A fracturing pump fluid end operation abnormality diagnosis system

Abstract

This invention provides a diagnostic system for abnormal operation of the hydraulic end of a fracturing pump, belonging to the field of fault diagnosis technology for oil and gas field fracturing equipment. The system divides the operating cycle using the crankshaft rotation phase as an anchor point, and eliminates the impact of data fluctuations caused by drastically changing operating conditions through normalization of operating conditions. A dual residual comparison chain of health status and trend is constructed through a real-time calibrated multi-physics field positive coupling simulation model. Abnormal extreme values ​​are identified through column-row dual-dimensional fitting analysis, and the anomaly confidence level is calculated by scenario-specific analysis based on fault mechanisms. The system then matches the fault mechanism library to output comprehensive diagnostic results. This invention can achieve early warning of hydraulic end degradation and precise fault location, with high diagnostic accuracy and low false positive rate, and is suitable for complex operating conditions in fracturing sites.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology for oil and gas field fracturing equipment, and in particular to a fracturing pump hydraulic end operation abnormality diagnosis system. Background Technology

[0002] Fracturing pumps are core equipment in oil and gas reservoir fracturing operations. Their function is to inject high-volume, high-viscosity fracturing fluid under pressure into the well, thereby creating fractures in the reservoir and increasing the productivity of oil and gas wells. The hydraulic end of the fracturing pump is the core pressure-bearing module that directly contacts the high-pressure fracturing fluid and completes the fluid intake-pressurization-discharge process. Operating under harsh conditions of high pressure, strong wear, and strong corrosion for extended periods, components such as the intake valve, discharge valve, plunger seal, and pump head are highly susceptible to wear, leakage, and fatigue cracks. Failure to diagnose and warn of these issues in a timely manner can lead to interruptions in fracturing operations and even safety accidents such as high-pressure fluid ejection.

[0003] Currently, the main technical deficiencies in the diagnosis of anomalies at the hydraulic end of fracturing pumps are as follows: Existing diagnostic methods mostly use fixed threshold alarms, which can only identify severe anomalies after a fault occurs, and cannot provide early warning of gradual deterioration. Moreover, fracturing pump field operations are subject to highly variable conditions, with displacement, pressure, and stroke frequency adjusted in real time. Fixed thresholds are prone to false positive alarms, resulting in extremely low diagnostic accuracy.

[0004] Some diagnostic methods based on simulation models have the problem of reversed simulation input and output logic. They directly input the measured operating response data into the simulation model without real-time model calibration. As a result, the simulation results have no clear physical meaning and cannot distinguish the data differences caused by operating condition fluctuations and equipment failures.

[0005] Existing diagnostic methods based on data comparison lack established period division and alignment rules based on phase anchoring. The operating parameters of the hydraulic end of the fracturing pump are strongly correlated with the crank angle phase. Without phase reference time alignment, cross-period data becomes incomparable, and the comparison results have no diagnostic value. At the same time, the comparison results of adjacent periods have logical contradictions in their attribution, making it impossible to construct a continuous comparison chain.

[0006] Existing fitting analysis methods suffer from serious dimensional conflicts, merging parameters with different physical dimensions for fitting, resulting in fitting results with no physical meaning; extreme value elements lack clear statistical criteria, and the calculation of anomaly confidence lacks fault mechanism and statistical support, leading to subjective assignment of values ​​and extremely high rates of missed and false faults, especially for the most common field faults such as valve group wear and seal leakage, which are progressive deterioration faults.

[0007] Therefore, this invention proposes a diagnostic system for abnormal operation of the hydraulic end of a fracturing pump. Summary of the Invention

[0008] This invention provides a diagnostic system for abnormal operation of the hydraulic end of a fracturing pump, in order to solve the aforementioned technical problems.

[0009] This invention provides a diagnostic system for abnormal operation of the hydraulic end of a fracturing pump, comprising: The data acquisition module is used to simultaneously acquire multi-source operation monitoring data and reference phase data from the hydraulic end of the fracturing pump, and to preprocess the acquired raw data to obtain a standardized time-series dataset. The period splitting module is used to split the standardized time series dataset into several sets of continuous and phase-synchronized running period data, and perform normalization processing to obtain a normalized period dataset. The simulation matrix construction module is used to perform real-time parameter calibration on the pre-built multi-physics positive coupling simulation model of the hydraulic end of the fracturing pump based on the boundary condition parameters of the normalized period dataset of the current operating cycle, so as to obtain the calibration simulation model of the current cycle. At this time, the phase-synchronized plunger displacement time series is used as the input of the calibration simulation model, and the output is the simulation working matrix corresponding to the current operating cycle. The comparison chain construction module is used to perform a first comparison between the simulation working matrix of the current period and the standard working matrix corresponding to the same standard working condition interval to obtain the health residual matrix. At the same time, it performs a second comparison between the simulation working matrix of the current period and the simulation working matrix of the previous adjacent running period of the same standard working condition interval to obtain the trend residual matrix, and constructs the comparison chain. The phase alignment module is used to split the comparison chain with the construction slug cycle of the fracturing pump as the reset cycle to obtain the split sub-chain corresponding to each reset cycle, and to perform phase alignment on all the running cycle data in the split sub-chain with the theoretical alignment time in the running cycle as the alignment benchmark to obtain the alignment sub-chain, to extract the residual data of all running cycles in the alignment sub-chain under the same theoretical alignment time, and to construct the combination matrix of the theoretical alignment time. The confidence determination module is used to perform column-dimensional trend fitting analysis on the combined matrix, and at the same time, perform row-dimensional feature fitting analysis on the combined matrix, and calculate the abnormal confidence level corresponding to each column element variable in the combined matrix according to different scenarios. The anomaly analysis module is used to traverse the anomaly confidence levels corresponding to all theoretical alignment times. Based on the temporal variation law of the anomaly confidence levels and the phase distribution characteristics of the corresponding health residual matrix, it matches the pre-built hydraulic end fault mechanism library and outputs the anomaly diagnosis results of the hydraulic end of the fracturing pump.

[0010] Preferably, each running cycle is used as a node, and the corresponding health residual matrix and trend residual matrix are combined into a comparison array for that cycle. All comparison arrays are then concatenated in the order of running time to construct a comparison chain.

[0011] Preferably, the confidence level determination module includes: The first extreme value identification submodule is used to perform column dimension trend fitting analysis on the combined matrix, obtain the column fitting residual corresponding to each column vector, and identify the first extreme value element that exceeds the first preset residual threshold. The second extreme value identification submodule is used to perform row dimension feature fitting analysis on the combined matrix, obtain the row fitting residual corresponding to each row vector, and identify the second extreme value element that exceeds the second preset residual threshold. The frequency statistics submodule is used to count the first frequency of variables in the same column being identified as first extreme value elements and the second frequency of variables being identified as second extreme value elements in the combination matrix. The scenario-based calculation submodule is used to calculate the anomaly confidence level corresponding to the element variable in the column based on the sum of the first frequency and the second frequency, and in combination with the physical quantity fault weight and extreme value deviation amplitude corresponding to the element variable in the column.

