Fault early warning method and electronic equipment

Abstract

The invention discloses a fault early warning method and electronic equipment, and relates to the technical field of optical modules, and the method comprises the steps: obtaining the latest monitoring values, historical monitoring values and alarm thresholds of various performance parameters of an optical module in a current period; and determining the sub-fault probability of the target performance parameter according to the latest monitoring value and the historical monitoring value of the target performance parameter, the alarm threshold and the fault probability conversion model. And determining the total fault probability by adopting the fault probability prediction model and the latest monitoring values of various performance parameters. And generating fault early warning information according to the sub-fault probabilities and the total fault probabilities of the various performance parameters. According to the scheme, the sub-fault probabilities are calculated from the dimensionality of independent performance parameters, the total fault probability is calculated from the perspective of correlation influence between different performances, and the sub-fault probabilities and the total fault probabilities of various performance parameters are combined, so that accurate fault early warning information can be generated, and the early warning accuracy is improved.

Description

Technical Field

[0001] This application relates to the field of optical module technology, and in particular to fault warning methods and electronic devices. Background Technology

[0002] In the field of optical module technology, the built-in Digital Diagnostic Monitoring (DDM) function of the optical module is generally used to monitor various performance parameters, such as temperature, drive voltage, and bias current. For any performance parameter, after the optical module obtains the monitored value, if the monitored value is within a preset range, the performance parameter of the optical module is determined to be normal; if the monitored value is outside the preset range, the performance parameter of the optical module is determined to be abnormal, and an early warning can be issued so that technicians can repair or replace the optical module in a timely manner.

[0003] However, this method has low accuracy in early warning, which in turn affects the normal operation of the optical module and even the normal operation of the cluster where the optical module is located. Summary of the Invention

[0004] This application provides fault early warning methods, devices, electronic equipment, storage media, and program products to address the problem of low accuracy in early warning in related technologies.

[0005] This application provides a fault early warning method, including: In the current cycle, obtain the latest monitoring value and alarm threshold of at least one performance parameter of the optical module, as well as the historical monitoring value of at least one performance parameter in previous cycles. Based on the latest monitoring value, alarm threshold, historical monitoring value of the target performance parameter, and the pre-built fault probability conversion model, determine the sub-fault probability corresponding to the target performance parameter, wherein the target performance parameter is any one of at least one performance parameter. The total failure probability of the optical module is determined based on the latest monitored value of at least one performance parameter and a pre-built failure probability prediction model. After determining the sub-fault probabilities corresponding to at least one performance parameter, fault warning information is generated based on the sub-fault probabilities and total fault probabilities corresponding to at least one performance parameter, in order to provide early warning for the optical module.

[0006] This application also provides a fault early warning device, including: The acquisition module is used to acquire, in the current cycle, the latest monitoring value and alarm threshold of at least one performance parameter of the optical module, as well as the historical monitoring value of at least one performance parameter in previous cycles. The determination module is used to determine the sub-fault probability corresponding to the target performance parameter based on the latest monitoring value, alarm threshold, historical monitoring value, and pre-built fault probability conversion model of the target performance parameter, wherein the target performance parameter is any one of at least one performance parameter; and to determine the total fault probability of the optical module based on the latest monitoring value of at least one performance parameter and the pre-built fault probability prediction model. The generation module is used to generate fault warning information based on the sub-fault probabilities corresponding to at least one performance parameter and the total fault probability, after determining the sub-fault probabilities corresponding to at least one performance parameter, so as to provide early warning for the optical module.

[0007] 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 fault warning methods when executing the computer program.

[0008] 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 fault warning methods.

[0009] 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 fault warning methods.

[0010] This application achieves the following: First, during the current period, both the latest and historical monitoring values ​​of various performance parameters of the optical module are obtained. Then, based on the latest and historical monitoring values ​​of the target performance parameter, along with alarm thresholds and a fault probability conversion model, the sub-fault probability of the target performance parameter is determined—that is, the probability that the optical module will fail in the performance dimension corresponding to the target performance parameter. Furthermore, considering the relationships between performance parameters, a fault probability prediction model and the latest monitoring values ​​of various performance parameters are used to determine the total fault probability. Finally, based on the sub-fault probabilities and the total fault probability of various performance parameters, fault warning information is generated. In the above process, this solution considers the dimension of individual performance parameters to calculate sub-fault probabilities and also considers the correlation between different performance parameters to calculate the total fault probability. Combining the sub-fault probabilities and the total fault probability of various performance parameters allows for the generation of accurate fault warning information, improving the accuracy of warnings. This enables technicians to take timely corresponding measures based on the fault warning information (i.e., shortening fault response time) and ensuring the normal operation of the optical module. Attached Figure Description

[0011] 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.

