Carrier module fault prediction system based on big data mining

Abstract

The present invention relates to the field of carrier fault prediction technology, and in particular, to a carrier module fault prediction system based on big data mining. It includes a multi-source data acquisition module, a dynamic interference analysis module, a fault correlation modeling module, and a prediction decision output module. The multi-source data acquisition module collects carrier waveform data, equipment working status data, and operating environment data in real time, and generates standardized data packets; the dynamic interference analysis module extracts spatiotemporal coupling features through a dynamic coupling algorithm, identifies interference patterns in combination with an environmental perception mechanism, and triggers an incremental learning optimization model based on a new type of interference, outputting an interference feature analysis report containing feature data, pattern judgment, and optimization suggestions. The present invention realizes dynamic coupling analysis of time-frequency domain features, solves the problem that traditional static models cannot identify new types of interference, significantly improves the accuracy and timeliness of 5G base station carrier module fault prediction, and has important engineering application value.

Description

Technical Field

[0001] The present invention relates to the technical field of carrier fault prediction, and in particular to a carrier module fault prediction system based on big data mining. Background Art

[0002] With the rapid deployment of 5G networks, the stability of base station carrier modules has become a key factor in ensuring communication quality. Traditional operation and maintenance models rely primarily on static threshold alarms and regular maintenance, making it difficult to effectively address potential failures in complex operating environments. To address this issue, existing technologies generally use LSTM-based time series prediction models to identify abnormal patterns by analyzing historical operating data, enabling monitoring and early warning of the operating status of carrier modules. However, in actual 5G base station operation and maintenance scenarios, transient anomalies in RF parameters caused by sudden network interference (such as surges in user traffic and conflicts between adjacent base station signals) still pose a severe challenge to existing prediction systems.

[0003] While existing technologies have achieved the prediction of common interference patterns through historical data analysis, they still have significant shortcomings in identifying new interference patterns. Specifically, when new interference patterns not covered by training data appear in the network (such as special signal conflicts caused by the access of new terminals), the prediction accuracy of existing models will drop significantly. Furthermore, the system needs to collect data again and complete full model training. This process often results in response delays of several weeks, which cannot meet the strict requirements of real-time operation and maintenance of 5G networks, seriously affecting the operating efficiency and maintenance costs of base stations. Summary of the Invention

[0004] The present invention provides a carrier module fault prediction system based on big data mining. The system generates an interference feature analysis report containing spatiotemporal coupling characteristics, interference pattern determination, and model optimization suggestions based on carrier waveform data, equipment operating status data, and operating environment data through a dynamic interference analysis module. The report is converted into an executable prediction decision through a fault correlation modeling module, thereby solving the problems raised in the above background technology, namely:

[0005] When new types of perturbations not covered by the training data are introduced into the network, the prediction accuracy of existing models will drop significantly.

[0006] To achieve the above objectives, the carrier module fault prediction system includes a multi-source data acquisition module, which collects carrier waveform data, equipment working status data, and operating environment data from the carrier module to generate a standardized data packet, and also includes:

[0007] A dynamic interference analysis module that receives standardized data packets and performs in-depth processing on them;

[0008] The dynamic interference analysis module extracts spatiotemporal coupling features from carrier waveform data using a dynamic coupling algorithm based on the equipment working status data;

[0009] The dynamic interference analysis module identifies new interference patterns from carrier signals based on spatiotemporal coupling characteristics and in combination with operating environment data and generates interference pattern determinations;

[0010] The dynamic interference analysis module triggers an incremental learning process based on new interference patterns to optimize the model and generate model optimization suggestions;

[0011] The dynamic interference analysis module generates an interference feature analysis report including spatiotemporal coupling features, interference pattern determination, and model optimization suggestions, and passes it to the fault correlation modeling module for fault prediction of the 5G base station carrier module.

[0012] In the above technical solution, the selection of carrier waveform data, equipment working status data, and operating environment data as the trinity analysis basis is a targeted improvement to the defects of isolated data analysis in the existing technology. If only carrier waveform data and equipment working status data are collected, the system will not be able to distinguish between inherent equipment failures and performance fluctuations caused by environmental factors. For example, it is impossible to tell whether the abnormality of high-frequency signals is caused by aging of the power amplifier or the influence of high temperature environment. If only carrier waveform and environmental data are analyzed, it is difficult to establish the correlation between interference characteristics and specific hardware components, resulting in a lack of targeted maintenance recommendations. This solution realizes full-dimensional modeling of "signal characteristics-hardware status-environmental background" through the spatiotemporal synchronous collection of three-source data.

