Fault diagnosis method and device for plunger pump

Abstract

The invention provides a fault diagnosis method and device for a plunger pump. The fault diagnosis method for the plunger pump comprises the steps that S1, pressure pulsation data of an oil port of the plunger pump are obtained; s2, dividing the pressure pulsation data into a plurality of period intervals, wherein the period intervals comprise a first period interval and other period intervals; s3, by taking the first period interval as a reference, respectively calculating interval similarity features between each of the other period intervals and the first period interval; and S4, based on the interval similarity features, fault diagnosis is carried out on the plunger pump. According to the fault diagnosis method for the plunger pump, fault diagnosis can be efficiently carried out on the plunger pump.

Description

Technical Field

[0001] The present invention relates to the field of plunger pumps, and particularly to a fault diagnosis method and device for a plunger pump. Background Art

[0002] Plunger pumps have the characteristics of high working pressure, high volumetric efficiency, and long service life, and are widely used in various construction machinery.

[0003] However, during the long-term operation of a plunger pump, faults may occur, and the construction machinery cannot be used normally. How to quickly and timely analyze the faults of the plunger pump is a problem to be solved. Summary of the Invention

[0004] Aiming at the above problems of the prior art, the purpose of the present invention is to provide a fault diagnosis method for a plunger pump, which can efficiently diagnose the faults of the plunger pump.

[0005] To solve the above problems, on the one hand, the present invention provides a fault diagnosis method for a plunger pump, and the fault diagnosis method for the plunger pump includes:

[0006] Step S1, obtaining the pressure pulsation data of the oil port of the plunger pump;

[0007] Step S2, dividing the pressure pulsation data into multiple cycle intervals, and the cycle intervals include a first cycle interval and other cycle intervals;

[0008] Step S3, taking the first cycle interval as a reference, and respectively calculating the interval similarity features between each of the other cycle intervals and the first cycle interval;

[0009] Step S4, performing fault diagnosis on the plunger pump based on the interval similarity features.

[0010] Further, the step S2 includes:

[0011] Step S21, analyzing the pressure pulsation data based on the fast Fourier transform, so as to divide out multiple preliminary cycle intervals;

[0012] Step S22, identifying the local peak points of the pressure pulsation data based on the significance analysis algorithm, and adjusting the preliminary cycle intervals based on the local peak points, so as to divide them into multiple precise cycle intervals.

[0013] Further, the significance analysis algorithm obtains the local peak points based on the significance calculation formula, and the significance formula is shown as the following formula (1):

[0014] Prominence = x[p] - max(v left , v right) (1)

[0015] Among them, x[p] is the height of the peak point, and v left is the height of the lowest point found to the left of the peak point, and v right is the height of the lowest point found to the right of the peak point. Prominence represents the significance of the peak point, and the point with the maximum significance is taken as the local peak point.

[0016] Furthermore, the plunger pump includes M plungers, and the M plungers reciprocate in sequence. M is an integer greater than or equal to 2.

[0017] In the step S2, each of the cycle intervals is divided into M sub-cycles, and the M sub-cycles correspond one-to-one to the M plungers.

[0018] In the step S3, the interval similarity feature of the cycle interval is determined through the similarity features of the corresponding sub-cycles between the other cycle interval and the first cycle interval.

[0019] Furthermore, the step S3 includes:

[0020] Data that does not contain complete sub-cycles in the initial pressure pulsation data is removed. The linear cross-correlation algorithm is used to calculate the similarity features of each sub-cycle in the other cycle interval and the corresponding sub-cycle in the first cycle interval, so as to obtain M similarity features, and the maximum value among the M similarity features is taken as the interval similarity feature.

[0021] Furthermore, the first cycle interval is expressed as: x = {x(0), x(1), … x(L - 1)}; the other cycle interval is expressed as: y = {y(0), y(1), … y(N - 1)}, N ≥ L, and the linear cross-correlation algorithm is as shown in the following formula (2):

[0022]

[0023] Among them, x(k) is the value of the first cycle interval at the k-th sampling point, the template length is L, and the range of k is 0 to L - 1; y(n) is the value of the other cycle interval at the n-th sampling point, and the range of n is 0 to N - 1; τ is the translation (time delay / hysteresis) index, and the range of τ is 0 to N - L. τ = 0 means that x is aligned with the first L points of y; j is the summation index; D(τ) is the sliding dot product under τ, which reflects the similarity.

[0024] Furthermore, the linear cross-correlation algorithm is normalized as shown in the following formula (3):

[0025]

[0026] ρ(τ) is the normalized correlation coefficient. A value of -1 indicates a perfect negative correlation, a value of +1 indicates a perfect positive correlation, and a value of 0 indicates no correlation.

