Self-adaptive calibration test method and system for vibration quantity of water pump for cooling AI server

Abstract

The invention relates to an AI server cooling water pump vibration quantity self-adaptive calibration test method and system, an intelligent test platform integrates a six-dimensional force sensor and a temperature compensation vibration exciter, the platform rigidity is automatically calibrated, a water pump-pipeline system transfer function is obtained, and a rotating speed-lift-modal frequency three-dimensional mathematical model is established; a three-axis MEMS accelerometer array is arranged at sensitive parts such as a water pump bearing seat and a motor shell, and vibration, current and pressure signals are synchronously collected; carrying out time-frequency domain signal processing by adopting variational mode decomposition in combination with self-adaptive S transformation, and extracting a 128-dimensional full-frequency domain feature vector containing a modulation side frequency band; effective values of vibration acceleration, speed and displacement are calculated through a frequency domain integration algorithm, and current harmonic interference is corrected through an electromagnetic vibration compensation model; and finally, a vibration health degree evaluation model containing 18 characteristic parameters is established based on a support vector machine, the test system comprises an intelligent test platform unit, a multi-source sensing unit, an edge calculation unit and a data management unit, and microsecond-level synchronous acquisition and GB-level data throughput are realized through a time sensitive network. And online incremental learning and automatic generation of a test report are supported. The method has the effect of improving the water pump vibration quantity test precision.

Description

TECHNICAL FIELD

[0001] The application relates to the field of water pump detection, in particular to an AI server cooling water pump vibration quantity self-adaptive calibration test method and system. BACKGROUND

[0002] With the rapid development of artificial intelligence technology, the operation scale of AI server clusters grows exponentially, and the high heat generated during operation puts high requirements on the cooling system. As the core power component of the AI server liquid cooling system, the stable operation of the water pump directly determines the cooling efficiency and server reliability. The water pump vibration quantity is not only a key indicator for measuring its operating state, but abnormal vibration may also cause sealing failure, bearing wear and other faults, leading to cooling liquid leakage and even server downtime. Therefore, accurately testing the water pump vibration quantity and achieving early fault warning have become necessary links to ensure the stable operation of AI data centers.

[0003] Currently, the traditional water pump vibration test method has significant limitations. In terms of test environment construction, the existing test platform mostly adopts a fixed stiffness structure, which cannot adapt to the dynamic characteristics of different types of water pumps, resulting in boundary condition error affecting the test results; in the signal acquisition link, single-point or low-precision sensor arrangement cannot fully capture the coupled vibration characteristics of multiple parts of the water pump, and lacks synchronous monitoring of multiple physical quantities such as current and pressure, which cannot separate interference factors such as electromagnetic vibration. In terms of data processing, conventional Fourier transform cannot analyze complex modulation signals, and simple threshold judgment cannot meet the high-precision diagnosis requirements of AI water pumps, resulting in large vibration quantity calculation error and low fault diagnosis accuracy.

[0004] To solve the above problems, the existing technical means cannot accurately test and evaluate the vibration quantity of AI server cooling water pumps. There is an urgent need for a test method and system with self-adaptive calibration capability and fusion of multi-source data, which can effectively improve the vibration quantity test precision and fault diagnosis reliability through intelligent sensing and advanced algorithms, and fill the technical gap in the field of AI server dedicated cooling equipment testing. SUMMARY

[0005] In order to improve the water pump vibration quantity test precision, the application provides an AI server cooling water pump vibration quantity self-adaptive calibration test method and system.

[0006] The AI server cooling water pump vibration quantity self-adaptive calibration test method and system provided by the application adopts the following technical solution: First aspect An AI server cooling water pump vibration quantity test method, comprising the following steps: S1, adaptive calibration start: install the water pump to be tested on the intelligent test platform integrated with six-dimensional force sensors, automatically calibrate the platform stiffness matrix through a closed-loop feedback control system, synchronously collect the initial modal parameters of the water pump in the no-load state, and establish a three-dimensional mathematical model containing the mapping relationship of rotational speed-lift-modal frequency; S2, multi-modal signal synchronous acquisition: at the radial and axial measuring points of the water pump bearing seat, the motor rotor shell and the impeller volute, arrange a three-axis MEMS accelerometer array in accordance with the ISO 1940 balance grade requirements, use a synchronous trigger data acquisition card to obtain vibration time domain signals at a sampling rate of not less than 20 kHz, and simultaneously synchronously collect water pump driving current waveforms and real-time data of inlet and outlet pressures; S3, time-frequency domain joint signal processing: perform variational modal decomposition on the original vibration signal, optimize the decomposition parameters in the modal 2-8 range through an adaptive particle swarm algorithm, screen the intrinsic modal functions with a kurtosis value greater than 2.5 in combination with the kurtosis criterion, construct a time-frequency energy distribution map based on an adaptive S transform, identify the fundamental frequency vibration component and extract the full-frequency domain feature vector containing the modulation side frequency band by using the three-dimensional mathematical model established in step S1; S4, multi-physical quantity fusion calculation: perform coordinate conversion on the time domain vibration signal according to the accelerometer sensitivity matrix, synchronously calculate the vibration acceleration effective value, vibration speed effective value and vibration displacement effective value by using a frequency domain integral algorithm, construct an electromagnetic vibration compensation model by establishing the cross-correlation function between the driving current spectrum and the vibration signal, generate a compensation coefficient matrix by least square fitting for the current harmonic components with a correlation coefficient greater than 0.3, and correct the calculation results.