[0012] Preferably, the scenario-based calculation submodule includes: The first calculation unit is used to calculate the maximum deviation amplitude and residual coefficient of variation of all extreme elements of the column variable when the sum of the first frequency and the second frequency is ≥2. Combined with the fault weights of the physical quantities corresponding to the column variable, it calculates the anomaly confidence level. : ,in, The anomaly confidence level has a value range of [0,1]. The fault weight of the physical quantity corresponding to the element variable in this column has a value range of [0,1]. That is, the correlation between the physical quantity corresponding to the element variable in this column and the fault is obtained based on the fault mechanism library. This is the proportionality coefficient; The maximum deviation is the sum of the absolute values ​​of the differences between the extreme elements and the fitted values ​​and the standard deviation of the residuals of that column vector. The ratio; Let be the coefficient of variation of the column vector residuals, and This is the mean of the residuals of the column vector; The second calculation unit is used to calculate the deviation multiple of the extreme value element when the sum of the first frequency and the second frequency is 1, and to match the deviation multiple with a preset multiple-confidence table to obtain the anomaly confidence level. Wherein, the deviation factor is the sum of the absolute value of the difference between the measured value and the corresponding fitted value of the extreme value element. The ratio of the threshold; Meanwhile, if the slope of the fitting trend of the column vector corresponding to the element variable of this column remains the same for n01 consecutive running cycles and the absolute value continues to increase, it is determined that there is a gradual deterioration trend, and the confidence is adjusted according to min(abnormal confidence + 0.2, 0.8), where 0.2 is the warning increment; The third calculation unit is used to determine if, when the sum of the first frequency and the second frequency is 0, the slope of the fitting trend of the column vector corresponding to the element variable in that column remains the same for n02 or more consecutive running cycles and the absolute value continues to increase, that there is an early gradual deterioration trend, and the abnormal confidence level is assigned to 0.2, triggering a yellow early warning; otherwise, it is determined to be a normal state, and the abnormal confidence level is assigned to 0.

[0013] Preferably, the preprocessing includes: validity verification, noise reduction, and phase resampling; The rule for dividing the operating cycle is as follows: the crank angle of 0° corresponding to the top dead center of the piston displacement in the reference phase data is taken as the starting point of the cycle, and the crank angle of 360° is taken as the ending point of the cycle. This corresponds to the piston completing one complete intake stroke and exhaust stroke. The crank angle of 0°-180° is the intake stroke, and the crank angle of 180°-360° is the exhaust stroke. The number of sampling points for phase resampling in the preprocessing is: The value is a positive integer, and each sampling point corresponds to a theoretical alignment time. The theoretical alignment time is a crank angle phase node that is equally divided within the operating cycle. The crank angle corresponding to the i-th theoretical alignment time is... .

[0014] Preferably, the periodic splitting module includes: The feature extraction submodule is used to extract the operating condition feature parameters of each set of operating cycle data, wherein the operating condition feature parameters include: the average number of strokes and the average discharge main pipe pressure within the operating cycle; The dimension construction submodule is used to construct a two-dimensional operating condition grid with the average number of strokes as the first dimension and the average discharge manifold pressure as the second dimension. Each grid corresponds to a standard operating condition range. The correction submodule is used to perform dimensionless correction on the pressure and vibration amplitude in the operating cycle data based on the parameter ratio between the current standard operating condition range and the rated operating condition range, so as to obtain a normalized cycle dataset.

[0015] Preferably, the multi-source operation monitoring data includes: single-cylinder pump head body high-pressure chamber pressure, suction main pipe pressure, discharge main pipe pressure, suction valve vibration acceleration, discharge valve vibration acceleration, and hydraulic oil temperature; The reference phase data includes: crankshaft angle signal and piston displacement signal.

[0016] Preferably, the measured peak cylinder pressure and valve group vibration frequency in the normalized periodic dataset of the current operating cycle are used as calibration targets. The weighted least squares method is used to optimize the parameters of the valve group damping coefficient, equivalent leakage coefficient, and fluid bulk modulus of the simulation model to obtain the calibration simulation model for the current cycle.

[0017] Compared with the prior art, the beneficial effects of this application are as follows: 1. This invention proposes a dual-benchmark data processing method of phase anchoring and operating condition normalization. The method uses the crank angle phase as the core benchmark for period division and alignment, which solves the industry pain point of incomparability of cross-period data under different operating conditions of fracturing pumps. Through operating condition interval division and dimensionless correction, the method effectively eliminates the data differences caused by normal operating condition fluctuations and significantly reduces the probability of false positive alarms.

[0018] 2. This invention constructs a feature extraction method based on real-time calibration forward simulation and dual residual matrix coupling. Based on a forward multiphysics field coupled simulation model, it performs real-time calibration using measured boundary parameters, ensuring the physical meaning and accuracy of the simulation results and solving the core defect of reversed simulation input and output logic in existing technologies. Through the coupled analysis of health residual matrix and trend residual matrix, it can identify both sudden faults and capture the trend characteristics of early progressive degradation, solving the problem of complete failure of existing technologies for progressive faults.

[0019] 3. This invention proposes a two-dimensional anomaly identification method based on column trend fitting and row feature fitting. Column fitting targets the cross-period variation trend of the same physical quantity, while row fitting targets the coupling relationship of multiple physical quantities at the same time. This avoids fitting conflicts between different physical quantities, and the fitting results have clear physical and statistical significance. The criteria for identifying extreme elements ensure the accuracy of anomaly identification.

[0020] 4. This invention constructs an anomaly confidence level calculation method based on fault mechanism and statistics, which abandons the defects of subjective assignment in the existing technology. It combines physical quantity fault weight, deviation amplitude, and trend characteristics to calculate confidence level in different scenarios. At the same time, it sets up a progressive deterioration early warning mechanism, which effectively reduces the fault missed detection rate. Through feature matching of the fault mechanism library, it realizes the whole process diagnosis from abnormal alarm to fault location, type identification, and severity assessment, which can guide on-site maintenance.

[0021] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0022] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0023] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a structural diagram of a fracturing pump hydraulic end operation anomaly diagnosis system according to an embodiment of the present invention; Figure 2 This is a structural diagram of the confidence determination module in an embodiment of the present invention. Detailed Implementation

[0024] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0025] This invention provides a diagnostic system for abnormal operation of the hydraulic end of a fracturing pump, such as... Figure 1 As shown, it includes: The data acquisition module is used to simultaneously acquire multi-source operation monitoring data and reference phase data from the hydraulic end of the fracturing pump, and to preprocess the acquired raw data to obtain a standardized time-series dataset. In this embodiment, the preprocessing includes: validity verification, noise reduction, and phase resampling. The validity verification process is as follows: Based on preset range thresholds (pressure sensor range 0-120MPa, accelerometer range 0-500g), invalid data exceeding the range is removed; based on data continuity thresholds, disconnected or lost data with no change for more than 10 consecutive sampling points is removed, and the removed data segments are completed using linear interpolation to ensure data continuity and temporal integrity.