[0012] Figure 1 This is a schematic diagram of the architecture of a fault early warning system provided in an embodiment of this application; Figure 2 A flowchart illustrating a fault early warning method provided in an embodiment of this application; Figure 3 A schematic diagram of a fault probability conversion model provided in an embodiment of this application; Figure 4 A schematic diagram of a performance parameter prediction model provided in an embodiment of this application; Figure 5 A schematic diagram of a fault probability prediction model provided in an embodiment of this application; Figure 6 A flowchart illustrating the process of determining the final failure probability, provided for an embodiment of this application; Figure 7 This is a schematic diagram of the structure of a fault early warning device provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0013] 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.

[0014] 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.

[0015] 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.

[0016] The fault early warning method provided in this application can be implemented by a fault early warning system, such as... Figure 1 As shown, a fault early warning system can include optical modules, fault early warning nodes, and multiple computing nodes. Fault early warning nodes can be computer devices, such as computers or servers. Computing nodes can be servers, such as servers equipped with Graphics Processing Units (GPUs), used to provide high-performance services. Optical modules are used to enable high-speed communication between different computing nodes; specifically, they convert electrical signals into optical signals and vice versa during communication between computing nodes. Fault early warning nodes monitor the optical modules and execute early warning operations.

[0017] In recent years, with the development of Artificial Intelligence Generated Content (AIGC) technology, the number of parameters and the scale of data in Large Language Models (LLMs) have shown a continuous growth trend. Training and using large language models requires significant computing and storage resources. To meet this demand, a distributed cluster deployment model is generally adopted when deploying large language models. In this model, multiple computing nodes work collaboratively to undertake complex computational tasks.

[0018] In the actual operation of a distributed cluster, frequent and high-speed data exchange operations are required between computing nodes to ensure the smooth progress of model training and inference tasks. These data exchange operations can include critical operations such as gradient synchronization, parameter updates, and the transmission of intermediate results. Due to this high-frequency, high-volume data interaction requirement, the performance and reliability of the cross-node interconnection network are particularly important, directly determining the overall system's computing power efficiency and task execution efficiency.

[0019] Currently, distributed clusters typically employ optical communication technologies for interconnection, such as InfiniBand and Remote Direct Memory Access over Converged Ethernet (RoCE). The core physical layer device in optical communication is the optical module, whose primary function is to convert electrical signals to optical signals, thereby enabling high-speed, long-distance data transmission. The performance of the optical module directly affects the overall transmission efficiency and stability of the network. However, in actual large-scale deployments and applications, optical modules inevitably experience a certain failure rate due to their inherent operating characteristics and complex physical environment factors. Optical module failures manifest in various forms, including performance degradation, complete failure, and link interruption. Once these failures occur, they can significantly increase communication latency between computing nodes, raise packet retransmission rates, and may even lead to a complete network connection outage.

[0020] During critical phases of training or inference in large language models, optical module failures can not only cause distributed computing jobs to stall or fail, resulting in a huge waste of computing resources, but also significantly extend task completion time and reduce overall training and inference efficiency. More seriously, the interruption of critical business operations can lead to significant economic losses and adverse effects.

[0021] To address the aforementioned technical problems, embodiments of this application provide a fault early warning method, which can be executed by a fault early warning node, such as... Figure 2 As shown, the specific processing steps of the fault early warning method may include: Step S201: In the current cycle, obtain the latest monitoring value and alarm threshold of at least one performance parameter of the optical module, as well as the historical monitoring value of at least one performance parameter in previous cycles.

[0022] Among them, performance parameters can be transmitted optical power, received optical power, bias current, driving voltage, temperature, etc.

[0023] Specifically, to enable timely early warning of optical modules, the fault warning node can monitor the optical modules, periodically acquiring the latest monitoring values ​​of various performance parameters and storing them in a preset storage location. In the current cycle, the fault warning node can collect the latest monitoring values ​​of various performance parameters from the optical modules using a preset method, and also read historical monitoring values ​​collected in previous cycles from the preset storage location. The number of historical cycles can be multiple, and the specific number can be pre-specified by technical personnel. The preset methods can include telemetry, SecureShell (SSH) protocol, Application Programming Interface (API), Simple Network Management Protocol (SNMP), etc.