[0007] By using the above technical scheme, adaptive calibration start is realized through the intelligent test platform integrated with six-dimensional force sensors, a three-dimensional mathematical model of water pump rotational speed-lift-modal frequency is established, the problem of large boundary condition error of the traditional test platform is solved, the dynamic characteristics of different water pumps can be accurately matched, multi-modal signal synchronous acquisition is combined with multi-physical quantity monitoring, vibration characteristics of the water pump are comprehensively captured and interference factors are separated, time-frequency domain joint signal processing and multi-physical quantity fusion calculation effectively analyze complex vibration signals and improve the vibration quantity calculation accuracy, and reliable data basis is provided for water pump health diagnosis.

[0008] Preferably, the intelligent test platform is configured with a temperature compensation exciter, a transfer function matrix of the water pump-pipeline system is obtained by 10-1000 Hz sweep excitation in the start-up stage, and is used for boundary condition correction of the subsequently collected vibration signals.

[0009] By adopting the technical scheme, the temperature compensation type exciter configured by the intelligent test platform obtains a water pump-pipeline system transfer function matrix through 10-1000Hz sweep excitation, can correct the boundary conditions of the vibration signal according to the environmental temperature change and the pipeline dynamic characteristics, eliminate the influence of external factors on the test result, and further improve the test accuracy and reliability.

[0010] Preferably, the installation positions of the accelerometer array are accurately calibrated in three-dimensional coordinates by a laser range finder, a sensor layout matrix containing position information is formed, and the installation angle error of each sensor is controlled within ±0.5°, meeting the ISO 20816 vibration measurement position accuracy requirement.

[0011] By adopting the technical scheme, the installation positions of the accelerometer array are accurately calibrated by the laser range finder, the installation angle error of the sensor is controlled within ±0.5°, the high-precision measurement requirement of the international standard is met, the spatial position accuracy of the collected vibration signal is ensured, signal distortion caused by sensor installation deviation is avoided, and the effectiveness of vibration data acquisition is improved.

[0012] Preferably, when the full-frequency-domain feature vector is extracted, a ±3 times frequency conversion analysis bandwidth centered on the fundamental frequency is defined, an improved Teager energy operator is used to calculate the instantaneous energy of each frequency component, and a 128-dimensional vibration state vector containing time domain, frequency domain and time-frequency domain features is constructed.

[0013] By adopting the technical scheme, when the full-frequency-domain feature vector is extracted, the analysis bandwidth is demarcated centered on the fundamental frequency, the instantaneous energy is calculated by using the improved Teager energy operator, and the 128-dimensional vibration state vector is constructed, the time domain, frequency domain and time-frequency domain features of the vibration signal can be completely extracted, the early weak fault features can be effectively identified, and the sensitivity and accuracy of fault diagnosis are improved.

[0014] Preferably, the vibration quantity calculation result includes: time domain characteristic quantity: peak value, effective value, kurtosis value, margin factor; frequency domain characteristic quantity: fundamental frequency amplitude ratio, sideband energy ratio, octave energy distribution; time-frequency domain characteristic quantity: energy barycenter frequency, root mean square frequency, frequency variance.

[0015] By adopting the technical scheme, multiple time domain, frequency domain and time-frequency domain characteristic quantities are defined as the vibration quantity calculation result, the vibration state of the water pump is fully characterized from multiple dimensions, compared with single index analysis, the running condition of the water pump can be more detailed and accurately reflected, and rich parameter basis is provided for subsequent fault diagnosis and health assessment.

[0016] Preferably, a vibration amount threshold self-learning module is set in the data processing stage, a support vector machine classifier is trained by historical test data, and a water pump vibration health degree evaluation model containing 18 characteristic parameters is established, including all the above-mentioned preferred characteristic quantities and water pump operating condition parameters.

