Server power aging test method and system

Abstract

This application discloses a server power supply aging test method and system, which relates to the field of power supply testing technology. The disclosed server power supply aging test method and system solves the problems of existing technologies such as inability to reproduce real-time dynamic load changes, insufficient data acquisition accuracy, and difficulty in fault location by acquiring basic electrical parameters, constructing personalized dynamic response models, injecting dynamic load disturbances, synchronously acquiring real-time electrical waveforms and performing comparative analysis. It can actively apply targeted dynamic load disturbances, support high-precision synchronous data acquisition, and realize accurate fault identification and propagation path analysis.

Description

Technical Field

[0001] This application relates to the field of power supply testing technology, and in particular to server power supply aging test methods and systems. Background Technology

[0002] As a core power supply component in data centers and high-performance computing facilities, the long-term reliability of server power supplies is crucial for ensuring system stability. During the power module manufacturing process, rigorous aging tests must be implemented to screen and eliminate products that fail early. Currently, the industry commonly uses a batch testing approach that places multiple power supplies under test in a high-temperature aging chamber, applying constant loads or low-frequency periodically changing loads. However, this method has significant technical limitations: the stress applied during testing is mainly limited to the steady-state or low-frequency dynamic range, failing to accurately reproduce the microsecond-level rapid load step changes caused by computing units such as the CPU and GPU during actual server operation. This results in a severely inadequate assessment of the power supply's dynamic response capabilities. More critically, transient and intermittent faults that may occur during testing, such as millisecond-level output voltage drops or false triggering of protection circuits under critical conditions, are difficult to capture and record effectively due to the low data acquisition rate of existing testing systems and the lack of precise time synchronization mechanisms between channels. Especially in cluster environments where multiple power supplies are tested in parallel, it is impossible to accurately locate the fault source and accurately analyze the fault propagation path. This not only causes some power supplies with latent defects to enter the market through testing, posing safety hazards for subsequent field operations, but also reduces the efficiency and accuracy of fault diagnosis.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide a server power supply aging test method and system, which aims to achieve accurate fault identification and propagation path analysis.

[0005] To achieve the above objectives, this application proposes a server power supply aging test method, the method comprising: Obtain the basic electrical parameters of the power supply of each server under test under steady-state aging conditions; Based on the aforementioned basic electrical parameters, a personalized dynamic response model for the power supply of each server under test is constructed. Based on the personalized dynamic response model, the dynamic load disturbance mode to be injected is determined to obtain the disturbance mode data to be injected, and the dynamic load disturbance mode is injected into the power supply of at least one selected server under test. At the synchronization moment of the dynamic load disturbance mode, the data acquisition units of all the power supplies of the servers under test are triggered to perform measurements and acquire the real-time electrical waveforms of all the power supplies of the servers under test during the disturbance. The real-time electrical waveform is compared and analyzed with the corresponding personalized dynamic response model to obtain the model deviation characteristics of each server power supply under test. Anomaly detection is performed based on the model deviation characteristics to identify abnormal power supplies and abnormal modes of the server under test.

[0006] In one embodiment, the step of constructing a personalized dynamic response model for each server under test power supply based on the basic electrical parameters includes: Under standard test conditions, a set of predefined dynamic characteristic scanning signals are applied to the power supply of each server under test, and scan response data is obtained. Extract a set of key parameters characterizing the dynamic performance of the power supply from the scan response data; The set of key parameters is associated with the model and specifications of the power supply of the server under test to construct the personalized dynamic response model.

[0007] In one embodiment, under standard test conditions, the step of applying a set of predefined dynamic characteristic scanning signals to each server under test power supply and acquiring scan response data includes: Apply microsecond-level load current step signals of different amplitudes and record the output voltage recovery waveform of the power supply of the server under test. Apply a small signal disturbance covering a preset frequency range and measure the frequency response of the control loop of the power supply of the server under test. When the server under test is powered on and switches between no-load and power-on states, the voltage or current waveforms of its internal preset monitoring points are captured; wherein, the output voltage recovery waveform, the control loop frequency response, and the voltage or current waveform constitute the scan response data.

[0008] In one embodiment, determining the dynamic load disturbance pattern to be injected based on the personalized dynamic response model to obtain the disturbance pattern data to be injected includes: Analyze the personalized dynamic response model to identify the key dynamic performance parameters of the power supply of at least one selected server under test; Based on the type of the key dynamic performance parameters, a load change pattern is generated to verify the dynamic response of the key dynamic performance parameters, which serves as the data for the disturbance pattern to be injected.

[0009] In one embodiment, the step of comparing and analyzing the real-time electrical waveform with the corresponding personalized dynamic response model to obtain the model deviation characteristics of each server power supply under test includes: From the real-time electrical waveform, extract the actual response characteristics of the first server under test power supply under the dynamic load disturbance mode, wherein the first server under test power supply is the power supply injected into the dynamic load disturbance mode. Based on the personalized dynamic response model, the expected response characteristics of the power supply of the first server under test are obtained under the dynamic load disturbance mode. The actual response characteristics are compared with the expected response characteristics to calculate the response deviation data of the power supply of the first server under test. The model deviation features are generated based on whether the response deviation data exceeds a preset deviation threshold.

[0010] In one embodiment, the data acquisition units of all the power supplies of the server under test are synchronized in time; the real-time electrical waveform is a waveform with a timestamp, and the deviation between the timestamps of the waveforms acquired by different acquisition units is less than a preset synchronization accuracy threshold. The anomaly detection based on the model deviation characteristics also includes: When at least two abnormal server power supplies are identified based on the model deviation characteristics, the start timestamps of the abnormal events in the real-time electrical waveforms of these abnormal server power supplies are extracted. Compare the order of the start timestamps of the abnormal events; The power supply of the server under test that first exhibits an abnormal event is identified as the primary fault source, and the fault propagation path is inferred based on the electrical connection relationship between the primary fault source and other abnormal power supplies; wherein, the abnormal mode includes the primary fault source and the fault propagation path.

[0011] In one embodiment, the method further includes: Based on the aforementioned anomaly pattern, a targeted enhanced perturbation test sequence is generated; A second round of dynamic load disturbance is applied to the power supply of the abnormal server under test according to the enhanced disturbance test sequence, and the second round of real-time electrical waveforms are acquired simultaneously. Analyze the second round of real-time electrical waveforms to determine the defect type corresponding to the abnormal mode and generate a defect verification conclusion.

[0012] In one embodiment, the step of generating a targeted reinforcement perturbation test sequence based on the anomalous pattern includes: Analyze the abnormal mode to determine the sensitive direction of the abnormal server power supply to the preset disturbance parameters in the dynamic load disturbance mode; Based on the sensitive direction and the preset perturbation parameters, at least one new perturbation mode is generated, wherein the new perturbation mode is obtained by adjusting at least one of the intensity, rate of change or frequency of the preset perturbation parameters; The at least one new perturbation mode is combined according to a preset logic to generate the enhanced perturbation test sequence.