[0027] Further, the step S4 includes:

[0028] Analyze the interval similarity and the pressure pulsation data based on the transformer model, so as to perform fault diagnosis on the plunger pump.

[0029] Further, the step S4 includes:

[0030] Step S41, form a two-dimensional array based on each of the cycle intervals and the corresponding interval similarity features;

[0031] Step S42, obtain a learnable fault category label and incorporate the fault category label into the two-dimensional array to form a data sequence;

[0032] Step S43, input the data sequence into the transformer network for processing and perform normalization processing through the softmax function, so as to perform fault diagnosis on the plunger pump.

[0033] On the other hand, the present invention provides a fault diagnosis device for a plunger pump, and the fault diagnosis device for the plunger pump includes:

[0034] An acquisition device for acquiring pressure pulsation data at the oil port of the plunger pump;

[0035] A cycle interval division device for dividing the pressure pulsation data into a plurality of cycle intervals, where the cycle intervals include a first cycle interval and other cycle intervals;

[0036] A similarity feature calculation device for calculating the interval similarity features between each of the other cycle intervals and the first cycle interval based on the first cycle interval;

[0037] A fault diagnosis device for performing fault diagnosis on the plunger pump based on the interval similarity features.

[0038] Due to the above technical solutions, the present invention has the following beneficial effects:

[0039] According to the plunger pump fault diagnosis method of the present invention, pressure pulsation data reflecting the hydrodynamic characteristics at the oil port of the plunger pump is acquired, the pressure pulsation data is divided into a plurality of cycle intervals through the periodicity of the plunger pump operation, the interval similarity features between other cycle intervals and the first cycle interval are calculated based on the well-operating first cycle interval, and fault diagnosis is performed on the plunger pump based on this. Thus, fault diagnosis of the plunger pump can be efficiently performed. Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0041] Figure 1 is a flowchart of a fault diagnosis method for a plunger pump according to an embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of a cycle interval according to an embodiment of the present invention;

[0043] Figure 3 is a schematic diagram of a normalized correlation coefficient according to an embodiment of the present invention;

[0044] Figure 4 is a schematic diagram of a fault diagnosis device for a plunger pump according to an embodiment of the present invention. Detailed Embodiments

[0045] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0046] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion.

[0047] Next, a fault diagnosis method for a plunger pump according to an embodiment of the present invention will be described.

[0048] As Figures 1 to 3 shown, the fault diagnosis method for a plunger pump according to an embodiment of the present invention includes:

[0049] Step S1, obtaining pressure pulsation data at the oil port of the plunger pump.

[0050] For example, by detecting the pressure at the oil port of the piston pump with a pressure sensor, pressure pulsation data can be obtained, thereby enabling the acquisition of the pressure pulsation data at the oil port of the piston pump.

[0051] After the piston pump has been operating for some time, when the internal components fail due to wear, the leakage flow pulsation will change, which in turn affects the overall flow pulsation. The flow pulsation is converted into pressure pulsation through the pipeline impedance and load, and the pressure pulsation data thus reflects the hydrodynamic characteristics of the piston pump, especially the high-pressure oil port. High-frequency flow pulsation data is often difficult to measure directly, while high-frequency pressure pulsation data is relatively easy to obtain. Based on this, the pressure pulsation data of the piston pump is used as the original input data for fault diagnosis in this application.

[0052] Step S2: Divide the pressure pulsation data into multiple cycle intervals, where the cycle intervals include a first cycle interval and other cycle intervals. Among them, the second cycle interval, the third cycle interval, the fourth cycle interval... are all called other cycle intervals.

[0053] The piston pump has reciprocating motion and is periodic. Each cycle interval corresponds to a complete action of the piston pump. By dividing the cycle intervals, a complete action of the piston pump can be analyzed, which is convenient for fault diagnosis from the complete action of the piston pump.

[0054] For example, the pressure pulsation data can be divided into multiple cycle intervals based on FFT (Fast Fourier Transform).

[0055] Step S3: Taking the first cycle interval as a reference, calculate the interval similarity features between each other cycle interval and the first cycle interval respectively.

[0056] When the new equipment of the piston pump is initially operating or after maintenance and debugging, the operating state of the first cycle interval is good, and this is used as the reference for this piston pump. By calculating the interval similarity features between each other cycle interval and the first cycle interval respectively, the changes of other cycle intervals relative to the first cycle interval after the piston pump has been operating for some time can be obtained.

