Permanent magnet direct current pump motor performance evaluation method and system based on cloud computing

Abstract

The invention relates to the technical field of motor performance evaluation, in particular to a permanent magnet direct current pump motor performance evaluation method and system based on cloud computing. Comprising the steps of obtaining current waveform data and vibration waveform data, and obtaining a first criterion based on the current waveform data and the vibration waveform data; obtaining a second criterion according to the vibration waveform data; obtaining a third criterion according to the vibration waveform data; obtaining a fourth criterion according to the current waveform data and the vibration waveform data; judging the causal correspondence between the current step and the output change of the control loop to obtain a fifth criterion; continuously observing the change trend of high-frequency components in the vibration waveform data to obtain a sixth criterion; based on the first criterion, the second criterion, the third criterion, the fourth criterion, the fifth criterion and the sixth criterion, obtaining a classification result; and confirming measures according to a classification result. The problems of repeated false alarm, invalid maintenance and system reliability evaluation deviation caused by the fact that a permanent magnet direct current pump motor performance evaluation system cannot distinguish real mechanical faults under a low-speed cruise working condition in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of motor performance evaluation, in particular to a cloud computing-based permanent magnet DC pump motor performance evaluation method and system. BACKGROUND

[0002] In a large commercial complex central air conditioning water circulation system, the spring damping base (natural frequency of about 3-5 Hz) is easy to coincide with the low-speed cruise excitation frequency during the night low-load microcirculation operation of the permanent magnet DC pump motor, causing structural resonance, amplifying the weak excitation into low-frequency large displacement shaking of the pump body and pipeline, and showing a low-frequency acceleration peak value and a small high-frequency component. The existing system only triggers a mechanical serious alarm according to the comparison between the instantaneous / short-time root mean square value of the pump body vibration sensor and the experience threshold, and does not distinguish the baselines in different working conditions; the cloud electrical model gives a higher score when there is no harmonic and no overcurrent, forming a logical conflict of "electrical health but mechanical seriousness". Moreover, due to the lack of unified judgment basis across the mechanical / electrical domain, the field operation and maintenance often starts from the reinforcement of basic parts or dynamic balancing, which is difficult to eliminate the shaking caused by the resonance of the damping base. The superimposed vibration judgment does not consider the frequency characteristics, the insufficient time synchronization accuracy of the edge and the cloud, and the problems that the electrical data is buffered or undersampled during the alarm period, and the system cannot effectively distinguish the structural vibration caused by the resonance amplification of the damping base from the real mechanical faults such as rotor imbalance, bearing failure and impeller damage under the low-speed cruise working condition, resulting in repeated false alarms, invalid maintenance and reliability evaluation deviation.

[0003] The existing technology needs to be improved in view of the above problems. SUMMARY

[0004] The application discloses a cloud computing-based permanent magnet DC pump motor performance evaluation method and system, aiming to solve the problem that the existing permanent magnet DC pump motor performance evaluation system cannot distinguish the structural vibration caused by the resonance amplification of the damping base from the real mechanical faults caused by the rotor imbalance, bearing failure or impeller damage and other rotating part defects under the low-speed cruise working condition, resulting in repeated false alarms, invalid maintenance and system reliability evaluation deviation.

[0005] The technical scheme of the application is as follows:

[0006] In the first aspect, the application discloses a cloud computing-based permanent magnet DC pump motor performance evaluation method, applied to a cloud computing platform, comprising:

[0007] When the vibration of the permanent magnet DC pump motor is detected to be abnormal, the current waveform data and the vibration waveform data of the permanent magnet DC pump motor are synchronously acquired;

[0008] Within a preset short-time analysis window, time-domain features are extracted from current waveform data and vibration waveform data to obtain a first criterion for preliminary judgment of vibration abnormal events. The preliminary judgment includes external hydraulic impact-dominated events, mechanical failure-prone events, or uncertain events.

[0009] Spectral analysis of vibration waveform data is performed to identify the matching relationship between the low-frequency main peak frequency of vibration and the natural frequency of the vibration damping base, and the proportion of low-frequency energy is calculated to obtain a second criterion reflecting the existence and intensity of background resonance.

[0010] When the first criterion indicates that the external hydraulic impact is the dominant event and the second criterion indicates that there is background resonance, the background resonance frequency band energy corresponding to the natural frequency of the vibration damping base is extracted from the total vibration energy of the vibration waveform data to obtain the residual vibration energy. Based on the change of the residual vibration energy within a preset period before and after the vibration abnormal event and the increase of the high-frequency energy of the vibration waveform data, a third criterion reflecting the non-resonance vibration increment and the high-frequency energy increment is obtained.

[0011] The cross-correlation function of the current waveform data and the vibration waveform data is calculated to determine the time delay and correlation coefficient between the two, so as to verify the phase relationship of the electromechanical response and obtain the fourth criterion.

[0012] The control loop output command value of the motor driver or the change information of its duty cycle to the permanent magnet DC pump motor drive signal is obtained to determine whether the current step has a causal relationship with the change in the control loop output, thus forming a fifth criterion that reflects the relationship between external load change and current response.

[0013] When the first criterion is an uncertain event, a delayed confirmation window is set. Within this delayed confirmation window, the energy change trend of the high-frequency components in the vibration waveform data over time is continuously observed to distinguish between transient external disturbances and mechanical damage with persistent characteristics, thus obtaining the sixth criterion.

[0014] Based on the first criterion, the second criterion, the third criterion, the fourth criterion, the fifth criterion, and the sixth criterion, the vibration abnormal events are classified according to the pre-set multidimensional deterministic numerical combination rules, and the classification results are obtained. The classification includes at least non-fault events caused by external transient hydraulic impact superimposed on background resonance, where there is a significant time lag between current step and vibration response, and / or real mechanical fault events caused by rotor jamming, bearing damage, or impeller failure, where current and vibration rise synchronously.

[0015] Based on the classification results, control whether to issue shutdown instructions to the operation and maintenance end and / or trigger local mitigation and delayed confirmation measures.

[0016] Secondly, this application also discloses a cloud computing-based permanent magnet DC pump motor performance evaluation system, including:

[0017] The data acquisition module is used to simultaneously acquire the current waveform data and vibration waveform data of the permanent magnet DC pump motor when abnormal vibration of the permanent magnet DC pump motor is detected.

[0018] The preliminary judgment module is used to extract time-domain features from current waveform data and vibration waveform data within a preset short-time analysis window to obtain a first criterion for preliminary judgment of vibration abnormal events. The preliminary judgment includes external hydraulic impact-dominated events, mechanical failure-prone events, or uncertain events.

[0019] The resonance identification module is used to perform spectral analysis on vibration waveform data, identify the matching relationship between the low-frequency main peak frequency of vibration and the natural frequency of the vibration damping base, and calculate the proportion of low-frequency energy to obtain a second criterion reflecting the existence and intensity of background resonance.

[0020] The energy stripping and evaluation module is used to strip the background resonance frequency band energy corresponding to the natural frequency of the vibration damping base from the total vibration energy of the vibration waveform data when the first criterion indicates that the external hydraulic impact is the dominant event and the second criterion indicates that there is background resonance, to obtain the remaining vibration energy, and based on the change of the remaining vibration energy within a preset period before and after the vibration abnormal event and the increase of the high-frequency energy of the vibration waveform data, to obtain a third criterion reflecting the non-resonance vibration increment and the high-frequency energy increment.

[0021] The phase verification module is used to calculate the cross-correlation function between current waveform data and vibration waveform data, determine the time delay and correlation coefficient between the two, and verify the phase relationship of electromechanical response to obtain the fourth criterion.

[0022] The control command reading module is used to obtain the output command value of the control loop of the motor driver or the change information of its duty cycle on the permanent magnet DC pump motor drive signal, and to determine whether the current step has a causal relationship with the change in the output of the control loop, forming a fifth criterion that reflects the relationship between external load sudden change and current response.

[0023] The delayed confirmation module is used to set a delayed confirmation window when the first criterion is an uncertain event. Within the delayed confirmation window, the energy change trend of the high-frequency components in the vibration waveform data over time is continuously observed to distinguish between transient external disturbances and mechanical damage with persistent characteristics, and to obtain the sixth criterion.

[0024] The classification and judgment module is used to classify abnormal vibration events based on the first criterion, the second criterion, the third criterion, the fourth criterion, the fifth criterion, and the sixth criterion, according to the pre-set multi-dimensional deterministic numerical combination rules, and to obtain the classification results. The classification includes at least non-fault events caused by external transient hydraulic impact superimposed on background resonance, where there is a significant time lag between current step and vibration response, and / or real mechanical fault events caused by rotor jamming, bearing damage, or impeller failure, where current and vibration rise synchronously.

[0025] The decision and instruction module is used to control whether to issue shutdown instructions to the operation and maintenance terminal and / or trigger local mitigation and delayed confirmation measures based on the classification results.

[0026] Beneficial Effects: The cloud computing-based performance evaluation method for permanent magnet DC pump motors disclosed in this application effectively solves the problem in existing technologies where structural vibrations caused by the resonance amplification of the vibration damping base under low-speed cruise conditions are difficult to distinguish from actual mechanical faults. By introducing innovative criteria such as dynamically calculating the natural frequency of the vibration damping base, stripping the background resonance frequency band energy, verifying the phase relationship of the electromechanical response, and distinguishing between transient interference and persistent damage, this application can avoid false alarms and ineffective repairs caused by existing evaluation systems that rely solely on vibration amplitude thresholds. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a cloud computing-based permanent magnet DC pump motor performance evaluation method provided in this application.

[0028] Figure 2 A flowchart of a cloud computing-based permanent magnet DC pump motor performance evaluation system is provided for this application.