[0026] The noise reduction process is as follows: a wavelet thresholding method with a 5-level decomposition based on the db5 wavelet basis is used, with the following settings: Wavelet basis selection criteria: The db5 wavelet is a Daubechies tightly supported orthogonal wavelet with a vanishing moment of 5. It has high time domain resolution for transient impact signals and can accurately preserve the impact characteristics of valve group opening and closing. At the same time, it has excellent filtering effect on electromagnetic interference and smooth noise transmitted from mechanical vibration at the power end. It is a universal preferred wavelet basis for noise reduction of reciprocating fluid machinery vibration signals. The decomposition level setting is based on the following: sampling frequency of 10kHz, Nyquist frequency of 5kHz, and frequency range of each band after 5-level wavelet decomposition: Level 1 detail factor: 2.5kHz~5kHz (high frequency electromagnetic interference noise); Level 2 detail factor: 1.25kHz~2.5kHz (noise transmitted by mechanical vibration at the power end); Level 3 detail factor: 0.625kHz~1.25kHz (high-frequency harmonics of valve assembly vibration); Level 4 detail factor: 0.3125kHz~0.625kHz (valve assembly vibration fundamental frequency harmonics); Level 5 detail coefficient + approximation coefficient: 0~0.3125kHz (valve opening and closing fundamental frequency, effective signal of cylinder pressure fluctuation); Five-layer decomposition can filter out high-frequency invalid noise, retain the effective characteristic signals of hydraulic end operation, and avoid signal distortion and excessive computation caused by too many decomposition layers.

[0027] The phase resampling process is as follows: using the 0°-360° rotation angle signal collected by the crankshaft angle encoder as a reference, the timing data is resampled at equal intervals. 360 sampling points are set in each operating cycle, with each 1° crank rotation angle being a sampling point, to obtain a standardized timing dataset with the same number of sampling points and one-to-one correspondence between phase nodes in each operating cycle.

[0028] In this embodiment, sensors installed on the fracturing pump synchronously collect multi-source operational monitoring data and reference phase data, with a sampling frequency set to 10kHz. The multi-source operational monitoring data includes: dynamic pressure signals from the high-pressure chambers of the five single-cylinder pump heads, pressure signals from the intake manifold and discharge manifold, vibration acceleration signals from the intake and discharge valves of each cylinder, and hydraulic oil temperature signals. The reference phase data includes: crankshaft angle signals and piston displacement signals. The sensor configuration includes: High-pressure chamber dynamic pressure sensors: 5, installed in the high-pressure chambers of 5 single-cylinder pump heads respectively, with a range of 0-120MPa, an accuracy of 0.2 grade, and a sampling frequency of 10kHz; Main pipe pressure sensors: 2, installed in the intake main pipe and the discharge main pipe respectively, with measuring ranges of 0-1MPa and 0-120MPa, accuracy of 0.2 grade, and sampling frequency of 10kHz; Triaxial accelerometer: 10 units, one installed on the valve cover of each cylinder's intake valve and exhaust valve, measuring range 0-500g, resolution 0.001g, sampling frequency 20kHz; Temperature sensors: 2, installed in the hydraulic oil chamber and the intake manifold respectively, with a range of -40℃ to 120℃, an accuracy of 0.5℃, and a sampling frequency of 1kHz; Absolute encoder: 1 unit, installed at the crankshaft power input end, with a resolution of 0.1° and a sampling frequency of 10kHz, used to acquire crankshaft angle signals; Five laser displacement sensors, each corresponding to one of the five plungers, with a range of 0-400mm, an accuracy of 0.1mm, and a sampling frequency of 10kHz, are used to acquire plunger displacement signals.

[0029] It should be noted that all sensors achieve hard synchronization acquisition through a synchronous acquisition card, ensuring the time consistency of multi-source data and phase data; the acquired data is transmitted in real time to the on-board diagnostic industrial control computer via industrial Ethernet, and the diagnostic system runs in real time within the industrial control computer.

[0030] The period splitting module is used to split the standardized time series dataset into several sets of continuous and phase-synchronized running period data, and perform normalization processing to obtain a normalized period dataset. In this embodiment, the operating cycle is divided as follows: the crank angle of 0° corresponding to the top dead center of the plunger displacement in the reference phase data is taken as the starting point of the cycle, and the crank angle of 360° is taken as the ending point of the cycle. This corresponds to the plunger completing one complete suction stroke and discharge stroke. Specifically, the crank angle of 0°-180° is the suction stroke, in which the plunger moves away from the working chamber to complete the liquid suction process; the crank angle of 180°-360° is the discharge stroke, in which the plunger moves deeper into the working chamber to complete the pressurized liquid discharge process.

[0031] The simulation matrix construction module is used to perform real-time parameter calibration on the pre-built multi-physics positive coupling simulation model of the hydraulic end of the fracturing pump based on the boundary condition parameters of the normalized period dataset of the current operating cycle, so as to obtain the calibration simulation model of the current cycle. At this time, the phase-synchronized plunger displacement time series is used as the input of the calibration simulation model, and the output is the simulation working matrix corresponding to the current operating cycle. In this embodiment, a multiphysics positive coupling simulation model of the hydraulic end of the fracturing pump is pre-constructed based on the actual structural parameters of the hydraulic end of the fracturing pump (cylinder diameter 100mm, plunger diameter 100mm, valve seat diameter 80mm, valve cavity volume 2.5L, etc.). It is constructed by combining fluid mechanics, rigid body dynamics, and fluid-structure interaction theory, and includes three core modules: a plunger motion rigid body dynamics module, a valve group opening and closing fluid-structure interaction module, and a fracturing fluid compressibility hydraulic module. The input parameters of this model include: plunger displacement timing, suction manifold pressure, discharge manifold pressure, fracturing fluid density, viscosity, and bulk modulus; the output parameters include: cylinder high-pressure chamber pressure timing, suction valve opening timing, discharge valve opening timing, suction valve vibration acceleration timing, discharge valve vibration acceleration timing, and plunger axial stress timing, totaling six core physical quantities.

[0032] In this embodiment, the steps for real-time parameter calibration are as follows: Boundary parameter input: The measured intake manifold pressure, discharge manifold pressure, and density and viscosity of fracturing fluid corresponding to the temperature in the normalized periodic dataset are used as the boundary input parameters of the simulation model; Calibration objectives: The calibration objectives are the measured peak pressure in the cylinder and the main frequency of valve group vibration, and the objective function is to minimize the mean square error between the simulation output and the measured data. Optimization parameters: The parameters to be optimized are the valve group damping coefficient c and the equivalent leakage coefficient. Fluid equivalent bulk modulus ; Optimization algorithm: The least squares method is used for parameter optimization, including: Constructing parameter vectors with optimization ; The calibration benchmark is based on the normalized measured data of the current operating cycle, which includes three core benchmark quantities: To obtain the timing data of the pressure in the high-pressure chamber inside the cylinder, This represents the number of phase resampling points within a single operating cycle, and ; The peak value of the measured cylinder pressure sequence is the maximum pressure value during the pressurization phase in a single cycle. The dominant frequency of the vibration signals of the intake and exhaust valves was extracted using Fast Fourier Transform (FFT). The parameter vector to be optimized The phase-synchronized plunger displacement timing and measured boundary condition parameters (suction manifold pressure, discharge manifold pressure, fracturing fluid properties) are input into the multiphysics forward coupling simulation model to obtain the corresponding simulation output: This is for simulating the timing data of in-cylinder pressure; The peak value of the simulated cylinder pressure timing; To simulate the main frequency of vibration signals from the intake and exhaust valves.

