Equipment early warning method and electronic equipment
By processing the server's performance parameter groups and fault risk indicators, the problem of inaccuracy in traditional alarm methods is solved, enabling accurate prediction and early warning of server health status and preventing faults from occurring.
Patent Information
- Application Number
- CN202610115466.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-28
- Publication Date
- 2026-03-03
AI Technical Summary
Traditional server alarm methods are inaccurate, leading to the problem that normal high temperatures under high load are falsely reported as faults, and minor anomalies under low load are ignored.
By acquiring the performance parameter set and fault risk index value of the target device, using time alignment and normalization to process multi-source heterogeneous operating data, the actual health index value is determined, and early warning operations are performed based on the fault risk index value and the predicted health index value.
It enables accurate prediction of the future health status of servers, provides early warning of potential faults, avoids serious faults, and improves the accuracy and efficiency of operation and maintenance.
Smart Images
Figure CN121597527A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of server technology, and in particular to device early warning methods and electronic devices. Background Technology
[0002] In the field of server technology, in order to ensure the stable operation of servers, servers are generally monitored. When an anomaly is detected, an alarm is immediately triggered so that maintenance personnel can perform maintenance operations in a timely manner.
[0003] Traditional methods typically use fixed thresholds to trigger alarms. For example, an alarm is triggered when the utilization of the server's central processing unit exceeds a preset threshold. However, this alarm method is inaccurate, and situations such as normal high temperatures under high load being falsely reported as faults and minor anomalies under low load being ignored may occur. Summary of the Invention
[0004] This application provides equipment early warning methods, devices, electronic equipment, storage media, and program products to solve the problem of inaccurate alarms in related technologies.
[0005] This application provides a device early warning method, including:
[0006] Obtain a set of performance parameters for at least one component of the target device and a failure risk index value for the target device, wherein the set of performance parameters includes the performance parameters of the component at at least one point in time; Obtain the baseline health indicator values of the target device at at least one time point; Based on the performance parameters of at least one component at the target time point, determine the actual health index value of the target device at the target time point, wherein the target time point is any one of at least one time points; After determining the actual health index value of the target device at at least one time point, the predicted health index value of the target device is determined based on the actual health index value of the target device at at least one time point, the baseline health index value, and the performance parameter group of at least one component. Based on the fault risk index value and the predicted health index value, an early warning operation is performed on the target equipment.
[0007] This application also provides a device warning system, comprising: The acquisition module is used to acquire a set of performance parameters of at least one component included in the target device and a failure risk index value of the target device, wherein the set of performance parameters includes the performance parameters of the component at at least one time point; and to acquire the baseline health index value of the target device at at least one time point. The determination module is used to determine the actual health index value of the target device at the target time point based on the performance parameters of at least one component at the target time point, wherein the target time point is any one of the at least one time points; after determining the actual health index value of the target device at the at least one time point, the module determines the predicted health index value of the target device based on the actual health index value of the target device at the at least one time point, the baseline health index value, and the performance parameter group of at least one component. The early warning module is used to perform early warning operations on target equipment based on fault risk index values and predicted health index values.
[0008] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for implementing the steps of any of the above-described device warning methods when executing the computer program.
[0009] This application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of any of the above-described device warning methods.
[0010] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above-described device warning methods.
[0011] This application first obtains the performance parameter sets of each component in the target device and the failure risk index values of the target device. The performance parameter sets include performance parameters at least one time point. To compare the degradation of the target device with theoretical conditions, a baseline health index value for the target device at at least one time point can also be obtained. Then, for each time point, the actual health index value of the target device at that time point can be determined based on the performance parameters of each component. Further, after determining the actual health index value at at least one time point, the predicted health index value of the target device can be determined based on the actual health index value, the baseline health index value, and the performance parameter sets at each time point. Finally, based on the failure risk index value and the predicted health index value, an early warning operation can be performed on the target device. In the above process, by predicting the future health index values of the target device in advance using the performance parameter sets of each component in the target device, the probability of future failure of the target device can be determined more accurately. Furthermore, the failure risk index value can indicate the magnitude of the risk faced by the target device when a failure occurs. Therefore, based on the failure risk index value and the predicted health index value, a more accurate early warning of whether the target device will fail can be provided. In addition, by taking appropriate early warning actions on the target equipment before a malfunction occurs, serious malfunctions of the target equipment can be avoided. Attached Figure Description
[0012] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 A schematic diagram of a computing cluster architecture provided for an embodiment of this application; Figure 2 A schematic flowchart illustrating a device early warning method provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a device early warning device provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0014] The technical solutions of the embodiments 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. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.
[0015] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.
[0016] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0017] The device early warning method provided in this application can be implemented by a computing cluster, such as... Figure 1 As shown, a computing cluster can include at least one node, which can be a server. The node can include monitoring nodes and worker nodes. The monitoring node can monitor the other worker nodes and perform relevant alerts.
[0018] Embodiments of this application provide a device early warning method, which can be executed by the aforementioned monitoring node, such as... Figure 2As shown, the specific processing steps of the equipment early warning method may include: Step S201: Obtain the performance parameter group of at least one component included in the target device and the failure risk index value of the target device.
[0019] The target device can be any node in the aforementioned computing cluster. Components can be a central processing unit, hard drive, power supply, fan, memory, etc. The performance parameter group can include the performance parameters of the components at at least one point in time. The performance parameters included in the performance parameter group can be normalized values within the range [0, 1]. The fault risk index value can be used to indicate the magnitude of the risk faced by the target device in the event of a fault.