[0057] Step S4: Based on the interval similarity features, conduct fault diagnosis on the piston pump.

[0058] For example, if the similarity difference is large, it indicates that the wear of the piston pump is very large and a fault has occurred. By analyzing the specific interval similarity features, various faults of the piston pump can be diagnosed.

[0059] The above plunger pump fault diagnosis method obtains pressure pulsation data reflecting hydrodynamic characteristics at the oil ports of the plunger pump, divides the pressure pulsation data into multiple periodic intervals according to the periodicity of the plunger pump operation, takes the first periodic interval with good operation as a reference, calculates the interval similarity characteristics between other periodic intervals and the first periodic interval, and diagnoses the faults of the plunger pump based on this. Thus, the fault diagnosis of the plunger pump can be efficiently carried out.

[0060] In some embodiments of the present invention, step S2 includes: step S21, analyzing the pressure pulsation data based on the fast Fourier transform to divide out multiple preliminary periodic intervals; step S22, identifying the local peak points of the pressure pulsation data based on the significance analysis algorithm, and adjusting the preliminary periodic intervals based on the local peak points to divide them into multiple accurate periodic intervals.

[0061] First, analyze the pressure pulsation data (signal spectrum) through the fast Fourier transform (FFT) to determine the main frequency components of the pressure pulsation data. Based on these main frequencies, multiple preliminary periodic intervals can be divided.

[0062] Next, among these preliminary periodic intervals, it is necessary to identify the local peak points of the pressure pulsation data, and these local peak points mark the start and end of the periodic intervals. To ensure that only one representative local peak point is selected in each periodic interval and to ensure the consistency of these local peak points in terms of period, mathematical significance analysis can be used. This analysis considers the relative height difference between each local peak point and its adjacent lowest point, as well as the distance between the maximum point and the adjacent maximum point, so as to exclude the errors introduced by noise or non-periodic fluctuations.

[0063] In addition, by performing frequency band filtering on the expected periodic frequencies, it is possible to further ensure that only the periodic components within the predefined frequency range are focused on, thereby improving the accuracy and repeatability of the division of the periodic intervals.

[0064] It should be noted that the above are only optional examples. It is also possible to directly identify the local peak points of the pressure pulsation data based on the significance analysis algorithm and divide the pressure pulsation data into multiple accurate periodic intervals based on the local peak points. All of these should be understood to be within the scope of the present invention.

[0065] Further, the significance analysis algorithm obtains the local peak points based on the significance calculation formula, and the significance formula is shown as formula (1) below:

[0066] Prominence=x[p]-max(v left , v right ) (1)

[0067] Wherein, x[p] is the peak point height, vleft is the lowest point height found to the left from the peak point, v right is the lowest point height found to the right from the peak point, Prominence represents the significance of the peak point, and the point with the maximum significance is taken as the local peak point.

[0068] Through the significance analysis algorithm, the local peak points can be accurately found, which is convenient for the accurate division of the cycle interval. For example, it forms Figure 2 the cycle interval.

[0069] In some embodiments of the present invention, the plunger pump includes M plungers, and the M plungers reciprocate in sequence. M is an integer greater than or equal to 2.

[0070] In step S2, each cycle interval is divided into M sub-cycles, and the M sub-cycles correspond one-to-one with the M plungers. In step S3, the interval similarity feature of the cycle interval is determined by the similarity features of the corresponding sub-cycles between other cycle intervals and the first cycle interval.

[0071] For example, a complete motion cycle of the plunger pump includes nine sub-cycles generated by the reciprocation of nine plungers. By analyzing the pressure pulsation data of these nine sub-cycles, the performance of a single plunger can be evaluated, and the correlation change of the overall system can also be analyzed.

[0072] It should be noted that the above are only optional examples, and the number of plungers of the plunger pump is not limited here.

[0073] Further, step S3 includes removing the data in the initial pressure pulsation data that does not contain complete sub-cycles, and calculating the similarity features of each sub-cycle in other cycle intervals and the corresponding sub-cycle in the first cycle interval by using the linear cross-correlation algorithm, so as to obtain M similarity features, and taking the maximum value among the M similarity features as the interval similarity feature.

[0074] The initial pressure pulsation data is often incomplete (does not include complete sub-cycles) and cannot reflect accurate periodic information. Therefore, these incomplete initial pressure pulsation data need to be discarded during the calculation process.

[0075] For example, the similarity features of the nine sub-cycles in other cycle intervals and the corresponding nine sub-cycles in the first cycle interval are calculated one by one to obtain nine similarity features, and the maximum value among the nine similarity features is used as the interval similarity feature. Thus, the interval similarity feature can be obtained quickly, and the situations of each plunger are comprehensively considered.

