Permanent magnet frequency conversion submersible pump test method and system

Abstract

The invention relates to the technical field of submersible pump testing, and discloses a permanent magnet variable frequency submersible pump testing method and system. The method comprises the following steps: acquiring an original response signal cluster of the submersible pump in multi-stage variable-speed operation by constructing an integrated test environment; decoupling and enhancing the signal cluster based on time sequence dependence, and separating out an independent hydraulic performance characteristic signal group and an independent electromagnetic vibration characteristic signal group; and inputting the characteristic signal group into a dynamic working condition simulator, and restoring an internal flow field pressure cloud picture and a rotor stress distribution picture of the submersible pump under the simulated working condition through an iterative approximation algorithm. The two are fused to deduce a potential failure mode and its evolution trajectory, and an enhanced test procedure is generated accordingly. According to the method, accurate separation of multi-physics field test signals and visualization of internal states are realized, and the accuracy of fault attribution and the advance of failure prediction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of submersible pump testing, in particular to a permanent magnet variable frequency submersible pump testing method and system. BACKGROUND

[0002] Currently, the performance test of permanent magnet variable frequency submersible pumps is usually carried out at fixed or a few operating points, and the final performance parameters such as flow, head, input electric power, etc. are mainly collected, and vibration, noise and other signals are independently monitored. However, during variable frequency operation, the state of the water pump is the result of dynamic coupling of electromagnetic, mechanical and hydraulic systems. The pressure, current, vibration and other signals obtained by actual testing are mixed responses under the excitation of the above-mentioned multiple physical fields. These signals are superimposed on each other in time and frequency domains, resulting in deep coupling of the dynamic characteristics of hydraulic performance and the state characteristics of electromagnetic machinery, which makes it difficult to effectively separate from the mixed data, thereby affecting the accurate positioning of efficiency fluctuations, vibration sources and other problems.

[0003] The limitation of the existing method is that the test and analysis remain at the level of external macroscopic parameters and mixed signals, and the quantitative mapping relationship between these external data and the internal key physical state of the water pump cannot be established. The test process cannot obtain the real-time pressure distribution and vortex structure of the internal flow field, nor can it monitor the dynamic stress state of the rotor and other components. Therefore, for potential failures caused by internal flow instability, pressure pulsation or local stress concentration, the existing technology mainly relies on empirical judgment after testing or disassembly analysis after failure occurs, and lacks the ability to predict and visualize the failure position and its evolution path based on the test process itself.

[0004] The existing test technology faces two key problems: how to separate the pure characteristics independently representing different physical processes from the mixed signals obtained by testing; how to use these external characteristics to reversely construct a visual model reflecting the internal dynamic flow field and structural stress state, so as to realize early prediction of potential failure. SUMMARY

[0005] The purpose of the present application is to provide a permanent magnet variable frequency submersible pump testing method and system to solve the problems raised in the background art.

[0006] To achieve the above-mentioned purpose, the present application provides a permanent magnet variable frequency submersible pump testing method, which comprises: An integrated test environment of the target submersible pump is constructed by a controllable load device and a multi-parameter sensing unit, and initial environmental parameters of the integrated test environment and baseline data of the submersible pump body are obtained; The output mode of the variable frequency drive unit is adjusted according to the initial environmental parameters, the target submersible pump is guided into a multi-stage variable speed operating state, and the original response signal cluster fed back by the multi-parameter sensing unit is captured synchronously; Performing time-dependent signal decoupling and data enhancement processing on the original response signal cluster to separate a characteristic signal group representing hydraulic performance and a characteristic signal group representing electromagnetic and vibration state; Inputting the characteristic signal group into a dynamic working condition simulator to restore internal flow field pressure cloud chart and rotor stress distribution chart of the target submersible pump under simulated working conditions through an iterative approximation algorithm; Fusing the internal flow field pressure cloud chart and the rotor stress distribution chart to deduce a set of potential failure modes of the target submersible pump, and labeling an evolution trajectory corresponding to each potential failure mode; Reversely correcting the baseline data of the submersible pump body according to the evolution trajectory to generate an enhanced test procedure for guiding physical entity testing.

[0007] Preferably, the time-dependent signal decoupling and data enhancement processing on the original response signal cluster to separate a characteristic signal group representing hydraulic performance and a characteristic signal group representing electromagnetic and vibration state includes: Performing stationarity test and detrending processing on the original response signal cluster to eliminate baseline drift components introduced by the integrated test environment; Using an adaptive filter bank to strip the processed signal of frequency bands, extract core frequency band signals directly related to impeller rotation frequency and its harmonics, and form a hydraulic performance original signal set; At the same time, using independent component analysis algorithm to perform blind source separation on wideband residual signals to decouple mutually orthogonal electromagnetic noise components and mechanical vibration components, and form an electromagnetic and vibration original signal set; Performing data enhancement processing on the hydraulic performance original signal set, the data enhancement processing including adding specific signal-to-noise ratio noise proportional to flow, and performing time domain stretching and compression transformation on pressure pulse signals; Performing data enhancement processing on the electromagnetic and vibration original signal set, the data enhancement processing including synthesizing vibration waveforms simulating different bearing gap states, and generating current harmonics simulating slight eccentricity conditions of the rotor; Marking the enhanced hydraulic performance original signal set as the characteristic signal group representing hydraulic performance, and marking the enhanced electromagnetic and vibration original signal set as the characteristic signal group representing electromagnetic and vibration state.

[0008] Preferably, the inputting the characteristic signal group into a dynamic working condition simulator to restore internal flow field pressure cloud chart and rotor stress distribution chart of the target submersible pump under simulated working conditions through an iterative approximation algorithm includes: The independent characteristic signal set representing hydraulic performance is mapped to the boundary condition driving source of the dynamic operating condition simulator. The boundary condition driving source includes the inlet pressure pulsation sequence and the outlet flow fluctuation sequence. The group of characteristic signals that independently characterize electromagnetic and vibration states is converted into an internal excitation loading source for the dynamic working condition simulator. The internal excitation loading source includes a time-varying electromagnetic force density field and a dynamic radial force vector. Initialize the fluid domain mesh and solid domain mesh in the dynamic working condition simulator, and set the material constitutive relation and fluid-structure interaction interface parameters; Using the boundary condition driving source and the internal excitation loading source as input, run a coupled solver for computational fluid dynamics and computational structural dynamics; After each computation time step, the residual between the virtual sensing signal output by the coupled solver and the corresponding set of feature signals is compared. Based on the magnitude and direction of the residual, the amplitude of the boundary condition driving source and the point of action of the internal excitation loading source are dynamically adjusted by the iterative approximation algorithm until the residual converges to a preset threshold. When the residual converges, the fluid domain pressure distribution, i.e. the internal flow field pressure cloud map, and the solid domain stress distribution, i.e. the rotor stress distribution map, are finally calculated by the dynamic operating condition simulator at the current time step.