[0020] Specifically, the target device records multi-source heterogeneous operational data during operation. This multi-source heterogeneous operational data can include various types of operational data such as sensor monitoring parameters, system log events, performance statistics, and firmware alarm records. For example, sensor monitoring parameters can include CPU temperature, memory voltage, hard drive spindle motor current, power supply voltage offset, and fan speed. System log events can include hardware error codes, error checking and correcting (ECC) error accumulation values, Peripheral Component Interconnect Express (PCIE) link retraining counts, and disk I / O timeout events recorded by the operating system or firmware. Performance statistics can include CPU instruction exception rate, memory access latency, disk queue depth, network packet loss rate, and hard drive remapped sectors. Firmware alarm records can include voltage limit violations, overheat warnings, and component self-test failure information reported by the Baseboard Management Controller (BMC) or Unified Extensible Firmware Interface (UEFI) firmware.
[0021] The aforementioned multi-source heterogeneous operational data can be collected in parallel by the management node through the out-of-band management channel and operating system interface of the target device. The out-of-band management channel can refer to the baseboard management controller of the target device. In this way, by collecting data from different sources in parallel through the out-of-band management channel and operating system interface, real-time monitoring of the server hardware status can be achieved. This ensures that critical fault precursor information can still be obtained under extreme conditions such as system-level failures or downtime, improving the integrity and reliability of data collection and providing a solid data foundation for subsequent fault diagnosis and prediction.
[0022] Because different types of operational data have different collection frequencies and inconsistent time scales, to subsequently determine the health of the target device by combining different types of operational data, Coordinated Universal Time (UTC) can be used as the base timestamp for each type of operational data, and time alignment operations can be performed on the different types of operational data. For example, sensor A collects data once per second, and sensor B collects data once every 5 seconds; the data collection times are not at the same point in time. Using the UTC timestamp as the base, for moments without actual sample values, the actual values before and after that moment can be used to estimate the value at that time through an interpolation algorithm. The interpolation algorithm can be linear interpolation, spline interpolation, etc. In this way, a time series matrix can be formed in which all data dimensions have values at the same series of equally spaced time points, with rows representing time points and columns representing different parameters.
[0023] Because different types of operational data have different dimensions and orders of magnitude—for example, voltage values are around 12V, temperature values are between 40-80 degrees Celsius, and sector error counts accumulate from zero to several hundred—to facilitate subsequent assessment of the target device's health based on different types of operational data, after completing the aforementioned time alignment operation, a normalization operation can be performed on each type of operational data. This scales all different types of operational data to the [0,1] interval to eliminate dimensional differences. For example, for any type of operational data, the maximum and minimum values can be determined first, and the first difference between the maximum and minimum values can be calculated. For any value at any given time point, the second difference between the value at that time point and the minimum value can be calculated. The ratio between the second difference and the first difference is then used as the normalized value for that time point.
[0024] After completing the above normalization operation, the monitoring node can also divide the running data into data segments based on the time window corresponding to the type of running data. The time window can be preset according to the degradation characteristics of different components. For example, a short time window is used for high-frequency changing parameters such as temperature and voltage, while a cumulative sliding window is used for slow degradation indicators such as the number of remapped sectors on the hard drive and the cumulative value of ECC errors. A cumulative sliding window is a window with a fixed starting point, but its ending point can continuously slide forward.
[0025] For each component included in the target device, the monitoring node can extract performance parameter sets corresponding to each component from the aforementioned multi-source heterogeneous operating data. For the central processing unit (CPU), the CPU temperature at at least one time point can be extracted from sensor monitoring parameters to generate a corresponding performance parameter set. For memory, the cumulative ECC error value of memory can be extracted from system log events. For the disk, the number of remapped sectors can be extracted from performance statistics. For the power supply, the voltage offset can be extracted from sensor monitoring parameters. For the fan, the fan speed at at least one time point can be extracted from sensor monitoring parameters to generate a corresponding performance parameter set.
[0026] This allows us to obtain the performance parameter set of each component included in the target device. Employing high-precision timestamp synchronization and dynamic range normalization effectively solves the inconsistency problem in time scale and physical dimensions of multi-source heterogeneous data, ensuring state alignment between different sensors and log sources. Combined with a time window strategy that adaptively divides based on degradation characteristics, it preserves the transient change characteristics of high-frequency parameters while capturing the long-term trends of slow degradation indicators, thus improving the ability of the time-series dataset to represent the evolution of real hardware states.
[0027] In addition, the monitoring node can directly obtain the fault risk index value of the target device. The fault risk index value can be preset or determined based on the service type of the target device.
[0028] Step S202: Obtain the baseline health indicator value of the target device at at least one time point.
[0029] Among them, the baseline health index value can be a theoretical health index value of the target device at the target time point under normal conditions.
[0030] Specifically, in order to compare the deviation between the actual degradation of the target device and the theoretical degradation, the monitoring node can obtain the benchmark health index values of the target device at each time point based on the various time points included in the performance parameter group.
[0031] Step S203: Determine the actual health index value of the target device at the target time point based on the performance parameters of at least one component at the target time point.
[0032] The target time point can be any one of at least one time point.
[0033] Specifically, after the aforementioned time alignment operation, the number of time points included in each performance parameter group can be consistent. Therefore, for each time point in the performance parameter group, the monitoring node can perform calculations on the performance parameters of different components at that time point to obtain the actual health index value of the target device at that time point. For example, the calculation operation can be a weighted summation operation, an averaging operation, etc.