[0017] By adopting the above technical solution, by setting the vibration amount threshold self-learning module, training the SVM classifier by using the historical test data, and establishing the water pump vibration health degree evaluation model containing multiple characteristic parameters, the autonomous learning and intelligent diagnosis of the test system are realized, the diagnostic threshold can be dynamically adjusted according to different water pump characteristics, and the adaptability and accuracy of fault diagnosis are improved.

[0018] The second aspect A system applied to the above-mentioned water pump vibration amount test for AI server cooling, comprising a. Intelligent test platform unit: integrating a six-dimensional force sensor array, a temperature compensation exciter and a laser calibration module, capable of constructing a dynamic coordinate system for water pump vibration test, automatically completing platform stiffness calibration and transfer function matrix measurement; b. Multi-source sensing unit: containing 12 channels of three-axis MEMS accelerometers, high-frequency current transformers and pressure transmitters, achieving microsecond-level time alignment through a synchronous clock module, with a sampling rate not less than 20kHz, meeting the ISO 13373 vibration signal acquisition standard; c. Edge computing unit: configuring a FPGA real-time signal processing module, built-in VMD noise reduction algorithm module, adaptive S transform engine and multi-physical quantity fusion calculation unit, supporting real-time time-frequency domain analysis and characteristic parameter calculation of vibration signals; d. Data management unit: based on a cloud computing platform to build a vibration characteristic database, integrating a health degree evaluation model training module and a test report automatic generation system, capable of realizing test data storage, model training and visual report output.

[0019] By adopting the above technical solution, through the collaborative work of the intelligent test platform unit, the multi-source sensing unit, the edge computing unit and the data management unit, a complete water pump vibration amount test system is constructed, forming a closed loop from hardware sensing to data processing, storage and analysis, realizing the automation and intelligentization of the test process, and improving the test efficiency and data management capability.

[0020] Preferably, time-sensitive network communication is adopted between the edge computing unit and the data management unit, the signal transmission delay is controlled within 50 microseconds, the data processing throughput is not less than 1GB / s, and the real-time and integrity of the test data are ensured.

[0021] By adopting the technical scheme, the time-sensitive network (TSN) communication is adopted, the signal transmission delay between the edge computing unit and the data management unit is controlled within 50 microseconds, the data processing throughput is ensured to be not less than 1 GB / s, the real-time transmission and processing requirements of a large amount of vibration data are met, the timeliness and integrity of the test data are ensured, and the real-time monitoring and rapid diagnosis are ensured.

[0022] Preferably, the sensitivity matrix of the triaxial MEMS accelerometer array is pre-calibrated by a national standard vibration calibration device, and the sensitivity error is controlled within ±1.5%, and the frequency response range covers 0.1Hz-10kHz.

[0023] By adopting the technical scheme, the triaxial MEMS accelerometer array is calibrated by a national standard vibration calibration device, the sensitivity error is controlled within ±1.5%, and the frequency response range covers 0.1Hz-10kHz, the high precision and wide frequency applicability of the sensor collected data are ensured, the accuracy and reliability of the vibration signal collection are improved, and a solid foundation is laid for subsequent data processing and analysis.

[0024] Preferably, the vibration health degree evaluation model training module supports online incremental learning, can dynamically update the compensation coefficient matrix and the SVM classifier parameters according to the real-time test data, and realizes self-calibration and freedom of the test system.

[0025] By adopting the technical scheme, the vibration health degree evaluation model training module supports online incremental learning, can dynamically update the compensation coefficient matrix and the SVM classifier parameters according to the real-time test data, and makes the test system have self-calibration and self-optimization capabilities, can adapt to the changes of the water pump operating state and the test environment, and continuously maintains high diagnostic precision and reliability.