[0009] Preferably, the step of fusing the internal flow field pressure cloud map and the rotor stress distribution map to deduce the set of potential failure modes of the target submersible pump includes: A spatiotemporal correlation matrix is ​​established between the high-pressure region in the internal flow field pressure cloud diagram and the high-stress region in the rotor stress distribution diagram. The spatiotemporal correlation matrix records the correspondence between the two in terms of spatial location and time phase. Scan the spatiotemporal correlation matrix to identify region pairs that meet a preset correlation strength threshold. These region pairs are defined as high-pressure-high-stress coupling risk zones. For each of the high-pressure-high-stress coupling risk zones, the material fatigue characteristic database is queried to obtain the predicted value of material crack initiation life under the current pressure-stress cycle characteristics; The low-frequency pressure pulsation patterns present in the internal flow field pressure cloud map are analyzed, and the periodic strain energy accumulation caused by them on the target submersible pump casing structure is calculated. Check whether there are any areas in the rotor stress distribution diagram where the stress concentration gradient exceeds the critical value, and assess the probability of brittle fracture by combining the statistical model of material micro-defects. By integrating the predicted crack initiation life of the material, the accumulated periodic strain energy, and the probability of brittle fracture, a set of potential failure modes including fatigue failure, excessive deformation failure, and brittle fracture failure is formed. For each failure mode in the set of potential failure modes, its load history in the dynamic operating condition simulator is traced back to generate the evolution trajectory describing its development from the initial state to the critical state.

[0010] Preferably, the step of reversely correcting the submersible pump body baseline data based on the evolution trajectory to generate an enhanced test procedure for guiding physical entity testing includes: Key load events and their timing are extracted from the evolution trajectory. The key load events include pressure peaks, stress peaks, and the times when specific vibration frequencies occur. The critical load events are compared with the standard operating parameters in the submersible pump body baseline data to identify load condition gaps not covered by the standard operating parameters. For each load condition gap, a supplementary test program is designed, which specifies the specific frequency modulation waveform of the frequency converter, the step change point of the controllable load device, and the sampling rate of data acquisition. The supplementary test procedure is integrated and arranged with the standard test procedure based on the submersible pump body baseline data to ensure that the order of application of test loads is consistent with the timing logic of the loads appearing in the evolution trajectory. To integrate each monitoring point in the orchestrated test process, a specific combination of sensors selected from the multi-parameter sensing unit and data recording trigger conditions are configured to form a complete set of instructions and parameters for the enhanced test procedure.

[0011] Preferably, the method further includes: The enhanced test procedure is executed to obtain high-fidelity real test data, and the high-fidelity real test data is calibrated to match the simulation results under the simulated working conditions. Based on the data stream after consistency calibration, a digital twin of the target submersible pump is constructed, and a preset accelerated degradation excitation signal is injected into the digital twin. Monitor the response sequence of the digital twin under the accelerated degradation excitation signal, and identify performance inflection points and structural weaknesses based on the response sequence; Based on the spatial coordinates of the performance inflection point and the structural weak link, a maintenance intervention instruction set with sequential execution logic is generated.

[0012] Preferably, the step of executing the enhanced test procedure, obtaining high-fidelity real-world test data, and calibrating the consistency between the high-fidelity real-world test data and the simulation results under the simulated working conditions includes: The target submersible pump is mounted on the physical test platform, and the variable frequency drive unit and the controllable load device are operated in accordance with the instructions and parameter set of the enhanced test procedure. The high-fidelity test data is obtained by synchronously recording the full-channel response of the target submersible pump during the execution of the enhanced test procedure using the calibrated multi-parameter sensing unit. Data segments corresponding to the simulated working conditions in terms of time scale and load type are extracted from the high-fidelity measured data and used as calibration comparison datasets. The hydraulic performance data in the calibration comparison dataset is compared with the statistical characteristics of the internal flow field pressure cloud map to calculate the pressure distribution similarity index and the flow prediction error index. The vibration and current data in the calibration comparison dataset are compared with the rotor stress distribution map and the derived signal of electromagnetic excitation to calculate the vibration mode correlation coefficient and the difference in current harmonic distortion. A set of model correction coefficients are generated based on the pressure distribution similarity index, the flow prediction error index, the vibration mode correlation coefficient, and the difference in current harmonic distortion. The model correction coefficients are applied to globally adjust the material property parameters and boundary condition models in the dynamic working condition simulator to complete the consistency calibration.

[0013] Preferably, constructing a digital twin of the target submersible pump based on the data stream after consistency calibration includes: Using the dynamic working condition simulator after the consistency calibration as the core computing engine, a physical behavior simulation layer of the digital twin is encapsulated. A data interface layer for the digital twin is established, which can receive high-fidelity test data from a physical test platform or historical database as input in real time. A state mapping layer is constructed for the digital twin, which converts the input real-time data into boundary conditions and initial conditions that can be recognized by the physical behavior simulation layer; A virtual sensing network is created in the digital twin, which corresponds completely to the multi-parameter sensing unit on the physical test platform in terms of spatial location and type, and is used to output virtual readings comparable to those of the physical sensors.

[0014] Preferably, the step of generating a maintenance intervention instruction set with sequential execution logic based on the spatial coordinates of the performance inflection point and the structural weak link includes: The operating time and load conditions corresponding to the performance inflection point are translated into a recommended preventive maintenance time window for physical submersible pumps. The spatial coordinates of the structural weak points are mapped onto the three-dimensional design drawings of the target submersible pump, and the geometric areas that need to be inspected or treated are identified. Based on the relative speed of development of different failure modes in the evolution trajectory, the preventive maintenance time window recommendations and the geometric regions are prioritized. For each high-priority maintenance item, specify the specific maintenance operation type, the required testing tools, and the acceptance criteria for determining whether the repair is satisfactory; All maintenance items, their corresponding priorities, operation types, testing tools, and acceptance criteria are encoded into a machine-readable set of maintenance intervention instructions in the order of execution.

[0015] Preferably, the present invention also includes a permanent magnet variable frequency submersible pump testing system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the permanent magnet variable frequency submersible pump testing method described above.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By performing time-dependent signal decoupling and data augmentation on the original response signal clusters, it is possible to separate characteristic signal groups that independently characterize hydraulic dynamics and electromechanical states from the mixed time-frequency data. This processing distinguishes responses from different physical mechanisms, enabling the independent identification of transient changes in hydraulic efficiency, specific electromagnetic harmonic components, and mechanical vibration modes, thereby improving the accuracy of fault attribution.

[0017] The separated characteristic signal group is input into a dynamic operating condition simulator, and an iterative approximation algorithm is used to reverse-engineer the pump's internal flow field pressure cloud map and rotor stress distribution map based on external data. This process establishes a quantitative correspondence between external measurements and internal physical fields, transforming the pressure gradient, vortex structure, and stress concentration areas of the impeller channel into directly observable and analyzable image data, thus achieving a quantitative assessment of the internal state.

