CN-121997048-A - Broadcast television network fault prediction and self-healing method based on AI

CN121997048ACN 121997048 ACN121997048 ACN 121997048ACN-121997048-A

Abstract

The invention relates to the technical field of learning methods, in particular to an AI-based broadcast television network fault prediction and self-healing method, which comprises the following steps: and extracting a channel quality difference value through edge detection, recombining the channel quality difference value to form a change trend set, performing differential analysis on a three-point window, identifying a confidence interval compression trend, extracting a prediction and detection symbol relation, identifying an error inversion direction, and finally generating a repair trigger signal label set. According to the invention, a dynamic trend structure of a prediction difference value is constructed, a trigger tag index is extracted by combining symbol consistency and direction reversal characteristics, aggregation and tag binding of time sequence characteristics are completed, a training sample set for learning is generated, the association recognition capability among the characteristics is enhanced, the perception precision of weak change signals is improved, the early warning range is expanded, abnormal misjudgment is reduced, a judgment basis based on trend evolution is formed, and reliable data support is provided for fault prediction and self-healing response.

Inventors

DENG HAITAO
Wo Shengwei

Assignees

山东广播电视台

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. The broadcast television network fault prediction and self-healing method based on the AI is characterized by comprising the following steps: S1, acquiring channel quality detection contents of continuous time points through edge detection in a broadcast television network, extracting difference values of prediction data and upper and lower limits of corresponding confidence intervals, and recombining the difference values according to time sequence to construct a prediction change trend set; S2, based on the predicted change trend set, selecting the difference value of every three time points to construct an analysis window, executing front-back difference on the difference value in the window, judging whether symbols are continuous and consistent, and generating a confidence interval compression trend feature; S3, extracting channel quality detection contents of a time point corresponding to the predicted data based on the predicted change trend set, comparing the two symbols, identifying whether symbols at adjacent moments are inverted or not, and constructing an error inversion direction characteristic; s4, synchronously comparing characteristic states according to time indexes based on the confidence interval compression trend characteristic and the error inversion direction characteristic, screening index positions meeting contraction and inversion conditions simultaneously, and generating a repair trigger signal label set; s5, selecting the prediction data in the prediction change trend set and the labels in the repair trigger signal label set to bind, constructing an input and output field pair, dividing training and verification data, and constructing a broadcast television network fault prediction and self-healing learning sample structure.
2. The AI-based broadcast television network fault prediction and self-healing method according to claim 1, wherein the prediction change trend set comprises a difference change amplitude, a difference change direction and a difference time sequence structure, the confidence interval compression trend feature comprises a symbol continuity mark, a trend direction consistency mark and a difference stability feature, the error reverse direction feature comprises a symbol deviation feature, a direction switching point and a reverse trend mark, the repair trigger signal tag set comprises a synchronization index position, a trend intersection mark and a trigger state tag, and the broadcast television network fault prediction and self-healing learning sample structure comprises a prediction data input feature, a repair trigger output tag and a time sequence division configuration.
3. The AI-based broadcast television network fault prediction and self-healing method according to claim 1, wherein the simultaneous satisfaction of the shrinkage and inversion conditions means that the signs of upper and lower limit differences of confidence intervals are continuously consistent at the time index to form a shrinkage trend, and the sign of the prediction error at the moment is inverted compared with the sign at the previous moment.
4. The AI-based broadcast television network failure prediction and self-healing method according to claim 1, wherein the identifying whether the sign of adjacent time is inverted is performed by determining the sign of the difference between the predicted value and the actual channel quality at two consecutive time points, and if the sign of the current time is opposite to the sign of the previous time, the sign is determined to be inverted.
5. The AI-based broadcast television network fault prediction and self-healing method according to claim 1, wherein the specific steps of S1 are: s101, collecting a plurality of continuous time point channel quality detection data frames of edge detection in a broadcast television network, respectively carrying out difference operation on a predicted value of each time point and the upper limit and the lower limit of a corresponding confidence interval, and sorting the obtained differences according to time sequence to generate a confidence interval difference sequence set; S102, calling the time point difference value in the confidence interval difference value sequence set, and carrying out sequence reconstruction processing in a time sequence through a sliding window merging and time index alignment mode to obtain a sequence matrix of a continuous structure, so as to generate a difference value sequence reconstruction matrix; S103, calculating the change rate according to the numerical change trend of the time points in the difference sequence reconstruction matrix, presetting a change rate threshold value, dividing and marking intervals, and aggregating trend intervals to obtain a predicted change trend set.