[0029] In the diagram: 1. Data acquisition module; 2. Preliminary judgment module; 3. Resonance identification module; 4. Energy stripping and evaluation module; 5. Phase verification module; 6. Control command reading module; 7. Delay confirmation module; 8. Classification judgment module; 9. Decision and command module. Detailed Implementation

[0030] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0031] Reference Figure 1 This application proposes a performance evaluation method for permanent magnet DC pump motors based on cloud computing. This method is applied to a cloud computing platform and includes:

[0032] S1000: When abnormal vibration of the permanent magnet DC pump motor is detected, the current waveform data and vibration waveform data of the permanent magnet DC pump motor are acquired simultaneously.

[0033] S2000: Within the preset short-time analysis window, time-domain features are extracted from current waveform data and vibration waveform data to obtain the first criterion for preliminary judgment of vibration abnormal events.

[0034] Preliminary assessments indicate that the events may be dominated by external hydraulic shock, mechanical failure, or uncertain events.

[0035] S3000: Performs spectral analysis on vibration waveform data, identifies the matching relationship between the low-frequency main peak frequency of vibration and the natural frequency of the vibration damping base, calculates the proportion of low-frequency energy, and obtains a second criterion reflecting the existence and intensity of background resonance;

[0036] S4000: When the first criterion indicates that the external hydraulic impact is the dominant event and the second criterion indicates that there is background resonance, the background resonance frequency band energy corresponding to the natural frequency of the vibration damping base is extracted from the total vibration energy of the vibration waveform data to obtain the remaining vibration energy. Based on the change of the remaining vibration energy within a preset period before and after the vibration abnormal event and the increase of the high-frequency energy of the vibration waveform data, a third criterion reflecting the non-resonance vibration increment and the high-frequency energy increment is obtained.

[0037] S5000: Calculate the cross-correlation function between current waveform data and vibration waveform data, determine the time delay and correlation coefficient between the two, and verify the phase relationship of electromechanical response to obtain the fourth criterion;

[0038] S6000: Obtain the control loop output command value of the motor driver or the change information of its duty cycle on the permanent magnet DC pump motor drive signal, determine whether the current step has a causal relationship with the change in the control loop output, and form the fifth criterion reflecting the relationship between external load sudden change and current response.

[0039] S7000: When the first criterion is an uncertain event, a delayed confirmation window is set. Within the delayed confirmation window, the energy change trend of the high-frequency components in the vibration waveform data over time is continuously observed to distinguish between transient external disturbances and mechanical damage with persistent characteristics, thus obtaining the sixth criterion.

[0040] S8000: Based on the first criterion, the second criterion, the third criterion, the fourth criterion, the fifth criterion, and the sixth criterion, the vibration anomaly events are classified according to the pre-set multidimensional deterministic numerical combination rules, and the classification results are obtained.

[0041] The classification includes at least non-fault events caused by external transient hydraulic impact superimposed on background resonance, where there is a significant time lag between current step and vibration response, and / or real mechanical fault events caused by rotor jamming, bearing damage, or impeller failure, where current and vibration rise synchronously.

[0042] S9000: Based on the classification results, control whether to issue a shutdown command to the operation and maintenance terminal and / or trigger local mitigation and delay confirmation measures.

[0043] The "permanent magnet DC pump motor" mentioned in this application refers to a DC motor that uses permanent magnets for excitation, typically used to drive water pumps for fluid transport in scenarios such as central air conditioning water circulation systems. Its characteristics include high efficiency, small size, and good speed regulation performance.

[0044] "Current waveform data" refers to the sequence data of the change of the operating current of the permanent magnet DC pump motor over time, which is collected in real time by a current sensor, reflecting the electrical load and operating status of the motor.

[0045] "Vibration waveform data" refers to the sequence data of vibration acceleration, velocity or displacement of a permanent magnet DC pump motor or its connected components that are collected in real time by vibration sensors, reflecting the mechanical operating status and potential faults of the motor.

[0046] A "vibration damping base" is a device used to support a permanent magnet DC pump motor and isolate its vibration from being transmitted to the foundation structure. It is usually made of elastic materials (such as rubber or springs) and has a specific natural frequency.

[0047] "Cloud computing platform" refers to a distributed computing environment that provides computing resources, storage resources and various services. In this application, it is used to receive, store, process and analyze various types of data uploaded by the permanent magnet DC pump motor and execute performance evaluation algorithms.

[0048] "Operations and maintenance terminal" refers to the terminal system or personnel responsible for the operation, maintenance and management of permanent magnet DC pump motors, receiving evaluation results and control instructions, and executing corresponding operations.

[0049] Specifically, when abnormal vibration of a permanent magnet DC pump motor is detected, it is necessary to simultaneously acquire both the current waveform data and the vibration waveform data of the motor. This can be achieved, for example, by installing current sensors and vibration sensors on the permanent magnet DC pump motor. The current sensor can be a Hall effect sensor or a current transformer to monitor changes in the motor's current in real time; the vibration sensor can be a piezoelectric accelerometer to monitor the motor's vibration in real time. These sensors convert the acquired analog signals into digital signals and upload them to a cloud computing platform via a data acquisition module. Synchronous data acquisition can be achieved through timestamp alignment to ensure strict temporal consistency between the current waveform data and the vibration waveform data for subsequent joint analysis.

[0050] Within a preset short-time analysis window, time-domain features are extracted from current and vibration waveform data to obtain a first criterion for preliminary judgment of abnormal vibration events. Preliminary judgment includes events dominated by external hydraulic impact, events with mechanical failure tendencies, or uncertain events. For example, statistical methods can be used to extract features from waveform data, such as calculating root mean square values, peak values, kurtosis, and margins. These features reflect the energy, impact, and periodicity of the signal. By setting different thresholds, these features can be combined to preliminarily determine the type of abnormal vibration event. For example, when vibration waveform data exhibits high amplitude and high kurtosis but short duration within a short period, it may be preliminarily judged as an event dominated by external hydraulic impact; when vibration waveform data exhibits sustained high amplitude, high margin, and a certain periodicity, it may be preliminarily judged as an event with mechanical failure tendencies. If the features are not obvious or fall between the two, it may be judged as an uncertain event.

[0051] Spectral analysis of vibration waveform data is performed to identify the matching relationship between the low-frequency main peak frequency of the vibration and the natural frequency of the vibration damping base, and the proportion of low-frequency energy is calculated to obtain a second criterion reflecting the existence and intensity of background resonance. For example, Fast Fourier Transform (FFT) can be used to perform spectral analysis on the vibration waveform data to obtain the frequency components of the vibration signal. By observing the spectrum, the low-frequency main peak frequency where the vibration energy is concentrated can be identified. The natural frequency of the vibration damping base can be obtained through design parameters or field testing. The low-frequency main peak frequency of the vibration is compared with the natural frequency of the vibration damping base to determine whether a matching relationship exists. At the same time, the energy of the vibration spectrum is integrated within a preset low-frequency band, and the proportion of energy in this frequency band to the total vibration energy is calculated to obtain the low-frequency energy proportion. When the low-frequency main peak frequency matches the natural frequency of the vibration damping base and the low-frequency energy proportion is high, it indicates the existence of background resonance and a high resonance intensity.

[0052] When the first criterion indicates an external hydraulic impact-dominated event and the second criterion indicates the presence of background resonance, the background resonance frequency band energy corresponding to the natural frequency of the vibration damping base is extracted from the total vibration energy of the vibration waveform data to obtain the residual vibration energy. Based on the change in residual vibration energy within a preset time period before and after the vibration anomaly event, and the increase in high-frequency energy in the vibration waveform data, a third criterion reflecting the increment of non-resonant vibration and high-frequency energy is obtained. For example, a band-stop filter can be used to filter the vibration waveform data, removing the frequency band energy corresponding to the natural frequency of the vibration damping base. The filtered signal is the residual vibration energy waveform. Within each preset time period before and after the vibration anomaly event, the energy of the residual vibration energy waveform is integrated, and the energy change before and after the event is compared to obtain the increase in residual vibration energy. Simultaneously, a fast Fourier transform is performed on the vibration waveform data, and energy integration is performed within a preset high-frequency band. The high-frequency energy change before and after the event is compared to obtain the increase in high-frequency energy. These increases together constitute the third criterion, used to assess the changes in non-resonant vibration and high-frequency energy.

[0053] The cross-correlation function is calculated between current waveform data and vibration waveform data to determine the time delay and correlation coefficient between them, thereby verifying the phase relationship of the electromechanical response and obtaining the fourth criterion. For example, cross-correlation analysis can be used to calculate the cross-correlation function between current waveform data and vibration waveform data. The peak position of the cross-correlation function can indicate the time delay between the two, and the peak value can indicate the correlation coefficient. By analyzing the time delay and correlation coefficient, the causal relationship and synchronicity between current changes and vibration changes can be determined. For example, when the current change leads the vibration change and the correlation coefficient is high, it may indicate that electrical factors are the cause of abnormal vibration.

[0054] The system acquires the output command value of the motor driver's control loop or the change in the duty cycle of its drive signal for the permanent magnet DC pump motor. It then determines whether the current step is causally related to the change in the control loop output, forming a fifth criterion reflecting the relationship between sudden changes in external load and current response. For example, the output command value of the control loop or the duty cycle of the PWM signal can be obtained by reading the motor driver's internal registers or communication interface. These changes in control signals are compared with the current step in the current waveform data to determine whether they are synchronized in time and have a logical causal relationship. For instance, if a step increase in current occurs when the control loop issues an acceleration command, it indicates that the current step is a normal load change response.

[0055] When the first criterion indicates an uncertain event, a delayed confirmation window is set. Within this window, the energy variation trend of the high-frequency components in the vibration waveform data over time is continuously observed to distinguish between transient external disturbances and persistent mechanical damage, thus obtaining the sixth criterion. For example, when initially determined to be an uncertain event, the system will not immediately alarm but will enter delayed confirmation mode. In this mode, the system will continuously collect vibration waveform data and perform energy analysis on the high-frequency components. If the high-frequency energy continues to rise or remains high within the delayed confirmation window, it may indicate persistent mechanical damage; if the high-frequency energy rises rapidly and then decays quickly, it may indicate transient external disturbance.