[0033] The basic objective function is constructed with minimizing the sum of squared residuals between the simulation output and the measured benchmark data as the core objective.

[0034]

[0035] in, The rated discharge pressure of the fracturing pump is used for dimensionless processing to eliminate the influence of pressure amplitude differences under different operating conditions on the fitting results and ensure the consistency of fitting weights under all operating conditions. This is the sub-item for the full-time fitting of in-cylinder pressure; This is a sub-term for fitting the peak pressure inside the cylinder; This is a sub-item for fitting the dominant frequency of valve group vibration; These are weighting coefficients, with values ​​of 0.7, 0.2, and 0.1 respectively. It should be noted that full-time matching of in-cylinder pressure is the core foundation for phase synchronization and residual matrix calculation, and is related to 100% of hydraulic end faults. The highest; the peak pressure inside the cylinder is the most sensitive characteristic of leakage and seal failure faults, and is one of the core calibration targets. Secondly, the main vibration frequency of the valve assembly is an auxiliary characteristic of valve assembly wear and jamming faults, used to supplement the calibration of the valve assembly's dynamic characteristics. Minimum.

[0036] Convergence criterion: During the iteration process, based on the objective function... Furthermore, when the goodness of fit between the simulated cylinder pressure timing and the measured data is greater than or equal to 0.95, the calibration is considered to have converged, and the calibration simulation model for the current cycle is obtained. If convergence is still not achieved after 20 iterations, the calibration parameters of the previous adjacent cycle under the same operating condition are used as the parameters for the current cycle, and a calibration anomaly message is output. It should be noted that if the maximum number of iterations exceeds 25, the single calculation time will exceed 1 second, which cannot meet the real-time requirements. If the maximum number of iterations is less than 15, the convergence success rate under extreme operating conditions will drop below 92%. Therefore, 20 iterations are set as the maximum number of iterations to balance the convergence success rate and real-time calculation efficiency.

[0037] The piston motion rigid body dynamics module describes the relationship between piston displacement, velocity, acceleration, and crank angle. The governing equations are:

[0038] in, This represents the piston displacement. Where is the crank radius. For crank angle, The length of the link. The speed of the piston movement. The crankshaft angular velocity, This is the acceleration of the piston movement.

[0039] The valve assembly opening and closing fluid-structure interaction module describes the variation of cylinder pressure with piston movement and valve assembly opening and closing. It is based on the fluid continuity equation, and the governing equation is: ,in, The pressure inside the cylinder. This is the equivalent bulk modulus of fracturing fluid. This refers to the clearance volume within the cylinder. The cross-sectional area of ​​the plunger. The flow rate at the suction valve inlet. Discharge valve flow rate This is the equivalent leakage flow rate between the valve assembly and the plunger seal.

[0040] The fracturing fluid compressibility hydraulic module describes the motion of the valve disc under the action of fluid forces and spring forces. It is based on rigid body dynamics equations, and the governing equations are: For valve disc quality, For valve opening, The pressure difference across the valve disc. This refers to the pressure-bearing area of ​​the valve disc. For valve spring stiffness, This is the pre-compression amount of the valve spring. This represents the damping coefficient of the valve assembly.

[0041] In this embodiment, the phase-synchronized plunger displacement timing is input into the calibration simulation model, and the simulation working matrix corresponding to the current operating cycle is output. The simulation working matrix is ​​a two-dimensional matrix with 360 rows and 6 columns. The rows correspond to 360 theoretical alignment times (1°-360° phase nodes), and the columns correspond to 6 physical quantities output by the simulation. The matrix element A(m5,n5) is the simulation output value of the m5th phase node and the n5th physical quantity.

[0042] It should be noted that if the current cycle is the first operating cycle within the construction block cycle, and there is no preceding adjacent cycle with the same working conditions, only the first comparison is performed to generate the health residual matrix. The trend residual matrix is ​​assigned a zero matrix of 360 rows and 6 columns. This zero matrix does not participate in subsequent extreme value identification, frequency statistics and confidence calculation. It is only used to store the simulation working matrix of the first cycle as the second comparison benchmark for the second operating cycle to ensure the complete construction of the comparison chain.

[0043] The comparison chain construction module is used to perform a first comparison between the simulation working matrix of the current period and the standard working matrix corresponding to the same standard working condition interval to obtain the health residual matrix. At the same time, it performs a second comparison between the simulation working matrix of the current period and the simulation working matrix of the previous adjacent running period of the same standard working condition interval to obtain the trend residual matrix, and constructs the comparison chain. In this embodiment, the dimension definition of the standard working matrix is: completely consistent with the simulation working matrix, which is a two-dimensional matrix of 360 rows × 6 columns. The rows correspond to 360 theoretical alignment times, and the columns correspond to: cylinder high pressure chamber pressure, intake valve opening, exhaust valve opening, intake valve vibration acceleration, exhaust valve vibration acceleration, and plunger axial stress. Generation rules: Before a new fracturing pump leaves the factory, a full-scale calibration test is conducted in each standard operating condition range. The measured boundary condition parameters under healthy conditions are collected, input into a multiphysics field forward coupling simulation model, and after parameter calibration, the standard working matrix of the corresponding standard operating condition range is output and stored in the diagnostic system. Update rules: After each pump head body overhaul, valve assembly replacement, and plunger seal replacement, a calibration test is performed again, and the standard working matrix for the corresponding standard operating condition range is updated to ensure the validity of the benchmark.

[0044] In this embodiment, the dimensions and physical meanings of rows and columns of the standard working matrix and the simulation working matrix are completely consistent. It is the reference matrix output by the calibration simulation model when the new hydraulic pump is in a healthy state at the factory under the corresponding standard working condition range.

[0045] Perform a double comparison operation for the current running cycle: First comparison: Compare the simulation working matrix of the current cycle with the standard working matrix corresponding to the same standard working condition interval, and calculate the difference between corresponding elements in the same dimension to obtain the health residual matrix. This matrix has the same dimensions as the simulation working matrix, where, in, This is the simulation working matrix for the current cycle. For the standard working matrix under the same working conditions, The corresponding element in the health residual matrix represents the degree of deviation between the current operating state and the health standard state.

[0046] The second comparison involves calculating the difference between corresponding elements in the same dimension of the simulation working matrix for the current cycle and the simulation working matrix for the previous adjacent cycle under the same standard operating conditions, thus obtaining the trend residual matrix. This matrix has the same dimensions as the simulation working matrix, where, , This is the simulation working matrix for the previous adjacent cycle under the same operating condition. The corresponding element in the trend residual matrix represents the trend of the current operating state relative to the previous period.