6. The AI-based broadcast television network fault prediction and self-healing method according to claim 1, wherein the specific steps of S2 are: S201, based on the predicted change trend set, selecting difference content corresponding to every three continuous time points to construct an analysis window, combining the analysis window into a difference triplet sequence according to time sequence, and carrying out index arrangement on the sequence to generate a difference sliding window sequence group; S202, calling a difference sequence in a window in the difference sliding window sequence group, sequentially executing front-back difference calculation between adjacent numerical values, constructing an adjacent difference change sequence according to a difference result, classifying and aggregating according to time indexes, and generating a continuous difference change sequence set; And S203, according to the continuous differential change sequence set, searching symbol information of adjacent items in the differential sequence, judging whether continuous paragraphs with consistent symbols exist in the same sequence, marking the continuous paragraphs as stable trend sections if a preset continuous consistent condition is met, and obtaining the compression trend characteristics of the confidence interval after aggregating all the marked sections.
7. The AI-based broadcast television network fault prediction and self-healing method according to claim 1, wherein the specific steps of S3 are: S301, based on the prediction change trend set, extracting prediction data corresponding to each time point and a channel quality detection data frame, matching according to time indexes, then comparing numerical symbol attributes of the two types of data, and integrating comparison results according to time sequence to generate a prediction detection symbol comparison sequence; S302, calling the predicted detection symbol comparison sequence, performing difference recognition on the symbols of the predicted data and the detection data in the time point comparison result, marking the symbols as direction difference items if the symbols are inconsistent, and recording all differences in time sequence to generate a direction difference marking sequence; s303, comparing the difference sign marks of adjacent time points item by item according to the direction difference sign sequence, identifying whether the sign change has inversion characteristics, extracting change fragments and performing sequence aggregation processing if the inversion relation is identified, so as to obtain error inversion direction characteristics.
8. The AI-based broadcast television network fault prediction and self-healing method according to claim 1, wherein the specific steps of S4 are: S401, extracting time indexes of the two types of features based on the confidence interval compression trend features and the error reversal direction features, constructing a joint index comparison table according to an index sequence, and respectively sorting feature state values at corresponding positions to generate an index feature mapping matrix; s402, calling the index feature mapping matrix, performing logic comparison operation on the compression trend state value and the inversion direction state value of each index position, marking the index positions when the state combination of the two accords with the bidirectional consistency condition, and aggregating all positions meeting the condition to generate a consistency matching index set; S403, according to the consistency matching index set, carrying out marking and coding operation on all index positions in the set, endowing each index with a corresponding label value, and constructing a label data frame structure to obtain a repair trigger signal label set.
9. The AI-based broadcast television network fault prediction and self-healing method according to claim 1, wherein the specific steps of S5 are: S501, extracting all predicted data as characteristic sources based on the predicted change trend set, searching index positions in the repair trigger signal label set, performing field binding on the corresponding predicted data and label values, and establishing a corresponding combination relation to generate a predicted label pairing structure set; s502, calling the predicted tag pairing structure set, sorting according to the time index of predicted data, dividing a training time period and a verification time period according to a sequence order, storing predicted values and tag pairs in a section into independent structures respectively, and generating a section division data pair set; and S503, constructing a mapping relation between an input characteristic value matrix and a label vector according to the structural content of the training section and the verification section in the interval division data pair set, and combining and packaging the data units into a unified format to obtain a broadcast television network fault prediction and self-healing learning sample structure.
10. The AI-based broadcast television network fault prediction and self-healing method of claim 9, wherein the joint encapsulation is to unify data units in a format, wherein the prediction feature matrix divided by time and the corresponding label vector are combined in a one-to-one correspondence manner, and are arranged into a unified input-output format which can be identified by a model.