[0018] By integrating flow field pressure cloud maps and rotor stress distribution maps for extrapolation, potential failure modes such as cavitation initiation and fatigue crack initiation can be systematically identified, and their evolution trajectories as operating conditions change can be quantitatively described. Based on this trajectory, an enhanced test procedure is generated by reverse-correcting baseline data. This guides physical testing to focus on the most sensitive operating condition ranges that are most prone to inducing failure, thereby improving the targeting of tests and the lead time for failure warnings. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the working principle of the permanent magnet variable frequency submersible pump testing method described in this invention. Figure 2 A flowchart for signal decoupling and data enhancement processing; Figure 3 Flowchart generated for enhanced test procedures; Figure 4 A scatter plot of three-dimensional stress distribution in the structurally weak points of a submersible pump body; Figure 5 This is a time-series graph showing the changes in multiple indicators such as flow rate, head, and efficiency during the operation of a submersible pump. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 This invention provides a testing method for a permanent magnet variable frequency submersible pump. The method includes: constructing an integrated test environment consisting of a controllable load device and a multi-parameter sensing unit; acquiring initial environmental parameters and baseline data of the target submersible pump within this environment; adjusting the output mode of the variable frequency drive unit based on the initial environmental parameters to guide the target submersible pump into a multi-stage variable speed operation state; and simultaneously capturing the raw response signal clusters fed back by the multi-parameter sensing unit during this process; performing time-dependent signal decoupling and data augmentation processing on the captured raw response signal clusters, which can separate two sets of characteristic signals, one independently characterizing hydraulic performance and the other independently characterizing electromagnetic and vibration states; inputting the processed characteristic signal set into a dynamic operating condition simulator and using an iterative approximation algorithm to reconstruct the internal flow field pressure cloud map and rotor stress distribution map of the target submersible pump under simulated operating conditions; and fusing the information contained in the internal flow field pressure cloud map and rotor stress distribution map to deduce a set of potential failure modes of the target submersible pump and marking the evolution trajectory from occurrence to development for each potential failure mode. Based on the obtained evolution trajectory, the previously acquired submersible pump body baseline data is reverse-corrected to generate a more targeted and predictive enhanced test procedure to guide subsequent physical entity testing.

[0022] Example 1: See Figure 2The original response signal set was subjected to stationarity testing and detrending processing to eliminate baseline drift components introduced by the integrated testing environment. An adaptive filter bank was used to strip the processed signals, extracting core frequency band signals directly related to the impeller rotation frequency and its harmonics; these signals constituted the original hydraulic performance signal set. Independent component analysis (ICA) was used to separate the broadband residual signals into blind sources, decoupling mutually orthogonal electromagnetic noise and mechanical vibration components; these components constituted the original electromagnetic and vibration signal set. Data augmentation processing was performed on the original hydraulic performance signal set, including adding noise with a specific signal-to-noise ratio proportional to the flow rate and performing time-domain stretching and compression transformations on the pressure pulse signal. Data augmentation processing was also performed on the original electromagnetic and vibration signal set, including synthesizing vibration waveforms simulating different bearing clearance states and generating current harmonics simulating slight misalignment between the stator and rotor. The data-augmented original hydraulic performance signal set was labeled as a feature signal group independently characterizing hydraulic performance, and the augmented original electromagnetic and vibration signal set was labeled as a feature signal group independently characterizing electromagnetic and vibration states.

[0023] In the specific implementation, the raw response signal cluster captured during the permanent magnet variable frequency submersible pump test undergoes time-dependent signal decoupling and data augmentation processing, which is achieved through a series of signal processing operations. The raw response signal cluster contains synchronous time-series data from pressure sensors, flow meters, vibration acceleration sensors, and current transformers. The processing begins with stationarity testing and detrending of the raw response signal cluster. The stationarity test uses the unit root test method to determine the time-varying nature of the signal mean and variance. For non-stationary signal segments, detrending processing is performed to eliminate baseline drift components introduced by the integrated test environment. The detrending processing uses a polynomial fitting method to remove slowly changing trend terms from the signal. In the specific implementation, a set of parallel adaptive filter banks is used to perform frequency band stripping on the detrending processed signal. The center frequency and bandwidth of the adaptive filter bank are dynamically configured according to the impeller's design speed and its harmonic frequencies, thereby extracting the core frequency band signals directly related to the impeller's rotational frequency and its harmonics. These extracted core frequency band signals constitute the hydraulic performance raw signal set, which mainly includes pressure and flow signal components related to head and flow pulsation. It is understandable that while extracting the original hydraulic performance signal set, it is necessary to process the broadband signal remaining after frequency band stripping. The independent component analysis algorithm is used to perform blind source separation on the broadband residual signal. The goal of the independent component analysis algorithm is to find a linear transformation that maximizes the statistical independence between the output signal components, thereby decoupling the mutually orthogonal electromagnetic noise components and mechanical vibration components. The decoupled electromagnetic noise components mainly come from the magnetic field harmonics of the permanent magnet synchronous motor, while the mechanical vibration components are related to mechanical components such as bearings and shafts. These separated components together constitute the original electromagnetic and vibration signal set.

[0024] In some embodiments, data augmentation processing is performed on the raw hydraulic performance signal set to expand the signal samples to cover a wider range of operating condition fluctuations. One augmentation process involves adding noise with a specific signal-to-noise ratio proportional to the instantaneous flow rate value to the pressure and flow sequences in the raw hydraulic performance signal set. The power of the noise is determined by a preset signal-to-noise ratio value and the instantaneous flow rate amplitude. Another augmentation process involves performing time-domain stretching and compression transformations on the pressure pulse signals in the raw hydraulic performance signal set. Time-domain stretching is achieved by reducing the sampling rate of the signal time axis through linear interpolation, while time-domain compression is achieved by extracting signal data points, thereby simulating changes in the system response speed. In some embodiments, data augmentation processing is performed on the raw electromagnetic and vibration signal sets, focusing on simulating potential state changes of physical components. One augmentation process involves synthesizing vibration waveforms simulating different bearing clearance states. This is based on a bearing dynamics model, generating simulated vibration signals with different characteristic frequencies and amplitudes by adjusting the radial clearance parameters in the model. Another enhancement process involves generating current harmonics under simulated slight stator-rotor misalignment conditions. This is achieved by injecting specific harmonic components into the three-phase current fundamental frequency. The amplitude and phase of these harmonic components are set according to a preset misalignment degree. In practice, the original hydraulic performance signal set after all enhancement processes are formally labeled as a feature signal set independently characterizing hydraulic performance, and the original electromagnetic and vibration signal sets after all enhancement processes are labeled as feature signal sets independently characterizing electromagnetic and vibration states. These two feature signal sets will be used for subsequent simulation analysis. A specific calculation formula for detrending processing is:

[0025] in: Indicates at a point in time The detrended signal value, Indicates at a point in time The original signal value, Indicates at a point in time The fitted polynomial trend value. It can be understood that the polynomial... The order of the algorithm needs to be selected based on the complexity of the baseline drift components in the original response signal cluster to avoid overfitting or underfitting.

[0026] Example 2: The process of inputting the characteristic signal set into a dynamic operating condition simulator and reconstructing the internal flow field pressure cloud map and rotor stress distribution map of the target submersible pump under simulated operating conditions using an iterative approximation algorithm is as follows: The characteristic signal set independently characterizing hydraulic performance is mapped to the boundary condition driving source of the dynamic operating condition simulator. This driving source includes the inlet pressure pulsation sequence and the outlet flow fluctuation sequence. The characteristic signal set independently characterizing electromagnetic and vibration states is converted into the internal excitation loading source of the dynamic operating condition simulator. This loading source includes a time-varying electromagnetic force density field and a dynamic radial force vector. The fluid domain mesh and solid domain mesh in the dynamic operating condition simulator are initialized, and the material constitutive relation and fluid-structure interaction interface parameters are set. Using the boundary condition driving source and the internal excitation loading source as input, the coupled solver of computational fluid dynamics and computational structural dynamics is run. After each calculation time step, the residual between the virtual sensing signal output by the coupled solver and the corresponding characteristic signal set is compared. Based on the magnitude and direction of the residual, the amplitude of the boundary condition driving source and the application point of the internal excitation loading source are dynamically adjusted using an iterative approximation algorithm until the residual converges to a preset threshold. When the residuals converge, the fluid domain pressure distribution, i.e., the internal flow field pressure cloud map, and the solid domain stress distribution, i.e., the rotor stress distribution map, are finally calculated by the dynamic operating condition simulator at the current time step.