[0034] Step S204: After determining the actual health index value of the target device at at least one time point, determine the predicted health index value of the target device based on the actual health index value of the target device at at least one time point, the baseline health index value, and the performance parameter group of at least one component.
[0035] Specifically, after determining the actual health indicator values of the target device at each time point, the monitoring node can determine the deviation of the target device from the theoretical degradation trend at each time point based on the actual health indicator values and the baseline health indicator values. Since the deviation represents the difference between the actual degradation characteristics and the theoretical degradation characteristics of the target device, and the performance parameter groups of each component can indicate the actual degradation characteristics of the target device, based on the deviation and the actual degradation characteristics, the predicted health indicator values of the target device in the future under the current degradation trend can be determined, which is more consistent with the actual degradation characteristics of the target device.
[0036] Step S205: Based on the fault risk index value and the predicted health index value, perform an early warning operation on the target equipment.
[0037] Specifically, the predicted health index value can represent the probability of the target equipment failing in the future, while the failure risk index value can represent the magnitude of the risk faced by the target equipment after a failure. Therefore, the monitoring node can perform early warning operations on the target equipment based on the failure risk index value and the predicted health index value.
[0038] The aforementioned early warning operations can be performed periodically, or they can be based on trigger commands. Trigger commands can be generated by operations and maintenance personnel after performing relevant operations on the operations and maintenance platform and then sent to the monitoring nodes.
[0039] The device early warning method of this application first obtains the performance parameter set of each component in the target device and the fault risk index value of the target device. The performance parameter set includes performance parameters at least one time point. To subsequently compare whether the degradation of the target device matches the theoretical condition, a baseline health index value of the target device at at least one time point can also be obtained. Then, for each time point, the actual health index value of the target device at that time point can be determined based on the performance parameters of each component at that time point. Further, after determining the actual health index value at at least one time point, the predicted health index value of the target device can be determined based on the actual health index value at each time point, the baseline health index value, and each performance parameter set. Finally, based on the fault risk index value and the predicted health index value, an early warning operation can be performed on the target device. In the above process, by predicting the future health index value of the target device in advance through the performance parameter set of each component in the target device, the probability of future failure of the target device can be determined more accurately. Furthermore, the fault risk index value can indicate the magnitude of the risk faced by the target device when a failure occurs. Therefore, based on the fault risk index value and the predicted health index value, an early warning of whether the target device will fail can be given more accurately. In addition, by taking appropriate early warning actions on the target equipment before a malfunction occurs, serious malfunctions of the target equipment can be avoided.
[0040] In some optional implementations, in step S202 above, the target device may determine the actual health indicator value of the target device at the target time point using the following specific steps: Step 1: Obtain the weight values corresponding to at least one component.
[0041] Step 2: Determine the initial health index value of the target device at the target time point based on the performance parameters and weight values of at least one component at the target time point.
[0042] Step 3: Using a pre-constructed nonlinear compression function, the initial health index values of the target device at the target time point are transformed nonlinearly to obtain the actual health index values of the target device at the target time point.
[0043] Specifically, the weight values of each component can be pre-set by technicians and adjusted in real time according to the actual operating conditions of the target device. Therefore, after determining the actual health indicator values of the target device at the target time point, the monitoring node can first obtain the weight values of each component at the target time point. For example, if the IO load indicator value of the target device's hard drive at the target time point is detected to be greater than a first preset threshold, the weight value corresponding to the hard drive will be increased by a preset value (e.g., 0.1). Similarly, if the temperature of the target device at the target time point is detected to be greater than a second preset threshold (e.g., 60 degrees Celsius), the weight values corresponding to the central processing unit and the fan will be increased by preset values. Furthermore, if the number of alarms from the fan at the target time point (e.g., 5 times) is detected to be greater than a third preset threshold, the weight value corresponding to the fan will be increased by a preset value. In this way, by adjusting the weight values in real time, the influence of the corresponding components on the actual health indicator values can be amplified, making the determined actual health indicator values more accurate. It should be noted that the sum of the weight values of all components included in the target device is 1.
[0044] Then, the monitoring node can perform a weighted summation of its various performance parameters at the target time point based on the weight values of each component, and obtain the initial health index value of the target device at the target time point.
[0045] Since the degradation of the target equipment is nonlinear, and the aforementioned weighted summation cannot reflect the nonlinear change characteristics of the equipment, this scheme can further use a nonlinear compression function to perform a nonlinear transformation on the initial health index value after determining the initial health index value, so as to obtain the actual health index value of the target equipment at the target time point, making the actual health index value reflect the nonlinear characteristics of the target equipment and more consistent with the actual operating conditions of the equipment.
[0046] For example, step two can be expressed as follows: (1) Step three can be expressed as follows: (2) in, For the target time point, This represents the initial health indicator value at the target time point. This represents the weight value of the i-th component in the target device at the target time point. This represents the performance parameters of the i-th component at the target time point, where n is the number of components included in the target device. This represents the shape parameter, used to control the curvature of the nonlinear compression function. This represents the offset parameter, used to determine the midpoint position of the nonlinear compression function. and The value can be set by the maintenance personnel based on the characteristics of the equipment.
[0047] In this way, by considering the performance parameters of different components, the determined health index values can reflect the comprehensive characteristics of the target equipment, making the determined health index values more accurate. Furthermore, this scheme uses a nonlinear function to perform a nonlinear transformation on the initial health index values, which can yield actual health index values that reflect the actual operating conditions, making them even more accurate.
