Machine learning-based core board fault prediction system, method and device and medium

Abstract

The invention discloses a core board fault prediction system, method and device based on machine learning and a medium. The system comprises a data acquisition module, a data preprocessing module, a model training module, a data fusion module, a fault prediction module and an alarm and feedback module. The data acquisition module acquires multi-modal data in the operation process of a core board, the data preprocessing module preprocesses the acquired data, the model training module constructs and optimizes a machine learning model based on historical fault data and classifies fault modes, the data fusion module fuses data of different sensors, and the data processing module processes the fused data. And a fault prediction module analyzes the operation data of the core board through a trained machine learning model and predicts the fault occurrence time and type, and an alarm and feedback module gives an alarm and generates a fault report when identifying the fault. The core board fault prediction accuracy and the core board fault response efficiency are improved, and the method can be applied to the technical field of fault detection.

Description

Technical Field

[0001] The present invention relates to the field of fault detection technology, and in particular to a core board fault prediction system, method, device and medium based on machine learning. Background Art

[0002] As the complexity of electronic devices continues to increase, especially in industrial control, communication equipment and embedded systems, the reliability of the core board, as an important part of the system, directly affects the stability and work efficiency of the entire system. The failure of the core board may not only cause the interruption of equipment operation, but also cause safety hazards and economic losses. Therefore, how to monitor the operating status of the core board in real time, accurately predict failures and take preventive measures in time has become a research hotspot in the current field of intelligent devices and industrial control.

[0003] However, existing fault prediction systems still face many technical challenges. First, multimodal data (such as temperature, power consumption, voltage, etc.) usually have large noise and inconsistency, resulting in low data quality, which affects the accuracy of fault prediction. Secondly, many systems lack effective data fusion methods and cannot fully utilize multi-source information from different sensors. In addition, most of the existing prediction models only target a single fault type and have difficulty handling complex and dynamically changing fault modes, especially when dealing with time series data during the operation of the core board. There is a lack of effective capture and processing capabilities for timing features. Finally, the alarm and feedback mechanism of the existing system is not intelligent enough to respond dynamically according to the severity of the fault. Summary of the invention

[0004] The purpose of the present invention is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0005] To this end, an object of an embodiment of the present invention is to provide a core board fault prediction system based on machine learning, which improves the accuracy of core board fault prediction and the response efficiency of core board faults.

[0006] Another object of an embodiment of the present invention is to provide a core board fault prediction method based on machine learning.

[0007] In order to achieve the above technical objectives, the technical solutions adopted by the embodiments of the present invention include:

[0008] In a first aspect, an embodiment of the present invention provides a core board fault prediction system based on machine learning, including a data acquisition module, a data preprocessing module, a model training module, a data fusion module, a fault prediction module, and an alarm and feedback module, wherein:

[0009] The data acquisition module is used to acquire multimodal data generated during the operation of the core board in real time, and the multimodal data includes temperature data, power consumption data, voltage data, clock signal data and onboard sensor data;

[0010] The data preprocessing module is used to perform denoising, normalization and feature enhancement processing on the multimodal data to obtain feature enhanced data;

[0011] The model training module is used to build and optimize the machine learning model based on the deep neural network and support vector machine algorithm according to the historical fault data, use the recursive neural network to capture the time series characteristics of the time series data of different fault modes, classify and label the fault modes through the supervised learning method, and obtain a trained fault prediction model;

[0012] The data fusion module is used to perform multi-level fusion on the feature enhancement data to obtain fused feature data;

[0013] The fault prediction module is used to predict the core board fault according to the fusion feature data and the fault prediction model to obtain a fault prediction result;

[0014] The alarm and feedback module is used to issue alarm information and generate a fault report according to the fault prediction result.

[0015] Furthermore, in one embodiment of the present invention, the data acquisition module specifically includes:

[0016] Embedded sensor interface management unit, used to interact with various sensors on the core board;

[0017] High-frequency sampling and dynamic adjustment module, used to dynamically adjust the sampling frequency according to the system load and sensor status;

[0018] The multimodal data acquisition unit is used to acquire the multimodal data in real time according to the sampling frequency.

[0019] Furthermore, in one embodiment of the present invention, the data preprocessing module specifically includes:

[0020] A multimodal data denoising unit, used for performing denoising processing on the multimodal data;

[0021] An outlier detection unit, used to identify and remove abnormal or outlier data;

[0022] A signal repair unit, used to repair incomplete data caused by data loss or sensor failure;

[0023] A normalization processing unit, used for normalizing the multimodal data to eliminate scale differences between different data types;

[0024] The feature enhancement unit is used to optimize the multimodal data through feature enhancement technology to make the feature information more prominent.

[0025] Furthermore, in one embodiment of the present invention, the model training module specifically includes:

[0026] A historical data processing unit, used for performing data preprocessing and data fusion on the historical fault data;

[0027] A deep neural network model building unit, used to build a deep neural network model based on a multi-layer perceptron;

[0028] A support vector machine classifier unit, as an auxiliary classifier of the deep neural network model, is used to classify the failure mode of the historical failure data;

[0029] Recurrent neural network model unit, used to process time series data in fault prediction and capture the time series characteristics of different fault modes;

[0030] Model optimization and verification unit, used for model optimization and verification.