[0027] In practice, the characteristic signal sets independently representing hydraulic performance and independently representing electromagnetic and vibration states are input into the dynamic operating condition simulator. An iterative approximation algorithm is used to reconstruct the internal flow field pressure cloud map and rotor stress distribution map of the target submersible pump under simulated operating conditions. The implementation process follows a defined calculation flow. The characteristic signal sets independently representing hydraulic performance are mapped to the boundary condition driving source of the dynamic operating condition simulator. The mapping relationship is that the pressure pulse signal sequence in the characteristic signal set, after amplitude calibration, is directly set as the inlet pressure pulsation sequence of the fluid domain, while the flow fluctuation signal sequence in the characteristic signal set is set as the outlet flow fluctuation sequence of the fluid domain. The characteristic signal sets independently representing electromagnetic and vibration states are converted into the internal excitation loading source of the dynamic operating condition simulator. The conversion process involves calculating the current harmonic components in the characteristic signal set into a time-varying electromagnetic force density field using the Maxwell stress tensor method, and calculating the vibration acceleration signal in the characteristic signal set into a dynamic radial force vector acting on the rotor surface after quadratic integration and directional decomposition. In the specific implementation, the fluid domain mesh and solid domain mesh in the dynamic working condition simulator are initialized. The fluid domain mesh generates a mixture of tetrahedral and prismatic elements based on the three-dimensional flow channel model of the pump, and the solid domain mesh generates a hexahedral dominant mesh based on the three-dimensional geometric model of the rotor and pump casing. The material constitutive relations and fluid-structure interaction interface parameters are set. The material constitutive relations include the turbulence model equations of the fluid and the elastic-plastic model of the solid. The fluid-structure interaction interface parameters define the interpolation method for data transfer and the time step synchronization strategy.

[0028] In some embodiments, a coupled solver of computational fluid dynamics and computational structural dynamics is run, using boundary condition driving sources and internal excitation loading sources as inputs. The coupled solver employs a partitioned strongly coupled algorithm, iteratively solving the fluid and solid equations at each physical time step until the displacement and force equilibrium conditions at the interface are satisfied. After each computational time step, the residual between the virtual sensing signal output by the coupled solver and the corresponding characteristic signal set is compared. The virtual sensing signal is data collected at virtual measurement points corresponding to the physical sensor positions within the dynamic condition simulator, including virtual pressure, virtual flow rate, virtual vibration acceleration, and virtual current signals. It can be understood that, based on the magnitude and direction of the residuals, the amplitude of the boundary condition driving source and the application point of the internal excitation loading source are dynamically adjusted through an iterative approximation algorithm. The adjustment strategy is based on the gradient information of the residual vector; the amplitude adjustment ratio of the driving source is proportional to the amplitude error of the pressure and flow rate components in the residuals, and the application point of the loading source is fine-tuned along the phase direction indicated by the vibration component in the residuals. In some embodiments, the iterative approximation process continues until the L2 norm of the residual vector converges to a preset threshold, which is determined based on the original measurement accuracy of the feature signal set. When the residual converges, the fluid domain pressure distribution, i.e., the internal flow field pressure cloud map, and the solid domain stress distribution, i.e., the rotor stress distribution map, calculated by the dynamic operating condition simulator at the current time step, are derived. The derived data is an array containing spatial coordinates and physical quantity values. The formula used to update the boundary condition driving source parameters in the iterative approximation algorithm is:

[0029] in: This represents the adjusted boundary condition driving source magnitude vector. This represents the boundary condition driving source magnitude vector before adjustment. This represents the learning rate coefficient. This represents the signal amplitude vector corresponding to the characteristic signal group that independently characterizes hydraulic performance. This represents the amplitude vector in the virtual sensing signal output by the coupled solver. Optional, the learning rate coefficient. The value of decays as the number of iterations increases. In practice, the adjustment of the application point of the internal excitation loading source follows a similar principle, but the direction of correction of the application point coordinates is determined by the phase difference between the virtual vibration signal and the characteristic vibration signal. It can be understood that through the above iterative process, the simulated working condition in the dynamic working condition simulator is continuously corrected, eventually making its output signal consistent with the input characteristic signal group in statistical characteristics. At this time, the internal state of the simulator is regarded as a high-fidelity reproduction of the real physical state.

[0030] Example 3: See Figure 3The following steps are taken to deduce the potential failure modes of the target submersible pump by integrating the internal flow field pressure cloud map and the rotor stress distribution map: First, a spatiotemporal correlation matrix is ​​established between the high-pressure region in the internal flow field pressure cloud map and the high-stress region in the rotor stress distribution map. This matrix records the spatial and temporal correspondence between the two. Second, the spatiotemporal correlation matrix is ​​scanned to identify region pairs that meet a preset correlation strength threshold; these region pairs are defined as high-pressure-high-stress coupling risk zones. Third, for each high-pressure-high-stress coupling risk zone, a material fatigue characteristic database is queried to obtain the predicted crack initiation life under the current pressure-stress cycle characteristics. Fourth, the low-frequency pressure pulsation patterns present in the internal flow field pressure cloud map are analyzed, and the periodic strain energy accumulation caused by these patterns on the target submersible pump casing structure is calculated. Fifth, the rotor stress distribution map is checked for locations where the stress concentration gradient exceeds a critical value, and the probability of brittle fracture is assessed using a statistical model of material micro-defects. Sixth, the predicted crack initiation life, periodic strain energy accumulation, and probability of brittle fracture are integrated to form a potential failure mode set including fatigue failure, excessive deformation failure, and brittle fracture failure. For each failure mode in the potential failure mode set, its load history in the dynamic condition simulator is traced back to generate an evolution trajectory describing its development from the initial state to the critical state.

[0031] The method for generating enhanced test procedures to guide physical entity testing involves reverse-correcting the submersible pump's baseline data based on its evolution trajectory. This involves extracting key load events and their timing from the evolution trajectory, including peak pressure, peak stress, and the occurrence of vibration frequencies. These key load events are then compared with standard operating parameters in the submersible pump's baseline data to identify load condition gaps not covered by the standard parameters. For each load condition gap, a supplementary test program is designed, specifying the inverter's specific frequency modulation waveform, the step change point of the controllable load device, and the data acquisition sampling rate. This supplementary test program is then integrated with the standard test program developed based on the submersible pump's baseline data to ensure that the order of load application matches the timing logic of load occurrences in the evolution trajectory. For each monitoring point in the integrated test procedure, a specific sensor combination selected from multi-parameter sensing units and data recording trigger conditions are configured, forming a complete set of instructions and parameters for the enhanced test procedure.