[0056] In this embodiment, all of the six criteria (first, second, third, fourth, fifth, and sixth) can be used to classify abnormal vibration events according to pre-defined multi-dimensional deterministic numerical combination rules, resulting in classification results. The classification includes at least non-fault events caused by external transient hydraulic impact superimposed with background resonance, where there is a significant time lag between the current step and the vibration response, and / or real mechanical fault events caused by rotor jamming, bearing damage, or impeller failure, where the current and vibration rise synchronously. For example, a decision tree or rule engine can be constructed, using the above six criteria as input, and making judgments according to preset logical rules. For instance, if the first criterion is that external hydraulic impact is the dominant event, the second criterion indicates background resonance, the third criterion shows that the non-resonant vibration increment is small but the high-frequency energy increase is small, the fourth criterion shows a significant time lag between the current and the vibration response, the fifth criterion shows that there is no causal correspondence between the current step and the control loop output change, and the sixth criterion shows that the high-frequency energy does not continuously increase, then it may be classified as a non-fault event caused by external transient hydraulic impact superimposed with background resonance. Conversely, if the first criterion indicates a mechanical failure-prone event, the third criterion shows that the increase in non-resonant vibration and the increase in high-frequency energy are both large, and the fourth criterion shows that the current and vibration rise synchronously, then it may be classified as a real mechanical failure event caused by rotor jamming, bearing damage, or impeller failure.

[0057] Based on the classification results, the system controls whether to issue a shutdown command to the operations and maintenance (O&M) department and / or trigger local mitigation and delayed confirmation measures. For example, when the classification result is a genuine mechanical failure event, the system can immediately issue a shutdown command to the O&M department to prevent the failure from escalating. When the classification result is a non-failure event or an uncertain event, the system can trigger local mitigation measures (such as adjusting operating parameters) or delayed confirmation measures, and send a low-priority alarm to the O&M department, suggesting further observation.

[0058] The method provided in this application works by using multi-dimensional, cross-domain data fusion and intelligent analysis to accurately classify and assess abnormal vibration events in permanent magnet DC pump motors. Traditional assessment systems often experience structural resonance during low-load nighttime periods because the natural frequency of the spring damping base in the pump unit's mechanical connection coincides with or is close to the mechanical excitation frequency of the permanent magnet DC pump motor during low-speed cruising. This amplifies the originally weak excitation force into a large-scale overall sway. This phenomenon is frequently misjudged as a serious mechanical fault in existing systems, leading to repeated false alarms and ineffective repairs.

[0059] This application achieves accurate classification of abnormal vibration events of permanent magnet DC pump motors by comprehensively applying multi-dimensional criteria and intelligent decision-making rules. It can effectively distinguish between non-fault events and real mechanical fault events, thereby avoiding repeated false alarms, ineffective maintenance, and deviations in system reliability evaluation, and significantly improving the accuracy and reliability of permanent magnet DC pump motor performance evaluation.

[0060] In another embodiment of this application, S2000 is further proposed to include:

[0061] S2100: Adaptive filtering is performed on current waveform data and vibration waveform data to retain the instantaneous changes reflecting the abrupt changes in electromechanical response characteristics while suppressing random noise and background interference, thus obtaining filtered current waveform data and vibration waveform data.

[0062] S2110: Perform differential operations on filtered current waveform data and vibration waveform data at multiple time scales to obtain the instantaneous slope of current and the instantaneous slope of vibration at each time scale.

[0063] S2120: Based on local extrema and their retention characteristics within a preset duration, determine the peak time of the instantaneous current slope from the instantaneous current slope, determine the peak time of the instantaneous vibration slope from the instantaneous vibration slope, and obtain the corresponding peak values ​​of the instantaneous current slope and the instantaneous vibration slope.

[0064] S2130: Based on the steady-state current and vibration level before the occurrence of the abnormal vibration event, the peak value of the instantaneous slope of the current and the peak value of the instantaneous slope of the vibration are dynamically normalized to obtain the normalized slope ratio.

[0065] S2140: The normalized slope ratio is combined with the peak time difference between the peak time of the instantaneous current slope and the peak time of the instantaneous vibration slope, and used as the first criterion.

[0066] Adaptive filtering aims to dynamically adjust filtering parameters based on the statistical characteristics of the signal and noise, thereby effectively removing noise while preserving transient features of the signal to the greatest extent possible, such as rapid changes in current or vibration signals caused by impacts or step jumps. Specifically, Kalman filtering, Wiener filtering, or wavelet transform-based denoising methods can be used. These methods can adjust the filtering intensity according to local signal characteristics, avoiding transient signal distortion that may be caused by traditional fixed-parameter filters.

[0067] Differential operations are performed on filtered current and vibration waveform data at multiple time scales to obtain the instantaneous current and vibration slopes at each time scale. Differential operations are used to capture the instantaneous rate of change of the signal, i.e., the signal's gradient over time. Using multiple time scales for differential operations allows the system to identify transient events of varying durations or intensities. For example, shorter time scales can capture rapid, sharp impacts, while longer time scales can reflect relatively slow but continuous trend changes.

[0068] Local extrema refer to the instantaneous maximum or minimum value of a signal at a certain point in time, while their persistence over a preset duration is used to distinguish true transient events from random noise spikes. Specifically, by setting an amplitude threshold and a duration threshold, a peak value is identified as valid only when the slope value exceeds the amplitude threshold and maintains this trend for a certain period of time, and its occurrence time is recorded.

[0069] The peak values ​​of the instantaneous current slope and the peak values ​​of the instantaneous vibration slope are dynamically normalized based on the steady-state current and vibration level before the occurrence of the abnormal vibration event, resulting in a normalized slope ratio. The purpose of dynamic normalization is to eliminate the influence of differences in the baseline level of the permanent magnet DC pump motor under different operating conditions (e.g., different loads and speeds) on the criterion, making abnormal events under different operating conditions comparable. For example, the peak value can be divided by the average steady-state value or root mean square value over a period of time before the event, thus obtaining a dimensionless slope ratio that more accurately reflects the relative intensity of the abnormal event.

[0070] The normalized slope ratio is combined with the peak time difference between the peak times of the instantaneous current slope and the instantaneous vibration slope, serving as the primary criterion. By combining these two features, the nature of the abnormal vibration event can be comprehensively assessed. The normalized slope ratio reflects the intensity of the event, while the peak time difference reveals the temporal relationship between the current response and the vibration response. For example, if the peak current slope significantly precedes the peak vibration slope, it may indicate external hydraulic shock; if they are almost synchronous and of high intensity, it may point to a mechanical failure.

[0071] The proposed solution, through the detailed steps outlined above, enables more precise capture of the electromechanical response characteristics of a permanent magnet DC pump motor during abnormal vibration events. Adaptive filtering ensures the integrity of key transient information in the original signal, preventing noise interference from misleading subsequent analysis. Multi-timescale differential operations allow the system to comprehensively perceive transient changes of different types and durations, thereby improving sensitivity to various abnormal events. By determining local extrema and their persistence characteristics, it effectively distinguishes between real anomalies and occasional noise, ensuring the reliability of the extracted peak features. Dynamic normalization eliminates the influence of changes in operating conditions, ensuring the validity and comparability of the criteria under different operating conditions. Finally, combining the normalized slope ratio with the peak time difference allows for a comprehensive assessment of the intensity of the abnormal event and the temporal relationship of the electromechanical response, providing a more comprehensive and accurate basis for the preliminary determination of external hydraulic shock-dominated events, mechanical failure-prone events, or uncertain events.

[0072] In another embodiment of this application, S3000 further includes the following steps:

[0073] S3100: Performs spectral analysis on vibration waveform data to identify the low-frequency main peak frequency of vibration;

[0074] S3110: Obtain the temperature data of the vibration damping base corresponding to the permanent magnet DC pump motor;

[0075] S3120: Estimate the elastic modulus of the vibration damping base at the current moment based on temperature data and preset temperature-elastic modulus characteristics of the vibration damping base material;

[0076] S3130: Based on the elastic modulus and structural parameters of the vibration damping base, dynamically calculate the natural frequency and drift range of the vibration damping base at the current moment;

[0077] S3140: Match the low-frequency main peak frequency of vibration with the natural frequency and its drift range to obtain the matching relationship between the low-frequency main peak frequency of vibration and the natural frequency of the vibration damping base.

[0078] S3150: Integrate the energy of the vibration waveform data within the preset low-frequency band, calculate the proportion of low-frequency energy to the total vibration energy, obtain the low-frequency energy ratio, and combine it with the matching relationship as the second criterion.

[0079] Specifically, spectral analysis of vibration waveform data aims to convert the time-domain signal into a frequency-domain signal, thereby revealing the frequency components and their energy distribution. Using methods such as Fourier transform or wavelet transform, the low-frequency main peak frequency in the vibration signal can be identified. This frequency is usually related to the overall vibration mode of the equipment or the low-frequency vibration characteristics of a specific component. A crucial step is obtaining the temperature data of the vibration damping base corresponding to the permanent magnet DC pump motor, as the material properties of the damping base, especially its elastic modulus, are affected by temperature. Temperature data can be acquired in real time using temperature sensors installed on the vibration damping base.

[0080] Furthermore, based on temperature data and the preset temperature-elastic modulus characteristics of the vibration damping base material, the elastic modulus of the vibration damping base at the current moment can be estimated. The elastic modulus of the vibration damping base material, such as rubber or polyurethane, decreases as temperature increases. Preset temperature-elastic modulus characteristic curves or mathematical models can be used for accurate estimation. Therefore, based on the elastic modulus and the structural parameters of the vibration damping base, the natural frequency and its drift range of the vibration damping base at the current moment can be dynamically calculated. The natural frequency of the vibration damping base is the frequency of its free vibration without external excitation. Its calculation is usually based on a structural mechanics model, considering the mass, stiffness (related to the elastic modulus), and geometry of the base. Since the elastic modulus changes with temperature, the natural frequency will also drift dynamically accordingly; therefore, a drift range needs to be calculated to cover the actual possible range of variation.