[0026] In summary, the present application includes at least one of the following beneficial technical effects: 1. By integrating the intelligent test platform of the six-dimensional force sensor and the temperature compensation type exciter, the platform stiffness matrix is automatically calibrated, and the water pump-pipeline system transfer function matrix is obtained, the problem that the traditional fixed stiffness platform cannot adapt to different water pump dynamic characteristics is solved, the boundary condition error is reduced by more than 60%. The no-load modal parameter acquisition and the three-dimensional mathematical model are established, the accurate mapping of the speed-lift-modal frequency is realized, a reliable dynamic coordinate system is provided for subsequent vibration signal analysis, and the universality and consistency of the test of different types of water pumps are significantly improved; 2.Using the variational mode decomposition (VMD) and the adaptive particle swarm optimization algorithm to optimize the decomposition parameters, combining the kurtosis criterion to screen effective modes, and suppressing noise interference while retaining early fault characteristics. Based on the time-frequency energy spectrum constructed by the adaptive S-transform, the visualization representation of the vibration signal in the time-frequency-energy three-dimensional space is realized, and the improved Teager energy operator is used to extract a 128-dimensional full-frequency domain feature vector. Compared with the traditional Fourier transform, the identification sensitivity of weak fault characteristics such as modulation sideband is improved by 40%, effectively solving the analysis problem of complex modulation signals; 3.The edge computing unit and the data management unit use time-sensitive network (TSN) communication, the signal transmission delay is controlled within 50 microseconds, the data throughput is ≥1GB / s, and the real-time processing and feature parameter calculation of multi-channel vibration signals under a 20kHz sampling rate are ensured (response time ≤200 microseconds). Combined with the database management and visual report generation system of the cloud computing platform, the whole process automation from signal acquisition, real-time analysis to health assessment is realized, which greatly improves the test efficiency and data management capability. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is an execution step block diagram of an AI server cooling water pump vibration quantity adaptive calibration test method according to an embodiment of the application.

[0028] Figure 2 is a time-frequency domain joint signal processing flow block diagram in an AI server cooling water pump vibration quantity adaptive calibration test method according to an embodiment of the application.

[0029] Figure 3 is a multi-physical quantity fusion calculation flow block diagram in an AI server cooling water pump vibration quantity adaptive calibration test method according to an embodiment of the application.

[0030] Figure 4 is a whole test principle block diagram of an AI server cooling water pump vibration quantity adaptive calibration test method and system according to an embodiment of the application. DETAILED DESCRIPTION

[0031] The following will be described in detail with reference to the accompanying drawings Figures 1-4 The application will be further described in detail.

[0032] The application discloses an AI server cooling water pump vibration quantity adaptive calibration test method and system.

[0033] Reference Figure 1 An AI server cooling water pump vibration quantity test method includes the following steps: S1, adaptive calibration start: install the water pump to be tested on the intelligent test platform integrated with six-dimensional force sensors, configure a temperature compensation exciter on the intelligent test platform, obtain the transfer function matrix of the water pump-pipeline system through 10-1000 Hz sweep excitation in the starting stage, use the transfer function matrix to correct the boundary conditions of the subsequently collected vibration signals, automatically calibrate the platform stiffness matrix through a closed-loop feedback control system, synchronously collect the initial modal parameters of the water pump in the no-load state, and establish a three-dimensional mathematical model containing the mapping relationship of rotational speed-head-modal frequency; Specifically, the collected rotational speed, head, and modal frequency data are imported into the Scikit-learn library of Python, a three-dimensional mathematical model containing the mapping relationship of rotational speed-head-modal frequency is established by using a multivariate linear regression algorithm for fitting, the model parameters are optimized through cross-validation to ensure that the determination coefficient R 2 of the model is greater than 0.98, and the accuracy and generalization ability of the model are ensured. In actual testing, the corresponding modal frequency of the water pump can be quickly predicted according to the real-time rotational speed and head of the water pump through the model, thereby providing a reference basis for subsequent vibration analysis.

[0034] S2, synchronous acquisition of multi-modal signals: refer to Figure 2 ISO 1940-1:2017, balance grade requirements, 1 accelerometer is arranged at each of the radial and axial measuring points of the bearing seat, 2 accelerometers are symmetrically arranged on the motor rotor shell, and 2 accelerometers are arranged near the blades of the impeller volute, so as to comprehensively monitor the vibration characteristics of the water pump at different positions, a three-axis MEMS accelerometer array is arranged in accordance with the ISO 1940 balance grade requirements, a synchronous trigger data acquisition card is used to obtain vibration time domain signals at a sampling rate of not less than 20 kHz, water pump driving current waveforms and real-time data of inlet and outlet pressures are synchronously collected, the installation positions of the accelerometer array are accurately calibrated by a laser range finder in three-dimensional coordinates to form a sensor layout matrix containing position information, the installation angle error of each sensor is controlled within ±0.5°, and the position accuracy requirement of ISO 20816 vibration measurement is met. Among them, the selection of the data acquisition card can use a synchronous trigger data acquisition card such as NIPXIe-4499, which has a multi-channel synchronous acquisition function, and the highest sampling rate can reach 102.4kS / s, meeting the requirement of a sampling rate of not less than 20 kHz. 16-24 analog input channels are configured, of which 12-18 channels are used to connect accelerometers (each three-axis accelerometer occupies 3 channels), 2-4 channels are used to collect current signals, and 2 channels are used to collect pressure signals, so as to ensure that various signals can be simultaneously connected to the acquisition system.