[0048] In some optional implementations, in step S204 above, each component's performance parameter group may include at least one sub-data group (the sub-data group being the aforementioned data segment), and the sub-data group may be a data group composed of performance parameters within a time window. The monitoring node may determine the predicted health indicator value of the target device at the target time point using the following specific steps: Step 1: Determine the offset of the target time point based on the actual health indicator value and the benchmark health indicator value at the target time point.
[0049] Step 2: After determining the offset corresponding to at least one time point in the target sub-data group, determine the abnormal disturbance index value corresponding to the target sub-data group based on the offset corresponding to at least one time point in the target sub-data group.
[0050] Step 3: After determining the abnormal disturbance index values corresponding to at least one sub-data group, determine the predicted health index values based on the abnormal disturbance index values corresponding to at least one sub-data group and at least one sub-data group.
[0051] The target subdata group can be any subdata group from at least one subdata group.
[0052] Specifically, the monitoring node can determine the difference between the actual health indicator value and the benchmark health indicator value at the target time point as the offset at the target time point. This offset can reflect the degree to which the target device deviates from the theoretical health value at the target time point, that is, it can reflect the health status of the target device.
[0053] For each sub-data group, the monitoring node can calculate the abnormal disturbance index value within the corresponding time window based on the offsets at each time point within that sub-data group, thus identifying any abnormalities in the target device within that time window. Taking a target sub-data group as an example, the monitoring node can calculate the average offset across all time points within the target sub-data group and use this average as the corresponding abnormal disturbance index value. Alternatively, the monitoring node can statistically analyze the distribution of offsets within the target sub-data group and determine the corresponding abnormal disturbance index value based on this distribution.
[0054] In a similar manner, after determining the abnormal disturbance index values corresponding to each sub-data group included in each performance parameter group, the monitoring node can determine the predicted health index value based on each sub-data group and its corresponding abnormal disturbance index value.
[0055] For example, step one can be expressed as follows: (3) in, This represents the offset from the target time point. This represents the actual health indicator value at the target time point. This represents the baseline health indicator value at the target time point.
[0056] In this way, by comparing the actual health condition with the theoretical health condition, unexpected degradation behaviors of the target equipment that deviate from the normal aging process can be identified. Furthermore, by calculating the abnormal disturbance index value through the offset, and then combining the abnormal disturbance index value with the predicted health index value, the final predicted health index value can reflect the current degree of health deviation of the target equipment, as well as information on the current abnormal external stress or potential failure causes of the target equipment, providing a quantitative basis for the early detection of latent faults.
[0057] In some optional implementations, in step two of step S204 above, the distribution may include the average offset and standard deviation of the target sub-data group. Accordingly, the monitoring node may use the following specific steps to determine the abnormal disturbance index value of the target sub-data group: Step 1: Determine the average offset and standard deviation based on the offset corresponding to at least one time point in the target sub-data group.
[0058] Step 2: Determine the abnormal perturbation index value corresponding to the target sub-data group based on the average offset, standard deviation, and pre-built abnormal perturbation index.
[0059] Specifically, the monitoring node can calculate the average offset of the target sub-data group at each time point, and determine the average offset as the mean. Then, based on the offset at each time point and the mean offset, the standard deviation of the target sub-data group's offset is calculated.
[0060] The monitoring node can pre-store the abnormal disturbance index determination model. Accordingly, the monitoring node can input the average offset and standard deviation of the target sub-data group into the abnormal disturbance index determination model to obtain the abnormal disturbance index value output by the abnormal disturbance index determination model.
[0061] For example, the model for determining the anomaly index can be expressed as follows: (4) in, This represents the abnormal disturbance index value corresponding to the target sub-data group. The preset weight value (for example, it could be 0.5). This represents the average offset of the target subgroup of data. This represents the standard deviation of the target subgroup of data. This represents the baseline mean of the instantaneous offset of the target device under steady-state operation. Indicates the baseline standard deviation. and The value can be set by the maintenance personnel based on the characteristics of the equipment.
[0062] In some optional implementations, in step three of step S204 above, the monitoring node may use the following specific steps to determine the predicted health indicator value of the target sub-data group: Step 1: Determine the time-sensitive factor corresponding to the target sub-data group based on the abnormal disturbance index value and preset gain coefficient corresponding to the target sub-data group, as well as the pre-acquired historical maximum abnormal disturbance index value.
[0063] Step 2: Based on the time-sensitive factor corresponding to the target sub-data group, adjust the performance parameters included in the target sub-data group to obtain the adjusted target sub-data group.
[0064] Step 3: After completing the adjustment operations for at least one sub-data group, input at least one sub-data group into the pre-built health indicator prediction model to obtain the predicted health indicator value output by the health indicator prediction model.
[0065] Specifically, step 1 can be expressed as follows: (5) in, It could be a time-sensitive factor. It can be a preset reference time sensitivity factor. The preset gain coefficient, This represents the abnormal disturbance index value. This is the historical maximum abnormal disturbance index value, used to normalize D.
[0066] For any set of performance parameters, the performance parameters of each time window can be adjusted using the time sensitivity factor of that time window. For example, the time sensitivity factor of the target sub-data set is multiplied by the performance parameters of each time point included in the target sub-data set to obtain the adjusted performance parameters corresponding to each time point. The adjusted performance parameters corresponding to each time point in the target sub-data set constitute the adjusted target sub-data set.
[0067] Similarly, after adjusting each sub-data set included in each performance parameter group, the adjusted sub-data sets can be input into the health indicator prediction model to obtain the predicted health indicator values. For example, the health indicator prediction model can be a neural network model, specifically a Long Short Term Memory (LSTM) model.