[0013] On this basis, the dynamic interference analysis module adopts a dynamic coupling algorithm to construct a time-frequency joint feature matrix, and models the association between the time domain fluctuation and frequency domain distribution of the carrier signal through the dynamic weighted time-frequency joint feature matrix.

[0014] In another technical solution, the dynamic interference analysis module dynamically adjusts the sensitivity weights of the feature extraction algorithm to the features of different frequency bands by monitoring the real-time changes of the power amplifier efficiency index, so as to optimize the extraction process of the spatiotemporal coupling features.

[0015] This technical solution is unable to capture the correlation between signal transient characteristics and spectral anomalies, as existing technologies often process time domain fluctuations and frequency domain distribution independently. For example, sudden interference may manifest as both a sudden change in time domain amplitude and energy diffusion in a specific frequency band. Traditional methods will split the analysis and miss the detection. The design of dynamically adjusting the frequency band sensitivity weights addresses the fatal weakness of fixed parameter feature extraction: when the power amplifier efficiency fluctuates, traditional fixed threshold feature extraction will simultaneously amplify noise and miss real anomalies. This solution associates device status with feature extraction parameters in real time, enabling the system to automatically strengthen feature extraction in abnormal frequency bands and weaken noise interference in normal frequency bands. This adaptive mechanism is particularly important under complex working conditions.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] 1. This invention achieves three breakthroughs in 5G base station carrier module fault prediction through the innovative design of a dynamic interference analysis module: First, the spatiotemporal coupling features extracted by a dynamic coupling algorithm address the inability of traditional static features to capture transient signal characteristics. Second, the multimodal fusion of environmental perception and device status feedback significantly reduces the misjudgment rate caused by single-dimensional analysis. Finally, an incremental learning mechanism enables the model to adapt to new interference types in real time, shortening the model update cycle from weeks to minutes. These technological innovations collectively build an intelligent prediction system with the ability to continuously evolve.

[0018] 2. The present invention achieves end-to-end intelligent processing from raw data to operation and maintenance decisions through a four-level closed-loop architecture consisting of multi-source data acquisition, dynamic interference analysis, fault correlation modeling, and predictive decision output. The three-dimensional correlation model constructed by the fault correlation modeling module realizes a comprehensive analysis of interference characteristics, equipment status, and environmental factors, while the multi-level decision-making mechanism of the predictive decision output module (real-time warning, preventive maintenance recommendations, and system optimization feedback) ensures the accuracy and foresight of operation and maintenance responses. This architectural design not only improves the prediction accuracy, but also forms a complete closed loop from feature analysis to decision execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the overall process structure of the present invention;

[0020] Figure 2 This is a flow chart of the multi-source data acquisition module of the present invention;

[0021] Figure 3 It is a flow chart of the dynamic interference analysis module of the present invention.

[0022] The meaning of each number in the figure is:

[0023] 100. Multi-source data acquisition module; 200. Dynamic interference analysis module; 300. Fault correlation modeling module; 400. Prediction decision output module. DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0025] At the same time, some technical terms are explained here:

[0026] Carrier waveform data is the original signal collected through the base station RF port, including time domain waveform and frequency domain spectrum characteristics;

[0027] The time domain waveform refers to the instantaneous characteristics of the signal amplitude changing over time, which can capture abnormal fluctuations caused by sudden interference;

[0028] Frequency domain spectrum characteristics refer to the distribution pattern of signal energy in different frequency bands, which can effectively identify steady-state anomalies such as spectrum distortion;

[0029] The spatiotemporal coupling feature refers to the simultaneous consideration of time series and spatial distribution characteristics when analyzing data, capturing the relationship between events or phenomena at the time points and geographical locations where they occur.

[0030] At present, when new interference that is not covered by the training data appears in the network, the prediction accuracy of the existing model will drop significantly. The present invention provides a carrier module fault prediction system based on big data mining, see Figure 1 As shown, it includes a multi-source data acquisition module 100, a dynamic interference analysis module 200, a fault correlation modeling module 300 and a prediction decision output module 400. By building a closed-loop intelligent analysis link, real-time and accurate prediction of 5G base station carrier module failures is achieved.