[0035] Further, the sampling rate of the data acquisition card is set to 24 kHz, which can ensure sufficient signal details and avoid data redundancy and processing pressure caused by too high sampling rate. The trigger mode of the acquisition card is set to external synchronization trigger, and through connecting a high-precision synchronization clock module (such as a GPS synchronization clock or a dedicated 10 MHz synchronization clock), the time alignment of all channel signals is realized to the microsecond level, ensuring the consistency of multi-modal signals in the time dimension.

[0036] Meanwhile, the whole system needs to be calibrated. After completing the installation of all sensors, the joint calibration of the whole multi-modal signal acquisition system is performed. By simulating the operating conditions of the water pump, standard vibration, current and pressure signals are input to check whether the output of each channel of the acquisition system is consistent with the input signal. The gain, offset and other parameters of the acquisition card are adjusted by using calibration software to ensure that the overall measurement accuracy of the system meets the requirements. At the same time, the synchronization effect of the synchronization clock module is verified, and by comparing the time difference of the same trigger signal collected by different channels, the synchronization parameters are adjusted to ensure accurate synchronization acquisition of multi-modal signals.

[0037] S3, joint signal processing in time and frequency domain: the original vibration signal is subjected to variational mode decomposition, the decomposition parameters are optimized in the mode 2-8 range by using an adaptive particle swarm algorithm, the intrinsic mode functions with a kurtosis value greater than 2.5 are screened by combining the kurtosis criterion, the time-frequency energy distribution map is constructed based on the adaptive S transform, and the three-dimensional mathematical model established in step S1 is used to identify the fundamental frequency vibration component and extract the full-frequency domain feature vector containing the modulation side frequency band; wherein, when the full-frequency domain feature vector is extracted, a ±3 times rotational frequency analysis bandwidth centered on the fundamental frequency is defined, the instantaneous energy of each frequency component is calculated by using an improved Teager energy operator, and a 128-dimensional vibration state vector containing time domain, frequency domain and time-frequency domain features is constructed.

[0038] Specifically, the original vibration signal collected often contains noise and interference, which is first subjected to detrending treatment to eliminate the direct current component and slow changing trend in the signal, so that the signal is more focused on the vibration characteristics. Then, a band-pass filter is used, and the filter passband is set to 0.1 Hz-10 kHz in combination with the vibration frequency characteristics of the water pump, so as to suppress low-frequency environmental noise and high-frequency electromagnetic interference and retain the effective vibration frequency band signal.

[0039] Further, the variational mode decomposition is the core step of decomposing the original signal into multiple intrinsic mode functions (IMF). VMD decomposes the complex signal into a limited number of bandwidth-limited modal components by constructing and solving a variational model. The core variational problem is represented as: wherein, u k is the set of modal functions obtained by decomposition, w kwhere f is the center frequency of each mode, s(t) is the original vibration signal.

[0040] To obtain the optimal decomposition effect, the adaptive particle swarm optimization (PSO) algorithm is used to optimize the decomposition mode number K (range set at 2-8) of VMD. The particle swarm optimization algorithm simulates the foraging behavior of bird flocks to search for the optimal solution in the solution space. In each iteration, the particle updates its position according to its own historical optimal position p best and the global optimal position g best , the update formula is as follows: v id (t+1) = ω·v id (t) + c1r 1d (t) [p id (t) - x i x id (t+1)] = x id (t+1) = x id (t) + v id (t+1) where v id is the particle velocity, x id is the particle position, ω is the inertia weight, c1 is the learning factor, r 1d is a random number.

[0041] The kurtosis value of each IMF after decomposition is taken as the optimization target, and the kurtosis calculation formula is as follows: where x i is the signal sample value, is the mean value, and σ is the standard deviation. The IMF with a kurtosis value greater than 2.5 is selected, and the modal component containing the fault characteristics is retained for subsequent analysis.

[0042] Further, for the adaptive S-transform and the signal analysis method with time-frequency localization characteristics, the definition is as follows: To adapt to the characteristics of the pump vibration signal, the adaptive S-transform is used to enhance the time-frequency resolution of different frequency components by optimizing the scale parameter of the S-transform. The selected IMF is subjected to adaptive S-transform to obtain a time-frequency matrix, with time as the horizontal axis, frequency as the vertical axis, and amplitude as the color depth. The time-frequency energy distribution map is constructed to intuitively display the energy distribution characteristics of the vibration signal in the time-frequency domain.