[0013] In one embodiment, the method further includes: After completing the aging test of the preset period, based on the accumulated basic electrical parameters, the real-time electrical waveforms and the model deviation characteristics, the change trend data of the key parameters of the power supply of the same server under test are statistically analyzed. Compare the trend data with a preset allowable aging drift threshold; Based on the comparison results, the expected values ​​or allowable deviation ranges of the corresponding parameters in the personalized dynamic response model are corrected to obtain an updated personalized dynamic response model and drift report.

[0014] In addition, to achieve the above objectives, this application also proposes a server power supply aging test system, which includes: a memory, a processor, and a server power supply aging test program stored in the memory and executable on the processor, wherein the server power supply aging test program is configured to implement the steps of the server power supply aging test method.

[0015] The server power supply aging test method and system proposed in this application solves the problems of existing technologies, such as the inability to reproduce real-time dynamic load changes, insufficient data acquisition accuracy, and difficulty in fault location, by acquiring basic electrical parameters, constructing personalized dynamic response models, injecting dynamic load disturbances, synchronously acquiring real-time electrical waveforms and performing comparative analysis. It can actively apply targeted dynamic load disturbances, support high-precision synchronous data acquisition, and achieve accurate fault identification and propagation path analysis. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an embodiment of the server power supply aging test method of this application. Figure 2 This is a schematic diagram of a structure provided for an embodiment of the server power supply aging test system of this application.

[0019] Explanation of icon numbers: 10. Memory; 20. Processor.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only 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 this application, but merely represents 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.

[0022] It should be understood that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] In existing technologies, server power supply aging tests primarily employ steady-state or low-frequency stress variations, which are insufficient to simulate the microsecond-level dynamic load impacts from CPU, GPU, and other components during actual server operation. Furthermore, existing testing systems suffer from low sampling rates and asynchronous data across channels, making it impossible to capture transient or intermittent faults. Moreover, they cannot pinpoint fault sources or determine fault propagation relationships within parallel testing clusters. This allows some power supplies with latent defects to pass the test, creating potential hazards for subsequent field operations and resulting in inefficient fault analysis.

[0024] Based on this, this application provides a server power supply aging test method, referring to... Figure 1 The server power supply aging test method includes steps S100 to S600, wherein: Step S100: Obtain the basic electrical parameters of each server power supply under steady-state aging conditions; Step S200: Construct a personalized dynamic response model for the power supply of each server under test based on the basic electrical parameters; Step S300: Determine the dynamic load disturbance mode to be injected based on the personalized dynamic response model to obtain the disturbance mode data to be injected, and inject the dynamic load disturbance mode into at least one selected server power supply under test. Step S400: At the synchronization moment of injecting the dynamic load disturbance mode, trigger the data acquisition units of all the power supplies of the servers under test to perform measurements and acquire the real-time electrical waveforms of all the power supplies of the servers under test during the disturbance. Step S500: Compare and analyze the real-time electrical waveform with the corresponding personalized dynamic response model to obtain the model deviation characteristics of each server power supply under test. Step S600: Based on the model deviation characteristics, perform anomaly judgment to identify abnormal power supplies and abnormal modes of the server under test.

[0025] In this embodiment, basic electrical parameters refer to fundamental electrical quantities such as voltage, current, power, efficiency, and ripple obtained through measurement under steady-state operation or specific static conditions of the server power supply. These parameters reflect the basic performance of the power supply under stable conditions. The personalized dynamic response model refers to a mathematical or simulation model established for each server power supply under test, based on testing and modeling its dynamic characteristics, capable of predicting its output response behavior under different dynamic load disturbances. This model exhibits individual differences and reflects the unique dynamic performance of each power supply. The dynamic load disturbance mode simulates rapid, transient load changes generated by components such as the CPU and GPU during actual server operation, such as current steps, pulses, and high-frequency oscillations. This mode is used to actively stimulate the dynamic response of the power supply to reveal its potential dynamic performance defects.

[0026] In this embodiment, the data of the disturbance mode to be injected refers to the digital instructions or waveform data that can be executed by the test equipment after the determined dynamic load disturbance mode is converted. This data includes specific parameters of the disturbance, such as amplitude, duration, rate of change, and frequency. The real-time electrical waveform refers to the instantaneous change curve of electrical quantities such as output voltage and current of the server power supply, which is synchronously measured and recorded by the high-speed data acquisition unit during the dynamic load disturbance injection. This waveform reflects the actual response process of the power supply under dynamic stress. The model deviation feature refers to the quantitative index of difference obtained by comparing the actual response (real-time electrical waveform) of the server power supply under dynamic load disturbance with the expected response predicted by the personalized dynamic response model. This feature can characterize the degree of deviation between the actual performance of the power supply and the ideal model. The abnormal pattern refers to the comprehensive description of the abnormal phenomenon formed by further analyzing its fault performance, fault propagation path, and possible causes after identifying the abnormal server power supply under test. This pattern helps to deeply understand the nature and root cause of the fault.

[0027] In this embodiment, the server power supply aging test method first acquires the basic electrical parameters of each server power supply under steady-state aging conditions. For example, this can be achieved by running the power supply under constant load for an extended period and periodically measuring parameters such as output voltage, output current, ripple voltage, and conversion efficiency. These parameters can be recorded using traditional multimeters, power meters, or low-speed data acquisition systems. As an optional implementation, these basic parameters can be recorded manually or through automated testing programs under typical steady-state conditions such as power-on, no-load, and full-load. These basic parameters provide benchmark data for subsequent dynamic performance analysis.

[0028] In this embodiment, a personalized dynamic response model for each server power supply under test is constructed based on the acquired basic electrical parameters. For example, a general power supply dynamic model template can be selected based on the power supply's rated power, output voltage range, and other basic parameters. The parameters in the model can then be manually input or adjusted to initially match the power supply's specifications. Alternatively, a simplified simulation model can be built using circuit simulation software based on the power supply's topology, and the component values ​​in the model can be initially set according to the basic electrical parameters. This model aims to initially characterize the dynamic behavior of the power supply.

[0029] In this embodiment, based on the constructed personalized dynamic response model, the dynamic load disturbance mode to be injected is determined to obtain the disturbance mode data, and this dynamic load disturbance mode is injected into at least one selected power supply of the server under test. For example, a current step signal with a fixed amplitude can be set according to the rated output current of the power supply, such as a jump from 50% load to 100% load, and this can be used as the disturbance mode. This disturbance mode can be converted into digital instructions and injected into the selected power supply under test through a programmable electronic load device. Alternatively, a sinusoidal current disturbance with a fixed frequency and amplitude can be set and injected into the power supply.