[0031] See also Figure 2 As shown, the multi-source data acquisition module 100 is the data foundation of the entire fault prediction system. This module uses multiple monitoring interfaces on the base station equipment to collect various key data generated during the operation of the carrier module in real time, providing comprehensive and accurate data support for subsequent analysis and processing. The multi-source data acquisition module 100 utilizes a distributed data acquisition architecture to ensure efficient and stable acquisition and transmission of various data types. This module primarily collects three core data types: carrier waveform data, equipment operating status data, and operating environment data.

[0032] To collect RF signal characteristic data, the module directly connects to the carrier module's signal monitoring port, acquiring real-time digitized carrier waveform data. This data fully records the signal's time-domain characteristics and frequency-domain distribution characteristics, with sampling accuracy meeting millisecond-level time resolution requirements. It is used to obtain complete signal data, including both time-domain waveform and frequency-domain spectral characteristics. To collect device operating status data, the module continuously collects power amplifier operating parameters through a dedicated data interface, including real-time output power values, operating efficiency indicators, and other key parameters reflecting the device's operating status. To comprehensively assess operating environment data, the module also integrates data from the base station's built-in environmental sensors, including physical environmental parameters such as temperature and humidity, as well as operating status indicators such as network load rate and channel occupancy. All collected data is precisely time-stamped to ensure time synchronization for subsequent analysis.

[0033] The multi-source data acquisition module 100 employs a unique data preprocessing process. First, the carrier waveform data is normalized to eliminate dimensional differences between channels. All acquired data is then timestamp aligned and outliers filtered to ensure data quality. The processed data is then packaged into standardized data packets, preserving the integrity of the original data while improving the processing efficiency of subsequent modules.

[0034] Preprocessed carrier waveform data is used for interference pattern analysis, equipment operating status data is used for fault probability assessment, and operating environment data provides context for the predictive model. This targeted data distribution strategy ensures that each functional module receives the most relevant data support.

[0035] Although traditional data acquisition and processing methods can obtain the operating data of the carrier module, there are significant deficiencies in practical applications: on the one hand, the static data analysis model cannot adapt to the complex and changeable interference characteristics in the 5G network environment; on the other hand, conventional feature extraction methods are difficult to effectively identify sudden new interference patterns. These limitations cause the prediction accuracy of existing systems to drop significantly when facing unknown interference, making it difficult to provide reliable decision support for base station operation and maintenance. It is precisely based on these key technical bottlenecks that the present invention introduces a dynamic interference analysis module 200, which realizes intelligent identification and dynamic tracking of various interference patterns by introducing adaptive feature extraction and online learning mechanisms. The standardized data packets pre-processed by the multi-source data acquisition module 100 are transmitted to the dynamic interference analysis module 200 in real time through a high-speed data channel, providing a high-quality data foundation for subsequent intelligent analysis.

[0036] The dynamic interference analysis module 200 is the core innovation of the present invention, which realizes dynamic identification and model optimization of interference characteristics through in-depth processing of standardized data packets. Figure 3As shown in the figure, this module first analyzes the spatiotemporal characteristics of the carrier waveform data received in the standardized data packets and uses a dynamic coupling algorithm to establish a joint time-frequency feature matrix to extract spatiotemporal coupling features. This processing dynamically adjusts the sensitivity threshold of feature extraction based on real-time reference to the power amplifier parameters in the device operating status data, ensuring that discriminative feature representations are obtained under different operating conditions. For example, when the power amplifier efficiency index fluctuates, the algorithm automatically increases its focus on the characteristics of the corresponding frequency band. This adaptive mechanism significantly improves the targeted feature extraction.

[0037] Based on the extracted spatiotemporal coupling features, the dynamic interference analysis module 200 intelligently identifies potential interference patterns in the carrier signal. The recognition process uses a pre-trained deep neural network model and introduces an environmental perception mechanism. Information such as temperature and humidity in the operating environment data is encoded as auxiliary features, which participate in model reasoning together with the spatiotemporal coupling features, and finally outputs the interference pattern determination result including the interference type and confidence score. This multimodal feature fusion method enables the model to distinguish between real interference signals and normal fluctuations caused by environmental factors, greatly reducing the false alarm rate. When the confidence level of the model output is lower than the preset threshold, the system will determine that a new interference pattern has been detected and automatically trigger the subsequent learning process.