[0032] In practical implementation, the internal flow field pressure cloud map and rotor stress distribution map are integrated to deduce the set of potential failure modes of the target submersible pump. This process is based on spatial and temporal correlation analysis. A spatiotemporal correlation matrix is ​​established between the high-pressure region in the internal flow field pressure cloud map and the high-stress region in the rotor stress distribution map. Each element of the spatiotemporal correlation matrix records the normalized product of the fluid pressure value and the solid stress value, as well as their phase difference, at a specific spatial grid cell and a specific time step, thereby quantifying the correspondence between the two in spatial location and temporal phase. The spatiotemporal correlation matrix is ​​scanned to identify region pairs that meet a preset correlation strength threshold. The correlation strength threshold is preset according to material properties and operating conditions. Any region pair whose spatiotemporal correlation matrix element value exceeds this threshold is defined as a high-pressure-high-stress coupling risk zone. For each high-pressure-high-stress coupling risk zone, a material fatigue characteristic database is queried. The database stores the SN curves and ε-N curves of the pump body and rotor materials to obtain the predicted value of material crack initiation life under the current pressure-stress cycle characteristics. This predicted value is calculated based on the modified Miner linear cumulative damage rule. It is understandable that the low-frequency pressure pulsation mode present in the internal flow field pressure cloud map is analyzed. The main frequency component of the low-frequency pressure pulsation mode is extracted by the spectrum analysis of the pressure signal, and the periodic strain energy accumulation caused by it on the target submersible pump casing structure is calculated. The strain energy accumulation is obtained by integrating the stress-strain curve within a complete pressure pulsation cycle.

[0033] In some embodiments, the rotor stress distribution map is checked for locations where the stress concentration gradient exceeds a critical value. The stress concentration gradient is obtained by calculating the rate of change of the stress field in space. The critical value is determined based on the fracture toughness of the material and the probability of brittle fracture is assessed using a statistical model of material micro-defects. This statistical model describes the defect size and distribution based on the Weibull distribution. The predicted crack initiation life, accumulated periodic strain energy, and probability of brittle fracture are integrated to form a set of potential failure modes, including fatigue failure, excessive deformation failure, and brittle fracture failure. Each failure mode entry includes the failure type, location coordinates, triggering conditions, and severity level. For each failure mode in the potential failure mode set, its load history in the dynamic operating condition simulator is traced back. The load history records the complete time series of physical quantities such as pressure, stress, and temperature acting on the location from the start of the simulation to the identification of the failure risk. An evolution trajectory describing its development from the initial state to the critical state is generated, expressed as a load-time curve or a damage parameter-time curve.

[0034] In some embodiments, the submersible pump body baseline data is reverse-corrected based on the evolution trajectory to generate an enhanced test procedure. Specifically, key load events and their timing are extracted from the evolution trajectory. Key load events include the occurrence times of pressure peaks, stress peaks, and vibration frequencies. The extraction rules are based on signal local extremum detection and characteristic frequency identification. The key load events are compared with standard operating parameters in the submersible pump body baseline data. Standard operating parameters typically include rated flow rate, rated head, and standard speed curves. Load condition gaps not covered by the standard operating parameters are identified. Load condition gaps refer to specific load combinations or change sequences that appear in the evolution trajectory but are not applied by the standard test procedure.

[0035] In practical implementation, a supplementary test program is designed for each load condition gap. This supplementary test program clearly specifies the specific frequency modulation waveform of the inverter, the step change point of the controllable load device, and the data acquisition sampling rate. The frequency modulation waveform is used to reproduce the speed fluctuations in the evolution trajectory, and the step change point is used to simulate sudden load changes. The supplementary test program is integrated with the standard test program developed based on the submersible pump's baseline data. This integration ensures that the application sequence of the test loads matches the temporal logic of the load occurrences in the evolution trajectory; for example, high-pressure peak events and high-stress peak events are applied in time. For each monitoring point in the integrated test process, a specific sensor combination selected from the multi-parameter sensing unit and data recording trigger conditions are configured. The data recording trigger conditions can be set to initiate high-speed recording when a channel signal exceeds a threshold, thus forming a complete set of instructions and parameters for the enhanced test procedure. Identification of the high-pressure-high-stress coupling risk zone involves calculating the spatiotemporal correlation strength, calculated using the following formula:

[0036] in: Represents spatial grid cells The spatiotemporal correlation strength value, This represents the total number of time steps. Indicates at time step Time unit The fluid pressure value, This represents the average pressure in the fluid domain. Indicates at time step The equivalent stress value of the element at the corresponding position. This represents the average equivalent stress in the solid domain. Indicates at time step The phase difference between pressure fluctuations and stress fluctuations. Optionally, when The value is greater than the preset association strength threshold. At that time, this spatial location was marked as a high-pressure-high-stress coupling risk zone. It can be understood that the generation of the evolution trajectory relies on in-depth analysis of the output data from the dynamic working condition simulator, while the reverse correction process enables physical testing to proactively explore potential risk points discovered in the virtual simulation.

[0037] Example 4: An enhanced test procedure is executed to acquire high-fidelity real-world test data, and the high-fidelity real-world test data is then calibrated against the simulation results under the given operating conditions. Based on the calibrated data stream, a digital twin of the target submersible pump is constructed, and a preset accelerated degradation excitation signal is injected into the digital twin. The response sequence of the digital twin under the accelerated degradation excitation signal is monitored, and performance inflection points and structural weaknesses are identified based on the response sequence. A maintenance intervention instruction set with sequential execution logic is generated based on the spatial coordinates of the performance inflection points and structural weaknesses.

[0038] The specific steps for executing the enhanced test procedure, acquiring high-fidelity measured data, and calibrating the consistency between the high-fidelity measured data and the simulation results include: loading the target submersible pump on the physical test platform and operating the variable frequency drive unit and controllable load device according to the instructions and parameter set of the enhanced test procedure. Through the calibrated multi-parameter sensing unit, the full-channel response of the target submersible pump during the execution of the enhanced test procedure is recorded synchronously to obtain high-fidelity measured data. Data segments corresponding to the simulation conditions in terms of time scale and load type are extracted from the high-fidelity measured data as a calibration comparison dataset. The statistical characteristics of the hydraulic performance data and the internal flow field pressure cloud map in the calibration comparison dataset are compared to calculate the pressure distribution similarity index and the flow prediction error index. The vibration and current data in the calibration comparison dataset are compared with the rotor stress distribution map and the derived signals of electromagnetic excitation to calculate the vibration mode correlation coefficient and the difference in current harmonic distortion. Based on the pressure distribution similarity index, the flow prediction error index, the vibration mode correlation coefficient, and the difference in current harmonic distortion, a set of model correction coefficients is generated. The model correction coefficients are applied to globally adjust the material property parameters and boundary condition models in the dynamic working condition simulator to complete the consistency calibration.