[0068] Since the abnormal disturbance index value reflects the degree of abnormality or instability of a time window, by adjusting the performance parameters of the corresponding time window through the time sensitivity factor calculated based on the abnormal disturbance index value, the performance parameters of the time window with a higher degree of abnormality can be more easily noticed by the health indicator prediction model, thereby enabling the health indicator prediction model to predict more accurate health indicator values.
[0069] In some alternative implementations, the total loss function used during the training phase of the health indicator prediction model can be as follows: (6) in, The mean squared error between the predicted and actual health indicator values at each time point is given. The correlation coefficient (e.g., Pearson correlation coefficient) between the predicted health indicator value evolution trend and the baseline health indicator value evolution trend at each time point is given, where A is the regularization weight, with a value between [0,1].
[0070] In some alternative implementations, the monitoring node may also calculate the composite acceleration factor of the target component (which may be any component included in the target device, such as the central processing unit) at various time points according to the following formulas (7) to (9), and then calculate the baseline health index value based on the composite acceleration factor: (7) in, Indicates the temperature acceleration factor. Indicates activation energy. Represents Boltzmann's constant. T represents the reference temperature, and T represents the actual operating temperature.
[0071] (8) in, Indicates the voltage acceleration factor. V represents the rated voltage, and V represents the actual operating voltage. This indicates the voltage acceleration index.
[0072] (9) in, The acceleration factor represents the combined acceleration due to temperature and voltage.
[0073] Specifically, at the target time point, the monitoring node can calculate the temperature acceleration factor at the target time point based on the actual operating temperature of the target component at the target time point and the above formula (7), and calculate the voltage acceleration factor of the target component based on the actual operating voltage of the target component at the target time point and the above formula (8). Then, the two are multiplied by formula (9) to obtain the composite acceleration factor at the target time point.
[0074] Furthermore, the monitoring node can calculate the baseline health index value of the target device at the target time point based on the mapping relationship between the acceleration factor, time, and baseline health index value, as well as the composite acceleration factor between the target time points. For example, this mapping relationship can take various forms such as linear functions and exponential functions. The exponential function can be... =exp(-B× (×t), where B can be 1e-5.
[0075] In this way, by taking into account the physical properties of key electronic components, the baseline health index values of the target equipment can be determined more accurately.
[0076] In some optional implementations, the health indicator prediction model may include a hidden layer. During training, the forget gate and input gate parameters of the hidden layer can be updated using a composite acceleration factor, specifically, the composite acceleration factor can be used as a scaling factor for each parameter. Thus, in When the value is greater than 1 (accelerated degradation), it can enhance forgetting and input, allowing health indicator prediction models to update the state more quickly. When ≈1 (normal degradation), maintain normal operation. When the value is less than 1 (degeneration slows down), forgetting and input can be reduced, maintaining a stable state.
[0077] In this way, by embedding the physical acceleration factor into the gating mechanism of the health indicator prediction model and introducing a physical consistency regularization term into the loss function, the health indicator prediction model is constrained by the law of hardware degradation during the prediction process. This not only improves the generalization ability of the health indicator prediction model in the small sample or early data stage, but also enhances the interpretability and physical rationality of the prediction results, effectively avoiding the erroneous predictions that may occur in pure data-driven models that violate physical laws.
[0078] In some optional implementations, in step S201 above, the fault risk index value can be preset by technicians or determined based on the real-time operation and maintenance information of the target equipment. If the fault risk index value is the latter, the monitoring node can obtain the fault risk index value of the target equipment using the following specific steps: Step 1: Obtain the target level, resource reserve information, and economic loss index value of the target business executed by the target device, as well as the economic loss index value caused by the shutdown in the event of a failure.
[0079] Step two: Determine the failure risk index value based on the target level, resource reserve information, and economic loss index value.
[0080] The rating system reflects the criticality of the business. Different levels of criticality correspond to different risks faced by the equipment. For example, a higher rating indicates a more critical business, and the impact of business downtime after equipment failure will differ. Additionally, the rating reflects the maintenance difficulty of the equipment. Higher ratings mean greater maintenance difficulty, requiring more time and manpower for repairs after a failure. Resource reserve information includes inventory information for various components used in the target equipment model and the scheduling information of maintenance personnel, such as the number of maintenance personnel in the current time period.
[0081] Specifically, the monitoring node can first obtain the service level agreement (SLA) information of the target device, and then obtain the target level corresponding to that SLA information. Next, the monitoring node can obtain the resource availability index value corresponding to the resource reserve information. Finally, the monitoring node can perform a weighted summation of the target level, the resource availability index value, and the economic loss index value to obtain the fault risk index value.
[0082] For example, step two can be expressed as follows: (10) in, The preset first weight value is used to indicate the importance of the level. This is a preset second weight value used to indicate the importance of resource availability metrics. This is a preset third weight value used to indicate the importance of the economic loss indicator value. C represents the level, and R represents the resource availability indicator value. This represents the value of the economic loss indicator.
[0083] In this way, by conducting a multi-dimensional joint assessment of fault risk level with business criticality, maintenance resources and downtime costs, a closed-loop transformation from technical status early warning to operation and maintenance decision support is achieved. The generated maintenance response mechanism not only takes into account the health status of the target equipment, but also considers business impact and operation and maintenance feasibility, improving the scientificity and practicality of maintenance strategies and effectively reducing the risk of unplanned downtime and overall operation and maintenance costs.