[0038] Based on the identified new interference patterns, the dynamic interference analysis module 200 triggers an incremental learning process to dynamically optimize the model and generate model optimization recommendations. The learning process is based on the previously extracted spatiotemporal coupling features, adjusting only the classifier parameters at the top level of the model while keeping the underlying feature extraction network unchanged. This design ensures that the model can quickly adapt to new interference patterns while avoiding the problem of feature representation drift caused by retraining the entire model. After each model update, the system automatically evaluates the performance of the new model on historical data to ensure that the update does not negatively affect the recognition performance of known interference patterns. The updated model parameters and performance evaluation reports are packaged and transmitted to downstream modules to form a complete processing closed loop.

[0039] After comprehensive analysis and processing, the dynamic interference analysis module 200 generates a unified interference signature analysis report consisting of three key components: a spatiotemporal coupling feature matrix detailing the signal's time-frequency distribution; interference pattern determination results identifying the type of identified or new interference and its confidence level; and model optimization recommendations indicating whether model parameters need adjustment. This report is ultimately transmitted via a high-speed data channel to the fault correlation modeling module 300, providing accurate feature input for subsequent fault prediction.

[0040] To ensure system reliability, the dynamic interference analysis module 200 also incorporates a comprehensive data verification mechanism. All output data is accompanied by an integrity check code, allowing downstream modules to verify the data's validity upon receipt. Furthermore, the dynamic interference analysis module 200 monitors key performance indicators during processing in real time, including feature extraction time and recognition accuracy. This operational status data is regularly aggregated to the system management module, providing a basis for decision-making on performance optimization. This closed-loop design not only enables accurate identification of interference signatures but also establishes a stable and reliable feature supply system for the entire prediction system.

[0041] Because existing technologies often analyze the relationship between interference signatures and fault phenomena in isolation, ignoring the combined influence of device operating status and environmental factors on fault mechanisms, the present invention introduces a fault correlation modeling module 300. This module receives the interference signature analysis report from the dynamic interference analysis module 200, which includes three core components: a spatiotemporal coupling signature matrix, interference pattern determination results, and model optimization recommendations. Based on this input data, the module first establishes a multidimensional correlation map, dynamically weighting the spatiotemporal coupling signatures with power amplifier parameters from device operating status data. It also incorporates operating environment data as a modulating factor, constructing a three-dimensional correlation model of interference signatures, device status, and environmental factors.

[0042] During model construction, the module employs a dynamic inference mechanism based on a graph neural network. A spatiotemporal coupling feature matrix serves as the node attributes of the graph, device operating status parameters define the connection weights between nodes, and environmental data serves as the adjustment coefficient for edge features. This design ensures the organic integration of data from different dimensions. For example, when the ambient temperature exceeds a threshold, the system automatically strengthens the correlation between high-frequency interference signatures and power amplifier failures. The module tracks the changing trends of interference pattern determination results in real time. When a new interference pattern is detected, it triggers an online update of the correlation graph to ensure the timeliness of the model.

[0043] After processing, the module generates a standardized fault correlation report containing three levels of analysis: a base layer records the static correlation strength between various interference characteristics and device parameters; a dynamic layer reflects the degree to which environmental factors affect these correlations; and a predictive layer outputs a multi-dimensional fault risk score based on the current state. This data is fully transmitted via a high-speed data bus to the predictive decision output module 400. The fault risk score is directly used for early warning decisions, while the correlation strength analysis provides operations and maintenance personnel with a basis for fault cause analysis. To ensure data consistency, all output results use the same spatiotemporal reference and encoding format as the upstream module, ensuring seamless parsing and processing by downstream modules.

[0044] The module also incorporates a comprehensive self-monitoring mechanism to continuously evaluate the accuracy and stability of the correlation model. Evaluation metrics include the sparsity of the correlation matrix and the consistency of the prediction results. These metrics are regularly fed back to the dynamic interference analysis module 200, providing a reference for optimizing the feature extraction algorithm. Through this closed-loop design, the fault correlation modeling module 300 not only accurately maps interference features to fault risks but also serves as a critical analytical hub connecting the entire prediction system.