[0039] In practice, the enhanced test procedure is executed to obtain high-fidelity measured data and calibrate it against the simulation results. This process begins with the preparation and operation of the physical test platform. On the physical test platform, the target submersible pump is loaded according to the instructions and parameter set of the enhanced test procedure. The variable frequency drive unit outputs a frequency modulation waveform, and the controllable load device is controlled to change the load torque at a preset step change point. The calibrated multi-parameter sensing unit synchronously records the full-channel response of the target submersible pump during the execution of the enhanced test procedure. The full-channel response includes inlet pressure, outlet pressure, flow rate, three-phase current, multi-directional vibration acceleration, and speed signals, thus obtaining high-fidelity measured data. Data segments corresponding to the simulated conditions in terms of time scale and load type are extracted from the high-fidelity measured data as a calibration comparison dataset. The extraction operation is performed according to the principles of timestamp alignment and load pattern matching. It is understandable that the statistical characteristics of the hydraulic performance data in the calibration comparison dataset are compared with those of the internal flow field pressure cloud map, and the pressure distribution similarity index and the flow prediction error index are calculated. The pressure distribution similarity index is obtained by comparing the correlation between the measured pressure profile and the pressure profile derived from the cloud map at corresponding spatial points, and the flow prediction error index is the root mean square error between the simulated predicted flow time series and the measured flow time series.

[0040] In some embodiments, the vibration and current data in the calibration comparison dataset are compared with the rotor stress distribution map and the derived signal of electromagnetic excitation to calculate the vibration mode correlation coefficient and the difference in current harmonic distortion. The vibration mode correlation coefficient is obtained by extracting the dominant frequency components of the measured vibration signal and the theoretical vibration signal derived from the dynamic response of the rotor stress distribution map, and calculating their spectral coherence. The difference in current harmonic distortion is the absolute difference between the total harmonic distortion of the measured current signal and the theoretical harmonic distortion calculated from the derived signal of electromagnetic excitation. Based on the pressure distribution similarity index, the flow prediction error index, the vibration mode correlation coefficient, and the difference in current harmonic distortion, a set of model correction coefficients is generated. The model correction coefficients are used to quantitatively adjust the physical parameters in the dynamic working condition simulator. The model correction coefficients are applied to globally adjust the material property parameters and boundary condition models in the dynamic working condition simulator. Material property parameters include the material elastic modulus and density, and boundary condition models include turbulence model constants and fluid-structure interaction interface damping coefficients, thereby completing the consistency calibration between the measured data and the simulation results with high fidelity. Refer to Table 1 for the quantitative calculation results of various indicators during the calibration process.

[0041] Table 1: Calculation Table of Consistency Calibration Indicators

[0042] Based on the data stream after consistency calibration, a digital twin of the target submersible pump is constructed, and a preset accelerated degradation excitation signal is injected into the digital twin. The accelerated degradation excitation signal is a load spectrum with enhanced amplitude or frequency, designed according to the evolution trajectory of a set of potential failure modes, used to accelerate the simulated performance degradation process of the submersible pump in virtual space. The response sequence of the digital twin under the accelerated degradation excitation signal is monitored. The response sequence includes virtual performance parameters and virtual structural state parameters output by the digital twin. Based on the response sequence, performance inflection points and structural weaknesses are identified. The performance inflection point is the turning point that marks the beginning of a nonlinear decline in key performance indicators such as efficiency or head. Structural weaknesses are local areas in the digital twin where stress, strain, or deformation continuously increases and approaches the allowable value of the material. In some embodiments, a maintenance intervention instruction set with sequential execution logic is generated based on the spatial coordinates of the performance inflection point and structural weaknesses, the spatial coordinates being associated with specific locations in the three-dimensional model of the digital twin. The formula for calculating the comprehensive calibration deviation and deriving the model correction coefficient is expressed as:

[0043] in: This represents the model correction coefficient for the elastic modulus of the material. This represents the weighting factor associated with the vibration mode. Represents the correlation coefficient of vibration modes. This represents the weighting factor related to current harmonic distortion. This indicates the difference in current harmonic distortion. Optional, weighting factor. and A multi-objective optimization algorithm is used to ensure that all calibration indicators converge collaboratively. In practice, the operation of generating maintenance intervention instruction sets is a direct output of the digital twin analysis function, mapping the risks identified in the virtual space to specific maintenance action sequences in the physical world. The entire process can be understood as starting with physical testing and verification, improving the fidelity of the digital model through a consistency calibration closed loop, then using a high-fidelity digital twin for in-depth life and failure analysis, and finally outputting explicit instructions to guide physical maintenance.

[0044] See Figure 4This is a 3D scatter plot of stress distribution in the structural weak points of a submersible pump body. It displays the stress distribution within the pump body and marks the weak areas corresponding to high stress. The stress value distribution ranges from 100-250 MPa. High-stress areas (yellow dots) are mainly concentrated in the central region of the pump body, at X=2-6 cm, Y=2-6 cm, and Z=2-6 cm. The structural weak points highly overlap with the high-stress areas, indicating that the stress in these areas is close to or exceeds the allowable value of the material, representing potential failure risk points. The stress values ​​in the edge areas of the pump body are lower, and the risk level is relatively lower. This chart is one of the core outputs of the digital twin degradation analysis of submersible pumps. By using spatial stress distribution and weak point marking, the structural risk areas of the pump body can be accurately located, providing spatial coordinates for subsequent maintenance interventions (such as local reinforcement and key monitoring).

[0045] Example 5: Using a dynamic operating condition simulator that has undergone consistency calibration as the core computing engine, a physical behavior simulation layer is encapsulated to form a digital twin. A data interface layer is established for the digital twin, capable of receiving high-fidelity measured data from the physical test platform or historical database as input in real time. A state mapping layer is constructed for the digital twin, converting the input real-time data into boundary and initial conditions recognizable by the physical behavior simulation layer. A virtual sensor network is created in the digital twin, completely corresponding in spatial location and type to the multi-parameter sensing units on the physical test platform, used to output virtual readings comparable to those of physical sensors. The method for generating a maintenance intervention instruction set with sequential execution logic based on the spatial coordinates of performance inflection points and structural weak points is as follows: The running time and load conditions corresponding to the performance inflection point are translated into a preventive maintenance time window suggestion for the physical submersible pump. The spatial coordinates of the structural weak points are mapped onto the 3D design drawing of the target submersible pump, identifying the geometric areas requiring enhanced inspection or treatment. Based on the relative speed of development of different failure modes in the evolution trajectory, the preventive maintenance time window suggestions and geometric areas are prioritized. For each high-priority maintenance item, specify the specific maintenance operation type, required testing tools, and acceptance criteria for determining whether the repair is satisfactory. Encode all maintenance items, their corresponding priorities, operation types, testing tools, and acceptance criteria into a machine-readable set of maintenance intervention instructions in the order of execution.

[0046] Using a dynamically calibrated simulator as the core computing engine, a physical behavior simulation layer is encapsulated to form the digital twin. This layer includes all calibrated fluid dynamics, structural mechanics, and electromagnetic field calculation modules, capable of performing simulations synchronously with or offline from the physical entity. A data interface layer is established for the digital twin, defining a standardized data communication protocol. This layer can receive high-fidelity measured data from the physical testing platform or historical database in real time as input. The input data stream is parsed immediately and sent to subsequent processing stages. A state mapping layer is constructed for the digital twin. This layer converts the high-fidelity measured data into boundary and initial conditions recognizable by the physical behavior simulation layer through feature extraction and parameter identification algorithms. For example, real-time current and rotational speed are mapped to input parameters of the electromagnetic force density field, and real-time inlet pressure is mapped to boundary conditions of the fluid domain. A virtual sensor network is created in the digital twin. The virtual sensor network corresponds completely to the multi-parameter sensor unit on the physical test platform in terms of spatial location coordinates, measurement direction and signal type. The virtual sensor network samples the calculation results from the physical behavior simulation layer and outputs virtual readings that are comparable to the physical sensor readings for subsequent comparative analysis and status diagnosis.