[0024] In addition, a corresponding alarm threshold can be set in the optical module for each performance parameter. The fault warning node can read the alarm threshold of the performance parameter from the optical module in advance for subsequent determination of the fault probability. Alternatively, the fault warning node can store the alarm thresholds corresponding to various performance parameters of the optical module, and then read the alarm thresholds corresponding to various performance parameters from the corresponding storage location when performing a warning operation in the current cycle.

[0025] Step S202: Determine the sub-fault probability corresponding to the target performance parameter based on the latest monitoring value of the target performance parameter, alarm threshold, historical monitoring value, and pre-built fault probability conversion model.

[0026] The target performance parameter can be any one of at least one performance parameter.

[0027] Specifically, considering a single performance parameter dimension, the latest monitoring value indicates the current performance of the optical module in that dimension, while historical monitoring values ​​indicate its past performance. In other words, the latest and historical monitoring values ​​can be used to analyze the trend changes of the optical module in that performance parameter dimension. The alarm threshold indicates the minimum acceptable value for the optical module's performance parameter, and the fault probability conversion model converts the monitoring values ​​into fault probabilities. Thus, taking a target performance parameter as an example, the fault warning node can use the fault probability conversion model to analyze the latest and historical monitoring values ​​of the target performance parameter, as well as the pre-acquired alarm threshold corresponding to the target performance parameter, to determine the sub-fault probability of the target performance parameter.

[0028] Step S203: Determine the total failure probability of the optical module based on the latest monitored value of at least one performance parameter and a pre-built failure probability prediction model.

[0029] Specifically, in order to more accurately determine whether the optical module will fail, this solution further adopts a failure probability prediction model, which considers multiple dimensions to determine whether the optical module will fail. That is, the failure warning node can input the latest monitoring values ​​of various performance parameters into the failure probability prediction model, and the failure probability prediction model will analyze and predict based on the latest monitoring values ​​of various input performance parameters and output the total failure probability.

[0030] Step S204: After determining the sub-fault probabilities corresponding to at least one performance parameter, generate fault warning information based on the sub-fault probabilities and total fault probabilities corresponding to at least one performance parameter, and use it to warn the optical module.

[0031] Specifically, after determining the sub-fault probabilities corresponding to various performance parameters in a manner similar to that described in step S202 above, the fault warning node can directly generate fault warning information by combining the sub-fault probabilities determined for each individual performance parameter dimension and the total fault probability determined from multiple dimensions. When the aforementioned fault warning node is a terminal device, it can directly display the fault warning information, allowing technicians to perform relevant maintenance operations based on the information. Alternatively, when the aforementioned fault warning node is a server, the server can forward the fault warning information to the terminal device where the technician is located, enabling the technician to perform relevant maintenance operations based on the information.

[0032] The fault warning method of this application first acquires both the latest and historical monitoring values ​​of various performance parameters of the optical module in the current cycle. Then, based on the latest and historical monitoring values ​​of the target performance parameter, as well as the alarm threshold and fault probability conversion model, the sub-fault probability of the target performance parameter is determined; that is, the probability that the optical module will fail in the performance dimension corresponding to the target performance parameter. Furthermore, considering the relationship between performance parameters, a fault probability prediction model and the latest monitoring values ​​of various performance parameters are used to determine the total fault probability. Finally, fault warning information is generated based on the sub-fault probabilities and the total fault probability of various performance parameters. In the above process, this solution considers the dimension of individual performance parameters to calculate the sub-fault probability, and also considers the correlation between different performance parameters to calculate the total fault probability. Combining the sub-fault probabilities and the total fault probability of various performance parameters, accurate fault warning information can be generated, improving the accuracy of the warning. This allows technicians to take timely corresponding measures based on the fault warning information (i.e., shorten the fault response time) to ensure the normal operation of the optical module.

[0033] In some optional implementations, the sub-fault probability of the above-mentioned target performance parameter can be one or more. When there are multiple sub-fault probabilities, the fault warning node can calculate the sub-fault probability in step S202 as follows: Step 1: Based on the alarm threshold and latest monitoring value of the target performance parameter, as well as the fault probability conversion model, determine the first sub-fault probability corresponding to the target performance parameter.

[0034] Step 2: Based on the historical monitoring values, latest monitoring values, alarm thresholds, and fault probability conversion model of the target performance parameters, determine the second sub-fault probability corresponding to the target performance parameters.

[0035] Among them, the first sub-fault probability and the second sub-fault probability together constitute the sub-fault probability of the target performance parameter.