[0081] Subsequently, the low-frequency peak frequency of the vibration is matched with the natural frequency and its drift range to determine whether resonance exists. If the low-frequency peak frequency falls within the drift range of the vibration damping base's natural frequency, it indicates the possible presence of background resonance. This matching relationship is an important basis for determining the presence or absence of resonance. As a preferred implementation, the vibration waveform data is integrated within a preset low-frequency band to calculate the proportion of low-frequency energy to the total vibration energy, thus obtaining the low-frequency energy percentage. This low-frequency band is typically set to include the vibration damping base's natural frequency and its drift range. The low-frequency energy percentage quantifies the intensity of background resonance, i.e., the share of resonance energy in the total vibration energy. Finally, the matching relationship and the low-frequency energy percentage are combined as a second criterion reflecting the presence and intensity of background resonance.

[0082] The solution proposed in this application achieves dynamic estimation of the natural frequency of the vibration damping base by introducing the temperature data of the vibration damping base and combining it with the temperature-elastic modulus characteristics of its material.

[0083] In another embodiment of this application, S4000 is further proposed to include:

[0084] S4100: Perform continuous wavelet transform on the vibration waveform data to obtain the time-frequency spectrum;

[0085] S4110: Identify energy concentration regions in the time-frequency spectrum and dynamically determine the instantaneous frequency and instantaneous bandwidth corresponding to the background resonance;

[0086] S4120: Construct an adaptive band-stop filter based on the instantaneous frequency and instantaneous bandwidth, and use the adaptive band-stop filter to perform real-time filtering on the vibration waveform data, stripping the background resonant frequency band energy to obtain the remaining vibration energy waveform;

[0087] S4130: During the process of stripping the background resonant frequency band energy, monitor the energy level of the remaining vibration energy waveform in the preset high frequency band. When the high frequency component shows abnormal attenuation relative to the reference level before the vibration abnormal event occurs, adjust the bandwidth and / or center frequency of the adaptive band-stop filter.

[0088] S4140: In each preset time period before and after the occurrence of an abnormal vibration event, the remaining vibration energy waveform is integrated, and the energy value in the preset time period after the abnormal vibration event is compared with the energy value in the preset time period before the abnormal vibration event to obtain the growth rate of the remaining vibration energy.

[0089] S4150: Performs fast Fourier transform on vibration waveform data, integrates energy within a preset high-frequency band, compares the high-frequency energy within a preset time period after the vibration abnormal event with the reference level of high-frequency energy within a preset time period before the vibration abnormal event, and obtains the increase in high-frequency energy.

[0090] S4160: The increase in residual vibration energy and the increase in high-frequency energy are used as the third criterion.

[0091] Specifically, performing continuous wavelet transform on vibration waveform data aims to decompose the original time-domain signal into the time-frequency domain, generating a time-frequency spectrum. This spectrum visually displays the distribution of vibration energy at different times and frequencies, and is particularly suitable for analyzing transient characteristics and frequency changes in non-stationary signals, providing refined time-frequency information for subsequent identification of background resonance. For example, Morlet wavelets and Mexican Hat wavelets can be selected as mother wavelets, and by adjusting the scaling and displacement parameters, analysis of the vibration signal at different frequency and time resolutions can be achieved.

[0092] Identifying energy concentration regions in a time-frequency spectrum and dynamically determining the instantaneous frequency and bandwidth corresponding to background resonances involves analyzing the time-frequency spectrum, such as using image processing techniques, threshold segmentation, or clustering algorithms, to locate regions where the energy is significantly higher than the background noise. These regions typically correspond to the main vibrational components of the signal, including background resonances. Dynamically determining the instantaneous frequency and bandwidth means that the process can adaptively adjust the identified resonance frequency center and its covered frequency range based on real-time changes in the energy concentration regions in the time-frequency spectrum, rather than using fixed parameters. For example, the instantaneous frequency can be determined by tracking the trajectory of energy peaks in the time-frequency spectrum, and the instantaneous bandwidth can be determined by analyzing the energy diffusion around the peaks.

[0093] In practical applications, an adaptive band-stop filter is constructed based on the aforementioned instantaneous frequency and bandwidth. This filter is then used to perform real-time filtering of the vibration waveform data, stripping away the background resonant frequency band energy to obtain the residual vibration energy waveform. An adaptive band-stop filter is a filter that can dynamically adjust its filtering parameters (such as center frequency and bandwidth) according to the characteristics of the input signal. Its purpose is to accurately filter out background resonant components and avoid inadvertently affecting non-resonant vibration components. For example, an IIR digital filter or an FIR digital filter can be used, whose coefficients are updated in real-time based on dynamically determined instantaneous frequency and bandwidth. In this way, even if the background resonant frequency drifts, the filter can effectively track and strip it, thus obtaining a purer residual vibration energy waveform that reflects non-resonant components.

[0094] Furthermore, during the process of stripping the background resonance frequency band energy, the energy level of the remaining vibration energy waveform within a preset high-frequency band is monitored. When the high-frequency components show abnormal attenuation relative to the reference level before the vibration anomaly occurred, the bandwidth and / or center frequency of the adaptive bandstop filter are adjusted. This step introduces a feedback mechanism to prevent over-filtering or misjudgment. For example, if the adaptive bandstop filter is set too aggressively, it may mistakenly identify some high-frequency fault features as background resonance and filter them out, leading to abnormal attenuation of the high-frequency components in the remaining vibration energy waveform. By monitoring the high-frequency energy level in real time, once abnormal attenuation is detected, the system can promptly adjust the filter parameters, such as appropriately reducing the bandwidth or fine-tuning the center frequency, to protect potential fault feature signals and ensure the accuracy of the stripping.

[0095] Furthermore, within each preset time period before and after the vibration anomaly event, the residual vibration energy waveform is integrated, and the energy value within the preset time period after the vibration anomaly event is compared with the energy value within the preset time period before the vibration anomaly event to obtain the increase in residual vibration energy. This step aims to quantify the degree of change in non-resonant vibration components after the event. Energy integration can be performed using the root mean square (RMS) value or the energy spectral density (PSD) integrated within a specific time window. By comparing the energy values ​​before and after the event, it is possible to accurately assess whether non-resonant vibration has increased significantly, which is crucial for distinguishing between external shocks and internal mechanical failures.

[0096] Simultaneously, a Fast Fourier Transform (FFT) is performed on the vibration waveform data, and energy integration is conducted within a preset high-frequency band. The high-frequency energy within a preset time period after the vibration anomaly event is compared with the reference high-frequency energy level within a preset time period before the vibration anomaly event to obtain the increase in high-frequency energy. The Fast Fourier Transform (FFT) is used to convert the time-domain signal to the frequency domain for frequency component analysis. The energy integration in the high-frequency band can reveal the presence of mechanical fault characteristics such as high-frequency impact or wear. By comparing it with the reference level before the event, the increase in high-frequency energy can be quantified, providing a basis for determining whether mechanical damage exists.

[0097] Finally, the increase in the remaining vibration energy and the increase in high-frequency energy were used as the third criterion. This criterion comprehensively considers the overall change in non-resonant vibration and the enhancement of high-frequency fault characteristics, providing a multi-dimensional evaluation index that can more comprehensively and accurately reflect the nature of abnormal vibration events.

[0098] The solution proposed in this application introduces continuous wavelet transform to perform time-frequency analysis on vibration waveform data, which can overcome the limitations of traditional Fourier transform in processing non-stationary signals and accurately capture the instantaneous frequency and bandwidth of background resonance.

[0099] In some preferred embodiments, suppose a permanent magnet DC pump motor suddenly detects abnormal vibration during normal operation. Based on preliminary judgment, the event is identified as a dominant event of external hydraulic impact, and the resonance identification module detects obvious background resonance, such as the natural frequency of the vibration damping base being excited.

[0100] At this point, the scheme of this application will initiate the calculation process of the third criterion. First, continuous wavelet transform is performed on the vibration waveform data at the time of the abnormal event to generate a refined time-frequency spectrum. In this spectrum, the system can clearly identify the energy concentration region of the background resonance and dynamically determine its instantaneous frequency (e.g., drifting from 50Hz to 52Hz) and instantaneous bandwidth.

[0101] Based on these dynamic parameters, the system constructs and adjusts an adaptive band-stop filter in real time to precisely filter out the dynamically changing background resonant frequency band energy, thereby obtaining the residual vibration energy waveform. During the filtering process, the system continuously monitors the energy level of the residual vibration energy waveform within a preset high-frequency band (e.g., 1kHz-5kHz). If abnormal attenuation of high-frequency energy is detected, which may indicate that the filter settings are too aggressive, the system will immediately adjust the filter bandwidth or center frequency to protect against potential high-frequency characteristic signals such as bearing damage or impeller failure.

[0102] Subsequently, the system performs energy integration on the residual vibration energy waveform after removing background resonance at preset time intervals before and after the vibration anomaly event (e.g., 5 seconds before and 5 seconds after the event), and compares the two to obtain the increase in residual vibration energy. Simultaneously, a Fast Fourier Transform is performed on the original vibration waveform data to calculate the energy increase in the high-frequency band.

[0103] Ultimately, these two increases are used as a third criterion. For example, if the increase in residual vibration energy is small and the increase in high-frequency energy is not significant, it may further support the conclusion that the event is a non-fault event caused by external transient hydraulic impact superimposed on background resonance. Conversely, if the increase in residual vibration energy is large and the increase in high-frequency energy is significant, it may indicate the existence of a real mechanical fault. Through this refined processing, this application can more accurately assess the nature of the event and provide a reliable basis for subsequent decision-making.

[0104] In another embodiment of this application, S5000 is further proposed to include:

[0105] S5100: Divides current waveform data and vibration waveform data into multiple overlapping short-time analysis windows;

[0106] S5110: Within each short-time analysis window, extract the current characteristic point sequence and vibration characteristic point sequence that characterize the electromechanical response from the current waveform data and vibration waveform data, respectively.