[0047] Using each running cycle as a node, the corresponding health residual matrix and trend residual matrix are combined into a comparison array for that cycle, and all comparison arrays are concatenated in chronological order. A comparison chain is constructed, where each comparison array uniquely corresponds to a running cycle, and the time order is completely consistent with the order of the running cycles.

[0048] The phase alignment module is used to split the comparison chain with the construction slug cycle of the fracturing pump as the reset cycle to obtain the split sub-chain corresponding to each reset cycle, and to perform phase alignment on all the running cycle data in the split sub-chain with the theoretical alignment time in the running cycle as the alignment benchmark to obtain the alignment sub-chain, to extract the residual data of all running cycles in the alignment sub-chain under the same theoretical alignment time, and to construct the combination matrix of the theoretical alignment time. In this embodiment, the slug injection command issued by the fracturing construction control system is used as the trigger signal. The start of slug injection is the starting point of the reset cycle, and the end of slug injection is the ending point of the reset cycle. If the injection time of a single slug exceeds 30 minutes, the reset cycle is divided into 30-minute intervals. It should be noted that the injection time of a single slug in fracturing construction is usually 5-30 minutes. Long slug injections exceeding 30 minutes are all continuous and stable sand addition operations, and the operating parameters are basically without significant fluctuations. Therefore, the 30-minute intervals can be adapted to the on-site construction process.

[0049] In this embodiment, a construction slug cycle contains 1200 consecutive operating cycles, corresponding to a typical construction condition of 120 strokes / min and slug duration of 10min. Using 360 theoretical alignment moments within the operating cycle as alignment benchmarks, phase alignment is performed on the 1200 operating cycle data within the split sub-chain: taking the crank angle of each operating cycle at 0° as the alignment starting point and the same phase node as the alignment benchmark, the residual data of all operating cycles are mapped onto a unified 0°-360° phase axis, achieving cross-cycle phase synchronization alignment and obtaining an aligned sub-chain. The residual data of all 1200 operating cycles within the aligned sub-chain at the same theoretical alignment moment (e.g., a 90° phase node) are extracted to construct the combination matrix corresponding to that theoretical alignment moment. This combination matrix is ​​a 1200-row, 6-column two-dimensional matrix, with rows corresponding to 1200 different operating cycles and columns corresponding to the residual values ​​of 6 physical quantities. For the first The first operating cycle, the first The residual value of a physical quantity at the current theoretical alignment time.

[0050] In this embodiment, the number of sampling points for phase resampling in preprocessing is: The value is a positive integer, and each sampling point corresponds to a theoretical alignment time. The theoretical alignment time is a crank angle phase node that is equally divided within the operating cycle. The crank angle corresponding to the i-th theoretical alignment time is... Each theoretical alignment moment corresponds to a 1° crank angle phase node, for a total of 360 theoretical alignment moments, i.e. .

[0051] The confidence determination module is used to perform column-dimensional trend fitting analysis on the combined matrix, and at the same time, perform row-dimensional feature fitting analysis on the combined matrix, and calculate the abnormal confidence level corresponding to each column element variable in the combined matrix according to different scenarios. In this embodiment, based on the correlation between each physical quantity in the fault mechanism library and typical faults, the preset weight values ​​are as follows, which can be corrected according to the fault statistics data of different models, as shown in Table 1:

[0052] The anomaly analysis module is used to traverse the anomaly confidence levels corresponding to all theoretical alignment times. Based on the temporal variation law of the anomaly confidence levels and the phase distribution characteristics of the corresponding health residual matrix, it matches the pre-built hydraulic end fault mechanism library and outputs the anomaly diagnosis results of the hydraulic end of the fracturing pump.

[0053] In this embodiment, the hydraulic end failure mechanism library is constructed based on the hydraulic end failure mechanism of fracturing pumps, laboratory failure simulation test data, and field failure statistics from over 1000 wells. It includes failure types, correlations of abnormal physical quantities, abnormal phase intervals of residual differences, characteristic patterns, severity classifications, and maintenance recommendations. The core typical failure feature library is shown in Table 2.

[0054] In this embodiment, the fault severity classification rules are as follows: Mild anomaly: 0.2≤C<0.4, early deterioration, yellow alert, monitoring is recommended; Moderate abnormality: 0.4≤C<0.7, fault development stage, orange warning, it is recommended to inspect and replace after construction is completed; Severe abnormality: C≥0.7, serious fault, red alert, immediate shutdown and maintenance recommended.

[0055] The system iterates through all 360 theoretical alignment moments corresponding to anomaly confidence levels. Based on the temporal variation of the anomaly confidence levels and the phase distribution characteristics of the corresponding health residual matrix, it matches the fault mechanism library and outputs the anomaly diagnosis results. In this embodiment, the in-cylinder pressure physical quantity of cylinder 3 is detected. The anomaly confidence level in the 0°-180° phase range is consistently ≥0.7, and the anomaly confidence level of the corresponding intake valve vibration acceleration is ≥0.6. Matching the fault mechanism library, the diagnosis result is: leakage in the intake valve of cylinder 3, the severity of the fault is moderate, the warning level is orange, and it is recommended to replace the intake valve assembly of cylinder 3 after construction. At the same time, for the progressive deterioration trend of the anomaly confidence level between 0.2 and 0.3, a yellow early warning is output, indicating that the corresponding cylinder valve seat has uniform wear, and it is recommended to strengthen monitoring.

[0056] This embodiment fully covers the 12 standard operating condition ranges described in this invention, encompassing typical operating conditions throughout the entire fracturing construction process, as detailed below: Stroke frequency dimension: three ranges: low stroke frequency (0-60 strokes / min), medium stroke frequency (60-120 strokes / min), and high stroke frequency (120-180 strokes / min); Pressure dimensions: four ranges: low pressure (0-26.25MPa), medium-low pressure (26.25-52.5MPa), medium-high pressure (52.5-78.75MPa), and high pressure (78.75-105MPa); The stable operating time for each working condition interval is ≥30 minutes, and the effective operating cycle under a single working condition is ≥1800, fully covering the entire process of fracturing construction, including replacement, circulation, pressure testing, fracture creation, sand-carrying injection, and high-pressure replacement.

[0057] The failures are categorized into three levels based on their development stage: mild early-stage degradation, moderate-stage development failure, and severe failure. The sample settings are shown in Table 3.

[0058] Health status control sample: brand new hydraulic end in fault-free condition, 1 set of each of the 12 standard operating condition ranges, a total of 12 sets, with each set having an effective operating cycle of ≥1800 cycles; Total effective sample size: 144 fault samples and 12 healthy control samples, fully covering all operating conditions, all fault types, and all fault development stages.

[0059] Three traditional diagnostic methods commonly used in fracturing operations were selected as parallel control groups and compared with the system of this invention under the same conditions, samples, and sequence, as follows: Control group 1: Fixed threshold alarm method (most commonly used method on site): Set fixed upper and lower thresholds based on the factory rated parameters of the equipment. When the monitored data exceeds the threshold, an alarm is triggered. Control group 2: Traditional offline simulation diagnosis method: Based on a multiphysics simulation model with fixed parameters, without real-time parameter calibration, fault diagnosis is achieved by directly comparing the measured operating data with the simulation results; Control group 3: Timing comparison method without phase alignment: cross-cycle data comparison is performed based on the time axis, without crank angle phase anchoring and operating condition normalization processing, and the timing data of adjacent cycles are directly compared.

