4, performing morphological filtering on the original signal, and determining the start and end time of a single impact of each sealing valve according to the number of sealing valves in each large cycle; S5. According to the start and end times, a periodic time-varying filter group is designed to filter the original signal to obtain a single sealing valve impact signal, calculate the characteristic index of the single sealing valve impact signal, and judge the health status of the sealing valve according to the combination of the characteristic indexes.

2. The data-driven method for extracting and analyzing a single impact signal of a plunger pump according to claim 1, characterized in that: The specific operations of step S1 are: S1-1, s(t)(t∈[0,T])) represents the original input signal, and firstly designs the adaptive envelope demodulation band search algorithm of the signal according to the maximum spectral kurtosis principle: construct a tree search structure, and the frequency band corresponding to each node in the structure is calculated according to the predetermined sampling frequency and the number of nodes in each layer; S1-2, then perform bandpass filtering on s(t) according to the filtering parameters of each node to obtain the filtered signal s corresponding to each node f (t); for s f (t) performing envelope demodulation to obtain the envelope signal corresponding to each node and calculating its kurtosis; S1-3. Select the envelope frequency band corresponding to all nodes with the largest kurtosis as the optimal filter, and perform bandpass filtering on the original signal to obtain the impact signal in the original signal.

3. The data-driven method for extracting and analyzing a single impact signal of a plunger pump according to claim 2, characterized in that: The specific operations of step S2 are: S2-1. Calculate the filtered signal s using Hilbert transform f (t) envelope signal: S2-2, perform Fourier transform to obtain the spectrum distribution of the envelope signal The envelope spectrum is further subjected to Fourier transform to obtain the inverse envelope spectrum P s (t); S2-3, Automatic search P s (t) of the (2n-1)th order harmonic energy and maximum value, where n is the number of cylinders of the plunger pump; the maximum harmonic energy and the corresponding fundamental period T r That is the real-time speed data at the pump output end.

4. The data-driven method for extracting and analyzing a single impact signal of a plunger pump according to claim 3 is characterized in that: The specific operations of step S3 are: S3-1. According to the speed data and the number of piston pump cylinders, the structural element expression is g(k)(k∈[0,T r / 20n]), where T r / 20n is the scale of the structural element; S3-2. The mathematical morphological expansion operation result of the original signal is calculated using the structural element to extract the single impact signal of each sealing valve in each large cycle.

5. The data-driven method for extracting and analyzing a single impact signal of a plunger pump according to claim 4, characterized in that: In step S3-2, a straight line structural element is used, and the expansion operation formula is: Where i represents the result of digital discretization of continuous signal time; based on morphological filtering, twice the signal amplitude corresponding to the maximum number of crossings is calculated as the optimal noise level N Level ; The signal sequence after morphological filtering (s⊕g)(i), all less than N Level Set the point of zero to get S g (i)((s⊕g)(i)>N Level ).

6. The data-driven method for extracting and analyzing a single impact signal of a plunger pump according to claim 5, characterized in that: The specific operations of step S4 are: S4-1. Calculate S g (i) The largest front T / T r The impact peak value is based on T r And the principle of the optimal main impact position, the peak position belonging to the main impact is obtained; S4-2. Determine the main shock starting time t of all time ranges T from the main shock peak value obtained. peak1 (j) and the end time t peak2 (j); S4-3, in each base period T r Inside, with T r / 2n is the search cycle, and S is searched in each search cycle. g (i) Maximum impact S g (i') as the effective impact in the search period, and obtain the corresponding time as the starting time t of the impact peak1 (i') and the end time t peak2 (i'); S4-4, thus each base period T r The internal search obtains 2n-1 impact start and end times, and all main impact times are combined to obtain the start and end times t of all single impacts of the sealing valves in the original signal. peak1 and t peak2 ; Where T is the signal period, T r is the base period.

7. The data-driven method for extracting and analyzing a single impact signal of a plunger pump according to claim 6, characterized in that: The specific operation of step S5 is: according to the start and end time information t peak1 and t peak2 , a time domain signal processing filter for a single sealing valve is designed. The filter is configured as follows: when the impact signal of one of the sealing valves is obtained, the filter is used to weaken the impact strength of other valves. The signal obtained by processing retains the impact characteristics of the current sealing valve. For the impact signal of each sealing valve, the characteristic index of the single sealing valve impact signal is calculated, and the health status of the sealing valve is judged according to the combination of characteristic indicators.

8. A data-driven method for extracting and analyzing a single impact signal of a plunger pump according to claim 1 or 7, characterized in that: The characteristic indicators include kurtosis, high-frequency energy and high-frequency frequency center of gravity.

Images