Description

Broadcast television network fault prediction and self-healing method based on AI Technical Field The invention relates to the technical field of learning methods, in particular to an AI-based broadcast television network fault prediction and self-healing method. Background The technical field of learning methods mainly relates to research and realization of algorithms, methods and strategies of artificial intelligent models, particularly neural network models in the training process, and core matters comprise key contents such as model parameter adjustment, training data processing, loss function optimization, learning rate control, overfitting prevention and control and the like. The field covers various machine learning paradigms such as supervised learning, unsupervised learning, reinforcement learning and the like, focuses on researching how to enable a computing model to have self-adjusting capability through historical data or simulation environment, thereby improving the adaptability and prediction capability of the computing model to new data, is widely applied to various technical directions such as voice recognition, image recognition, natural language processing, intelligent control, fault diagnosis and the like, and is tightly combined with the fields such as computer science, control engineering, statistics and the like to form a interdisciplinary technical system. The traditional broadcast television network fault prediction and self-healing method is a technical scheme for analyzing possible fault phenomena in the operation process of a broadcast television network system and taking countermeasures after faults occur to recover the normal operation of the system, generally relies on a static rule base and expert experience to perform fault judgment, monitors the network operation state through a manually set threshold detection means, and performs processing flows such as local restarting, path switching or equipment replacement and the like by means of manual inspection or a preset flow after faults occur so as to complete fault positioning and system recovery. The existing broadcast television network fault processing mode relies on a fixed threshold value to carry out state judgment, details fluctuation in the channel quality change process can not be perceived, fault identification is highly dependent on specific rule matching, state evolution under different operation scenes is difficult to dynamically reflect, the processing flow is mainly based on manual inspection and static means, an early response mechanism to abnormal evolution trend is lacking, prediction capability is limited to simple mapping of a known mode, key signals are difficult to capture in time at the early stage of fault evolution, a systematic evaluation basis for abnormal states is not formed, response is lagged when facing a changeable network environment, fault positioning and response measures lack of data-driven accurate guidance, the system is wholly lack of a sustainable update and adaptive novel feedback path, and intelligent and prepositive network maintenance requirements are difficult to support. Disclosure of Invention In order to achieve the above purpose, the present invention adopts the following technical scheme, and an AI-based broadcast television network fault prediction and self-healing method comprises the following steps: S1, acquiring channel quality detection contents of continuous time points through edge detection in a broadcast television network, extracting difference values of prediction data and upper and lower limits of corresponding confidence intervals, and recombining the difference values according to time sequence to construct a prediction change trend set; S2, based on the predicted change trend set, selecting the difference value of every three time points to construct an analysis window, executing front-back difference on the difference value in the window, judging whether symbols are continuous and consistent, and generating a confidence interval compression trend feature; S3, extracting channel quality detection contents of a time point corresponding to the predicted data based on the predicted change trend set, comparing the two symbols, identifying whether symbols at adjacent moments are inverted or not, and constructing an error inversion direction characteristic; s4, synchronously comparing characteristic states according to time indexes based on the confidence interval compression trend characteristic and the error inversion direction characteristic, screening index positions meeting contraction and inversion conditions simultaneously, and generating a repair trigger signal label set; s5, selecting the prediction data in the prediction change trend set and the labels in the repair trigger signal label set to bind, constructing an input and output field pair, dividing training and verification data, and constructing a broadcast television network fault prediction and self-