[0107] S5120: Within each short-time analysis window, calculate the cross-correlation function between the current characteristic point sequence and the vibration characteristic point sequence to obtain the local time delay and maximum correlation coefficient corresponding to each short-time analysis window;

[0108] S5130: Perform trend analysis on the local time delay and maximum correlation coefficient of each short-time analysis window arranged in chronological order to obtain the characteristics of the phase relationship of electromechanical response changing with time;

[0109] S5140: Based on the results of trend analysis, the leading or lagging relationship of current relative to vibration and its stability are used as the fourth criterion.

[0110] Specifically, dividing current waveform data and vibration waveform data into multiple overlapping short-time analysis windows means segmenting continuously acquired current waveform data and vibration waveform data into a series of segments of fixed length that overlap with each other. This segmentation method aims to capture the local characteristics and transient behavior of the signal, while ensuring data continuity and avoiding information loss through overlap. For example, the length of each short-time analysis window can be set according to the operating frequency of the permanent magnet DC pump motor or a specific fault characteristic frequency, while the overlap rate can be set to 50% or 75% to balance computational efficiency and information integrity.

[0111] Within each of the aforementioned short-time analysis windows, current feature point sequences and vibration feature point sequences characterizing the electromechanical response are extracted from the current waveform data and vibration waveform data, respectively. These feature point sequences are key events or patterns in the signal that reflect the electromechanical coupling behavior of the motor; for example, they can be signal peaks, valleys, energy abrupt changes, or envelope peaks of specific frequency components. Extraction methods may include, but are not limited to, threshold detection, wavelet transform, Hilbert transform, etc., with the aim of condensing the key information in the original waveform data into a series of discrete feature points.

[0112] Within each of the aforementioned short-time analysis windows, the cross-correlation function between the current characteristic point sequence and the vibration characteristic point sequence is calculated, yielding the local time delay and maximum correlation coefficient for each short-time analysis window. The cross-correlation function is a mathematical tool that measures the similarity between two signals at different time offsets. By calculating the cross-correlation function, the time offset that achieves the best match between the current characteristic point sequence and the vibration characteristic point sequence—the local time delay—and the degree of similarity at this point, the maximum correlation coefficient, can be identified. The local time delay indicates the temporal order between the current response and the vibration response, while the maximum correlation coefficient quantifies the strength and reliability of this relationship.

[0113] Trend analysis is performed on the aforementioned local time delays and maximum correlation coefficients of each short-time analysis window arranged in chronological order to obtain the characteristics of the electromechanical response phase relationship changing over time. Since the operating state of the motor may change dynamically over time, the local results within a single short-time analysis window may be affected by transient noise or interference. Therefore, by performing trend analysis on a series of continuous local time delays and maximum correlation coefficients, such as using moving averages, exponential smoothing, or statistical regression, the long-term evolution pattern and stability of the electromechanical response phase relationship can be revealed, thereby more accurately determining the persistence of anomalies.

[0114] Based on the trend analysis results, the leading or lagging relationship between current and vibration, and its stability, are used as the fourth criterion. Ultimately, the fourth criterion is formed based on the temporal sequence of current and vibration revealed by the trend analysis (i.e., whether current leads or lags behind vibration) and the stability of this relationship (i.e., the sustained high or low correlation coefficient and the volatility of the time delay). For example, a sustained current leading vibration may indicate a sudden change in external load, while a synchronous increase in current and vibration with a high correlation may indicate an internal mechanical failure.

[0115] The proposed solution meticulously divides continuous current waveform data and vibration waveform data into overlapping short-time analysis windows, and extracts a sequence of feature points characterizing the electromechanical response based on these windows. This allows for a more precise capture of transient changes and local characteristics during the operation of the permanent magnet DC pump motor.

[0116] In another embodiment of this application, S5110 further includes:

[0117] S5111: Within each short-time analysis window, perform multi-resolution analysis on current waveform data and vibration waveform data to identify local energy concentration areas at different resolution levels;

[0118] S5112: Perform time-domain envelope extraction on the current waveform data and vibration waveform data corresponding to each local energy concentration area to obtain the current envelope signal and vibration envelope signal;

[0119] S5113: Perform peak detection on the current envelope signal and vibration envelope signal, and filter the detected peaks by combining the preset peak duration threshold condition and amplitude threshold. The peak times that meet the peak duration threshold condition and amplitude threshold are respectively used as the current feature point sequence and vibration feature point sequence.

[0120] Within each short-time analysis window, multi-resolution analysis is performed on the current waveform and vibration waveform data. The aim is to capture the transient characteristics and energy distribution of the signal at different frequencies and time scales, thereby more comprehensively identifying local energy concentration regions related to the electromechanical response. For example, wavelet packet decomposition can be used to decompose the original signal into multiple frequency sub-bands, and the energy characteristics of each sub-band can be analyzed to accurately locate key information in the signal.

[0121] Furthermore, time-domain envelope extraction is performed on the current waveform and vibration waveform data corresponding to each local energy concentration region. The purpose is to eliminate the influence of the carrier frequency and highlight the modulation information and transient impact characteristics of the signal. Through envelope extraction, current envelope signals and vibration envelope signals can be obtained, which can more intuitively reflect the amplitude variation trend and impact energy of the signal.

[0122] Based on this, peak detection is performed on the current envelope signal and the vibration envelope signal, and the detected peaks are filtered by combining preset peak duration threshold conditions and amplitude thresholds. Specifically, peak detection aims to identify significant transient events in the envelope signal, while the peak duration threshold condition and amplitude threshold condition are used to filter out random noise and unimportant fluctuations, ensuring that the selected peaks are real and representative electromechanical response events. Peak moments that meet these conditions are determined as current feature point sequences and vibration feature point sequences. These sequences will be used as inputs for subsequent cross-correlation function calculations to accurately analyze the phase relationship of the electromechanical response.

[0123] The scheme in this application can effectively decompose the original current waveform data and vibration waveform data into different frequencies and time scales through multi-resolution analysis, thereby identifying local energy concentration regions related to electromechanical response in complex signals.

[0124] In some embodiments of this application, multi-resolution analysis of current waveform data and vibration waveform data within each short-time analysis window is proposed to identify local energy concentration areas. However, in practical applications, due to the complex operating environment of permanent magnet DC pump motors, the signals may contain background noise, transient interference, and weak fault characteristic signals, making it difficult for simple multi-resolution analysis to accurately and robustly identify the local energy concentration areas that truly reflect the electromechanical response. If the above problems are not addressed, misjudgments or omissions may occur, affecting the accuracy of subsequent feature point sequence extraction and thus reducing the reliability of the entire performance evaluation method. In another embodiment of this application, it is further proposed to optimize the multi-resolution analysis process by refining wavelet packet decomposition, energy distribution calculation, targeted weight adjustment, high-frequency noise and fault feature differentiation, and time-domain characteristic confirmation to more accurately identify local energy concentration areas.

[0125] In another embodiment of this application, step S5111 further includes:

[0126] S5111-1: Perform wavelet packet decomposition on the current waveform data and vibration waveform data in each short-time analysis window to obtain multiple wavelet packet frequency sub-bands covering different frequency ranges;

[0127] S5111-2: Calculate the energy distribution of each wavelet packet frequency sub-band;

[0128] S5111-3: Based on the energy distribution and the frequency range of each wavelet packet frequency sub-band, preliminary identification of candidate local energy concentration sub-bands is performed;

[0129] S5111-4: Adjust the candidate local energy concentration sub-band with low energy but whose frequency falls within the preset fault characteristic frequency neighborhood to increase its weight in subsequent judgment.

[0130] S5111-5: Combining the energy variation trend of adjacent wavelet packet frequency sub-bands, the adjusted candidate local energy concentration sub-bands are jointly determined to obtain the preliminary local energy concentration region;

[0131] S5111-6: For wavelet packet frequency sub-bands within a preset high-frequency band, dynamically adjust their energy thresholds based on background noise levels and / or instantaneous signal-to-noise ratios, distinguish high-frequency noise from high-frequency fault characteristics in each wavelet packet frequency sub-band, and incorporate wavelet packet frequency sub-bands identified as having high-frequency fault characteristics into the initial local energy concentration region to form an updated local energy concentration region.

[0132] S5111-7: Combining the sparsity and / or pulse characteristics of the signals in the time domain within each wavelet packet frequency sub-band, the updated local energy concentration region is finally confirmed.

[0133] Specifically, wavelet packet decomposition is a method that decomposes a signal into different frequency sub-bands. Compared to traditional wavelet transform, it can further decompose the high-frequency components, thus providing finer frequency resolution. Through wavelet packet decomposition, the original current waveform data and vibration waveform data are decomposed into a series of wavelet packet frequency sub-bands, each representing the energy distribution of the signal within a specific frequency range. Calculating the energy distribution on each wavelet packet frequency sub-band quantifies the activity level of each sub-band, providing a basis for subsequent identification of energy concentration areas.

[0134] Based on the energy distribution and the frequency range of each wavelet packet frequency sub-band, sub-bands with relatively high energy or specific frequency characteristics can be preliminarily identified as candidate local energy concentration sub-bands. In practical applications, when a permanent magnet DC pump motor experiences an early fault, its characteristic signal may have low energy, but its frequency often falls within the specific fault characteristic frequency neighborhood. Therefore, these candidate local energy concentration sub-bands with low energy but potential fault indication significance need to be adjusted, for example, by assigning higher weights, to ensure that these weak but important features are not ignored.

[0135] Furthermore, by combining the energy variation trends of adjacent wavelet packet frequency sub-bands, the adjusted candidate local energy concentration sub-bands can be jointly determined. For example, if the energy of a sub-band is high, and the energy of its adjacent sub-bands also shows an increasing or concentrated trend, the confidence that the region is a local energy concentration region can be strengthened, thus obtaining a preliminary local energy concentration region.