[0030] In this embodiment, at the synchronization moment of injecting the dynamic load disturbance mode, the data acquisition units of all power supplies under test are triggered to perform measurements, acquiring the real-time electrical waveforms of all power supplies during the disturbance. For example, the recording function of an oscilloscope or data acquisition card can be activated manually at the same time as the electronic load begins to inject the disturbance to capture the output voltage and current waveforms of the power supply. Alternatively, a simple trigger signal can be used to simultaneously activate all acquisition devices when the disturbance begins. This approach aims to record the instantaneous response of the power supply under dynamic stress.

[0031] In this embodiment, the acquired real-time electrical waveforms are compared and analyzed with the corresponding personalized dynamic response models to obtain the model deviation characteristics of each power supply under test. For example, key indicators such as peak overshoot, voltage drop, and recovery time in the real-time electrical waveforms can be compared with the predicted values ​​of the personalized dynamic response model under the same disturbance, and the absolute or percentage differences of these indicators can be calculated. These differences can be used as components of the model deviation characteristics. Finally, anomaly judgment is performed based on the obtained model deviation characteristics to identify abnormal power supplies under test and their abnormal patterns. For example, a fixed threshold can be set; if a certain model deviation characteristic of the power supply (such as voltage drop) exceeds the threshold, the power supply is determined to be abnormal. The abnormal pattern can be simply described as "excessive voltage drop" or "excessive recovery time." This judgment method aims to initially screen out power supplies whose performance deviates from expectations.

[0032] In this embodiment, by constructing a personalized dynamic response model for each server power supply under test and injecting targeted dynamic load disturbances based on this model, it is possible to proactively stimulate and capture transient and intermittent faults that the server power supply may encounter under microsecond-level dynamic load impacts during actual operation. By synchronously acquiring real-time electrical waveforms and comparing and analyzing them with the personalized model, hidden defects that are difficult to detect using traditional testing methods can be identified. This improves the fault detection accuracy of aging tests, reduces the probability of potentially risky power supplies entering the actual operating environment, and provides a data foundation for subsequent fault analysis.

[0033] In one feasible implementation, the step of constructing a personalized dynamic response model for each server power supply under test based on the basic electrical parameters includes: applying a set of predefined dynamic characteristic scanning signals to each server power supply under test under standard test conditions to obtain scanning response data; extracting a set of key parameters characterizing the dynamic performance of the power supply from the scanning response data; and associating the set of key parameters with the model and specification parameters of the server power supply under test to construct the personalized dynamic response model.

[0034] In this embodiment, under standard test conditions, a set of predefined dynamic characteristic scanning signals are applied to each power supply under test (PST) to acquire scan response data. The aim is to actively stimulate the PST to obtain behavioral data under different dynamic conditions. Standard test conditions typically refer to testing in a controllable and stable environment with rated input voltage, rated output voltage, and specific ambient temperature to ensure the repeatability and comparability of test results. The predefined dynamic characteristic scanning signals can cover various types. For example, sinusoidal perturbation signals of different frequencies and amplitudes can be applied to probe the power supply's frequency response characteristics; transient signals such as step load changes, ramp load changes, or pulse load changes can also be applied to evaluate the power supply's transient response speed, overshoot / undershoot amplitude, and settling time. Through these scanning signals, the dynamic response characteristics of the power supply can be comprehensively probed. The acquired scan response data typically includes waveform data of the power supply's output voltage, output current, and voltage or current changes over time at key internal points.

[0035] In this embodiment, a set of key parameters characterizing the dynamic performance of the power supply is extracted from the scan response data. The purpose is to transform the raw, large amount of scan response data into refined, quantifiable dynamic performance indicators. These key parameters may include, but are not limited to: voltage drop, voltage overshoot, recovery time, output impedance, control loop bandwidth, phase margin, gain margin, output ripple noise characteristics, startup time, and shutdown time during the load transient response. The extraction method can employ various signal processing techniques, such as using Fourier transform to perform frequency analysis on the waveform data to obtain frequency response characteristics; using peak detection algorithms to identify overshoot / undershoot in the transient response; and calculating the recovery time using threshold crossover time. These parameters can intuitively and quantitatively reflect the performance of the power supply under dynamic operating conditions.

[0036] In this embodiment, the set of key parameters is associated with the model and specification parameters of the power supply of the server under test to construct the personalized dynamic response model. The aim is to combine the extracted dynamic performance parameters with the inherent properties of the power supply to form a mathematical or simulation model capable of predicting its dynamic behavior. The model and specification parameters may include the power supply's rated power, input voltage range, output voltage, output current, internal topology, and main component types. The association can be data-driven, for example, using machine learning algorithms (such as regression analysis or neural networks) to establish a mapping relationship between key parameters and model coefficients; or using a physical model-based approach, mapping key parameters to the values ​​of model components through a parameterized circuit model (such as a SPICE model). The constructed personalized dynamic response model can predict the power supply's output response based on given input conditions (such as load changes), and due to its personalized characteristics, it can reflect the subtle dynamic behavior differences of a single power supply under the influence of factors such as manufacturing tolerances and component variations, rather than simply the average performance of a general-purpose model.

[0037] In this embodiment, by employing the aforementioned technical solution, the construction of a personalized dynamic response model for a server power supply no longer relies solely on basic electrical parameters. Instead, it actively acquires detailed response data of the power supply under dynamic operating conditions by applying predefined dynamic characteristic scanning signals under standard test conditions. From this scanning response data, a set of key parameters characterizing the power supply's dynamic performance is extracted and correlated with the power supply's model and specifications, thereby constructing a more accurate and comprehensive personalized dynamic response model. This method can deeply uncover the unique dynamic behavior characteristics of each server power supply under test, such as its transient response speed and stability margin, making the subsequent dynamic load disturbance patterns determined based on this model more targeted. Furthermore, when comparing real-time electrical waveforms, it can more accurately identify subtle performance drifts or potential defects in the power supply, improving the accuracy and effectiveness of aging tests.

[0038] In one feasible implementation, under standard test conditions, the step of applying a set of predefined dynamic characteristic scanning signals to each power supply under test (PST) to obtain scanning response data includes: applying microsecond-level load current step signals of different amplitudes and recording the output voltage recovery waveform of the PST; applying small signal disturbances covering a preset frequency range and measuring the control loop frequency response of the PST; and capturing the voltage or current waveforms of preset monitoring points within the PST during startup and no-load state switching. The output voltage recovery waveform, the control loop frequency response, and the voltage or current waveform constitute the scanning response data.

[0039] In this embodiment, microsecond-level load current step signals of different amplitudes are applied, and the output voltage recovery waveform of the power supply under test (PST) is recorded. This technique aims to evaluate the dynamic response capability of the PST in the face of sudden large load changes. By using a programmable electronic load, load current step signals with microsecond-level rise or fall times are applied to the output of the PST. The amplitude of these step signals can be adjusted according to test requirements, such as transitions from light to heavy loads, or switching between different load points, to simulate various sudden load changes that a server may encounter in actual operation. Simultaneously with the application of the step signals, a high-bandwidth oscilloscope or high-speed data acquisition system is used to accurately record the voltage change waveform at the output of the PST, i.e., the output voltage recovery waveform. This waveform can intuitively reflect the voltage drop, overshoot, oscillation, and the time required for recovery to a steady state. These parameters are key indicators for measuring the transient response performance of the power supply.