[0036] Specifically, by comparing the latest monitored value of the target performance parameter with the alarm threshold, the absolute health level of the target performance parameter can be determined. Therefore, after comparing the alarm threshold and the latest monitored value of the target performance parameter, the fault warning node can further employ a fault probability transformation model to convert the comparison result into the first sub-fault probability.

[0037] By comparing the historical and latest monitoring values ​​of the target performance parameter, it can be determined whether the monitoring value of the target performance parameter has undergone a sudden change. Therefore, after comparing the historical monitoring values, the latest monitoring values, and the alarm threshold of the target performance parameter, the fault early warning node can use a fault probability transformation model to convert the comparison result into a second sub-fault probability.

[0038] In this way, since the health status of an optical module does not only deteriorate when its various performance parameters reach alarm thresholds, this solution uses alarm thresholds as a reference. Based on the latest monitoring values, or on a combination of historical and latest monitoring values, it calculates the probability of optical module failure, converting the failure status into a probability output. This is more accurate, allowing technicians to understand the health level or failure severity of the optical module, rather than simply providing a health or failure status as presented in related technologies. Furthermore, providing comprehensive health status information of the optical module using only probability values ​​is quite convenient.

[0039] In some optional implementations, in step one above, the fault warning node can specifically calculate the probability of the first sub-fault using the following steps: Step 1: Based on the latest monitored values ​​of the target performance parameters and the alarm threshold, determine the normalized value corresponding to the latest monitored values ​​of the target performance parameters.

[0040] Step 2: Using a fault probability conversion model, the normalized value corresponding to the latest monitored value of the target performance parameter is converted into the first sub-fault probability.

[0041] The fault probability conversion model can be a mapping relationship between normalized values ​​and probability values. The mapping relationship can be a mapping relationship table or a functional relationship.

[0042] Specifically, alarm thresholds can include a maximum alarm threshold and a minimum alarm threshold. Furthermore, the performance of optical modules from different manufacturers, batches, deployment environments, and lifecycle stages varies. Using only a constant alarm threshold to predict whether an optical module is faulty results in low accuracy. Therefore, to avoid this problem, this solution normalizes the monitored values ​​using alarm thresholds.

[0043] During the normalization process, the fault warning node can calculate the first difference between the highest and lowest alarm thresholds, and the second difference between the latest monitored value of the target performance parameter and the lowest alarm threshold. The ratio of the second difference to the first difference is then determined as the normalized value corresponding to the latest monitored value of the target performance parameter. In other words, the normalized value can be expressed as follows: (1) Where X represents the normalized value, and x represents the latest monitored value of the target performance parameter. This indicates that the monitored values ​​of the target performance parameters are normalized. This represents the lowest alarm threshold for the target performance parameter. This indicates the highest alarm threshold for the target performance parameter.

[0044] After normalization, the fault warning node can input the determined normalized value into the fault probability conversion model. The model can then convert the normalized value and output the first sub-fault probability. Specifically, if the fault probability conversion model is a mapping table, the fault warning node can find the probability value matching the normalized value in the mapping table as the first sub-fault probability. Alternatively, if the fault probability conversion model uses a multi-segment function, the fault warning node can first determine the target interval to which the normalized value belongs from multiple preset intervals, and then use the function corresponding to the target interval to calculate the first sub-fault probability corresponding to the normalized value.

[0045] In this way, by using the alarm threshold as a benchmark to normalize the latest monitoring values, the comparison is made regardless of the different standards for setting the alarm threshold or between different types of performance parameters, making it convenient to compare the health status of different performance parameters.

[0046] For example, the functional relationship can be expressed as follows: (2) in, Normalized value Let be the probability, and h be a preset hyperparameter whose value ranges from (0, 0.5). The scale factor is expressed as , The scale parameter is expressed as This parameter means that the high and low thresholds can be set in advance, and the monitored values ​​that deviate from the threshold (1-2h) / 2 can be considered to be in a 100% healthy state.

[0047] The curve indicated by formula (2) can be as follows: Figure 3 As shown, the probability value of 0.5 indicated by the dashed line can be the threshold for judging whether the performance parameter is healthy. That is, the monitoring value corresponding to a probability value less than 0.5 indicates that the optical module is in a healthy state in this performance parameter dimension.

[0048] In some optional implementations, in step two above, the fault warning node can calculate the second sub-fault probability using the following specific steps: Step 1: Input the historical monitoring values ​​of the target performance parameters into the pre-built performance parameter prediction model to obtain the predicted values ​​of the target performance parameters in the current period output by the performance parameter prediction model.