[0043] Correspondingly, the three-dimensional mathematical model established in step S1 is used to identify the fundamental frequency vibration component of the vibration signal in combination with the time-frequency energy spectrum. A ±3 times rotational frequency analysis bandwidth centered on the fundamental frequency is defined, and an improved Teager energy operator is used to calculate the instantaneous energy of each frequency component in the range. Ψ[x(n)]=x 2 (n)-x(n-1)x(n+1) By calculating the instantaneous energy, a 128-dimensional vibration state vector containing time domain (such as peak value, effective value), frequency domain (such as fundamental frequency amplitude, sideband energy), and time-frequency domain (such as energy center frequency) characteristics is extracted, fully representing the vibration characteristics of the water pump and providing rich feature information for fault diagnosis.

[0044] S4, multi-physical quantity fusion calculation: refer to Figure 3 , the time domain vibration signal is converted according to the accelerometer sensitivity matrix, the vibration acceleration effective value, the vibration speed effective value and the vibration displacement effective value are calculated synchronously by using the frequency domain integral algorithm, the electromagnetic vibration compensation model is constructed by establishing the cross-correlation function between the driving current spectrum and the vibration signal, the compensation coefficient matrix is generated by least square fitting for the current harmonic components with a correlation coefficient greater than 0.3, and the calculation results are corrected.

[0045] Based on the accelerometer sensitivity matrix, the voltage signal output by the sensor is converted into the actual physical quantity (acceleration), and the specific steps are as follows: S401. Sensitivity matrix calibration The sensitivity matrix of each three-axis MEMS accelerometer is: Wherein, s x , s y , s z are the sensitivities of the three axes (unit: m·s -2 / v), which are calibrated by a national standard vibration calibration device (such as the vibration calibration system of China Metrology Science and Technology Institute) to ensure that the sensitivity error is controlled within ±1.5% and the frequency response range covers 0.1Hz-10kHz.

[0046] Further, the coordinate conversion formula is that the original voltage signal collected by the sensor is V(t)=[V x (t),V y (t),V z (t)] T , and the actual acceleration signal is a(t)=S·V(t).

[0047] Wherein, a(t)=[a x (t),a y (t),a z (t)]T is a triaxial acceleration time-domain signal.

[0048] S402. For the frequency-domain integration factor, where, vibration velocity frequency-domain signal: (first integration, j is the imaginary unit); vibration displacement frequency-domain signal: (second integration).

[0049] Subsequently, through inverse Fourier transform and time-domain effective value calculation, inverse Fourier transform of V(f) and D(f) is carried out to obtain time-domain signals v(t) and d(t), and effective values are calculated: acceleration effective value: velocity effective value: displacement effective value: where T is the signal collection time.

[0050] S403. Further, for the electromagnetic vibration compensation model construction, the following steps are required: S4031. Current signal spectrum analysis: FFT is performed on the driving current signal i(t) to obtain the current frequency I(f), and each harmonic component I k (f) = I(kf0), where f0 is the fundamental frequency (the electrical frequency corresponding to the water pump speed).

[0051] S4032. Cross-correlation coefficient calculation: the cross-correlation coefficient of the current harmonic component and the vibration signal (taking acceleration as an example) is calculated: where i k (t) is the time-domain signal of the kth harmonic, Cov is the covariance, and σ a , σ ik are standard deviations, and harmonic components with |p k |>0.3 are selected as the main interference sources.

[0052] S4033. Least squares fitting compensation parameters Let the compensation coefficient matrix be: K = [k 1, k 2, …, k m ], m is the number of effective interference harmonics, and the linear compensation model is established: a 补偿 (t) = a(t) - K·i 干扰 (t) By minimizing the kurtosis value (or mean square error) of the compensated signal, the optimal K is solved: S4034. For the frequency domain integral filter setting, a 50-stage Butterworth low-pass filter is adopted, with a cutoff frequency of (f s = 24 kHz as the sampling rate), to suppress the high-frequency noise introduced in the integral process.

[0053] S4035. Compensation effect verification: Take a certain type of water pump as an example: the effective value of the vibration acceleration without compensation is 1.2 m·s -2 , the cross-correlation coefficient of the 5th harmonic (f = 250 Hz) is 0.42, and after compensation, the effective value is reduced to 0.9 m·s -2 , the kurtosis value is reduced from 3.2 to 2.1, and the surface electromagnetic interference is effectively suppressed.

[0054] Based on the above process, the vibration quantity calculation results include: Time domain characteristic quantity: peak value, effective value, kurtosis value, margin factor; Frequency domain characteristic quantity: fundamental frequency amplitude ratio, sideband energy ratio, octave energy distribution; Time-frequency domain characteristic quantity: energy barycenter frequency, root mean square frequency, frequency variance.