[0040] In this embodiment, a small signal disturbance covering a preset frequency range is applied, and the frequency response of the control loop of the power supply under test (PST) is measured. This technique is used to analyze the stability and dynamic characteristics of the internal control loop of the PST in depth. Specifically, a frequency response analyzer or network analyzer is typically used to inject a small AC disturbance signal covering a preset frequency range (e.g., from tens of hertz to hundreds of kilohertz) into the power supply's control loop, for example, through an injection point in the feedback loop. Simultaneously, the response signal at a specific point in the control loop is measured, and the gain and phase difference between the disturbance signal and the response signal are calculated. By scanning different frequencies, a Bode plot of the power supply control loop can be plotted, from which key parameters such as phase margin, gain margin, and crossover frequency can be extracted. These parameters are important bases for evaluating the stability and bandwidth of the power supply control loop and are crucial for predicting the power supply's behavior under complex load conditions.

[0041] In this embodiment, during the startup and no-load state switching of the power supply under test (UTP), the voltage or current waveforms of its internal preset monitoring points are captured. This technical feature aims to capture the transient behavior of the UTP during critical operating mode transitions, which often reveal potential design flaws or signs of aging in the power supply. Specifically, when the power supply starts from a completely power-off state to stable output, and when switching between no-load and load states, a multi-channel high-speed data acquisition device is used to synchronously capture the voltage or current waveforms of preset monitoring points (e.g., the drive voltage of the primary-side switching transistor, the secondary-side rectified output voltage, the current of the PFC (Power Factor Correction) circuit, etc.) within the power supply. These monitoring points are typically nodes of critical functional modules within the power supply, and their waveforms can reflect the internal operating state of the power supply under extreme conditions such as startup shocks and sudden load changes, such as the presence of overshoot, undershoot, oscillation, or abnormal current spikes, thereby providing important data for evaluating the reliability and stability of the power supply.

[0042] In this embodiment, the output voltage recovery waveform, control loop frequency response, and voltage or current waveforms collectively constitute the scan response data. By integrating the data obtained from the three different testing methods—the output voltage recovery waveform reflecting transient load response, the frequency response data characterizing control loop stability, and the voltage or current waveforms at internal monitoring points revealing key mode transition behaviors—a comprehensive, multi-dimensional scan response dataset is formed. This integrated data acquisition method ensures a comprehensive characterization of the server power supply's dynamic performance, laying a solid foundation for subsequently building a high-precision, highly robust, and personalized dynamic response model.

[0043] In this embodiment, by applying microsecond-level load current step signals of different amplitudes and recording the output voltage recovery waveform through the above-described technical solution, the transient response characteristics of the power supply under test (PST) when facing sudden large load changes can be accurately captured, revealing key dynamic indicators such as voltage drop, overshoot, and recovery time. Simultaneously, by applying small-signal disturbances covering a preset frequency range and measuring the control loop frequency response, the stability margin, bandwidth, and other performance parameters of the power supply's internal control loop can be analyzed in depth, thereby evaluating its ability to suppress high-frequency disturbances and its overall stability. Furthermore, capturing the voltage or current waveforms at preset monitoring points during power supply startup and no-load state switching can effectively reveal potential abnormal behaviors or design defects during critical operating mode transitions. These multi-dimensional, high-precision scan response data collectively form the basis for comprehensively characterizing the power supply's dynamic performance, enabling the constructed personalized dynamic response model to more accurately reflect the power supply's true dynamic characteristics. This provides solid data support for subsequent aging tests and anomaly detection, improving the accuracy and reliability of aging tests.

[0044] In one feasible implementation, determining the dynamic load disturbance pattern to be injected based on the personalized dynamic response model to obtain the disturbance pattern data to be injected includes: analyzing the personalized dynamic response model to identify key dynamic performance parameters of the at least one selected power supply of the server under test; and generating a load change pattern for verifying the dynamic response of the key dynamic performance parameters, based on the type of the key dynamic performance parameters, as the disturbance pattern data to be injected.

[0045] In this embodiment, analyzing the personalized dynamic response model to identify the key dynamic performance parameters of at least one selected power supply under test (PST) refers to an in-depth analysis of the personalized dynamic response model of each pre-built PST. This model encompasses the dynamic behavior characteristics of the power supply under different operating conditions. Key dynamic performance parameters are those performance indicators that are crucial to the stability and reliability of the power supply and may experience significant drift or degradation during aging. These parameters may include, but are not limited to, transient response time, voltage overshoot / undershoot amplitude, load regulation, ripple noise, control loop bandwidth, phase margin, and gain margin. The identification process can be based on the power supply's design specifications, historical fault data, industry standards, or through sensitivity analysis of the model. For example, if the model shows a large voltage undershoot under a specific load step, the voltage undershoot amplitude is identified as a key dynamic performance parameter.

[0046] In this embodiment, load variation patterns are generated based on the type of the key dynamic performance parameters to verify the dynamic response of the key dynamic performance parameters. These load variation patterns, used as the data to be injected perturbation patterns, refer to the targeted design and generation of one or more load variation patterns after identifying the key dynamic performance parameters. These load variation patterns are not generic load steps, but are customized according to the characteristics of the identified key parameters. For example, if the key parameter is transient response time, the generated load variation pattern might be a current step signal with a specific rise / fall rate and amplitude; if the key parameter is the stability of the control loop (such as phase margin), the load variation pattern might be a series of AC current perturbations with specific frequencies and amplitudes, or a complex waveform capable of exciting potential oscillation modes. In this way, it is ensured that the injected dynamic load perturbation can maximize the excitation and verification of the power supply's dynamic response on specific key performance parameters, thereby more effectively revealing aging effects.

[0047] In this embodiment, through the above technical solution, this application can accurately identify the key dynamic performance parameters that may degrade during the aging process for each server power supply under test, based on its personalized dynamic response model. Based on the types of these key parameters, a highly targeted dynamic load disturbance pattern is systematically generated. This customized disturbance pattern can more effectively stimulate and verify the specific dynamic performance of the power supply, thereby exposing potential defects and performance degradation trends earlier and more accurately during aging tests, avoiding test blind spots or inefficiencies that may result from using general disturbance patterns. This makes the aging test process more efficient and in-depth, enabling a more comprehensive evaluation of the long-term reliability of the server power supply.