[0049] Step 2: Determine the second sub-fault probability based on the latest monitored values, predicted values, alarm thresholds, and fault probability conversion model of the target performance parameters.

[0050] Specifically, the fault early warning node can first input historical monitoring values ​​of the target performance parameter into the performance parameter prediction model. The performance parameter prediction model analyzes and predicts based on the input historical monitoring values, and then outputs the predicted value of the target performance parameter. For example, the performance parameter prediction model can be an Exponentially Weighted Moving Average (EWMA) algorithm. Since historical monitoring values ​​reflect the normal evolution trend of the target performance parameter, and the predicted value is also predicted according to this evolution trend, and the latest monitoring value reflects the actual value of the target performance parameter in the current period, the fault early warning node can compare the latest monitoring value, the predicted value, and the alarm threshold, and then use a fault probability conversion model to convert the comparison result and determine the second sub-fault probability.

[0051] refer to Figure 4 For any performance parameter, the historical monitoring value can be expressed as t represents the period identifier of the current period. For the target performance parameter, after prediction by the performance parameter prediction model, the predicted value for the current period can be obtained. .

[0052] In some optional implementations, in step 2 of step two above, the fault warning node can calculate the second sub-fault probability using the following specific steps: Step a: Determine the prediction deviation value based on the latest monitored and predicted values ​​of the target performance parameters.

[0053] Step b: Determine the normalized value corresponding to the prediction deviation value based on the prediction deviation value and the alarm threshold.

[0054] Step c involves using a fault probability conversion model to convert the normalized value corresponding to the prediction deviation value into the second sub-fault probability.

[0055] Specifically, the fault prediction node can determine the prediction deviation value as the absolute value of the difference between the latest monitored value and the predicted value of the target performance parameter. The prediction deviation value can be expressed as: , This indicates the latest monitoring value. This represents the predicted value. Similarly, the fault prediction node can determine the ratio of the prediction deviation value to the first difference (which can be the difference between the highest alarm threshold and the lowest alarm threshold) as the normalized value corresponding to the prediction deviation value. Finally, the fault prediction node can input the prediction deviation value into the fault probability conversion model. After the fault probability conversion model converts the normalized value corresponding to the prediction deviation value, it obtains the second sub-fault probability output by the fault probability conversion model (the specific process can be referred to the conversion process of the first sub-fault probability above, which will not be repeated here).

[0056] In contrast, while related technologies employ post-event alarms, this solution generates predicted values ​​from historical monitoring values. This allows us to determine the monitoring values ​​that should have been monitored in the current period based on the original development trend. The latest monitoring value is the actual value of the target performance parameter in the current period. If the actual value deviates slightly from the predicted value, it indicates that the optical module is performing normally in terms of the target performance parameter. If the actual value deviates significantly from the predicted value, it indicates that the optical module is performing abnormally in terms of the target performance parameter. By identifying the trend of change, we can determine whether there is a fault and provide early warnings, enabling technicians to intervene in a timely manner and avoid more serious fault problems.

[0057] For example, in the first four cycles, the monitored value of the target performance parameter is 0.8, and the alarm threshold is 0.5. In the fifth cycle, based on the monitored values ​​of the previous four cycles, the performance parameter prediction model predicts that the target performance parameter will be 0.8 in the current cycle, while the actual monitored value is 0.6, indicating a significant change. This suggests that the optical module is highly likely to experience a sudden failure. If the traditional threshold comparison method is used, 0.6 is considered greater than 0.5, indicating a healthy state, and no alarm will be triggered. However, this solution addresses this situation by determining the prediction deviation value and its corresponding probability value. After a probability transformation operation, a second sub-fault probability is obtained. This second sub-fault probability is used in the fault warning information generation process, thus achieving an early warning function.

[0058] In some optional implementations, in step S203 above, the fault warning node can calculate the total fault probability using the following specific steps: Step 1: Input the latest monitoring value of at least one performance parameter into the neural network model to obtain the target output vector output by the neural network model.

[0059] The target output vector may include at least two elements.

[0060] Step 2: Determine the total failure probability based on the failure probability calculation model and the elements included in the target output vector.

[0061] Specifically, refer to Figure 5 A neural network model can include an input layer, hidden layers, and an output layer. After normalizing the latest monitoring values ​​of various performance parameters, the fault warning node sequentially inputs these values ​​into the input layer, hidden layer, and output layer to obtain the target output vector from the output layer. For example, it can be represented as... .