[0047] In some embodiments, a set of maintenance intervention instructions with sequential execution logic is generated based on the spatial coordinates of performance inflection points and structural weak points. This process transforms virtual analysis conclusions into executable physical operation guidelines. The operating time and load conditions corresponding to the performance inflection point are translated into preventive maintenance time window recommendations for the physical submersible pump. The translation process uses the position of the performance inflection point on the accelerated degradation time axis and back-calculates the actual operating time using a time scaling ratio. The spatial coordinates of structural weak points are mapped onto the 3D design drawings of the target submersible pump. The mapping operation is achieved through coordinate transformation, identifying geometric areas that require enhanced inspection or treatment, such as highlighting the blade root area or bearing housing area on the 3D drawing. It can be understood that the preventive maintenance time window recommendations and geometric areas are prioritized according to the relative development speed of different failure modes in the evolution trajectory, with faster-developing failure modes corresponding to higher maintenance priorities. For each high-priority maintenance item, a specific maintenance operation type, required testing tools, and acceptance criteria for judging whether the repair is acceptable are specified. Maintenance operation types include, but are not limited to, non-destructive testing, dynamic balancing, and clearance adjustment. Acceptance criteria are set according to industry specifications or manufacturer technical conditions.

[0048] In practice, all maintenance items, their corresponding priorities, operation types, testing tools, and acceptance criteria are encoded into a machine-readable maintenance intervention instruction set according to their execution order. The encoding format adopts a structured data format such as JSON or XML. The instruction set includes a unique task identifier, a list of execution steps, and logical dependencies between steps. The decision formula for calculating the priority of maintenance tasks can be expressed as:

[0049] in: Indicates the first The overall priority score of each maintenance task. This represents an adjustment coefficient related to time urgency. This represents the remaining time from the current moment until the predicted performance inflection point or failure threshold. This represents an adjustment coefficient related to the severity of structural risk. This indicates the severity score of the structural weaknesses. (Optional, severity score) It can be quantified by the ratio of the maximum equivalent stress output by the digital twin to the material's yield strength, or by the cumulative fatigue damage value. In some embodiments, an adjustment coefficient... and The value is determined through optimization using domain expert knowledge or historical maintenance data. It can be understood that the generated maintenance intervention instruction set can be directly integrated into the enterprise's computerized maintenance management system, driving the automatic generation and dispatch of maintenance work orders. The continuous operation of the digital twin and the generation of maintenance instructions form a closed loop from status monitoring and intelligent diagnosis to predictive maintenance, enabling the submersible pump testing and maintenance strategy to shift from a fixed-cycle-based model to a precise model based on actual conditions and predicted risks.

[0050] See Figure 5This is a time-series graph showing the changes in multiple indicators (flow rate, head, and efficiency) of a submersible pump during operation, illustrating the dynamic correlation of core performance parameters over a 100-minute operating period. In the initial phase (0-60 minutes): flow rate and head show a synchronous fluctuation trend, while efficiency fluctuates slightly with both (maintaining between 70% and 85%), indicating a strong correlation between the parameters. In the later phase (60-100 minutes): the fluctuation range of flow rate and head increases, and efficiency shows a significant decrease (from 80% to around 65%), with a weakening synchronicity with the flow rate / head ratio, indicating that equipment performance is beginning to degrade. The head generally remains in the 80-120m range, and the flow rate is concentrated in the 40-120mL / h range, but the fluctuation range of both in the later phase corresponds to the decrease in efficiency, a typical manifestation of abnormal performance. This type of graph is a core tool for monitoring the operating status of submersible pumps. Through multi-parameter time-series correlation analysis, early signs of performance degradation (such as a weakened correlation between efficiency and flow rate / head) can be identified in a timely manner, providing a basis for equipment maintenance or operating condition adjustments.

[0051] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A testing method for a permanent magnet variable frequency submersible pump, characterized in that, The method includes: An integrated test environment for the target submersible pump is constructed using a controllable load device and a multi-parameter sensing unit, and the initial environmental parameters of the integrated test environment and the baseline data of the submersible pump body are obtained. The output mode of the variable frequency drive unit is adjusted according to the initial environmental parameters to guide the target submersible pump into a multi-stage variable speed operation state, and the original response signal clusters fed back by the multi-parameter sensing unit are captured simultaneously. The original response signal cluster is subjected to time-dependent signal decoupling and data augmentation processing to separate the feature signal group that independently characterizes hydraulic performance and the feature signal group that independently characterizes electromagnetic and vibration states. The characteristic signal group is input into a dynamic working condition simulator, and the internal flow field pressure cloud map and rotor stress distribution map of the target submersible pump under simulated working conditions are restored by an iterative approximation algorithm. By integrating the internal flow field pressure cloud map and the rotor stress distribution map, the set of potential failure modes of the target submersible pump is deduced, and the evolution trajectory corresponding to each potential failure mode is marked. The baseline data of the submersible pump body is reverse-corrected based on the evolution trajectory to generate an enhanced test procedure to guide the physical entity test.

2. The test method for permanent magnet variable frequency submersible pump according to claim 1, characterized in that, The process of performing time-dependent signal decoupling and data augmentation on the original response signal cluster separates the characteristic signal groups that independently characterize hydraulic performance and the characteristic signal groups that independently characterize electromagnetic and vibration states, including: The original response signal cluster is subjected to stationarity testing and detrending processing to eliminate baseline drift components introduced by the integrated test environment; An adaptive filter bank is used to strip the frequency band of the processed signal and extract the core frequency band signal that is directly related to the impeller rotation frequency and its harmonics to form the original hydraulic performance signal set. Meanwhile, the independent component analysis algorithm is used to perform blind source separation on the broadband residual signal, decouple the mutually orthogonal electromagnetic noise components and mechanical vibration components, and form a set of original electromagnetic and vibration signals; Data augmentation processing is performed on the raw hydraulic performance signal set, including adding noise with a specific signal-to-noise ratio proportional to the flow rate and performing time-domain stretching and compression transformation on the pressure pulse signal; Data augmentation processing is performed on the original electromagnetic and vibration signal set. The data augmentation processing includes synthesizing vibration waveforms simulating different bearing clearance states and generating current harmonics simulating slight misalignment of stator and rotor. The enhanced hydraulic performance original signal set is labeled as the independent characteristic signal group representing hydraulic performance, and the enhanced electromagnetic and vibration original signal set is labeled as the independent characteristic signal group representing electromagnetic and vibration states.