[0060] Based on the disassembly and testing results, and using industry-standard quantitative evaluation indicators, the definitions are as follows: Diagnostic accuracy: Number of correctly diagnosed fault samples / Total number of fault samples × 100%. A correct diagnosis requires that the fault type, fault location, fault severity and disassembly results are consistent. False positive rate: (Number of healthy control samples that falsely triggered the fault alarm / Total number of healthy control samples) × 100%; Early degradation identification rate: Number of correctly identified mild early degradation samples / Total number of mild early degradation samples × 100%.

[0061] The experimental results are shown in Table 4:

[0062] The beneficial effects of the above technical solution are: by phase anchoring and working condition normalization, the influence of changing working conditions is effectively eliminated; by dual residual comparison and dual-dimensional fitting, the simultaneous identification of sudden failures and gradual degradation is achieved; by matching the fault mechanism library, accurate fault location is achieved, and the on-site verification diagnostic accuracy rate is ≥95%, and the false positive rate is ≤3%, which meets the diagnostic needs of fracturing construction sites.

[0063] This invention provides a diagnostic system for abnormal operation of the hydraulic end of a fracturing pump, wherein the confidence determination module, such as... Figure 2 As shown, it includes: The first extreme value identification submodule is used to perform column dimension trend fitting analysis on the combined matrix, obtain the column fitting residual corresponding to each column vector, and identify the first extreme value element that exceeds the first preset residual threshold. In this embodiment, for each column vector of the combined matrix, with the running cycle number as the independent variable and the residual value as the dependent variable, a weighted least squares method is used to perform linear trend fitting, obtaining the fitting function of each column vector and the corresponding column fitting residual; based on Statistical criteria: Calculate the mean of the residuals of the column vector. with standard deviation The first preset residual threshold is The element whose column fit residual exceeds the threshold is identified as the first extreme value element, and the slope of the fit trend of the column vector is output. Used for subsequent trend feature determination.

[0064] The second extreme value identification submodule is used to perform row dimension feature fitting analysis on the combined matrix, obtain the row fitting residual corresponding to each row vector, and identify the second extreme value element that exceeds the second preset residual threshold. For each row vector of the combined matrix, based on the coupling relationship of six physical quantities under healthy conditions, principal component analysis is used for feature fitting. Principal components with a cumulative contribution rate ≥95% are extracted, and the fitted value of the row vector is reconstructed. The row fitting residual is then calculated. Statistical criteria: Calculate the mean of the residuals of the row vector. with standard deviation The second preset residual threshold is Elements whose row fitting residuals exceed this threshold are identified as the second extreme value elements.

[0065] The frequency statistics submodule is used to count the first frequency of variables in the same column being identified as first extreme value elements and the second frequency of variables being identified as second extreme value elements in the combination matrix. The scenario-based calculation submodule is used to calculate the anomaly confidence level corresponding to the element variable in the column based on the sum of the first frequency and the second frequency, and in combination with the physical quantity fault weight and extreme value deviation amplitude corresponding to the element variable in the column.

[0066] Preferably, the scenario-based calculation submodule includes: The first calculation unit is used to calculate the maximum deviation amplitude and residual coefficient of variation of all extreme elements of the column variable when the sum of the first frequency and the second frequency is ≥2. Combined with the fault weights of the physical quantities corresponding to the column variable, it calculates the anomaly confidence level. : ,in, The anomaly confidence level has a value range of [0,1]. The fault weight of the physical quantity corresponding to the element variable in this column has a value range of [0,1]. That is, the correlation between the physical quantity corresponding to the element variable in this column and the fault is obtained based on the fault mechanism library. is the proportionality coefficient; D is the maximum deviation magnitude, which is the absolute value of the difference between the extreme element and the fitted value and the standard deviation of the column vector residual. The ratio; Let be the coefficient of variation of the column vector residuals, and This is the mean of the residuals of the column vector; The second calculation unit is used to calculate the deviation multiple of the extreme value element when the sum of the first frequency and the second frequency is 1, and to match the deviation multiple with a preset multiple-confidence table to obtain the anomaly confidence level. Wherein, the deviation factor is the sum of the absolute value of the difference between the measured value and the corresponding fitted value of the extreme value element. The ratio of the threshold; Meanwhile, if the slope of the fitting trend of the column vector corresponding to the element variable of this column remains the same for n01 consecutive running cycles and the absolute value continues to increase, it is determined that there is a gradual deterioration trend, and the confidence is adjusted according to min(abnormal confidence + 0.2, 0.8), where 0.2 is the warning increment; The third calculation unit is used to determine if, when the sum of the first frequency and the second frequency is 0, the slope of the fitting trend of the column vector corresponding to the element variable in that column remains the same for n02 or more consecutive running cycles and the absolute value continues to increase, that there is an early gradual deterioration trend, and the abnormal confidence level is assigned to 0.2, triggering a yellow early warning; otherwise, it is determined to be a normal state, and the abnormal confidence level is assigned to 0.

[0067] In this embodiment, the proportional coefficient is obtained by fitting and optimizing on-site fault data.

[0068] In this embodiment, Its core function is to measure dimensionless abnormal deviations. By using the classical exponential failure distribution model of reciprocating machinery, and mapping it to the standardized confidence interval of [0,1], the fault identification sensitivity and false alarm rate are balanced, and the formula adopts... It is a simplified form of the Weibull distribution, based on full-condition simulation tests of three core faults: suction valve leakage, discharge valve wear, and plunger seal failure, to obtain the abnormal deviation degree. The correspondence between the measured failure probability and the optimal scaling factor is determined using the least squares method. The fitted data are shown in Table 5:

[0069] When k0=0.5, the goodness of fit is the highest and the mean square error is the smallest.

[0070] Based on 5-fold cross-validation of 12,460 healthy samples and 237 labeled faulty samples, with k=0.5: The average accuracy rate of fault diagnosis is 94.3%; the false positive alarm rate is 2.7%; and the early deterioration missed detection rate is 1.8%, achieving the goal of early warning of progressive faults.

[0071] In this embodiment, a sum of ≥2 represents multiple extreme value anomalies, indicating a high probability of failure; a sum of 1 represents a single extreme value anomaly, indicating an occasional anomaly; and a sum of 1 represents no extreme value anomalies, indicating a normal state / early degradation.

[0072] In this embodiment, n01=5, that is, the slope of the fitted trend remains the same for 5 or more consecutive running cycles and the absolute value continues to increase, which is judged as a significant gradual deterioration trend; n02=10, that is, the slope of the fitted trend remains the same for 10 or more consecutive running cycles and the absolute value continues to increase, which is judged as an early deterioration trend. In this embodiment, the conventional operating stroke rate of the fracturing pump is 60-180 strokes / min, and the single operating cycle duration is 0.33-1s. n01 = 5 cycles, corresponding to a duration of 1.65-5s, covering more than 3 complete valve group opening and closing cycles, which can eliminate occasional interference from single-cycle operating condition fluctuations and ensure the reliability of trend judgment. n02=10 cycles correspond to a duration of 3.3-10s. For early degradation without extreme value anomalies, extending the judgment cycle can further reduce false alarms caused by fluctuations in normal operating conditions.