[0048] In one feasible implementation, the step of comparing and analyzing the real-time electrical waveform with the corresponding personalized dynamic response model to obtain the model deviation characteristics of each server power supply under test includes: extracting the actual response characteristics of the first server power supply under test under the dynamic load disturbance mode from the real-time electrical waveform, wherein the first server power supply under test is the power supply injected into the dynamic load disturbance mode; obtaining the expected response characteristics of the first server power supply under test under the dynamic load disturbance mode based on the personalized dynamic response model; comparing the actual response characteristics with the expected response characteristics to calculate the response deviation data of the first server power supply under test; and generating the model deviation characteristics based on whether the response deviation data exceeds a preset deviation threshold.

[0049] In this embodiment, extracting the actual response characteristics of the first server under test power supply under dynamic load disturbance mode from the real-time electrical waveform refers to identifying and quantifying the key dynamic characteristics exhibited by the actual output electrical behavior of the power supply from the original voltage and current waveforms using signal processing algorithms such as peak detection, transient response analysis, or Fourier transform under dynamic load disturbance mode. These characteristics may include, but are not limited to, the peak value of voltage drop or overshoot, recovery time, oscillation frequency, and damping ratio. The purpose is to simplify the complex real-time electrical waveform into a set of quantifiable parameters for subsequent comparison with model predictions. For example, for a load step disturbance, the transient drop amplitude, the time required to recover to the steady-state value, and the overshoot during the recovery process can be extracted from the output voltage waveform.

[0050] Based on this, obtaining the expected response characteristics of the first server under test power supply under the dynamic load disturbance mode based on a personalized dynamic response model refers to obtaining the electrical behavior characteristics that the power supply should exhibit theoretically or ideally under the same dynamic load disturbance mode by running or simulating the personalized dynamic response model of the power supply. The personalized dynamic response model is constructed based on the power supply's basic electrical parameters and dynamic characteristic scan data, and can predict the power supply's response under specific disturbances. For example, if the model is a transfer function or state-space model, the disturbance signal can be input into the model, its predicted voltage or current response waveform can be calculated and output, and then dynamic parameters of the same type as the actual response characteristics can be extracted from the predicted waveform.

[0051] In this embodiment, the actual response characteristics are compared with the expected response characteristics to calculate the response deviation data of the first server power supply under test. The response deviation data is an indicator that quantifies the difference between the actual response and the expected response. By comparing the extracted actual response characteristics with the expected response characteristics predicted by the model one by one, the numerical difference between them can be calculated. This comparison can be at the parameter level, i.e., comparing the differences or ratios of various dynamic parameters; or it can be at the waveform level, i.e., quantifying the overall deviation by calculating waveform similarity. The calculated deviation data can intuitively reflect the degree of deviation between the actual performance and the ideal performance of the power supply.

[0052] In this embodiment, the model deviation feature is generated based on whether the response deviation data exceeds a preset deviation threshold. The model deviation feature is a clearly indicative marker ultimately used for anomaly detection. After calculating the response deviation data, it needs to be compared with a preset deviation threshold. These thresholds are typically determined based on power supply design specifications, industry standards, or historical data experience, representing the acceptable normal fluctuation range of power supply performance. If one or more sets of response deviation data exceed the corresponding deviation threshold, a model deviation feature indicating the deviation can be generated, such as a Boolean flag, a deviation level, or structured data containing specific deviation information. This feature serves as a direct basis for subsequent anomaly detection, indicating that the actual behavior of the power supply has deviated from the expected range of its personalized model, potentially indicating aging or a potential fault.

[0053] In this embodiment, through the above technical solution, this application provides a structured and quantitative method for comparing real-time electrical waveforms with personalized dynamic response models. First, key actual response features are extracted from complex real-time electrical waveforms, transforming the raw waveform data into a comparable parameter set, effectively reducing the complexity of data processing. Second, corresponding expected response features are obtained based on the personalized dynamic response model, ensuring that the comparison benchmark is customized for the unique performance of each power supply, improving the accuracy of the comparison. Furthermore, by calculating the response deviation data between the actual response features and the expected response features, the degree of deviation in power supply performance can be accurately quantified. Finally, model deviation features are generated based on whether the response deviation data exceeds a preset deviation threshold, making the anomaly judgment process more objective and automated, avoiding errors that may be caused by subjective judgment. This method can clearly identify subtle performance changes of the power supply under dynamic load disturbances, providing a reliable and highly discriminative basis for subsequent anomaly judgment, thereby improving the accuracy and efficiency of server power supply aging tests.

[0054] In one feasible implementation, the data acquisition units of all power supplies under test are synchronized in time; the real-time electrical waveform is a waveform with a timestamp, and the deviation between the timestamps of the waveforms acquired by different acquisition units is less than a preset synchronization accuracy threshold; the anomaly judgment based on the model deviation characteristics further includes: when at least two abnormal power supplies under test are identified according to the model deviation characteristics, extracting the start timestamps of the abnormal events in the real-time electrical waveforms of these abnormal power supplies under test; comparing the order of the start timestamps of the abnormal events; determining the power supply under test that first exhibits an abnormal event as the primary fault source, and inferring the fault propagation path based on the electrical connection relationship between the primary fault source and other abnormal power supplies; wherein, the abnormal mode includes the primary fault source and the fault propagation path.

[0055] In this embodiment, to ensure that all data acquisition units of the server under test can accurately capture the moment of event occurrence, these data acquisition units are designed to be time-synchronized. This means that all acquisition units share a unified time reference, for example, calibrated through a high-precision clock source (such as GPS timing or an NTP server). Therefore, each acquired real-time electrical waveform is accompanied by a precise timestamp, and the deviation between the timestamps of waveforms acquired by different acquisition units is strictly controlled within a preset synchronization accuracy threshold. This threshold is typically set at the microsecond or even nanosecond level to ensure that even rapidly changing electrical events can be accurately time-aligned.

[0056] In this embodiment, when the system identifies at least two power supplies of the server under test exhibiting abnormalities based on model deviation characteristics, in order to analyze the correlation between these abnormalities in depth, it is necessary to extract the start timestamps of the abnormal events from the real-time electrical waveforms of these abnormal power supplies. The start timestamp of an abnormal event refers to the precise point in time when an electrical parameter (such as voltage, current, ripple, etc.) first deviates from its normal range or expected behavior in the real-time electrical waveform. This can be achieved through signal processing algorithms, such as threshold detection, change point detection, or pattern recognition, to accurately mark the starting point of the abnormality. After acquiring all relevant start timestamps of abnormal events, the system compares these timestamps to determine the chronological order of the abnormal events. Due to the time synchronization of the data acquisition unit and the high precision of the timestamps, this comparison can reliably reveal which power supply's abnormal event occurred earliest.

[0057] In this embodiment, the power supply of the server under test that first exhibits an anomaly is identified as the primary fault source. This is because in multi-power supply systems, the initial anomaly is often the root cause of a series of subsequent chain reactions. To further understand the fault propagation mechanism, the system infers the fault propagation path based on the known electrical connections between the primary fault source and other anomalous power supplies. This typically involves analyzing the topology of the server power supply system, such as parallel connections between power modules, shared power buses, or cascaded power supply relationships. In this way, a fault propagation chain from the primary fault source to other anomalous power supplies can be constructed. Ultimately, the anomaly pattern not only includes the anomaly information of a single power supply but also expands to a comprehensive description including the primary fault source and the fault propagation path, providing comprehensive guidance for subsequent fault diagnosis and repair.