[0055] Among them, in the data processing stage, set the vibration quantity threshold self-learning module, through the historical test data training support vector machine classifier, establish the water pump vibration health degree evaluation model containing 18 characteristic parameters, the characteristic parameters include time domain characteristic quantity, frequency domain characteristic quantity and time-frequency domain characteristic quantity, the vibration health degree evaluation model training module supports online incremental learning, can dynamically update the compensation coefficient matrix and SVM classifier parameters according to real-time test data, realize the self-calibration and freedom of test system.

[0056] Further, referring to Figure 4 A test system for the above-mentioned AI server cooling water pump vibration quantity test method, comprising the following modules: a. Intelligent test platform unit: integrating six-dimensional force sensor array, temperature compensation exciter and laser calibration module, can build a dynamic coordinate system for water pump vibration test, automatically complete platform stiffness calibration and transfer function matrix measurement; b. Multi-source sensing unit: containing 12 channels of three-axis MEMS accelerometer, high-frequency current transformer and pressure transmitter, achieving microsecond-level time alignment through synchronous clock module, with a sampling rate not less than 20 kHz, meeting the ISO 13373 vibration signal acquisition standard; c.Edge computing unit: configure FPGA real-time signal processing module, built-in VMD noise reduction algorithm module, adaptive S transform engine and multi-physical quantity fusion calculation unit, according to the sensor layout matrix and sensitivity matrix, convert each measuring point vibration signal into acceleration, speed and displacement signal in global coordinate system, use frequency domain integration method (accuracy ≤0.5%) to calculate effective value, and correct current harmonic interference through electromagnetic vibration compensation model (response time ≤200 microseconds), support real-time time-frequency domain analysis and characteristic parameter calculation of vibration signal; Time-sensitive network communication is used between edge computing unit and data management unit, signal transmission delay is controlled within 50 microseconds, data processing throughput is not less than 1GB / s, ensuring the real-time and integrity of test data.

[0057] d. Data management unit: build vibration feature database based on cloud computing platform, integrate health degree evaluation model training module and test report automatic generation system, which can realize storage, model training and visual report output of test data, report format supports PDF / A-3 (long-term archiving) and HTML5 (Web interactive), including electronic signature and timestamp (complying with ISO17025 standard). The implementation principle of an embodiment of the AI server cooling water pump vibration quantity adaptive calibration test method and system is as follows: the intelligent test platform unit uses a six-dimensional force sensor, a temperature compensation type exciter and a laser calibration module to build a dynamic coordinate system and automatically calibrate the platform stiffness, obtain the water pump-pipeline system transfer function matrix, and provide accurate mechanical boundary conditions for testing; the multi-source sensing unit reasonably arranges three-axis MEMS accelerometers and other sensors at key parts of the water pump, and realizes synchronous acquisition and time alignment of multi-modal signals by combining IEEE 1588 protocol, and comprehensively obtains vibration, current, pressure and other data; the edge computing unit is based on FPGA hardware architecture, and uses VMD, adaptive S transform and other algorithms to pre-process, extract features and perform multi-physical quantity fusion calculation on the collected signals, and quickly and accurately generates 128-dimensional vibration state vectors and other results; the data management unit uses InfluxDB and other storage technologies to build a vibration feature database, uses TensorFlow to train an SVM classifier to realize offline training and online incremental learning of the health degree evaluation model, and automatically generates test reports that comply with standards. Each unit cooperates through communication technologies such as time-sensitive network, data is sequentially transmitted and feedback optimized between units, and finally realizes high-precision and intelligent testing and health evaluation of the vibration quantity of the AI server cooling water pump.

[0058] The above are preferred embodiments of the present application, and do not limit the protection scope of the present application, therefore: any equivalent changes made according to the structure, shape, principle of the present application shall be covered within the protection scope of the present application.