[0045] The prediction decision output module 400 serves as the final decision-making layer of the system and receives the standardized fault correlation report from the fault correlation modeling module 300 as the core input data. The report contains the static correlation data of the base layer, the environmental adjustment coefficient of the dynamic layer, and the multi-dimensional risk score of the prediction layer, which provides a comprehensive basis for decision-making. The module first performs a multi-dimensional fusion analysis on the input fault risk score, combines the key indicators in the real-time working status data of the equipment (such as power amplifier efficiency, standing wave ratio, etc.), and calculates the comprehensive fault probability through a weighted decision algorithm. In the decision-making process, the module innovatively introduces a dynamic threshold mechanism, which will automatically adjust according to parameters such as temperature and humidity in the operating environment data to ensure that the early warning decision can adapt to the differences in fault performance under different environmental conditions.

[0046] Based on the analysis results, the module generates three levels of decision output: the primary output is a real-time warning instruction. When the comprehensive failure probability exceeds the dynamic threshold, an alarm signal containing the fault type and location information is immediately sent to the operation and maintenance system. The intermediate output is a preventive maintenance recommendation, which proposes equipment inspection or parameter adjustment plans based on the predicted fault development trend. The advanced output is system optimization feedback, which transmits model deviations or emerging fault mode characteristics discovered during the decision-making process back to the dynamic interference analysis module 200, forming a closed-loop optimization mechanism. All output data is encapsulated in a unified structured format and transmitted to the base station operation and maintenance management system via a dedicated communication protocol to ensure accurate execution of the instructions.

[0047] To ensure decision reliability, the module incorporates a multi-level validation mechanism: During the data input phase, the integrity and timeliness of fault-related reports are verified; during the decision-making calculation phase, a redundant computing architecture is employed to ensure result consistency; and during the output phase, compliance checks are performed on instruction content. Furthermore, the module continuously records key indicators of the decision-making process, including performance data such as early warning response latency and decision accuracy. This data will be used for ongoing system optimization and operational performance evaluation. Through this refined design, the Predictive Decision Output Module 400 not only achieves a complete closed-loop from feature analysis to operational decision-making, but also provides the entire prediction system with traceable and verifiable intelligent decision-making capabilities.

[0048] This approach extracts spatiotemporal coupling features from carrier waveform data, combines them with environmental perception mechanisms to identify interference patterns, and ultimately generates an interference signature analysis report containing signature data, pattern determinations, and optimization recommendations. This approach addresses the inability of traditional models to identify new types of interference. Through a closed-loop process of feature extraction, pattern recognition, and model optimization, it achieves the dual goals of accurate prediction and continuous optimization.