[0073] When n01=5, the accuracy of progressive fault identification is 96.2%, and the false alarm rate is ≤2%. If the cycle is shortened to 3, the false alarm rate rises to 7.8%. If the cycle is extended to 8, the fault warning lag time exceeds 10 seconds, thus losing the value of early warning. When n02=10, the accuracy of early degradation warning is 92.5%, and the false alarm rate is ≤1.5%; if the period is shortened to 5 cycles, the false alarm rate increases to 6.2%; if the period is extended to 15 cycles, the early degradation missed detection rate increases to 5.7%.

[0074] It should be noted that the same slope sign means that the slope is positive in consecutive periods (the residual continues to increase and the deterioration intensifies), and the continuous increase in absolute value means that the difference between the slopes of adjacent periods is ≥5%, which excludes stable fluctuations with no change in slope and further improves the reliability of the judgment.

[0075] In this embodiment, a preset multiple-confidence mapping table is used, based on... The criteria and field failure statistics are constructed as shown in Table 6:

[0076] In this embodiment, in min(anomaly confidence + 0.2, 0.8), the compensation increment of 0.2 is used to reasonably weight early anomalies with a continuous deterioration trend but weak extreme value characteristics, in order to identify progressive faults such as valve group wear and seal aging. This value comprehensively considers the signal amplitude characteristics, statistical significance level, and field early warning classification requirements of typical hydraulic end deterioration faults. It can effectively improve the identification sensitivity of progressive deterioration without causing false alarms due to overcompensation, which is in line with the engineering experience value range for early fault warning of high-pressure reciprocating pump equipment. 0.8 is the critical value for severe fault judgment commonly used in the field of fracturing equipment fault diagnosis. Above this value, it usually corresponds to severe faults that may endanger construction safety, such as valve jamming, seal failure, and plunger damage, which require immediate shutdown and handling. Below this value, it corresponds to early or moderate anomalies that can continue to operate and be repaired as appropriate. Setting an upper limit of 0.8 can ensure that the warning level matches the degree of fault danger and avoid unnecessary construction interruption.

[0077] The beneficial effects of the above technical solution are: it not only ensures the sensitivity of identifying sudden serious faults, but also realizes the early detection of gradual deterioration such as valve group wear and seal aging, effectively reducing the fault missed detection rate to below 2%, and greatly improving the reliability of diagnostic results and on-site guidance.

[0078] This invention provides a diagnostic system for abnormal operation of the hydraulic end of a fracturing pump, wherein the cycle segmentation module includes: The feature extraction submodule is used to extract the operating condition feature parameters of each set of operating cycle data, wherein the operating condition feature parameters include: the average number of strokes and the average discharge main pipe pressure within the operating cycle; The dimension construction submodule is used to construct a two-dimensional operating condition grid with the average number of strokes as the first dimension and the average discharge manifold pressure as the second dimension. Each grid corresponds to a standard operating condition range. In this embodiment, the division of the third interval of the stroke dimension is based on the stroke operation pattern of the fracturing pump in the field: fracturing construction is divided into three typical working conditions: 0-1 / 3 of the rated stroke is the low stroke interval, used for displacement, circulation, and pressure testing stages; 1 / 3-2 / 3 of the rated stroke is the medium stroke interval, which is the most commonly used continuous sand addition construction condition in the field, accounting for more than 80% of the total construction time; 2 / 3-1 of the rated stroke is the high stroke interval, used for high-displacement fracture creation stage, with the largest load and the highest failure rate; the division of the fourth interval of the pressure dimension is based on the following: fracturing construction is divided into circulation pressure testing (0-1 / 4 of the rated pressure), pre-fracturing with fluid (1 / 4-1 / 2 of the rated pressure), and sand-carrying fluid injection (1 / 2-3 / 4 of the rated pressure). The high-pressure replacement (3 / 4-1 rated pressure) has four standard stages. Therefore, the standard operating condition intervals are divided as follows: Firstly, based on the average number of strokes within the operating cycle, the intervals are divided into three sections according to the rated strokes of the fracturing pump: 0-1 / 3, 1 / 3-2 / 3, and 2 / 3-1. Secondly, based on the average discharge manifold pressure, the intervals are divided into four sections according to the rated pressure of the fracturing pump: 0-1 / 4, 1 / 4-1 / 2, 1 / 2-3 / 4, and 3 / 4-1. A 3×4 two-dimensional operating condition grid is constructed, resulting in 12 standard operating condition intervals. Each grid corresponds to a unique standard operating condition interval. Taking a rated stroke of 180 strokes / min and a rated pressure of 105 MPa as an example, the specific division of the standard operating condition intervals is shown in Table 7.

[0079] The correction submodule is used to perform dimensionless correction on the pressure and vibration amplitude in the operating cycle data based on the parameter ratio between the current standard operating condition range and the rated operating condition range, so as to obtain a normalized cycle dataset.

[0080] In this embodiment, the general correction formula is: The data is normalized after dimensionless correction. These are the original measured data. For normalization correction coefficients, The rated number of strokes for the fracturing pump. This represents the average number of strokes in the current operating cycle. The rated discharge pressure of the fracturing pump. This represents the average discharge manifold pressure during the current operating cycle.

[0081] If the current operating conditions have an average stroke rate of 120 strokes / min and an average discharge pressure of 60 MPa, and a rated stroke rate of 180 strokes / min and a rated pressure of 105 MPa, the normalization coefficient is (180 / 120) × (105 / 60) = 2.625. Multiply the measured pressure amplitude by this coefficient to eliminate the amplitude difference caused by the fluctuation of the operating conditions.

[0082] The beneficial effects of the above technical solution are: it solves the problem of frequent false positive alarms caused by the fixed threshold of the existing technology due to the fluctuation of operating conditions, while ensuring the comparability of cross-cycle data under different operating conditions, providing a unified and effective data benchmark for subsequent comparison of dual residual matrices and extraction of abnormal features, and keeping the false positive alarm rate of the system stably controlled within 3% under strong changing operating conditions.