[0076] Further, the first period interval is represented as: x = {x(0), x(1), … x(L-1)}; other period intervals are represented as: y = {y(0), y(1), … y(N-1)}, N ≥ L, and the linear cross-correlation algorithm is as shown in Equation (2) below:

[0077]

[0078] Among them, x(k) is the value of the first period interval at the k-th sampling point, the template length is L, and the range of k is 0 to L-1; y(n) is the value of the period interval at the n-th sampling point, and the range of n is 0 to N-1; τ is the translation (time delay / hysteresis) index, and the range of τ is 0 to N-L. τ = 0 means that the first L points of x and y are aligned; k is the summation index (the summation index within the window, and the summation is performed for the template length L in a loop); D(τ) is the sliding dot product under τ, which reflects the similarity (reflects the linear similarity between the template and the window signal, and the value changes with the amplitude).

[0079] L is the number of sample points (window length) of the first period interval x of the template, L << N; N is the number of sample points of other period intervals to be detected (the total length of the pressure pulsation data).

[0080] Thus, the similarity characteristics of each sub-period in other period intervals corresponding to the sub-period of the first period interval can be accurately calculated.

[0081] Further, the linear cross-correlation algorithm is normalized as shown in Equation (3) below:

[0082]

[0083] ρ(τ) is the normalized correlation coefficient. Its value equal to -1 indicates complete negative correlation, its value equal to +1 indicates complete positive correlation, and its value equal to 0 indicates no correlation.

[0084] Thus, the similarity characteristics can be clearly and explicitly represented. After obtaining the normalized result, to extract the interval similarity characteristics and simplify the input of the deep learning model (such as the transformer model below), a single correlation coefficient is extracted from the normalized array to represent the overall correlation between other period intervals and the first period interval. The maximum value method is selected to represent the maximum correlation between the two period intervals at the best alignment (i.e., the best relative time delay of the period intervals).

[0085] In some embodiments of the present invention, step S4 includes analyzing the interval similarity and pressure pulsation data based on the transformer model, so as to perform fault diagnosis on the plunger pump.

[0086] Due to the design of its self-attention mechanism, the Transformer model is very suitable for processing time series data. In the fault diagnosis of piston pumps, the periodic interval (motion cycle signal) of each piston is a type of time series data, and the Transformer can effectively capture the long-distance dependencies in this sequence.

[0087] Further, step S4 includes: step S41, forming a two-dimensional array based on each periodic interval and the corresponding interval similarity feature; step S42, obtaining a learnable fault category label and incorporating the fault category label into the two-dimensional array to form a data sequence; step S43, inputting the data sequence into the Transformer network for processing and performing normalization through the softmax function to diagnose the faults of the piston pump.

[0088] For example, first, form a two-dimensional array with each periodic interval and the corresponding interval similarity feature, and map the two-dimensional array to a high-dimensional embedding space through a linear projection layer. Then, obtain a learnable fault category label (class token) and merge it with the two-dimensional array embedding vector to form an unordered token embedding data sequence containing the fault category label. Next, input the data sequence into the Transformer network, then pass through a fully connected layer, and then enter the softmax function to perform the normalization operation to obtain the fault category result.

[0089] It should be noted that since the input of the Transformer itself does not contain sequence order information, position encoding is introduced to distinguish the positions of the elements in the sequence. The position encoding is implemented using alternating sine and cosine functions.

[0090] Next, the fault diagnosis device 1000 of the piston pump according to the embodiment of the present invention will be described.

[0091] As Figure 4 shown, the fault diagnosis device of the piston pump according to the embodiment of the present invention includes an acquisition device 1001, a periodic interval division device 1002, a similarity feature calculation device 1003, and a fault diagnosis device 1004. The acquisition device 1001 is used to acquire the pressure pulsation data of the oil port of the piston pump. The periodic interval division device 1002 is used to divide the pressure pulsation data into multiple periodic intervals, and the periodic intervals include a first periodic interval and other periodic intervals. The similarity feature calculation device 1003 is used to calculate the interval similarity features between each other periodic interval and the first periodic interval based on the first periodic interval. The fault diagnosis device 1004 is used to diagnose the faults of the piston pump based on the interval similarity features.

[0092] It should be noted that, when the device provided in the above embodiments realizes its functions, only the division of the above-mentioned functional modules is used for illustration. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device provided in the above embodiments and the corresponding method embodiments belong to the same concept. For the specific implementation process, please refer to the corresponding method embodiments and will not be elaborated here.