Claims

1. A method for testing the vibration of a cooling water pump for an AI server, characterized in that: The following steps are involved: S1. Adaptive calibration start: The water pump to be tested is installed on an intelligent test platform with an integrated six-dimensional force sensor. The platform stiffness matrix is ​​automatically calibrated through a closed-loop feedback control system. The initial modal parameters of the water pump in the no-load state are simultaneously collected, and a three-dimensional mathematical model containing the speed-head-modal frequency mapping relationship is established. S2. Synchronous multimodal signal acquisition: A triaxial MEMS accelerometer array is deployed at radial and axial measurement points on the water pump bearing seat, motor rotor housing, and impeller volute in accordance with ISO 1940 balance grade requirements. A synchronously triggered data acquisition card is used to acquire vibration time-domain signals at a sampling rate of no less than 20kHz. Simultaneously, the water pump drive current waveform and real-time inlet and outlet pressure data are collected. S3. Joint signal processing in the time-frequency domain: performing variational modal decomposition on the original vibration signal, optimizing the decomposition parameters within the mode range of 2-8 using an adaptive particle swarm algorithm, screening intrinsic mode functions with a kurtosis value greater than 2.5 using the kurtosis criterion, constructing a time-frequency energy distribution map based on the adaptive S-transform, and using the three-dimensional mathematical model established in step S1 to identify the fundamental frequency vibration component and extract the full-frequency domain eigenvector including the modulation sidebands; S4. Multi-physical quantity fusion calculation: The time domain vibration signal is transformed into coordinates according to the accelerometer sensitivity matrix, and the frequency domain integration algorithm is used to synchronously calculate the effective value of vibration acceleration, vibration velocity and vibration displacement. The electromagnetic vibration compensation model is constructed by establishing the cross-correlation function between the driving current spectrum and the vibration signal. The current harmonic components with correlation coefficients greater than 0.3 are fitted with the least squares method to generate a compensation coefficient matrix, and the calculation results are corrected.

2. The testing method according to claim 1, wherein: The intelligent test platform is equipped with a temperature-compensated vibration exciter, and obtains the transfer function matrix of the water pump-pipeline system through 10-1000 Hz sweep frequency excitation during the startup phase, which is used to correct the boundary conditions of the vibration signals collected subsequently.

3. The testing method according to claim 1, wherein: The installation position of the accelerometer array is precisely calibrated in three dimensions using a laser rangefinder to form a sensor layout matrix containing position information. The installation angle error of each sensor is controlled within ±0.5°, meeting the ISO 20816 vibration measurement position accuracy requirements.

4. The testing method according to claim 1, wherein: When extracting the full-frequency domain feature vector, a ±3-fold frequency analysis bandwidth centered on the fundamental frequency is defined, and an improved Teager energy operator is used to calculate the instantaneous energy of each frequency component to construct a 128-dimensional vibration state vector containing time domain, frequency domain and time-frequency domain features.

5. The testing method according to claim 1, wherein: The vibration amount calculation results include: Time domain characteristics: peak value, effective value, kurtosis value, margin factor; Frequency domain characteristics: fundamental frequency amplitude ratio, sideband energy ratio, octave energy distribution; Time-frequency domain feature quantities: energy center frequency, root mean square frequency, and frequency variance.

6. The testing method according to any one of claims 1 to 5, characterized in that: In the data processing stage, a vibration threshold self-learning module is set up, and a support vector machine classifier is trained through historical test data to establish a water pump vibration health assessment model containing 18 characteristic parameters. The characteristic parameters include all the characteristic quantities and water pump operating condition parameters described in claim 5.

7. The AI ​​server cooling water pump vibration measurement test system according to any one of claims 1 to 6, characterized in that: include: a. Intelligent test platform unit: integrated with six-dimensional force sensor array, temperature compensation exciter and excitation The cursor calibration module can build a dynamic coordinate system for water pump vibration testing and automatically complete platform stiffness calibration and transfer function matrix measurement; b. Multi-source sensing unit: This unit includes a 12-channel triaxial MEMS accelerometer, a high-frequency current transformer, and a pressure transmitter. It uses a synchronous clock module to achieve microsecond-level time alignment, with a sampling rate of no less than 20kHz, meeting the ISO 13373 vibration signal acquisition standard. c. Edge computing unit: Equipped with an FPGA real-time signal processing module, it features a built-in VMD noise reduction algorithm module, an adaptive S-transform engine, and a multi-physics fusion computing unit, supporting real-time time-frequency domain analysis and characteristic parameter calculation of vibration signals. d. Data management unit: Builds a vibration feature database based on the cloud computing platform, integrates a health assessment model training module and an automatic test report generation system, and can realize test data storage, model training and visual report output.

8. The test system according to claim 7, wherein: Time-sensitive network communication is adopted between the edge computing unit and the data management unit, the signal transmission delay is controlled within 50 microseconds, and the data processing throughput is not less than 1GB / s, ensuring the real-time and integrity of the test data.

9. The test system according to claim 7, wherein: The sensitivity matrix of the three-axis MEMS accelerometer array is pre-calibrated using a national standard vibration calibration device, with the sensitivity error controlled within ±1.5% and the frequency response range covering 0.1Hz-10kHz.

10. The test system according to claim 7, wherein: The vibration health assessment model training module supports online incremental learning and can dynamically update the compensation coefficient matrix and SVM classifier parameters according to real-time test data, thereby achieving self-calibration and liberalization of the test system.

Images