[0136] Furthermore, during the operation of a permanent magnet DC pump motor, high-frequency noise and high-frequency fault characteristics may coexist within the high-frequency band. To accurately distinguish between the two, this application proposes dynamically adjusting the energy threshold of wavelet packet frequency sub-bands within a preset high-frequency band based on the background noise level and / or the instantaneous signal-to-noise ratio. For example, when the background noise level is high, the energy threshold can be increased accordingly to avoid misjudging noise as fault characteristics; similar adjustments can be made when the instantaneous signal-to-noise ratio is low. Through this dynamic adjustment, high-frequency noise and high-frequency fault characteristics can be distinguished more effectively, and wavelet packet frequency sub-bands identified as high-frequency fault characteristics can be incorporated into the initial local energy concentration region, forming an updated local energy concentration region.

[0137] Finally, the updated local energy concentration region is definitively confirmed by combining the sparsity and / or impulse characteristics of the signal within each wavelet packet frequency sub-band in the time domain. For example, genuine fault characteristic signals often exhibit sparsity or impulse characteristics in the time domain, while random noise typically does not possess these characteristics. By analyzing the sparsity or impulse intensity of the signal, misidentified noise components can be further eliminated, thereby ensuring that the finally confirmed local energy concentration region is genuine and diagnostically valuable.

[0138] The scheme in this application introduces wavelet packet decomposition to achieve refined analysis of current waveform data and vibration waveform data at different frequency scales, thereby enabling a more comprehensive capture of the energy distribution in the signal.

[0139] In some preferred embodiments, a specific example is given below. Suppose that during the operation of a permanent magnet DC pump motor, the bearings experience premature wear, resulting in a weak high-frequency vibration signal. In traditional methods for identifying local energy concentration areas, this high-frequency vibration signal is easily ignored because of its low energy and potential masking by background noise.

[0140] The proposed solution first performs wavelet packet decomposition on the acquired current waveform data and vibration waveform data within a short-time analysis window. Through decomposition, the signal is subdivided into multiple wavelet packet frequency sub-bands; for example, the vibration signal is decomposed into sub-bands such as 0-50Hz, 50-100Hz, and 100-200Hz. Subsequently, the energy distribution of each sub-band is calculated.

[0141] Suppose a low energy is detected in the 2000-3000Hz sub-band, but this frequency range happens to coincide with the characteristic frequency neighborhood of early wear in this type of bearing. In this case, the solution of this application will adjust the weight of the 2000-3000Hz sub-band according to preset rules, increasing its importance in subsequent judgments, even if its energy is not high.

[0142] Next, by combining the energy change trends of adjacent sub-bands, the adjusted candidate sub-bands are jointly judged to preliminarily identify the energy concentration area containing bearing wear characteristics.

[0143] Furthermore, assuming that some energy concentrations are also detected in the 5000-8000Hz high-frequency band, the system dynamically adjusts the energy threshold for this band based on the current background noise level and instantaneous signal-to-noise ratio to distinguish whether this energy is high-frequency noise or a genuine fault characteristic. For example, if the background noise level is high, the threshold will be increased, and only signals with energy significantly higher than the noise level will be considered fault characteristics. In this way, high-frequency noise interference in the environment can be effectively eliminated.

[0144] Finally, time-domain analysis is performed on the updated local energy concentration regions, such as the 2000-3000Hz and 5000-8000Hz regions (if the 5000-8000Hz region is identified as a fault characteristic). If the signals in these regions exhibit significant sparsity or impulsive characteristics, such as periodic impulse signals, these regions are ultimately confirmed as local energy concentration regions reflecting the electromechanical response of the permanent magnet DC pump motor. In this way, even weak early bearing wear signals can be accurately identified, avoiding missed diagnoses and providing strong support for early fault diagnosis of permanent magnet DC pump motors.

[0145] In another embodiment of this application, S5112 further includes:

[0146] S5112-1: Perform time-frequency analysis on the current waveform data and vibration waveform data corresponding to each local energy concentration area to determine the instantaneous frequency and instantaneous amplitude change trend of the local energy concentration area within the corresponding short-time analysis window;

[0147] S5112-2: Dynamically adjust the algorithm parameters used for envelope extraction based on the trend of instantaneous frequency and instantaneous amplitude changes. The algorithm parameters shall include at least the preprocessing analysis window length and / or the window width of the smoothing filter and its weight coefficients.

[0148] S5112-3: The dynamically adjusted envelope extraction algorithm is applied to process the current waveform data and vibration waveform data corresponding to each local energy concentration region to obtain the current envelope signal and vibration envelope signal.

[0149] This study involves time-frequency analysis of the current and vibration waveforms corresponding to various local energy concentration regions to reveal the energy distribution characteristics of the signal in both time and frequency dimensions. Specifically, wavelet transform, short-time Fourier transform, or Hilbert-Huang transform can be used to obtain the instantaneous frequency and amplitude information of the signal. The instantaneous frequency reflects the main frequency components of the signal at a given moment, while the trend of instantaneous amplitude variation characterizes the change of signal energy over time. This information allows for a more comprehensive understanding of the dynamic characteristics of local energy concentration regions.

[0150] Furthermore, based on the determined trends in instantaneous frequency and amplitude changes, the algorithm parameters used for envelope extraction can be dynamically adjusted. For example, when a drastic change in the instantaneous frequency or a large fluctuation in the instantaneous amplitude of the signal is detected, the length of the preprocessing analysis window can be shortened accordingly to improve the response speed to transient changes. Simultaneously, the window width and weighting coefficients of the smoothing filter can be adjusted; for instance, a wider window can be used to obtain a smoother envelope when the signal changes gradually, while a narrower window with higher weights can be used to preserve details when the signal contains sharp pulses. This dynamic adjustment mechanism ensures that the envelope extraction algorithm can adaptively adapt to different signal characteristics, avoiding over-smoothing or under-smoothing problems that may occur with fixed parameters.

[0151] Therefore, by applying a dynamically adjusted envelope extraction algorithm to process the current waveform data and vibration waveform data corresponding to each local energy concentration region, accurate current envelope signals and vibration envelope signals reflecting the signal modulation characteristics can be obtained. These envelope signals can effectively highlight the impulse components or periodic modulation information in the signal, providing high-quality input for subsequent feature point sequence extraction.

[0152] The solution proposed in this application first performs time-frequency analysis on the current waveform data and vibration waveform data of the local energy concentration area to obtain the instantaneous frequency and instantaneous amplitude change trends, thereby enabling a deeper understanding of the inherent dynamic characteristics of the signal.

[0153] In another embodiment of this application, S5113 further includes:

[0154] S5113-1: Obtain the current operating status of the motor and divide the operating status into the starting stage, steady-state operation stage and shutdown stage;

[0155] S5113-2: During the startup phase, a preset low amplitude threshold and a preset long peak duration threshold are used to detect peak values ​​in the current envelope signal and vibration envelope signal, and the detected peak values ​​are filtered in combination with the preset peak duration threshold and amplitude threshold.

[0156] S5113-3: During the shutdown phase, a preset low amplitude threshold and a preset short peak duration threshold are used to detect peak values ​​in the current envelope signal and vibration envelope signal, and the detected peak values ​​are filtered in combination with the preset peak duration threshold and amplitude threshold.

[0157] S5113-4: During the steady-state operation phase, a preset high amplitude threshold and a preset medium peak duration threshold are used to detect peak values ​​in the current envelope signal and vibration envelope signal, and the detected peak values ​​are filtered in combination with the preset peak duration threshold and amplitude threshold.

[0158] S5113-5: Obtain the motor speed and its rate of change, fine-tune the peak duration threshold corresponding to the current operating stage in real time according to the rate of change, and complete peak detection and screening under the common constraints of the fine-tuned peak duration threshold and the corresponding amplitude threshold.

[0159] Obtaining the current operating status of the motor refers to dividing the entire operating cycle of the motor into the starting stage, steady-state operation stage, and shutdown stage by monitoring parameters such as the motor's starting current, speed changes, and running time. These stages have different signal characteristics. For example, the starting stage is usually accompanied by large transient currents and vibrations, while the steady-state operation stage is relatively stable.

[0160] During the startup phase, as the motor accelerates from a standstill to its rated speed, current and vibration signals may exhibit large transient changes and low initial amplitudes. Therefore, using a preset low amplitude threshold can ensure that even early abnormal signals with low amplitudes can be captured, while a preset long peak duration threshold helps identify potential fault characteristics that last longer but whose amplitudes may not have reached steady-state levels.

[0161] During the shutdown phase, as the motor decelerates from running to a stop, the signal amplitude typically decays gradually. Using a preset low amplitude threshold can detect weak abnormal signals that may appear before the motor stops, while a preset short peak duration threshold is suitable for capturing rapidly decaying transient events, avoiding misjudging normal decay processes as persistent faults.

[0162] During steady-state operation, the motor runs at a relatively stable speed, and the current and vibration signals are usually quite stable, but periodic operating noise may exist. Using a preset high amplitude threshold helps filter out normal background noise and minor fluctuations, thus more effectively identifying significant peaks that truly indicate a fault. A preset medium peak duration threshold is suitable for the typical fault characteristic duration during this phase.

[0163] Furthermore, the motor speed and its rate of change are acquired, and the peak duration threshold corresponding to the current operating stage is fine-tuned in real time based on the rate of change. The purpose is to finely adjust the peak detection duration threshold. Changes in motor speed, especially the rate of change, directly affect the frequency and duration of certain characteristics in the current and vibration signals. For example, when the speed rises or falls rapidly, the duration of certain fault characteristics may shorten or lengthen. By monitoring the rate of change of speed in real time and fine-tuning the peak duration threshold accordingly, the peak detection algorithm can better adapt to the dynamic operating conditions of the motor, improving detection accuracy during speed fluctuations or load changes.

[0164] The solution proposed in this application achieves adaptive matching of signal characteristics for different operating stages by dynamically adjusting the amplitude threshold and peak duration threshold of peak detection according to the motor's operating state (start-up, steady state, shutdown).