[0058] In this embodiment, through the above technical solution, this application can accurately capture the chronological order of abnormal events occurring on different power supplies of the server under test. When multiple abnormal power supplies are identified, by comparing the start timestamps of these abnormal events, the power supply that first exhibited an abnormality can be accurately determined as the primary fault source. Based on this, combined with the internal electrical connections of the system, the propagation path of the fault can be further deduced. This method effectively solves the technical challenge of distinguishing between independent faults and abnormalities caused by fault propagation in complex multi-power supply systems, improves the accuracy and efficiency of fault diagnosis, provides key information for quickly locating the root cause of the fault and taking targeted measures, avoids misjudgment of secondary faults, and thus optimizes the aging test and fault analysis process of server power supplies.

[0059] In one feasible implementation, the method further includes: generating a targeted enhanced perturbation test sequence based on the abnormal pattern; applying a second round of dynamic load perturbation to the power supply of the abnormal server under test according to the enhanced perturbation test sequence, and simultaneously acquiring the second round of real-time electrical waveforms; analyzing the second round of real-time electrical waveforms, determining the defect type corresponding to the abnormal pattern, and generating a defect verification conclusion.

[0060] In this embodiment, the step of generating targeted enhanced disturbance test sequences based on the identified anomaly patterns aims to design more diagnostic test schemes. Anomaly patterns may indicate weaknesses in the power supply under specific dynamic conditions, such as slow output voltage recovery during load transients or instability under specific frequency disturbances. Targeted generation of enhanced disturbance test sequences means analyzing these anomaly patterns to identify key dynamic performance parameters or sensitive operating points that cause the anomalies. For example, if the anomaly pattern manifests as excessive voltage drops, the enhanced disturbance test sequence can be designed to perform tests under larger load steps, faster load change rates, or more extreme operating temperatures. This sequence is not a general-purpose test but is specifically tailored to stimulate and amplify specific anomalous behaviors in order to more clearly reveal potential defects.

[0061] In this embodiment, after generating a targeted enhanced perturbation test sequence, a second round of dynamic load perturbation is applied to the abnormal server power supply according to the enhanced perturbation test sequence, and the second round of real-time electrical waveforms is acquired simultaneously. This step practically applies the sequence to the previously identified abnormal server power supply. This is typically achieved through a high-precision, programmable electronic load capable of accurately simulating various dynamic load changes. While applying these enhanced perturbations, high-speed data acquisition equipment, such as a digital oscilloscope or a high sampling rate data acquisition card, is needed to simultaneously measure the key electrical parameters of the abnormal server power supply (such as output voltage, output current, internal control signals, etc.), and the simultaneous acquisition of the second round of real-time electrical waveforms ensures precise time alignment between the perturbation input and the power supply response.

[0062] In this embodiment, the second round of real-time electrical waveforms is analyzed to determine the defect type corresponding to the abnormal pattern and generate a defect verification conclusion. This step is the core of the diagnostic process. By conducting in-depth analysis of the second round of real-time electrical waveforms, the specific defect type can be further inferred from the surface abnormal pattern. The analysis methods may include, but are not limited to: waveform feature extraction, such as measuring overshoot, undershoot, settling time, ripple, and noise characteristics; frequency domain analysis, identifying abnormal frequency components or resonant points through Fourier transform; and detailed comparison with the reference waveform or theoretical model of a healthy power supply. For example, if the enhanced disturbance test sequence reveals a significant decrease in the control loop gain or phase margin of the power supply at a specific frequency, it may indicate aging or failure of the control loop components. Finally, based on the analysis results, a clear defect verification conclusion can be drawn, such as "output filter capacitor capacity decay," "PWM controller parameter drift," or "power MOSFET switching characteristic degradation," thereby providing a direct basis for fault location and repair.

[0063] In this embodiment, after initially identifying the abnormal patterns of the server power supply using the above technical solution, a targeted enhanced disturbance test sequence can be generated based on these abnormal patterns. This sequence is specifically designed to excite and amplify the identified abnormal behavior, thereby capturing more diagnostically valuable power response data when applying a second round of dynamic load disturbance and simultaneously acquiring a second round of real-time electrical waveforms. In-depth analysis of these second-round real-time electrical waveforms allows the testing method to move from superficial abnormal patterns to specific defect types, such as identifying whether it is due to aging of specific components, parameter drift, or other functional failures. This improves the diagnostic accuracy and efficiency of server power supply aging tests, moving from "knowing there is a problem" to "knowing exactly what the problem is," providing a clear and actionable basis for subsequent fault location, repair, or preventative maintenance, thereby effectively extending the service life of the server power supply and improving system reliability.

[0064] In one feasible implementation, the step of generating a targeted enhanced disturbance test sequence based on the abnormal pattern includes: parsing the abnormal pattern to determine the sensitive direction of the abnormal server power supply to a preset disturbance parameter in the dynamic load disturbance pattern; generating at least one new disturbance pattern based on the sensitive direction and the preset disturbance parameter, wherein the new disturbance pattern is obtained by adjusting at least one of the intensity, rate of change, or frequency of the preset disturbance parameter; and combining the at least one new disturbance pattern according to a preset logic to generate the enhanced disturbance test sequence.

[0065] In this embodiment, when analyzing the abnormal patterns and determining the sensitivity direction of the abnormal power supply to the preset disturbance parameters in the dynamic load disturbance mode, this application aims to gain a deeper understanding of the failure characteristics of the abnormal power supply. By analyzing previously identified abnormal patterns, such as analyzing model deviation characteristics, abnormal event types and occurrence conditions in real-time electrical waveforms, it is possible to infer which external stimuli (i.e., preset disturbance parameters in the dynamic load disturbance mode) the abnormal power supply exhibits higher sensitivity to. For example, if the abnormality manifests as a slow recovery from an output voltage drop, it may be sensitive to the step amplitude or rate of change of the load current; if it manifests as excessive output ripple, it may be sensitive to load disturbances of a specific frequency. Determining the sensitivity direction, i.e., clarifying which parameter dimension (such as intensity, rate of change, or frequency) adjustment can more effectively trigger or amplify the abnormal phenomenon.

[0066] In this embodiment, when generating at least one new disturbance mode based on the sensitive direction and the preset disturbance parameters, this application will generate a targeted new disturbance mode according to the determined sensitive direction of the abnormal power source and the preset disturbance parameters. Specifically, one or more preset disturbance parameters in the original dynamic load disturbance mode can be adjusted according to the sensitive direction. For example, if the power source is found to be sensitive to the step amplitude of the load current, a new disturbance mode with a larger step amplitude can be generated; if it is sensitive to the rate of change of the load current, a new disturbance mode with a faster rise or fall rate can be generated; if it is sensitive to disturbances at a specific frequency, a new disturbance mode that scans around that frequency or applies a larger amplitude can be generated. This adjustment can be to increase the parameter value to strengthen the stimulus, or to decrease the parameter value to observe the critical point, or to change the change law of the parameter in order to more clearly expose or verify the defect.