[0062] After obtaining the target output vector from the neural network model, the fault warning node can input the elements included in the target output vector into the fault probability calculation model. The fault probability calculation model calculates the total fault probability by processing each element in the target output vector. When the target output vector is a two-dimensional vector, it can be represented as... Accordingly, the failure probability calculation model can be expressed as follows: (3) in, This represents the total failure probability. and These are all elements in the target output vector.

[0063] Conversely, when it is necessary to calculate the probability of health, it can be calculated as follows: (4) in, For the probability of health, .

[0064] In this way, the different performance characteristics of the optical modules affect each other. The neural network model can combine fully connected linear algorithms and nonlinear activation functions to fit the input features and learn the correlation between different performance parameters, thereby outputting a target output vector. Since the target output vector contains the correlation between different performance parameters, the total failure probability determined based on the target output vector is more accurate.

[0065] In some optional implementations, in step S204 above, the fault warning node can generate fault warning information using the following specific steps: Step 1: Determine the final failure probability of the optical module based on the total failure probability and the sub-failure probabilities corresponding to at least one performance parameter.

[0066] Step 2: Generate fault warning information corresponding to the final fault probability based on the final fault probability.

[0067] Specifically, such as Figure 6 As shown, in order to more easily determine the health status of the optical module, the fault warning node can compare the total fault probability and the sub-fault probabilities corresponding to various performance parameters, and select one probability value from these multiple probability values ​​as the final fault probability of the optical module. This single probability value indicates the health status of the optical module. Furthermore, by using the final fault probability as the fault warning information, technicians can easily determine whether the optical module needs to be maintained and adjusted.

[0068] For example, a fault warning node can be selected as the final fault probability from the total fault probability and the sub-fault probabilities corresponding to at least one performance parameter, which can be expressed as: (5) in, and Both can be arrays. Each element in the array can be the first sub-fault probability of a performance parameter. Each element in the array can be a second sub-fault probability, which is a performance parameter. This represents the total failure probability. The maximum failure probability could be... The sub-fault probability of a certain performance parameter can also be The sub-fault probability of a certain performance parameter can also be the total fault probability.

[0069] In this way, by calculating multiple probability values ​​and selecting the highest probability value as the final failure probability, the fault warning node can issue warnings in a more conservative manner, allowing for early maintenance of the optical module and preventing optical module failures. Furthermore, both the two different sub-fault probabilities and the total failure probability represent the health status of the optical module from different perspectives. Selecting the probability value from these allows the fault warning node to identify the probability value corresponding to the factor with the greatest impact on the health status of the optical module, resulting in a more accurate determination of the probability value and enabling precise early warning actions to be taken for the optical module.

[0070] In some optional implementations, when the fault warning node is a terminal device, the fault warning node can generate fault warning information corresponding to the final fault probability based on the final fault probability. For example, the fault warning node can determine whether the final fault probability is greater than a preset fault probability threshold. If so, an alarm flag can be added to the fault warning information for display to perform an alarm operation on the optical module. If not, only the fault warning information can be displayed to prompt the technicians for judgment.

[0071] In this way, alarms can be issued in an emergency, which can promptly instruct technicians to resolve the faults in the optical module and avoid affecting its normal operation.

[0072] In some optional implementations, the alarm threshold for the aforementioned target performance parameter can be obtained in the following manner: Different performance parameters reflect the optical module's performance in various aspects, and these performance characteristics gradually change over time. Using a fixed alarm threshold consistently would lead to frequent false alarms, even though the optical module could actually function normally. Therefore, to address the issue of inaccurate fault warnings caused by these changes, the fault warning node can update the alarm threshold for the target performance parameter every preset number of cycles (the preset number can be determined based on the target performance parameter). Specifically, when the total number of target operating cycles corresponding to the current cycle reaches an integer multiple of the preset number, the fault warning node can obtain an alarm threshold table corresponding to the target performance parameter. This alarm threshold table can include alarm thresholds corresponding to multiple total operating cycles. Accordingly, the fault warning node can determine the alarm threshold corresponding to the total number of target operating cycles from the alarm threshold table and set it as the alarm threshold for the target performance parameter.

[0073] In this way, different performance parameters correspond to different alarm threshold tables, and the alarm thresholds are updated every certain period of time. This allows subsequent fault warning operations to be based on the latest alarm thresholds, which are related to the usage time of the optical module and can more accurately reflect the fault status of the optical module. Therefore, the final fault probability is more accurate, which means that it is possible to more accurately determine whether the optical module has failed.