[0084] In some optional implementations, in step S205 above, the target device may perform a warning operation according to the following specific steps: Step 1: Based on the predicted health indicator value and multiple preset health indicator value ranges, determine the target health indicator value range to which the predicted health indicator value belongs.
[0085] Step 2: Based on the target health indicator value range, determine the target risk level corresponding to the target health indicator value range from among several preset risk levels.
[0086] Step 3: Based on the fault risk index value, determine the target fault risk index interval to which the fault risk index value belongs from among the preset multiple fault risk index intervals.
[0087] Step 4: Based on the target health indicator value range and the target failure risk indicator range, perform early warning operations on the target equipment.
[0088] Specifically, the monitoring node can pre-store a mapping table between health indicator value ranges and risk levels, as shown in Table 1.
[0089] Table 1
[0090] For example, H1 can be 0.3. It can be 0.8.
[0091] Accordingly, the monitoring node can first determine the target health indicator value range to which the predicted health indicator value belongs from the multiple health indicator value ranges included in the mapping relationship table. Then, it determines the risk level corresponding to the target health indicator value range from the multiple risk levels included in the mapping relationship table as the target risk level. Similarly, the monitoring node can pre-store multiple fault risk indicator ranges, and accordingly, it can determine the target fault risk indicator range to which the fault risk indicator value belongs from these ranges. Finally, the monitoring node can select the corresponding maintenance method based on the target health indicator value range and the target fault risk indicator range, and perform early warning operations on the target equipment.
[0092] In this way, selecting the appropriate early warning method under different circumstances can reduce operation and maintenance costs, accurately execute early warning operations, and avoid serious failures.
[0093] In some optional implementations, in step four of step S205, the monitoring node may perform an early warning operation on the target device using the following specific steps: Step 1: If the target risk level is determined to be the first risk level and the target fault risk index range is the first fault risk index range, generate alarm information corresponding to the target device.
[0094] The operation of generating alarm information is called the early warning operation.
[0095] Alternatively, in step 2, if the target risk level is determined to be the second risk level and the target fault risk index range is the second fault risk index range, maintenance task information corresponding to the target equipment is generated.
[0096] Among them, the second risk level is lower than the first risk level, the third risk level is lower than the second risk level, and the risk level corresponding to the second fault risk index range is lower than the risk level corresponding to the first fault risk index range. The operation of generating maintenance task information is the early warning operation.
[0097] Alternatively, in step 3, if the target risk level is determined to be the third risk level, information on changes in abnormal disturbance indicators is obtained.
[0098] Step 4: If the change information of the abnormal disturbance index is determined to be the target change information, the monitoring mode of the target device is adjusted from the first monitoring mode to the second monitoring mode.
[0099] The second monitoring mode has a higher monitoring intensity than the first monitoring mode, and adjusting the monitoring mode constitutes an early warning operation. Target change information is used to indicate that abnormal disturbance indicators maintain an upward trend across multiple consecutive time points.
[0100] Specifically, in step 1, when the target risk level is determined to be the first risk level (e.g., high risk in Table 1), and the target fault risk index range is within the first fault risk index range, it indicates that the target device has a high risk. In this case, the highest level of alarm operation can be performed on the target device. For example, the monitoring node can send alarms to the terminal devices of the maintenance personnel through various means such as telephone, SMS, and health platform pop-ups. In this way, the maintenance personnel can immediately log in to the target device remotely to confirm the fault risk. If the target business allows, the service will be migrated to the backup server immediately, and an engineer will be arranged to replace the faulty hardware (such as memory, hard drive, or power supply) in the data center within 2 hours. If migration is not possible, the business side will be coordinated to shut down the system for emergency repair within the next 30 minutes to minimize the impact on the target device.
[0101] In step 2, if the target risk level is determined to be the second risk level (e.g., medium risk in Table 1) and the target failure risk index range is within the second failure risk index range, maintenance task information can be generated to notify maintenance personnel to perform maintenance operations on the target device within a specified time period. Specifically, a maintenance task can be automatically created in the maintenance work order system and the maintenance personnel can be notified. The maintenance personnel can then negotiate with the business department for an off-peak time (e.g., 2:00 AM next Sunday) to shut down the device, replace components with minor abnormalities but not yet failed (such as fans or aging hard drives), and simultaneously perform a firmware upgrade and dust cleaning. After completing the above operations, the maintenance personnel can close the work order after verifying the stability of the target device.
[0102] In step 3, if the target risk level is determined to be the third risk level (e.g., low risk in Table 1), the monitoring node can first acquire information on changes in abnormal disturbance indicators. Then, if the abnormal disturbance indicators maintain an upward trend for a predetermined number of consecutive time points, the monitoring mode of the target device can be adjusted from the first monitoring mode to the second monitoring mode for more intensive monitoring to prevent target device failure. For example, the data acquisition frequency of the target device can be increased from once per minute to once every 10 seconds, or the data acquisition items can be increased; originally, only the temperature of the central processing unit was monitored, but now the voltage fluctuation of the central processing unit is also monitored. Additionally, the monitoring node can tag the monitoring information of the target device with a "highlighted" label and transmit it to the monitoring screen. The monitoring screen can then display the health information of the tagged target devices, allowing maintenance personnel to focus on the target devices. Alternatively, the monitoring node can automatically generate a special health report for the target device daily and send it to the device where the on-duty engineer is located for manual review.
[0103] In this way, different maintenance measures can be taken for different situations, which can not only avoid waste of resources, but also avoid risks in advance.
[0104] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.