[0083] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A diagnostic system for abnormal operation of the hydraulic end of a fracturing pump, characterized in that, include: The data acquisition module is used to simultaneously acquire multi-source operation monitoring data and reference phase data from the hydraulic end of the fracturing pump, and to preprocess the acquired raw data to obtain a standardized time-series dataset. The period splitting module is used to split the standardized time series dataset into several sets of continuous and phase-synchronized running period data, and perform normalization processing to obtain a normalized period dataset. The simulation matrix construction module is used to perform real-time parameter calibration on the pre-built multi-physics positive coupling simulation model of the hydraulic end of the fracturing pump based on the boundary condition parameters of the normalized period dataset of the current operating cycle, so as to obtain the calibration simulation model of the current cycle. At this time, the phase-synchronized plunger displacement time series is used as the input of the calibration simulation model, and the output is the simulation working matrix corresponding to the current operating cycle. The comparison chain construction module is used to perform a first comparison between the simulation working matrix of the current period and the standard working matrix corresponding to the same standard working condition interval to obtain the health residual matrix. At the same time, it performs a second comparison between the simulation working matrix of the current period and the simulation working matrix of the previous adjacent running period of the same standard working condition interval to obtain the trend residual matrix, and constructs the comparison chain. The phase alignment module is used to split the comparison chain with the construction slug cycle of the fracturing pump as the reset cycle to obtain the split sub-chain corresponding to each reset cycle, and to perform phase alignment on all the running cycle data in the split sub-chain with the theoretical alignment time in the running cycle as the alignment benchmark to obtain the alignment sub-chain, to extract the residual data of all running cycles in the alignment sub-chain under the same theoretical alignment time, and to construct the combination matrix of the theoretical alignment time. The confidence determination module is used to perform column-dimensional trend fitting analysis on the combined matrix, and at the same time, perform row-dimensional feature fitting analysis on the combined matrix, and calculate the abnormal confidence level corresponding to each column element variable in the combined matrix according to different scenarios. The anomaly analysis module is used to traverse the anomaly confidence levels corresponding to all theoretical alignment times. Based on the temporal variation law of the anomaly confidence levels and the phase distribution characteristics of the corresponding health residual matrix, it matches the pre-built hydraulic end fault mechanism library and outputs the anomaly diagnosis results of the hydraulic end of the fracturing pump.

2. The fracturing pump hydraulic end operation anomaly diagnosis system according to claim 1, characterized in that, Using each running cycle as a node, the corresponding health residual matrix and trend residual matrix are combined into a comparison array for that cycle. All comparison arrays are concatenated in the order of running time to construct a comparison chain.

3. The fracturing pump hydraulic end operation anomaly diagnosis system according to claim 1, characterized in that, The confidence level determination module includes: The first extreme value identification submodule is used to perform column dimension trend fitting analysis on the combined matrix, obtain the column fitting residual corresponding to each column vector, and identify the first extreme value element that exceeds the first preset residual threshold. The second extreme value identification submodule is used to perform row dimension feature fitting analysis on the combined matrix, obtain the row fitting residual corresponding to each row vector, and identify the second extreme value element that exceeds the second preset residual threshold. The frequency statistics submodule is used to count the first frequency of variables in the same column being identified as first extreme value elements and the second frequency of variables being identified as second extreme value elements in the combination matrix. The scenario-based calculation submodule is used to calculate the anomaly confidence level corresponding to the element variable in the column based on the sum of the first frequency and the second frequency, and in combination with the physical quantity fault weight and extreme value deviation amplitude corresponding to the element variable in the column.

4. The fracturing pump hydraulic end operation anomaly diagnosis system according to claim 3, characterized in that, The scenario-specific calculation submodule includes: The first calculation unit is used to calculate the maximum deviation amplitude and residual coefficient of variation of all extreme elements of the column variable when the sum of the first frequency and the second frequency is ≥2. Combined with the fault weights of the physical quantities corresponding to the column variable, it calculates the anomaly confidence level. : ,in, The anomaly confidence level has a value range of [0,1]. The fault weight of the physical quantity corresponding to the element variable in this column has a value range of [0,1]. That is, the correlation between the physical quantity corresponding to the element variable in this column and the fault is obtained based on the fault mechanism library. is the proportionality coefficient; D is the maximum deviation magnitude, which is the absolute value of the difference between the extreme element and the fitted value and the standard deviation of the column vector residual. The ratio; Let be the coefficient of variation of the column vector residuals, and , This is the mean of the residuals of the column vector; The second calculation unit is used to calculate the deviation multiple of the extreme value element when the sum of the first frequency and the second frequency is 1, and to match the deviation multiple with a preset multiple-confidence table to obtain the anomaly confidence level. Wherein, the deviation factor is the sum of the absolute value of the difference between the measured value and the corresponding fitted value of the extreme value element. The ratio of the threshold; Meanwhile, if the slope of the fitting trend of the column vector corresponding to the element variable of this column remains the same for n01 consecutive running cycles and the absolute value continues to increase, it is determined that there is a gradual deterioration trend, and the confidence is adjusted according to min(abnormal confidence + 0.2, 0.8), where 0.2 is the warning increment; The third calculation unit is used to determine if, when the sum of the first frequency and the second frequency is 0, the slope of the fitting trend of the column vector corresponding to the element variable in that column remains the same for n02 or more consecutive running cycles and the absolute value continues to increase, that there is an early gradual deterioration trend, and the abnormal confidence level is assigned to 0.2, triggering a yellow early warning; otherwise, it is determined to be a normal state, and the abnormal confidence level is assigned to 0.

5. The fracturing pump hydraulic end operation anomaly diagnosis system according to claim 1, characterized in that, The preprocessing includes: validity verification, noise reduction, and phase resampling; The rule for dividing the operating cycle is as follows: the crank angle of 0° corresponding to the top dead center of the piston displacement in the reference phase data is taken as the starting point of the cycle, and the crank angle of 360° is taken as the ending point of the cycle. This corresponds to the piston completing one complete intake stroke and exhaust stroke. The crank angle of 0°-180° is the intake stroke, and the crank angle of 180°-360° is the exhaust stroke. The number of sampling points for phase resampling in the preprocessing is: The value is a positive integer, and each sampling point corresponds to a theoretical alignment time. The theoretical alignment time is a crank angle phase node that is equally divided within the operating cycle. The crank angle corresponding to the i-th theoretical alignment time is... .

6. The fracturing pump hydraulic end operation anomaly diagnosis system according to claim 1, characterized in that, The periodic splitting module includes: The feature extraction submodule is used to extract the operating condition feature parameters of each set of operating cycle data, wherein the operating condition feature parameters include: the average number of strokes and the average discharge main pipe pressure within the operating cycle; The dimension construction submodule is used to construct a two-dimensional operating condition grid with the average number of strokes as the first dimension and the average discharge manifold pressure as the second dimension. Each grid corresponds to a standard operating condition range. The correction submodule is used to perform dimensionless correction on the pressure and vibration amplitude in the operating cycle data based on the parameter ratio between the current standard operating condition range and the rated operating condition range, so as to obtain a normalized cycle dataset.

7. The fracturing pump hydraulic end operation anomaly diagnosis system according to claim 1, characterized in that, The multi-source operation monitoring data includes: single-cylinder pump head body high-pressure chamber pressure, suction main pipe pressure, discharge main pipe pressure, suction valve vibration acceleration, discharge valve vibration acceleration, and hydraulic oil temperature; The reference phase data includes: crankshaft angle signal and piston displacement signal.

8. The fracturing pump hydraulic end operation anomaly diagnosis system according to claim 1, characterized in that, Using the measured peak cylinder pressure and valve group vibration frequency in the normalized periodic dataset of the current operating cycle as calibration targets, the weighted least squares method is used to optimize the parameters of the valve group damping coefficient, equivalent leakage coefficient, and fluid bulk modulus of the simulation model to obtain the calibration simulation model for the current cycle.

Images