[0074] 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.

[0075] Embodiments of this application also provide a fault early warning device, such as... Figure 7 As shown, it includes: The acquisition module 710 is used to acquire, in the current cycle, the latest monitoring value and alarm threshold of at least one performance parameter of the optical module, as well as the historical monitoring value of at least one performance parameter in previous cycles. The determination module 720 is used to determine the sub-fault probability corresponding to the target performance parameter based on the latest monitoring value, alarm threshold, historical monitoring value, and pre-built fault probability conversion model of the target performance parameter, wherein the target performance parameter is any one of at least one performance parameter; and to determine the total fault probability of the optical module based on the latest monitoring value of at least one performance parameter and the pre-built fault probability prediction model. The generation module 730 is used to generate fault warning information based on the sub-fault probabilities corresponding to at least one performance parameter and the total fault probability after determining the sub-fault probabilities corresponding to at least one performance parameter, so as to provide warnings to the optical module.

[0076] In some alternative implementations, the determining module 720 is specifically used for: Based on the alarm threshold and latest monitoring value of the target performance parameter, as well as the fault probability conversion model, determine the first sub-fault probability corresponding to the target performance parameter; Based on the historical monitoring values, latest monitoring values, alarm thresholds, and fault probability conversion model of the target performance parameters, determine the second sub-fault probability corresponding to the target performance parameters; Here, the first sub-fault probability and the second sub-fault probability are both sub-fault probabilities of the target performance parameter.

[0077] In some alternative implementations, the determining module 720 is specifically used for: Based on the latest monitored values ​​and alarm thresholds of the target performance parameters, determine the normalized values ​​corresponding to the latest monitored values ​​of the target performance parameters; A fault probability conversion model is used to convert the normalized value corresponding to the latest monitored value of the target performance parameter into the first sub-fault probability.

[0078] In some alternative implementations, the determining module 720 is specifically used for: The historical monitoring values ​​of the target performance parameters are input into the pre-built performance parameter prediction model to obtain the predicted values ​​of the target performance parameters in the current period. The second sub-fault probability is determined based on the latest monitored values, predicted values, alarm thresholds, and fault probability conversion model of the target performance parameters.

[0079] In some alternative implementations, the determining module 720 is specifically used for: The prediction deviation value is determined based on the latest monitored and predicted values ​​of the target performance parameters; Based on the prediction deviation value and the alarm threshold, determine the normalized value corresponding to the prediction deviation value; A fault probability conversion model is used to convert the normalized value corresponding to the prediction deviation value into the second sub-fault probability.

[0080] In some optional implementations, the fault probability prediction model includes a neural network model and a fault probability calculation model; the determination module 720 is specifically used for: The latest monitored value of at least one performance parameter is input into a neural network model to obtain a target output vector output by the neural network model, wherein the target output vector includes at least two elements; The total failure probability is determined based on the failure probability calculation model and the elements included in the target output vector.

[0081] In some alternative implementations, the generation module 730 is specifically used for: The final failure probability of the optical module is determined based on the total failure probability and the sub-failure probabilities corresponding to at least one performance parameter. Based on the final failure probability, generate a fault warning message corresponding to the final failure probability.

[0082] In some alternative implementations, the generation module 730 is specifically used for: The highest failure probability is selected from the total failure probability and the sub-failure probabilities corresponding to at least one performance parameter as the final failure probability.

[0083] In some optional implementations, the fault probability conversion model uses the following expression:

[0084] in, Normalized value Let be the probability, and h be a preset hyperparameter whose value ranges from (0, 0.5). The scale factor is expressed as , The scale parameter is expressed as .

[0085] For a description of the features in the embodiment corresponding to the fault warning device, please refer to the relevant description of the embodiment corresponding to the fault warning method, which will not be repeated here.

[0086] Embodiments of this application also provide an electronic device, such as... Figure 8 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 fault warning method embodiments.

[0087] 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 fault warning method embodiments when running.

[0088] 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.

[0089] 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 fault warning method embodiments.

[0090] 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 fault warning method embodiments.

[0091] 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.