[0067] In this embodiment, to comprehensively verify the defects of the abnormal power supply, a combination of multiple new perturbation modes is typically required. Therefore, this application combines the at least one new perturbation mode according to preset logic to generate the enhanced perturbation test sequence. The preset logic may include, but is not limited to: applying different types of new perturbation modes sequentially (e.g., step followed by scan), or triggering certain perturbation modes under specific conditions (e.g., applying high-frequency perturbation when the power supply reaches a specific temperature), or repeating certain perturbation modes in a cyclical manner. This combination aims to simulate more complex, more stringent, or more likely to excite specific defects in the operating environment, thereby ensuring the comprehensiveness and effectiveness of defect verification.

[0068] In this embodiment, through the above technical solution, this application can accurately determine the sensitive direction of the abnormal server power supply to specific dynamic load disturbance parameters based on the identified abnormal patterns. Based on this, by selectively adjusting the intensity, rate of change, or frequency of these preset disturbance parameters, new disturbance patterns that can effectively stimulate and verify defects are generated, and these patterns are combined into a reinforced disturbance test sequence according to preset logic. This method avoids blind or generalized testing, making the defect verification process more targeted and efficient. It can more clearly expose the potential defects of abnormal power supplies, improve the accuracy and efficiency of defect identification, thereby providing a solid foundation for subsequent defect type judgment and verification conclusion generation, and improving the diagnostic capability and testing efficiency of server power supply aging tests.

[0069] In one feasible implementation, the method further includes: after completing the aging test of a preset period, based on the accumulated basic electrical parameters, the real-time electrical waveform and the model deviation characteristics, statistically analyzing the changing trend data of the key parameters of the power supply of the same server under test; comparing the changing trend data with a preset aging drift allowable threshold; and, based on the comparison result, correcting the expected value or allowable deviation range of the corresponding parameter in the personalized dynamic response model to obtain an updated personalized dynamic response model and drift report.

[0070] In this embodiment, after completing the aging test for a preset period, this step aims to ensure that the server power supply has undergone sufficient aging time before model correction and drift analysis, allowing the aging effect to fully manifest. The preset period can be determined based on the power supply's design life, expected aging rate, or industry standards; for example, it could be hundreds or thousands of hours of continuous operation, or completing a specific number of start-stop cycles, load cycles, etc. After completing the preset period, the system will trigger subsequent data statistics, comparison, and model correction processes.

[0071] In this embodiment, to comprehensively assess the aging state of the power supply, this method is based on the accumulated basic electrical parameters, the real-time electrical waveforms, and the model deviation characteristics. The basic electrical parameters refer to the power supply's static performance indicators, such as output voltage accuracy, ripple, and efficiency, acquired at the start of the aging test or periodically. Accumulating these parameters reflects the long-term performance changes of the power supply under steady-state conditions. The real-time electrical waveforms are the transient response data of the power supply during dynamic load disturbances, such as voltage dips, recovery times, and overshoots. Accumulating these waveforms reveals the changes in the power supply's dynamic performance over time. The model deviation characteristics are the deviation information obtained by comparing the real-time electrical waveforms with a personalized dynamic response model; they directly reflect the difference between the power supply's actual behavior and its expected behavior. Accumulating these deviation characteristics can track the degree and pattern of the power supply's performance deviating from its initial model. By comprehensively analyzing this accumulated data, a more accurate understanding of the overall aging trend of the power supply can be obtained.

[0072] Based on this, this method statistically analyzes the changing trend data of key parameters of the same server power supply under test. After accumulating sufficient data, the system performs statistical analysis on the key performance parameters of each server power supply under test to identify their changing trends over time. Key parameters may include, but are not limited to, the average or standard deviation of output voltage, the peak-to-peak value of ripple voltage, the recovery time of transient response, overshoot / undershoot amplitude, and efficiency. The changing trend data can be obtained through various statistical methods, such as performing linear regression analysis on historical data to determine the drift rate of parameters, or using methods such as moving averages and exponential smoothing to smooth the data and highlight long-term trends. This step aims to quantify the degradation or drift of power supply performance.

[0073] In this embodiment, the trend data is compared with a preset aging drift allowable threshold. To determine whether the power supply's aging is within an acceptable range, the statistically analyzed trend data is compared with the preset aging drift allowable threshold. The aging drift allowable threshold is set based on the power supply's design specifications, reliability requirements, or historical experience, and is used to define the boundary between normal aging and abnormal degradation. For example, the output voltage drift can be set to not exceed ±1%, and the ripple voltage increase can be set to not exceed 20%. By comparing, it can be determined whether the power supply's performance drift is still within the expected range, or whether it has exceeded the normal aging range, indicating potential failure risks.

[0074] In this embodiment, based on the comparison results, the expected values ​​or allowable deviation ranges of the corresponding parameters in the personalized dynamic response model are corrected. If the comparison results show that the key parameters of the power supply have experienced aging drift within an acceptable range, the personalized dynamic response model of the power supply needs to be corrected. Correction may include adjusting the expected values ​​of the corresponding parameters in the model to match the current aging state of the power supply. For example, if the average value of a parameter drifts slightly, the expected value of the model is adjusted accordingly. Simultaneously, the allowable deviation range of the parameter can also be corrected. As the power supply ages, its performance may become slightly unstable; therefore, appropriately widening the allowable deviation range can avoid misjudging normal, slight fluctuations as abnormalities. This dynamic correction ensures that the model can always accurately reflect the current performance state of the power supply, thereby improving the accuracy of subsequent anomaly detection.

[0075] In this embodiment, the final result is an updated personalized dynamic response model and drift report. After the above corrections, the system generates an updated personalized dynamic response model. This updated model will be used for subsequent aging tests and anomaly detection to ensure the continued effectiveness of the tests. Simultaneously, the system will also generate a drift report, detailing the trends of key parameters of the power supply within a preset period, comparison results with the allowable aging drift threshold, and the specific content of the model corrections. This report provides important evidence for the long-term performance evaluation, reliability analysis, and predictive maintenance of the power supply.

[0076] In this embodiment, the above technical solution effectively solves the problem of inaccurate initial models caused by the natural drift of power supply performance over time in traditional aging tests. By periodically statistically analyzing the changing trends of key parameters and comparing them with preset allowable aging drift thresholds, this method can intelligently correct the expected values ​​or allowable deviation ranges in the personalized dynamic response model. This allows the model to dynamically adapt to the aging state of the power supply, thereby avoiding misjudging normal performance drift as abnormal and improving the accuracy and robustness of aging tests. Furthermore, the generated drift report provides valuable data support for long-term reliability assessment and fault prediction of the power supply, contributing to more precise power supply health management and maintenance strategies.

