Search

CN-122027502-A - Hidden ageing risk pre-judging method and system for light cat equipment

CN122027502ACN 122027502 ACN122027502 ACN 122027502ACN-122027502-A

Abstract

The invention provides a hidden ageing risk pre-judging method and system for light cat equipment, comprising the following steps of collecting a plurality of core operation indexes of the light cat equipment, constructing a feature vector representing the health state of the equipment, calculating an anomaly score, namely inputting the feature vector into a pre-trained unsupervised anomaly detection model, calculating to obtain an equipment health anomaly score, constructing a time sequence, screening risk equipment, namely screening potential risk equipment with the score not exceeding a dominant fault threshold value and in a monotonically increasing situation of the sequence based on the time sequence, calculating the global average change rate of the time sequence of the potential risk equipment as ageing trend intensity of the equipment, and identifying high-risk hidden ageing equipment according to the ageing trend intensity. The method aims at converting the operation and maintenance mode from passive response to active early warning, and realizes early discovery and accurate intervention on the recessive ageing equipment.

Inventors

  • LIN LIWEN
  • XIE JIE
  • HUANG HONGJIAN
  • ZHENG SHENGHUA
  • CHEN SHUXI

Assignees

  • 中邮科通信技术股份有限公司

Dates

Publication Date
20260512
Application Date
20260205

Claims (8)

  1. 1. The hidden ageing risk pre-judging method for the light cat equipment is characterized by comprising the following steps of: A feature vector construction step of collecting a plurality of core operation indexes of the cat equipment and constructing a feature vector representing the health state of the equipment; the abnormal score calculating step, namely inputting the feature vector into a pre-trained unsupervised abnormal detection model, and calculating to obtain the equipment health abnormal score; Continuously calculating the equipment health anomaly score in a fixed period to form an equipment health risk score time sequence; Screening potential risk equipment with scores not exceeding a dominant fault threshold value and in a monotonically increasing state based on the time sequence; calculating the global average change rate of the time sequence of the potential risk equipment as the aging trend intensity of the equipment; And a risk pre-judging step, namely identifying high-risk hidden ageing equipment according to the ageing trend intensity.
  2. 2. The method for pre-judging hidden ageing risk of a photo cat device according to claim 1, wherein the core operation index comprises at least five of CPU occupancy, memory occupancy, device temperature, transmission light power and CRC error rate.
  3. 3. The method for pre-judging hidden ageing risk of a photo-cat device according to claim 1, wherein the unsupervised anomaly detection model is an isolated forest model which is trained by using historical data of the photo-cat device with known health, and the device health anomaly score is calculated based on the path length of the device to be evaluated in a model isolation tree.
  4. 4. The method for pre-judging hidden ageing risk of a photo cat device according to claim 1, wherein the step of screening the risk device specifically comprises: Screening out equipment with all scores lower than a preset dominant fault threshold value in the time sequence; And (3) relative trend criteria, further screening out equipment with monotonous increasing time sequence overall in the screening result to form a subset of potential risk equipment.
  5. 5. The method for pre-judging hidden ageing risk of a light cat device according to claim 1, wherein the global average change rate is obtained by calculating a mean value of first-order discrete differences of the time sequence, and the calculation formula is as follows: Wherein s= [ S 1 ,S 2 ,…,S t ] is a time sequence, t is a sequence length, and a is an aging trend intensity.
  6. 6. The method for pre-judging hidden ageing risk of a photo cat device according to claim 1, wherein the step of pre-judging risk comprises the steps of arranging the potential risk devices in descending order according to the ageing trend intensity of the potential risk devices, and selecting Top-N devices which are ranked at the front to be identified as high-risk hidden ageing devices.
  7. 7. The method for pre-determining hidden ageing risk of a light cat device according to claim 1, further comprising the steps of model persistence and heat loading: model persistence, namely serializing the unsupervised anomaly detection model object after training, and storing the model object to a disk; And model hot loading, namely loading the saved model files through deserialization when the system is started or the model needs to be updated, so that the quick multiplexing of the model is realized.
  8. 8. A hidden ageing risk pre-judging system for a photo cat device, configured to implement a hidden ageing risk pre-judging method for a photo cat device according to any one of claims 1 to 7, comprising: The collecting layer is used for collecting operation index data of the cat equipment; the learning training layer is used for training an unsupervised anomaly detection model based on the acquired data; The evaluation detection layer is used for calculating the health anomaly score of the equipment by using the trained model and executing risk prejudging logic; and the display layer is used for generating and outputting a hidden ageing risk pre-judging report.