[0165] In some preferred embodiments, it is assumed that during the startup phase of the permanent magnet DC pump motor, the amplitude of its current and vibration signals is typically low and transient. In this case, the system employs, for example, an amplitude threshold 50% lower than that of the steady-state operation phase and a peak duration threshold twice as long as that of the steady-state operation phase to ensure that any weak but persistent abnormal vibrations or current pulses that may occur during motor startup are captured. Once the motor enters the steady-state operation phase, the system automatically switches to a higher amplitude threshold (e.g., 50% higher than during startup) and a moderate peak duration threshold to filter out normal operating fluctuations and focus on detecting fault characteristics with significant amplitude and specific duration. If, during steady-state operation, the motor speed fluctuates by ±10% due to load changes, the system fine-tunes the peak duration threshold in real time based on the rate of speed change. For example, when the speed decreases, the duration threshold is appropriately extended by 5-10% to accommodate potentially wider vibration pulses, thereby ensuring accurate identification of characteristic points even under dynamic operating conditions.

[0166] Reference Figure 2 The specific implementation of this application also discloses a cloud computing-based permanent magnet DC pump motor performance evaluation system, including:

[0167] Data acquisition module 1 is used to simultaneously acquire the current waveform data and vibration waveform data of the permanent magnet DC pump motor when abnormal vibration of the permanent magnet DC pump motor is detected.

[0168] The preliminary judgment module 2 is used to extract time-domain features from current waveform data and vibration waveform data within a preset short-time analysis window to obtain a first criterion for preliminary judgment of vibration abnormal events. The preliminary judgment includes external hydraulic impact-dominated events, mechanical failure-prone events, or uncertain events.

[0169] The resonance identification module 3 is used to perform spectral analysis on the vibration waveform data, identify the matching relationship between the low-frequency main peak frequency of the vibration and the natural frequency of the vibration damping base, and calculate the proportion of low-frequency energy to obtain a second criterion reflecting the existence and intensity of background resonance.

[0170] The energy stripping and evaluation module 4 is used to strip the background resonance frequency band energy corresponding to the natural frequency of the vibration base from the total vibration energy of the vibration waveform data when the first criterion indicates that the external hydraulic impact is the dominant event and the second criterion indicates that there is background resonance, to obtain the remaining vibration energy, and based on the change of the remaining vibration energy within a preset period before and after the vibration abnormal event and the increase of the high-frequency energy of the vibration waveform data, to obtain a third criterion reflecting the non-resonance vibration increment and the high-frequency energy increment.

[0171] Phase verification module 5 is used to calculate the cross-correlation function between current waveform data and vibration waveform data, determine the time delay and correlation coefficient between the two, and verify the phase relationship of electromechanical response to obtain the fourth criterion;

[0172] The control command reading module 6 is used to obtain the output command value of the control loop of the motor driver or the change information of its duty cycle on the permanent magnet DC pump motor drive signal, and to determine whether the current step has a causal relationship with the change in the output of the control loop, forming a fifth criterion that reflects the relationship between the sudden change in external load and the current response.

[0173] The delayed confirmation module 7 is used to set a delayed confirmation window when the first criterion is an uncertain event. Within the delayed confirmation window, the energy change trend of the high-frequency components in the vibration waveform data over time is continuously observed to distinguish between transient external disturbances and mechanical damage with persistent characteristics, and to obtain the sixth criterion.

[0174] The classification and judgment module 8 is used to classify abnormal vibration events based on the first criterion, the second criterion, the third criterion, the fourth criterion, the fifth criterion, and the sixth criterion, according to the pre-set multi-dimensional deterministic numerical combination rules, and to obtain the classification results. The classification includes at least non-fault events caused by external transient hydraulic impact superimposed on background resonance, where there is a significant time lag between current step and vibration response, and / or real mechanical fault events caused by rotor jamming, bearing damage, or impeller failure, where current and vibration rise synchronously.

[0175] The decision and instruction module 9 is used to control whether to issue a shutdown instruction to the operation and maintenance terminal and / or trigger local mitigation and delayed confirmation measures based on the classification results.

[0176] This application proposes a cloud computing-based permanent magnet DC pump motor performance evaluation system, which aims to solve the problems of false alarms caused by vibration damping base resonance, inability to distinguish resonance from actual mechanical faults, and electromechanical evaluation logic conflicts in traditional evaluation systems under low-speed operation conditions of permanent magnet DC pump motors.

[0177] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A cloud computing-based permanent magnet DC pump motor performance evaluation method, characterized in that, The application is applied to a cloud computing platform, comprising: When detecting vibration abnormality of the permanent magnet DC pump motor, synchronously acquiring current waveform data and vibration waveform data of the permanent magnet DC pump motor; In a preset short-time analysis window, time domain feature extraction is performed on the current waveform data and the vibration waveform data to obtain a first criterion for preliminary determination of the vibration abnormality event, and the preliminary determination includes an external hydraulic impact dominant event, a mechanical failure tendency event or an uncertain event; Spectrum analysis is performed on the vibration waveform data to identify a matching relationship between a vibration low-frequency main peak frequency and a natural frequency of a vibration reduction base, and to calculate a low-frequency energy proportion, thereby obtaining a second criterion reflecting existence and intensity of background resonance; When the first criterion represents the external hydraulic impact dominant event and the second criterion represents existence of the background resonance, background resonance frequency band energy corresponding to the natural frequency of the vibration reduction base is stripped from total vibration energy of the vibration waveform data to obtain residual vibration energy, and based on changes in the residual vibration energy and an increase in high-frequency energy of the vibration waveform data within a preset time period before and after the vibration abnormality event, a third criterion reflecting non-resonance vibration increment and high-frequency energy increment is obtained; A cross-correlation function is calculated for the current waveform data and the vibration waveform data to determine time delay and correlation coefficient therebetween, thereby verifying a phase relationship of electromechanical response to obtain a fourth criterion; Control loop output instruction values of a motor driver or change information of duty cycles of driving signals of the permanent magnet DC pump motor are acquired to determine whether current steps have a causal corresponding relationship with changes in the control loop output, thereby forming a fifth criterion reflecting a relationship between external load mutation and current response; When the first criterion is the uncertain event, a delay confirmation window is set, and changes in energy of high-frequency components in the vibration waveform data over time are continuously observed within the delay confirmation window to distinguish transient external interference from mechanical damage with persistent characteristics, thereby obtaining a sixth criterion; Based on the first criterion, the second criterion, the third criterion, the fourth criterion, the fifth criterion and the sixth criterion, a vibration abnormality event is classified according to a pre-set multi-dimensional certainty numerical combination rule to obtain a classification result, and the classification at least includes a non-failure event caused by superposition of external transient hydraulic impact and background resonance, a significant time lag between current steps and vibration response and / or a real mechanical failure event caused by rotor jamming, bearing damage or impeller failure, and synchronous rise of current and vibration; According to the classification result, whether to issue a shutdown instruction to an operation and maintenance end and / or trigger local mitigation and delay confirmation measures is controlled.

2. The cloud computing-based permanent magnet DC pump motor performance evaluation method according to claim 1, characterized in that, In a preset short-time analysis window, time domain feature extraction is performed on the current waveform data and the vibration waveform data to obtain a first criterion for preliminary determination of the vibration abnormality event, including: Adaptive filtering processing is performed on the current waveform data and the vibration waveform data to suppress random noise and background interference while retaining instantaneous changes reflecting mechanical and electrical response mutation characteristics, thereby obtaining filtered current waveform data and vibration waveform data; Differential operations are performed on the filtered current waveform data and the vibration waveform data at multiple time scales to obtain current transient slope and vibration transient slope at each time scale; Based on local extreme values and their retention characteristics within a preset duration, current transient slope peak time is determined from the current transient slope, vibration transient slope peak time is determined from the vibration transient slope, and corresponding current transient slope peak and vibration transient slope peak are obtained; The current transient slope peak and the vibration transient slope peak are dynamically normalized according to the steady-state current and vibration level before the occurrence of the vibration abnormal event to obtain a normalized slope ratio; The normalized slope ratio is combined with the peak time difference between the current transient slope peak time and the vibration transient slope peak time as the first criterion. 3.The cloud-computing-based permanent magnet DC pump motor performance evaluation method according to claim 1, characterized in that, Frequency spectrum analysis is performed on the vibration waveform data to identify the matching relationship between the vibration low-frequency main peak frequency and the inherent frequency of the vibration isolation base, and the low-frequency energy proportion is calculated to obtain the second criterion reflecting the existence and intensity of background resonance, including: Frequency spectrum analysis is performed on the vibration waveform data to identify the vibration low-frequency main peak frequency; Temperature data of the vibration isolation base corresponding to the permanent magnet DC pump motor are obtained; According to the temperature data and the preset temperature-elastic modulus characteristics of the vibration isolation base material, the elastic modulus of the vibration isolation base at the current time is estimated; According to the elastic modulus and the structural parameters of the vibration isolation base, the inherent frequency and its drift interval of the vibration isolation base at the current time are dynamically calculated; The vibration low-frequency main peak frequency is matched with the inherent frequency and its drift interval to obtain the matching relationship between the vibration low-frequency main peak frequency and the inherent frequency of the vibration isolation base; The vibration waveform data are energy integrated within a preset low-frequency frequency band to calculate the proportion of low-frequency energy in total vibration energy, and the low-frequency energy proportion is obtained and combined with the matching relationship as the second criterion.