[0105] Embodiments of this application also provide an equipment maintenance device, such as... Figure 3 As shown, it includes: The acquisition module 310 is used to acquire a performance parameter group of at least one component included in the target device and a failure risk index value of the target device, wherein the performance parameter group includes the performance parameters of the component at at least one time point; and to acquire the baseline health index value of the target device at at least one time point. The determination module 320 is used to determine the actual health index value of the target device at the target time point based on the performance parameters of at least one component at the target time point, wherein the target time point is any one of the at least one time points; after determining the actual health index value of the target device at the at least one time point, the predicted health index value of the target device is determined based on the actual health index value of the target device at the at least one time point, the baseline health index value, and the performance parameter group of at least one component. The early warning module 330 is used to perform early warning operations on the target equipment based on the fault risk index value and the predicted health index value.
[0106] In some alternative implementations, the determining module 320 is specifically used for: Obtain the weight value corresponding to at least one component; Based on the performance parameters and weight values of at least one component at the target time point, determine the initial health index value of the target device at the target time point; By using a pre-constructed nonlinear compression function, the initial health index values of the target device at the target time point are transformed nonlinearly to obtain the actual health index values of the target device at the target time point.
[0107] In some alternative implementations, each component's performance parameter group includes at least one sub-data group; the determination module 320 is specifically used for: Determine the offset of the target time point based on the actual health indicator value and the benchmark health indicator value at the target time point; After determining the offset corresponding to at least one time point in the target sub-data group, the abnormal disturbance index value corresponding to the target sub-data group is determined based on the offset corresponding to at least one time point in the target sub-data group. The target sub-data group is any sub-data group among at least one sub-data group. After determining the abnormal disturbance index values corresponding to at least one sub-data group, the predicted health index values are determined based on the abnormal disturbance index values corresponding to at least one sub-data group and at least one sub-data group.
[0108] In some alternative implementations, the determining module 320 is specifically used for: Based on the offset corresponding to at least one time point in the target sub-data set, determine the mean offset and standard deviation respectively; Based on the average offset, standard deviation, and pre-built anomaly perturbation index determination model, the anomaly perturbation index value corresponding to the target sub-data group is determined.
[0109] In some alternative implementations, the determining module 320 is specifically used for: Based on the abnormal disturbance index value and preset gain coefficient corresponding to the target sub-data group, as well as the pre-acquired historical maximum abnormal disturbance index value, determine the time-sensitive factor corresponding to the target sub-data group; Based on the time-sensitive factor corresponding to the target sub-data set, the performance parameters included in the target sub-data set are adjusted to obtain the adjusted target sub-data set. After completing the adjustment operations for at least one sub-data group, the at least one sub-data group is input into the pre-built health indicator prediction model to obtain the predicted health indicator value output by the health indicator prediction model.
[0110] In some optional implementations, the acquisition module 310 is specifically used for: Obtain the target level, resource reserve information, and economic loss index value caused by the shutdown when a failure occurs for the target business executed by the target device; The failure risk index value is determined based on the target level, resource reserve information, and economic loss index value.
[0111] In some optional implementations, an early warning operation is performed on the target equipment based on the fault risk indicator value and the predicted health indicator value, including: Based on the predicted health indicator values and multiple preset health indicator value ranges, determine the target health indicator value range to which the predicted health indicator value belongs. Based on the target health indicator value range, determine the target risk level corresponding to the target health indicator value range from multiple preset risk levels; Based on the fault risk index value, determine the target fault risk index interval to which the fault risk index value belongs from among multiple preset fault risk index intervals; Based on the target health indicator value range and the target failure risk indicator range, perform early warning operations on the target equipment.
[0112] In some alternative implementations, the early warning module 330 is specifically used for: When the target risk level is determined to be the first risk level and the target failure risk index range is the first failure risk index range, an alarm message corresponding to the target device is generated. The operation of generating alarm message is the early warning operation. Alternatively, if the target risk level is determined to be the second risk level and the target fault risk index range is the second fault risk index range, maintenance task information corresponding to the target equipment is generated. In this case, the second risk level is lower than the first risk level, and the risk level corresponding to the second fault risk index range is lower than the risk level corresponding to the first fault risk index range. The operation of generating maintenance task information is called early warning operation.
[0113] In some alternative implementations, the early warning module 330 is specifically used for: When the target risk level is determined to be the third risk level, information on the changes in abnormal disturbance indicators is obtained, where the third risk level is lower than the second risk level. When the change information of the abnormal disturbance index is determined to be the target change information, the monitoring mode of the target device is adjusted from the first monitoring mode to the second monitoring mode. The monitoring intensity of the second monitoring mode is higher than that of the first monitoring mode. The adjustment operation of the monitoring mode is the early warning operation.
[0114] For a description of the features in the embodiment corresponding to the equipment early warning device, please refer to the relevant description of the embodiment corresponding to the equipment early warning method, which will not be repeated here.
[0115] Embodiments of this application also provide an electronic device, such as... Figure 4 As shown, it includes a memory 10 and a processor 20. The memory 10 stores a computer program, and the processor 20 is configured to run the computer program to perform the steps in any of the above-described device warning method embodiments.
[0116] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described device warning method embodiments when it is run.
[0117] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.
[0118] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described device warning method embodiments.
[0119] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above-described device warning method embodiments.
[0120] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0121] The foregoing has provided a detailed description of the device early warning method, apparatus, electronic device, storage medium, and program product provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.