[0049] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A carrier module fault prediction system based on big data mining, comprising a multi-source data acquisition module (100), wherein the multi-source data acquisition module (100) acquires carrier waveform data, equipment working status data, and operating environment data from the carrier module to generate a standardized data packet, characterized in that: Also includes: A dynamic interference analysis module (200), the dynamic interference analysis module (200) receiving a standardized data packet and performing in-depth processing on the data packet; The dynamic interference analysis module (200) extracts spatiotemporal coupling features from carrier waveform data using a dynamic coupling algorithm based on equipment operating status data; The dynamic interference analysis module (200) identifies a new interference pattern from a carrier signal based on spatiotemporal coupling characteristics and in combination with operating environment data, and generates an interference pattern determination; The dynamic interference analysis module (200) triggers an incremental learning process based on a new interference pattern to perform model optimization and generate model optimization suggestions; The dynamic interference analysis module (200) generates an interference feature analysis report including spatiotemporal coupling features, interference pattern determination, and model optimization suggestions, and transmits the report to the fault correlation modeling module (300) for fault prediction of the 5G base station carrier module; Carrier waveform data is the original signal collected through the base station RF port, including time domain waveform and frequency domain spectrum characteristics; The multi-source data acquisition module (100) continuously acquires the operating parameters of the power amplifier through a dedicated data interface, including real-time output power value and working efficiency index, which are key parameters reflecting the operating status of the equipment, as equipment operating status data; The multi-source data acquisition module (100) integrates the environmental sensor data built into the base station equipment, including physical environmental parameters of temperature and humidity, as well as network load rate and channel occupancy status indicators, as operating environment data; First, the carrier waveform data in the received standardized data packets is analyzed for spatiotemporal characteristics. A dynamic coupling algorithm is used to establish a joint time-frequency feature matrix, thereby extracting spatiotemporal coupling features. This process uses real-time reference to the power amplifier parameters in the device operating status data to dynamically adjust the sensitivity threshold of feature extraction to ensure that discriminative feature representations are obtained under different operating conditions. The dynamic interference analysis module (200) intelligently identifies potential interference patterns in the carrier signal; the identification process uses a pre-trained deep neural network model, introduces an environmental perception mechanism, encodes temperature and humidity information in the operating environment data as auxiliary features, and participates in model reasoning together with the spatiotemporal coupling features, and finally outputs an interference pattern determination result including the interference type and confidence score; When the confidence level output by the dynamic interference analysis module (200) is lower than a preset threshold, the system determines that a new interference pattern has been detected; Based on the identified new interference pattern, the dynamic interference analysis module (200) triggers an incremental learning process to dynamically optimize the model and generate model optimization suggestions; The learning process is based on the previously extracted spatiotemporal coupling features, and only adjusts the classifier parameters at the top layer of the model, while keeping the underlying feature extraction network unchanged; The fault correlation modeling module (300) receives the interference feature analysis report from the dynamic interference analysis module (200), including three core contents: a time-space coupling feature matrix, an interference pattern determination result, and a model optimization suggestion; First, a multidimensional correlation map is established, dynamically weighting the spatiotemporal coupling characteristics with the power amplifier parameters in the equipment operating status data. At the same time, operating environment data is introduced as a regulating factor to construct a three-dimensional correlation model of interference characteristics, equipment status, and environmental factors. In the process of constructing the three-dimensional correlation model, the fault correlation modeling module (300) adopts a dynamic reasoning mechanism based on a graph neural network; the spatiotemporal coupling feature matrix is ​​used as the node attribute of the graph, the equipment working state parameters define the connection weights between nodes, and the environmental data is used as the adjustment coefficient of the edge feature; After the processing is completed, the fault correlation modeling module (300) generates a standardized fault correlation report, which contains three levels of analysis results: the basic layer records the static correlation strength of various interference characteristics and equipment parameters; the dynamic layer reflects the degree of influence of environmental factors on the correlation relationship; the prediction layer outputs a multi-dimensional fault risk score based on the current state; The prediction decision output module (400) receives the standardized fault correlation report from the fault correlation modeling module (300) as core input data; The prediction decision output module (400) first performs a multi-dimensional fusion analysis on the input fault risk score, combines the key indicators in the real-time working status data of the equipment, and calculates the comprehensive fault probability through a weighted decision algorithm; in the decision process, a dynamic threshold mechanism is introduced, and the threshold is automatically adjusted according to the temperature and humidity parameters in the operating environment data; Based on the analysis results, the prediction decision output module (400) generates decision outputs including three levels: the primary output is a real-time warning instruction, which immediately sends an alarm signal containing the fault type and location information to the operation and maintenance system when the comprehensive fault probability exceeds the dynamic threshold; the intermediate output is a preventive maintenance suggestion, which proposes equipment inspection or parameter adjustment plans based on the fault development trend prediction; the high-level output is system optimization feedback, which transmits the model deviation or newly emerged fault mode characteristics found in the decision-making process back to the dynamic interference analysis module (200), forming a closed-loop optimization mechanism.

2. The carrier module fault prediction system based on big data mining according to claim 1, characterized in that: The multi-source data acquisition module (100) eliminates the dimension differences between channels through normalization processing, and generates the standardized data packet after timestamp alignment and outlier filtering.

3. The carrier module fault prediction system based on big data mining according to claim 1, characterized in that: The dynamic interference analysis module (200) monitors the real-time changes of the power amplifier efficiency index and dynamically adjusts the sensitivity weights of the feature extraction algorithm to the features of different frequency bands, thereby optimizing the extraction process of the spatiotemporal coupling features.

4. The carrier module fault prediction system based on big data mining according to claim 1, characterized in that: The dynamic interference analysis module (200) verifies the model after incremental learning using historical data, so as to evaluate the impact of the model update on the ability to recognize known interference patterns.

5. The carrier module fault prediction system based on big data mining according to claim 1, characterized in that: The prediction decision output module (400) automatically adjusts the dynamic threshold of the early warning decision according to the operating environment data, and generates a decision output including real-time early warning instructions, preventive maintenance suggestions and system optimization feedback.

Images