4. The cloud-computing-based permanent magnet DC pump motor performance evaluation method according to claim 1, characterized in that, When the first criterion represents an external hydraulic impact dominant event and the second criterion represents the existence of background resonance, the background resonance frequency band energy corresponding to the inherent frequency of the vibration isolation base is stripped from the total vibration energy of the vibration waveform data to obtain the residual vibration energy, and based on the change of the residual vibration energy within a preset period before and after the vibration abnormal event and the increase of high-frequency energy of the vibration waveform data, a third criterion reflecting the non-resonant vibration increment and high-frequency energy increment is obtained, including: Continuous wavelet transform is performed on the vibration waveform data to obtain a time-frequency spectrum; In the time-frequency spectrum, an energy concentration area is identified, and the instantaneous frequency and instantaneous bandwidth corresponding to the background resonance are dynamically determined; An adaptive band-stop filter is constructed according to the instantaneous frequency and the instantaneous bandwidth, and the vibration waveform data are real-time filtered using the adaptive band-stop filter to strip the background resonance frequency band energy and obtain the residual vibration energy waveform; During the process of stripping the background resonance frequency band energy, the energy level of the residual vibration energy waveform within a preset high-frequency frequency band is monitored, and when the high-frequency component abnormally decays relative to the reference level before the occurrence of the vibration abnormal event, the bandwidth and / or center frequency of the adaptive band-stop filter are adjusted. integrate the residual vibration energy waveforms in each preset time period before and after the vibration abnormal event, compare the energy value in the preset time period after the vibration abnormal event with the energy value in the preset time period before the vibration abnormal event, and obtain the growth amplitude of the residual vibration energy; perform fast Fourier transform on the vibration waveform data, integrate the energy in a preset high-frequency frequency band, compare the high-frequency energy in the preset time period after the vibration abnormal event with a high-frequency energy reference level in the preset time period before the vibration abnormal event, and obtain the growth amplitude of the high-frequency energy; use the growth amplitude of the residual vibration energy and the growth amplitude of the high-frequency energy as the third criterion.

5. The cloud-computing-based permanent magnet DC pump motor performance evaluation method according to claim 1, characterized in that, calculate a cross-correlation function of the current waveform data and the vibration waveform data, determine a time delay and a correlation coefficient between the current waveform data and the vibration waveform data, verify a phase relationship of the electromechanical response, and obtain a fourth criterion, including: divide the current waveform data and the vibration waveform data into a plurality of short-time analysis windows that overlap with each other; extract a current feature point sequence and a vibration feature point sequence representing the electromechanical response from the current waveform data and the vibration waveform data in each short-time analysis window; calculate a cross-correlation function between the current feature point sequence and the vibration feature point sequence in each short-time analysis window, and obtain a local time delay and a maximum correlation coefficient corresponding to each short-time analysis window; perform trend analysis on the local time delay and the maximum correlation coefficient of each short-time analysis window arranged in time sequence, and obtain a feature of a change of the electromechanical response phase relationship with time; use the leading or lagging relationship and stability of the current relative to the vibration as the fourth criterion according to the result of the trend analysis.

6. The cloud-computing-based permanent magnet DC pump motor performance evaluation method according to claim 5, characterized in that, extract a current feature point sequence and a vibration feature point sequence representing the electromechanical response from the current waveform data and the vibration waveform data in each short-time analysis window, including: perform multi-resolution analysis on the current waveform data and the vibration waveform data in each short-time analysis window, and identify local energy concentration regions at different resolution levels; extract time-domain envelopes of the current waveform data and the vibration waveform data corresponding to each local energy concentration region, and obtain a current envelope signal and a vibration envelope signal; perform peak value detection on the current envelope signal and the vibration envelope signal, and screen the detected peak values by combining a preset peak value duration threshold condition and an amplitude threshold, and use the peak time meeting the peak value duration threshold condition and the amplitude threshold as the current feature point sequence and the vibration feature point sequence, respectively.

7. The cloud-computing-based permanent magnet DC pump motor performance evaluation method according to claim 6, characterized in that, perform multi-resolution analysis on the current waveform data and the vibration waveform data in each short-time analysis window, and identify local energy concentration regions at different resolution levels, including: perform wavelet packet decomposition on the current waveform data and the vibration waveform data in each short-time analysis window, and obtain a plurality of wavelet packet frequency subbands covering different frequency ranges; calculate energy distribution of each wavelet packet frequency subband; preliminarily identify candidate local energy concentration subbands according to the energy distribution and the frequency range of each wavelet packet frequency subband; For the candidate local energy concentrated sub-band with low energy but its frequency falls into the preset fault feature frequency neighborhood, the weight in the subsequent judgment is adjusted and increased; The adjusted candidate local energy concentrated sub-band is jointly judged according to the energy change trend of the adjacent wavelet packet frequency sub-band, and a preliminary local energy concentrated region is obtained; For the wavelet packet frequency sub-band in the preset high-frequency frequency band, the energy threshold is dynamically adjusted according to the background noise level and / or the instantaneous signal-to-noise ratio, the high-frequency noise and the high-frequency fault feature are distinguished for each wavelet packet frequency sub-band, and the wavelet packet frequency sub-band judged as the high-frequency fault feature is integrated into the preliminary local energy concentrated region to form an updated local energy concentrated region; The updated local energy concentrated region is finally confirmed in combination with the sparsity and / or pulse characteristics of the signal in the time domain in each wavelet packet frequency sub-band. 8.The cloud-computing-based permanent magnet DC pump motor performance evaluation method according to claim 6, characterized in that, The current envelope signals and the vibration envelope signals are obtained by performing time-domain envelope extraction on the current waveform data and the vibration waveform data corresponding to each local energy concentrated region, including: The instantaneous frequency and the instantaneous amplitude change trend of the local energy concentrated region in the corresponding short-time analysis window are determined by performing time-frequency analysis on the current waveform data and the vibration waveform data corresponding to each local energy concentrated region; According to the instantaneous frequency and the instantaneous amplitude change trend, the algorithm parameters for envelope extraction are dynamically adjusted, and the algorithm parameters at least include a pretreatment analysis window length and / or a window width of a smoothing filter and a weight coefficient thereof; The current envelope signals and the vibration envelope signals are obtained by applying the envelope extraction algorithm dynamically adjusted to the current waveform data and the vibration waveform data corresponding to each local energy concentrated region. 9.The cloud-computing-based permanent magnet DC pump motor performance evaluation method according to claim 6, characterized in that, The peak values of the current envelope signals and the vibration envelope signals are detected, and the detected peak values are screened in combination with a preset peak duration threshold condition and an amplitude threshold value, and the peak time meeting the peak duration threshold condition and the amplitude threshold value is taken as the current feature point sequence and the vibration feature point sequence, including: The current running state of the motor is obtained, and the running state is divided into a starting stage, a steady-state running stage and a shutdown stage; In the starting stage, the current envelope signals and the vibration envelope signals are detected by adopting a preset lower amplitude threshold value and a preset longer peak duration threshold value, and the detected peak values are screened in combination with a preset peak duration threshold condition and an amplitude threshold value; In the shutdown stage, the current envelope signals and the vibration envelope signals are detected by adopting a preset lower amplitude threshold value and a preset shorter peak duration threshold value, and the detected peak values are screened in combination with a preset peak duration threshold condition and an amplitude threshold value; In the steady-state running stage, the current envelope signals and the vibration envelope signals are detected by adopting a preset higher amplitude threshold value and a preset medium peak duration threshold value, and the detected peak values are screened in combination with a preset peak duration threshold condition and an amplitude threshold value; The motor speed and its change rate are acquired, the peak duration threshold corresponding to the current operation stage is adjusted in real time according to the change rate, and the peak detection and screening are completed under the joint constraint of the adjusted peak duration threshold and the corresponding amplitude threshold.

10. A cloud computing-based permanent magnet DC pump motor performance evaluation system, characterized in that, Comprise: A data acquisition module is configured to, when detecting vibration abnormality of a permanent magnet DC pump motor, synchronously acquire current waveform data and vibration waveform data of the permanent magnet DC pump motor; A preliminary determination module is configured to, within a preset short-time analysis window, perform time domain feature extraction on the current waveform data and the vibration waveform data to obtain a first criterion for preliminary determination of the vibration abnormality event, the preliminary determination including an external hydraulic impact dominant event, a mechanical fault tendency event or an uncertain event; A resonance identification module is configured to perform frequency spectrum analysis on the vibration waveform data, identify a matching relationship between a vibration low-frequency main peak frequency and a natural frequency of a vibration reduction base, and calculate a low-frequency energy proportion to obtain a second criterion reflecting existence and intensity of background resonance; An energy stripping and evaluation module is configured to, when the first criterion represents the external hydraulic impact dominant event and the second criterion represents existence of the background resonance, strip background resonance frequency band energy corresponding to the natural frequency of the vibration reduction base from total vibration energy of the vibration waveform data to obtain residual vibration energy, and based on changes in the residual vibration energy within a preset time period before and after the vibration abnormality event and an increase in high-frequency energy of the vibration waveform data, obtain a third criterion reflecting non-resonance vibration increment and high-frequency energy increment; A phase verification module is configured to calculate a cross-correlation function of the current waveform data and the vibration waveform data to determine a time delay and a correlation coefficient therebetween, verify a phase relationship of an electromechanical response, and obtain a fourth criterion; A control instruction reading module is configured to acquire a control loop output instruction value of a motor driver or change information of a duty cycle of a driving signal of the permanent magnet DC pump motor, determine whether a current step has a causal corresponding relationship with a change in the control loop output, and form a fifth criterion reflecting a relationship between an external load mutation and a current response; A delay confirmation module is configured to, when the first criterion is the uncertain event, set a delay confirmation window, continuously observe a change trend of energy of a high-frequency component in the vibration waveform data over time within the delay confirmation window to distinguish between transient external interference and mechanical damage with persistent characteristics, and obtain a sixth criterion; A classification determination module is configured to, based on the first criterion, the second criterion, the third criterion, the fourth criterion, the fifth criterion and the sixth criterion, perform classification on the vibration abnormality event according to a pre-set multi-dimensional certainty numerical combination rule to obtain a classification result, the classification at least including a non-fault event caused by superposition of external transient hydraulic impact and background resonance, and / or a real mechanical fault event caused by rotor jamming, bearing damage or impeller failure, and synchronous rise of current and vibration; A decision and instruction module is configured to, according to the classification result, control whether to issue a shutdown instruction to an operation and maintenance end and / or trigger local mitigation and delay confirmation measures.

Images