Claims
1. A method for early warning of equipment, characterized in that, include: Obtain a set of performance parameters for at least one component of the target device and a failure risk index value for the target device, wherein the set of performance parameters includes the performance parameters of the component at at least one point in time; Obtain the baseline health indicator value of the target device at at least one of the time points; Based on the performance parameters of at least one of the components at a target time point, the actual health index value of the target device at the target time point is determined, wherein the target time point is any one of the at least one time point; After determining the actual health index value of the target device at at least one of the time points, the predicted health index value of the target device is determined based on the actual health index value and the baseline health index value of the target device at at least one of the time points, as well as the performance parameter group of at least one of the components. Based on the fault risk index value and the predicted health index value, an early warning operation is performed on the target device.
2. The equipment early warning method according to claim 1, characterized in that, Determining the actual health indicator value of the target device at the target time point based on the performance parameters of at least one of the components at the target time point includes: Obtain the weight value corresponding to at least one of the components; Based on the performance parameters and weight values of at least one of the components at the target time point, the initial health index value of the target device at the target time point is determined; By using a pre-constructed nonlinear compression function, the initial health index value of the target device at the target time point is transformed nonlinearly to obtain the actual health index value of the target device at the target time point.
3. The equipment early warning method according to claim 1 or 2, characterized in that, Each of the component's performance parameter groups includes at least one sub-data group; After determining the actual health indicator value of the target device at at least one of the time points, the step of determining the predicted health indicator value of the target device based on the actual health indicator value of the target device at at least one of the time points, the baseline health indicator value, and the performance parameter group of at least one of the components includes: Based on the actual health indicator values and the benchmark health indicator values at the target time point, determine the offset of the target time point; After determining the offset corresponding to at least one time point in the target sub-data group, the abnormal disturbance index value corresponding to the target sub-data group is determined based on the offset corresponding to at least one time point in the target sub-data group, wherein the target sub-data group is any one of the at least one sub-data group; After determining the abnormal disturbance index values corresponding to at least one of the sub-data groups, the predicted health index value is determined based on the abnormal disturbance index values corresponding to at least one of the sub-data groups and at least one of the sub-data groups.
4. The equipment early warning method according to claim 3, characterized in that, After determining the offset corresponding to at least one time point in the target sub-data group, the step of determining the abnormal disturbance index value corresponding to the target sub-data group based on the offset corresponding to at least one time point in the target sub-data group includes: Based on the offset corresponding to at least one of the time points in the target sub-data group, the average offset and standard deviation are determined respectively. Based on the average offset, the standard deviation, and the pre-constructed abnormal disturbance index determination model, the abnormal disturbance index value corresponding to the target sub-data group is determined.
5. The equipment early warning method according to claim 3, characterized in that, After determining the abnormal disturbance index values corresponding to at least one of the sub-data groups, the step of determining the predicted health index value based on the abnormal disturbance index values corresponding to at least one of the sub-data groups and at least one of the sub-data groups includes: Based on the abnormal disturbance index value and preset gain coefficient corresponding to the target sub-data group, as well as the pre-acquired historical maximum abnormal disturbance index value, determine the time-sensitive factor corresponding to the target sub-data group; Based on the time-sensitive factor corresponding to the target sub-data group, the performance parameters included in the target sub-data group are adjusted to obtain the adjusted target sub-data group. After completing the adjustment operations corresponding to at least one of the sub-data groups, the at least one of the sub-data groups is input into the pre-built health indicator prediction model to obtain the predicted health indicator value output by the health indicator prediction model.
6. The equipment early warning method according to claim 1 or 2, characterized in that, Obtaining the fault risk index value of the target device includes: Obtain the target level, resource reserve information, and economic loss index value caused by the shutdown when a failure occurs for the target service executed by the target device; The fault risk index value is determined based on the target level, the resource reserve information, and the economic loss index value.
7. The equipment early warning method according to claim 3, characterized in that, The step of performing an early warning operation on the target device based on the fault risk index value and the predicted health index value includes: Based on the predicted health indicator value and multiple preset health indicator value ranges, determine the target health indicator value range to which the predicted health indicator value belongs; Based on the target health indicator value range, a target risk level corresponding to the target health indicator value range is determined from a set of preset risk levels; Based on the fault risk index value, the target fault risk index interval to which the fault risk index value belongs is determined from a set of multiple preset fault risk index intervals. Based on the target health indicator value range and the target fault risk indicator range, an early warning operation is performed on the target equipment.
8. The equipment early warning method according to claim 7, characterized in that, The step of performing an early warning operation on the target device based on the target health indicator value range and the target fault risk indicator range includes: When the target risk level is determined to be the first risk level and the target fault risk index range is the first fault risk index range, an alarm message corresponding to the target device is generated, wherein the operation of generating the alarm message is the early warning operation. Alternatively, if the target risk level is determined to be the second risk level and the target fault risk index range is the second fault risk index range, maintenance task information corresponding to the target equipment is generated, wherein the second risk level is lower than the first risk level, the risk level corresponding to the second fault risk index range is lower than the risk level corresponding to the first fault risk index range, and the operation of generating the maintenance task information is the early warning operation.
9. The equipment early warning method according to claim 8, characterized in that, The method further includes: If the target risk level is determined to be the third risk level, the change information of the abnormal disturbance index is obtained, wherein the third risk level is lower than the second risk level; If the change information of the abnormal disturbance index is determined to be the target change information, the monitoring mode of the target device is adjusted from the first monitoring mode to the second monitoring mode, wherein the monitoring intensity of the second monitoring mode is higher than that of the first monitoring mode, and the adjustment operation of the monitoring mode is the early warning operation.
10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the device warning method as described in any one of claims 1 to 9 when executing the computer program.