Description

Hidden ageing risk pre-judging method and system for light cat equipment Technical Field The invention relates to the technical field of IT and software development, in particular to a hidden ageing risk pre-judging method and system for light cat equipment. Background The optical modem is used as a user side terminal of the optical broadband network, and the stability of long-term operation of the optical modem is a key for guaranteeing user experience. Currently, the equipment operation and maintenance system of telecom operators is almost completely built on the paradigm of "dominant fault diagnosis", and is embodied in two mature but limited modes: static threshold alert mechanism The network management system presets a static threshold of the key performance index. The method can trigger an alarm only when the equipment index is deteriorated to obvious symptoms, and can not capture the slow performance decline trend of the equipment in early aging, so that the method has no prediction capability on hidden aging risks. Periodic batch change strategy And replacing the cutter according to the theoretical service life of the equipment. The strategy completely ignores the difference of the device individuals in the operation environment and the use load, so that a large number of seriously aged devices cannot be replaced in time due to 'short of the year', and meanwhile, a plurality of devices with good states are replaced prematurely, so that huge waste of operation and maintenance resources is caused. With the popularization of high-quality internet applications, the tolerance of users to slight quality differences such as instantaneous network jamming and video buffering is becoming lower and lower. The root of these problems is often the implicit aging of the light cat, i.e. the performance of the core components of the device (such as the optical module, the motherboard capacitor and the heat dissipation system) has already begun to decline, so that the key indexes (such as the error rate, the temperature baseline and the transmitted optical power) of the light cat show irreversible degradation trend, but the instantaneous value of the light cat does not break through the traditional operation and maintenance threshold value. The prior art includes complex multidimensional analysis models, and the technical center of gravity is used for diagnosing and positioning abnormal quality differences or quality differences reaching a severity degree. In other words, they can answer "which device is bad", but simply fail to answer the more prospective question of "which device is about to get bad". Therefore, the prior art system has a clear blank that an effective means capable of early and accurately predicting the hidden ageing risk of the photo cat is lacked. Disclosure of Invention Therefore, the invention aims to provide a hidden ageing risk pre-judging method and system for light cat equipment, which aim to change an operation and maintenance mode from passive response to active early warning so as to realize early discovery and accurate intervention of the hidden ageing equipment. In order to achieve the purpose, the invention adopts the following technical scheme that the hidden ageing risk pre-judging method of the light cat equipment comprises the following steps: A feature vector construction step of collecting a plurality of core operation indexes of the cat equipment and constructing a feature vector representing the health state of the equipment; the abnormal score calculating step, namely inputting the feature vector into a pre-trained unsupervised abnormal detection model, and calculating to obtain the equipment health abnormal score; Continuously calculating the equipment health anomaly score in a fixed period to form an equipment health risk score time sequence; Screening potential risk equipment with scores not exceeding a dominant fault threshold value and in a monotonically increasing state based on the time sequence; calculating the global average change rate of the time sequence of the potential risk equipment as the aging trend intensity of the equipment; And a risk pre-judging step, namely identifying high-risk hidden ageing equipment according to the ageing trend intensity. In a preferred embodiment, the core operation index includes at least five of CPU occupancy, memory occupancy, device temperature, transmit optical power, and CRC error rate. In a preferred embodiment, the unsupervised anomaly detection model is an isolated forest model trained using known health photo-cat equipment history data, and the equipment health anomaly score is calculated based on the path length of the equipment to be evaluated in the model isolation tree. In a preferred embodiment, the risk device screening step specifically includes: Screening out equipment with all scores lower than a preset dominant fault threshold value in the time sequence; And (3) relative trend criteria, further screening out equi