[0077] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the server power supply aging test method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0078] This application also provides a server power supply aging test system, for reference... Figure 2 The server power supply aging test system includes: a memory 10, a processor 20, and a server power supply aging test program stored on the memory 10 and executable on the processor 20. The server power supply aging test program is configured to implement the steps of the server power supply aging test method.

[0079] The server power supply aging test system provided in this application, employing the server power supply aging test method in the above embodiments, can achieve accurate fault identification and propagation path analysis. Compared with the prior art, the beneficial effects of the server power supply aging test system provided in this application are the same as those of the server power supply aging test method provided in the above embodiments, and other technical features of the server power supply aging test system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0080] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0081] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. All equivalent structural transformations made under the technical concept of this application using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included within the scope of patent protection of this application.

Claims

1. A method for aging test of server power supply, characterized in that, The method includes: Obtain the basic electrical parameters of the power supply of each server under test under steady-state aging conditions; Based on the aforementioned basic electrical parameters, a personalized dynamic response model for the power supply of each server under test is constructed. Based on the personalized dynamic response model, the dynamic load disturbance mode to be injected is determined to obtain the disturbance mode data to be injected, and the dynamic load disturbance mode is injected into the power supply of at least one selected server under test. At the synchronization moment of the dynamic load disturbance mode, the data acquisition units of all the power supplies of the servers under test are triggered to perform measurements and acquire the real-time electrical waveforms of all the power supplies of the servers under test during the disturbance. The real-time electrical waveform is compared and analyzed with the corresponding personalized dynamic response model to obtain the model deviation characteristics of each server power supply under test. Anomaly detection is performed based on the model deviation characteristics to identify abnormal power supplies and abnormal modes of the server under test.

2. The server power supply aging test method as described in claim 1, characterized in that, The steps for constructing a personalized dynamic response model for the power supply of each server under test based on the aforementioned basic electrical parameters include: Under standard test conditions, a set of predefined dynamic characteristic scanning signals are applied to the power supply of each server under test, and scan response data is obtained. Extract a set of key parameters characterizing the dynamic performance of the power supply from the scan response data; The set of key parameters is associated with the model and specifications of the power supply of the server under test to construct the personalized dynamic response model.

3. The server power supply aging test method as described in claim 2, characterized in that, Under standard testing conditions, the steps for applying a set of predefined dynamic characteristic scanning signals to the power supply of each server under test and obtaining the scan response data include: Apply microsecond-level load current step signals of different amplitudes and record the output voltage recovery waveform of the power supply of the server under test. Apply a small signal disturbance covering a preset frequency range and measure the frequency response of the control loop of the power supply of the server under test. When the server under test is powered on and switches between no-load and power-on states, the voltage or current waveforms of its internal preset monitoring points are captured; wherein, the output voltage recovery waveform, the control loop frequency response, and the voltage or current waveform constitute the scan response data.

4. The server power supply aging test method as described in claim 1, characterized in that, The process of determining the dynamic load disturbance pattern to be injected based on the personalized dynamic response model to obtain the disturbance pattern data to be injected includes: Analyze the personalized dynamic response model to identify the key dynamic performance parameters of the power supply of at least one selected server under test; Based on the type of the key dynamic performance parameters, a load change pattern is generated to verify the dynamic response of the key dynamic performance parameters, which serves as the data for the disturbance pattern to be injected.

5. The server power supply aging test method as described in claim 1, characterized in that, The steps of comparing and analyzing the real-time electrical waveform with the corresponding personalized dynamic response model to obtain the model deviation characteristics of each server power supply under test include: From the real-time electrical waveform, extract the actual response characteristics of the first server under test power supply under the dynamic load disturbance mode, where the first server under test power supply is the power supply injected into the dynamic load disturbance mode. Based on the personalized dynamic response model, the expected response characteristics of the power supply of the first server under test are obtained under the dynamic load disturbance mode. The actual response characteristics are compared with the expected response characteristics to calculate the response deviation data of the power supply of the first server under test. The model deviation features are generated based on whether the response deviation data exceeds a preset deviation threshold.

6. The server power supply aging test method as described in claim 1, characterized in that, All data acquisition units of the power supply of the server under test are synchronized in time; the real-time electrical waveform is a waveform with a timestamp, and the deviation between the timestamps of the waveforms acquired by different acquisition units is less than the preset synchronization accuracy threshold. The anomaly detection based on the model deviation characteristics also includes: When at least two abnormal server power supplies are identified based on the model deviation characteristics, the start timestamps of the abnormal events in the real-time electrical waveforms of these abnormal server power supplies are extracted. Compare the order of the start timestamps of the abnormal events; The power supply of the server under test that first exhibits an abnormal event is identified as the primary fault source, and the fault propagation path is inferred based on the electrical connection relationship between the primary fault source and other abnormal power supplies; wherein, the abnormal mode includes the primary fault source and the fault propagation path.

7. The server power supply aging test method as described in claim 1, characterized in that, The method further includes: Based on the aforementioned anomaly pattern, a targeted enhanced perturbation test sequence is generated; A second round of dynamic load disturbance is applied to the power supply of the abnormal server under test according to the enhanced disturbance test sequence, and the second round of real-time electrical waveforms are acquired simultaneously. Analyze the second round of real-time electrical waveforms to determine the defect type corresponding to the abnormal mode and generate a defect verification conclusion.

8. The server power supply aging test method as described in claim 7, characterized in that, The steps for generating targeted reinforcement perturbation test sequences based on the aforementioned anomaly patterns include: Analyze the abnormal mode to determine the sensitive direction of the abnormal server power supply to the preset disturbance parameters in the dynamic load disturbance mode; Based on the sensitive direction and the preset perturbation parameters, at least one new perturbation mode is generated, wherein the new perturbation mode is obtained by adjusting at least one of the intensity, rate of change, or frequency of the preset perturbation parameters; The at least one new perturbation mode is combined according to a preset logic to generate the enhanced perturbation test sequence.

9. The server power supply aging test method as described in claim 1, characterized in that, The method further includes: After completing the aging test of the preset period, based on the accumulated basic electrical parameters, the real-time electrical waveforms and the model deviation characteristics, the change trend data of the key parameters of the power supply of the same server under test are statistically analyzed. Compare the trend data with a preset allowable aging drift threshold; Based on the comparison results, the expected values ​​or allowable deviation ranges of the corresponding parameters in the personalized dynamic response model are corrected to obtain an updated personalized dynamic response model and drift report.

10. A server power supply aging test system, characterized in that, The server power supply aging test system includes: a memory, a processor, and a server power supply aging test program stored in the memory and executable on the processor, wherein the server power supply aging test program is configured to implement the steps of the server power supply aging test method as described in any one of claims 1 to 9.

Images