3. The test method for permanent magnet variable frequency submersible pump according to claim 2, characterized in that, The step of inputting the characteristic signal set into a dynamic operating condition simulator and reconstructing the internal flow field pressure cloud map and rotor stress distribution map of the target submersible pump under simulated operating conditions through an iterative approximation algorithm includes: The independent characteristic signal set representing hydraulic performance is mapped to the boundary condition driving source of the dynamic operating condition simulator. The boundary condition driving source includes the inlet pressure pulsation sequence and the outlet flow fluctuation sequence. The group of characteristic signals that independently characterize electromagnetic and vibration states is converted into an internal excitation loading source for the dynamic working condition simulator. The internal excitation loading source includes a time-varying electromagnetic force density field and a dynamic radial force vector. Initialize the fluid domain mesh and solid domain mesh in the dynamic working condition simulator, and set the material constitutive relation and fluid-structure interaction interface parameters; Using the boundary condition driving source and the internal excitation loading source as input, run a coupled solver for computational fluid dynamics and computational structural dynamics; After each computation time step, the residual between the virtual sensing signal output by the coupled solver and the corresponding set of feature signals is compared. Based on the magnitude and direction of the residual, the amplitude of the boundary condition driving source and the point of action of the internal excitation loading source are dynamically adjusted by the iterative approximation algorithm until the residual converges to a preset threshold. When the residual converges, the fluid domain pressure distribution, i.e. the internal flow field pressure cloud map, and the solid domain stress distribution, i.e. the rotor stress distribution map, are finally calculated by the dynamic operating condition simulator at the current time step.

4. The test method for permanent magnet variable frequency submersible pump according to claim 3, characterized in that, The method of integrating the internal flow field pressure cloud map and the rotor stress distribution map to deduce the set of potential failure modes of the target submersible pump includes: A spatiotemporal correlation matrix is ​​established between the high-pressure region in the internal flow field pressure cloud diagram and the high-stress region in the rotor stress distribution diagram. The spatiotemporal correlation matrix records the correspondence between the two in terms of spatial location and time phase. Scan the spatiotemporal correlation matrix to identify region pairs that meet a preset correlation strength threshold. These region pairs are defined as high-pressure-high-stress coupling risk zones. For each of the high-pressure-high-stress coupling risk zones, the material fatigue characteristic database is queried to obtain the predicted value of material crack initiation life under the current pressure-stress cycle characteristics; The low-frequency pressure pulsation patterns present in the internal flow field pressure cloud map are analyzed, and the periodic strain energy accumulation caused by them on the target submersible pump casing structure is calculated. Check whether there are any areas in the rotor stress distribution diagram where the stress concentration gradient exceeds the critical value, and assess the probability of brittle fracture by combining the statistical model of material micro-defects. By integrating the predicted crack initiation life of the material, the accumulated periodic strain energy, and the probability of brittle fracture, a set of potential failure modes including fatigue failure, excessive deformation failure, and brittle fracture failure is formed. For each failure mode in the set of potential failure modes, its load history in the dynamic operating condition simulator is traced back to generate the evolution trajectory describing its development from the initial state to the critical state.

5. The test method for permanent magnet variable frequency submersible pump according to claim 1, characterized in that, The step of reversely correcting the baseline data of the submersible pump body based on the evolution trajectory to generate an enhanced test procedure for guiding physical entity testing includes: Key load events and their timing are extracted from the evolution trajectory. The key load events include pressure peaks, stress peaks, and the times when specific vibration frequencies occur. The critical load events are compared with the standard operating parameters in the submersible pump body baseline data to identify load condition gaps not covered by the standard operating parameters. For each load condition gap, a supplementary test program is designed, which specifies the specific frequency modulation waveform of the frequency converter, the step change point of the controllable load device, and the sampling rate of data acquisition. The supplementary test procedure is integrated and arranged with the standard test procedure based on the submersible pump body baseline data to ensure that the order of application of test loads is consistent with the timing logic of the loads appearing in the evolution trajectory. To integrate each monitoring point in the orchestrated test process, a specific combination of sensors selected from the multi-parameter sensing unit and data recording trigger conditions are configured to form a complete set of instructions and parameters for the enhanced test procedure.

6. The test method for permanent magnet variable frequency submersible pump according to claim 1, characterized in that, The method further includes: The enhanced test procedure is executed to obtain high-fidelity real test data, and the high-fidelity real test data is calibrated to match the simulation results under the simulated working conditions. Based on the data stream after consistency calibration, a digital twin of the target submersible pump is constructed, and a preset accelerated degradation excitation signal is injected into the digital twin. Monitor the response sequence of the digital twin under the accelerated degradation excitation signal, and identify performance inflection points and structural weaknesses based on the response sequence; Based on the spatial coordinates of the performance inflection point and the structural weak link, a maintenance intervention instruction set with sequential execution logic is generated.

7. The test method for permanent magnet variable frequency submersible pump according to claim 6, characterized in that, The process of executing the enhanced test procedure, obtaining high-fidelity real-world test data, and calibrating the consistency between the high-fidelity real-world test data and the simulation results under the simulated operating conditions includes: The target submersible pump is mounted on the physical test platform, and the variable frequency drive unit and the controllable load device are operated in accordance with the instructions and parameter set of the enhanced test procedure. The high-fidelity test data is obtained by synchronously recording the full-channel response of the target submersible pump during the execution of the enhanced test procedure using the calibrated multi-parameter sensing unit. Data segments corresponding to the simulated working conditions in terms of time scale and load type are extracted from the high-fidelity measured data and used as calibration comparison datasets. The hydraulic performance data in the calibration comparison dataset is compared with the statistical characteristics of the internal flow field pressure cloud map to calculate the pressure distribution similarity index and the flow prediction error index. The vibration and current data in the calibration comparison dataset are compared with the rotor stress distribution map and the derived signal of electromagnetic excitation to calculate the vibration mode correlation coefficient and the difference in current harmonic distortion. A set of model correction coefficients are generated based on the pressure distribution similarity index, the flow prediction error index, the vibration mode correlation coefficient, and the difference in current harmonic distortion. The model correction coefficients are applied to globally adjust the material property parameters and boundary condition models in the dynamic working condition simulator to complete the consistency calibration.

8. The test method for permanent magnet variable frequency submersible pump according to claim 7, characterized in that, The construction of a digital twin of the target submersible pump based on the data stream after consistency calibration includes: Using the dynamic working condition simulator after the consistency calibration as the core computing engine, a physical behavior simulation layer of the digital twin is encapsulated. A data interface layer for the digital twin is established, which can receive high-fidelity test data from a physical test platform or historical database as input in real time. A state mapping layer is constructed for the digital twin, which converts the input real-time data into boundary conditions and initial conditions that can be recognized by the physical behavior simulation layer; A virtual sensing network is created in the digital twin, which corresponds completely to the multi-parameter sensing unit on the physical test platform in terms of spatial location and type, and is used to output virtual readings comparable to those of the physical sensors.

9. The test method for permanent magnet variable frequency submersible pump according to claim 8, characterized in that, The step of generating a maintenance intervention instruction set with sequential execution logic based on the spatial coordinates of the performance inflection point and the structural weak link includes: The operating time and load conditions corresponding to the performance inflection point are translated into a recommended preventive maintenance time window for physical submersible pumps. The spatial coordinates of the structural weak points are mapped onto the three-dimensional design drawings of the target submersible pump, and the geometric areas that need to be inspected or treated are identified. Based on the relative speed of development of different failure modes in the evolution trajectory, the preventive maintenance time window recommendations and the geometric regions are prioritized. For each high-priority maintenance item, specify the specific maintenance operation type, the required testing tools, and the acceptance criteria for determining whether the repair is satisfactory; All maintenance items, their corresponding priorities, operation types, testing tools, and acceptance criteria are encoded into a machine-readable set of maintenance intervention instructions in the order of execution.

10. A permanent magnet variable frequency submersible pump testing system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the permanent magnet variable frequency submersible pump testing method according to any one of claims 1 to 9.

Images