[0093] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A fault diagnosis method for a plunger pump, characterized in that, The described fault diagnosis method includes: Step S1, obtaining the pressure pulsation data of the oil port of the piston pump; Step S2, dividing the pressure pulsation data into multiple cycle intervals, where the cycle intervals include a first cycle interval and other cycle intervals; Step S3, taking the first cycle interval as a reference, calculating the interval similarity features between each of the other cycle intervals and the first cycle interval respectively; Step S4, performing fault diagnosis on the piston pump based on the interval similarity features.

2. The fault diagnosis method of the plunger pump according to claim 1, characterized in that The step S2 includes: Step S21, analyzing the pressure pulsation data based on the fast Fourier transform to divide out multiple preliminary cycle intervals; Step S22, identifying the local peak points of the pressure pulsation data based on the significance analysis algorithm, and adjusting the preliminary cycle intervals based on the local peak points to divide them into multiple precise cycle intervals.

3. The fault diagnosis method of the plunger pump according to claim 2, wherein, The significance analysis algorithm obtains the local peak points based on the significance calculation formula, and the significance formula is shown as the following formula (1): Prominence=x[p]-max(v left ,v rig ) (1) Among them, x[p] is the peak point height, v left is the lowest point height found to the left from the peak point, v rig is the lowest point height found to the right from the peak point, Prominence represents the significance of the peak point, and the point with the maximum significance is taken as the local peak point.

4. The fault diagnosis method of the plunger pump according to claim 1, characterized in that The piston pump includes M pistons, and the M pistons reciprocate in sequence, where M is an integer greater than or equal to 2. In the step S2, each cycle interval is divided into M sub-cycles, and the M sub-cycles correspond one-to-one with the M pistons. In the step S3, the interval similarity feature of the cycle interval is determined through the similarity features of the corresponding sub-cycles between the other cycle interval and the first cycle interval.

5. The fault diagnosis method of the plunger pump according to claim 4, wherein, The step S3 includes: Removing the data in the initial pressure pulsation data that does not contain complete sub-cycles, and using the linear cross-correlation algorithm to calculate the similarity features of each sub-cycle in the other cycle interval and the corresponding sub-cycle in the first cycle interval, thereby obtaining M similarity features, and taking the maximum value among the M similarity features as the interval similarity feature.

6. The fault diagnosis method of the plunger pump according to claim 5, characterized in that, The first cycle interval is represented as: x = {x(0), x(1), … x(L - 1)}; the other cycle interval is represented as: y = {y(0), y(1), … y(N - 1)}, N ≥ L, and the linear cross-correlation algorithm is shown as the following formula (2): Where x(k) is the value of the first cycle interval at the k-th sampling point, the template length is L, and the range of k is 0 to L - 1; y(n) is the value of the other cycle interval at the n-th sampling point, and the range of n is 0 to N - 1; τ is the translation (time delay / hysteresis) index, and the range of τ is 0 to N - L, τ = 0 means that the first L points of x and y are aligned; k is the summation index; D(τ) is the sliding dot product under τ, reflecting the similarity.

7. The fault diagnosis method of the plunger pump according to claim 6, characterized in that, Normalize the linear cross-correlation algorithm, as shown in the following formula (3): ρ(τ) is the normalized correlation coefficient, whose value equal to -1 indicates complete negative correlation, whose value equal to +1 indicates complete positive correlation, and whose value equal to 0 indicates no correlation.

8. The fault diagnosis method of the plunger pump according to claim 2, wherein, The step S4 includes: Analyzing the interval similarity and the pressure pulsation data based on the transformer model to perform fault diagnosis on the piston pump.

9. The fault diagnosis method of the plunger pump according to claim 8, characterized in that, The step S4 includes: Step S41: Form a two-dimensional array based on each of the cycle intervals and the corresponding interval similarity features; Step S42: Obtain learnable fault class labels and incorporate the fault class labels into the two-dimensional array to form a data sequence; Step S43: Input the data sequence into a transformer network for processing and perform normalization processing through a softmax function to diagnose faults of the plunger pump.

10. A fault diagnosis device for a plunger pump, characterized in that, The fault diagnosis device includes: An acquisition device for acquiring pressure pulsation data at the oil port of the plunger pump; A cycle interval division device for dividing the pressure pulsation data into a plurality of cycle intervals, where the cycle intervals include a first cycle interval and other cycle intervals; A similarity feature calculation device for calculating, based on the first cycle interval, the interval similarity features between each of the other cycle intervals and the first cycle interval; A fault diagnosis device for diagnosing faults of the plunger pump based on the interval similarity features.

Images