[0092] The foregoing has provided a detailed description of a fault 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 its core ideas. 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 fault early warning method, characterized in that, include: In the current cycle, the latest monitoring value and alarm threshold of at least one performance parameter of the optical module are obtained, as well as the historical monitoring value of at least one of the performance parameters in previous cycles. Based on the latest monitoring value, alarm threshold, historical monitoring value of the target performance parameter, and the pre-built fault probability conversion model, the sub-fault probability corresponding to the target performance parameter is determined, wherein the target performance parameter is any one of at least one of the performance parameters. The total failure probability of the optical module is determined based on the latest monitored value of at least one of the performance parameters and a pre-built failure probability prediction model. After determining the sub-fault probabilities corresponding to at least one of the performance parameters, fault warning information is generated based on the sub-fault probabilities corresponding to at least one of the performance parameters and the total fault probability, in order to provide a warning to the optical module.

2. The fault early warning method according to claim 1, characterized in that, The step of determining the sub-fault probability corresponding to the target performance parameter based on the latest monitored value, alarm threshold, historical monitored value, and pre-built fault probability conversion model of the target performance parameter includes: Based on the alarm threshold and latest monitoring value of the target performance parameter, and the fault probability conversion model, determine the first sub-fault probability corresponding to the target performance parameter; Based on the historical monitoring values, latest monitoring values, alarm thresholds, and fault probability conversion model of the target performance parameter, determine the second sub-fault probability corresponding to the target performance parameter; Wherein, the first sub-fault probability and the second sub-fault probability are both sub-fault probabilities of the target performance parameter.

3. The fault early warning method according to claim 2, characterized in that, The step of determining the first sub-fault probability corresponding to the target performance parameter based on the alarm threshold and the latest monitoring value of the target performance parameter, and the fault probability conversion model, includes: Based on the latest monitoring value and alarm threshold of the target performance parameter, determine the normalized value corresponding to the latest monitoring value of the target performance parameter; Using the aforementioned fault probability conversion model, the normalized value corresponding to the latest monitored value of the target performance parameter is converted into the first sub-fault probability.

4. The fault early warning method according to claim 2 or 3, characterized in that, The step of determining the second sub-fault probability corresponding to the target performance parameter based on the historical monitoring value, the latest monitoring value, the alarm threshold, and the fault probability conversion model of the target performance parameter includes: The historical monitoring values ​​of the target performance parameter are input into the pre-built performance parameter prediction model to obtain the predicted value of the target performance parameter in the current period output by the performance parameter prediction model. The second sub-fault probability is determined based on the latest monitored value, predicted value, alarm threshold of the target performance parameter, and the fault probability conversion model.

5. The fault early warning method according to claim 4, characterized in that, The step of determining the second sub-fault probability based on the latest monitored value, predicted value, alarm threshold of the target performance parameter, and the fault probability conversion model includes: The prediction deviation value is determined based on the latest monitored and predicted values ​​of the target performance parameters; Based on the prediction deviation value and the alarm threshold, a normalized value corresponding to the prediction deviation value is determined; Using the aforementioned fault probability conversion model, the normalized value corresponding to the prediction deviation value is converted into the second sub-fault probability.

6. The fault early warning method according to any one of claims 1 to 3, characterized in that, The failure probability prediction model includes a neural network model and a failure probability calculation model; determining the total failure probability of the optical module based on the latest monitored value of at least one of the performance parameters and the pre-built failure probability prediction model includes: The latest monitored value of at least one of the performance parameters is input into the neural network model to obtain the target output vector output by the neural network model, wherein the target output vector includes at least two elements; The total failure probability is determined based on the failure probability calculation model and the elements included in the target output vector.

7. The fault early warning method according to any one of claims 1 to 3, characterized in that, After determining the sub-fault probabilities corresponding to at least one of the performance parameters, the step of generating fault warning information based on the sub-fault probabilities corresponding to at least one of the performance parameters and the total fault probability includes: The final failure probability of the optical module is determined based on the total failure probability and the sub-failure probabilities corresponding to at least one of the performance parameters. Based on the final failure probability, generate a fault warning message corresponding to the final failure probability.

8. The fault early warning method according to claim 7, characterized in that, Determining the final failure probability of the optical module based on the total failure probability and the sub-failure probabilities corresponding to at least one of the performance parameters includes: The highest failure probability is selected from the total failure probability and the sub-failure probabilities corresponding to at least one of the performance parameters as the final failure probability.

9. The fault early warning method according to claim 3, characterized in that, The fault probability conversion model uses the following expression: in, The normalized value, Let be the probability, and h be a preset hyperparameter whose value ranges from (0, 0.5). The scale factor is expressed as , The scale parameter is expressed as .

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the fault warning method as described in any one of claims 1 to 9 when executing the computer program.

Images