[0031] Furthermore, in one embodiment of the present invention, the data fusion module specifically includes:

[0032] A Bayesian reasoning fusion unit, used to perform probability fusion on the feature enhancement data from different sources by using a Bayesian reasoning method to obtain first fused data;

[0033] A Kalman filter fusion unit, used for real-time fusion and optimization of the dynamically changing feature enhancement data to obtain second fused data;

[0034] a unified feature vector generating unit, configured to generate the fused feature data according to the first fused data and the second fused data;

[0035] The real-time status combining unit is used to dynamically adjust the weight of data fusion according to the real-time running status of the core board.

[0036] Furthermore, in one embodiment of the present invention, the fault prediction module specifically includes:

[0037] A fault prediction unit, used to input the fused feature data into the fault prediction model to predict a potential fault occurrence time and a corresponding fault type;

[0038] The dynamic prediction threshold adjustment unit is used to dynamically adjust the prediction threshold of the fault prediction model according to the probability of fault occurrence.

[0039] Furthermore, in one embodiment of the present invention, the alarm and feedback module specifically includes:

[0040] An alarm signal triggering unit, used to send out an alarm signal when the fault prediction module identifies a potential fault, wherein the alarm signal includes a low-level alarm, a medium-level alarm and a high-level alarm;

[0041] An alarm signal sending unit, used to send the generated alarm signal to relevant personnel through different channels;

[0042] A fault report generating unit is used to automatically generate a fault report after the fault prediction module identifies a potential fault, wherein the fault report includes the fault type, predicted fault probability, fault occurrence time, original multimodal data, and repair suggestions.

[0043] In a second aspect, an embodiment of the present invention provides a core board fault prediction method based on machine learning, comprising the following steps:

[0044] According to historical fault data, a machine learning model is built and optimized based on deep neural network and support vector machine algorithm. Recurrent neural network is used to capture the time series characteristics of time series data of different fault modes. The fault modes are classified and labeled through supervised learning method to obtain a trained fault prediction model.

[0045] Acquire multimodal data generated during the operation of the core board in real time, wherein the multimodal data includes temperature data, power consumption data, voltage data, clock signal data, and onboard sensor data;

[0046] Performing denoising, normalization and feature enhancement processing on the multimodal data to obtain feature enhanced data;

[0047] Performing multi-level fusion on the feature enhancement data to obtain fused feature data;

[0048] Perform core board fault prediction based on the fused feature data and the fault prediction model to obtain a fault prediction result;

[0049] An alarm message is issued and a fault report is generated according to the fault prediction result.

[0050] In a third aspect, an embodiment of the present invention provides a core board fault prediction device based on machine learning, comprising:

[0051] at least one processor;

[0052] at least one memory for storing at least one program;

[0053] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned core board fault prediction method based on machine learning.

[0054] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, which stores a program executable by a processor, and the program executable by the processor is used to execute the above-mentioned core board fault prediction method based on machine learning when executed by the processor.

[0055] The advantages and beneficial effects of the present invention will be partly given in the following description, partly become apparent from the following description, or be understood through the practice of the present invention:

[0056] The present invention ensures that the multimodal data (such as temperature, power consumption, voltage, clock signal, etc.) collected from the core board in real time has high accuracy and consistency before being input into the machine learning model through the multi-level optimization design of data collection, preprocessing and fusion. The data is fused through Bayesian reasoning and Kalman filtering algorithms, which effectively improves the robustness of the data and reduces the influence of noise and sensor errors. The feature extraction and enhancement algorithm ensures the accurate capture of key features (such as current fluctuations and power consumption anomalies), and provides a reliable data basis for subsequent model training and fault prediction, thereby greatly improving the system's ability to recognize complex fault modes, improving the accuracy of core board fault prediction and the response efficiency of core board faults; through the collaborative work of deep neural networks, support vector machines and recursive neural networks, the system can accurately model fault modes from multimodal historical data and real-time data, especially in capturing timing fault characteristics. Excellent, so it can predict the time and type of potential failures in real time, and dynamically adjust the prediction threshold according to the probability of failure, to ensure accurate warning in the early stage of failure. This intelligent prediction mechanism effectively improves the efficiency of fault handling, helps operation and maintenance personnel take measures in advance, and reduces downtime and losses; the alarm and feedback module provides a multi-level alarm trigger mechanism and automatic fault report generation function, which can promptly notify operation and maintenance personnel and provide detailed fault analysis and repair suggestions, and can issue alarms through multiple channels such as sound, vision and remote notification to ensure timely response in different environments; in addition, the system integrates a feedback mechanism, which continuously optimizes the system's fault prediction model and alarm strategy through the processing feedback of operation and maintenance personnel, and enhances the system's adaptability. With the accumulation of feedback information, the system can more accurately adjust the model and response mechanism when dealing with complex and new faults, further improving the accuracy of fault prediction and system stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solution in the embodiments of the present invention, the following introduction is made to the drawings required for use in the embodiments of the present invention. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solution of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0058] Figure 1 A structural block diagram of a core board fault prediction system based on machine learning provided in an embodiment of the present invention;

[0059] Figure 2 A flowchart of a core board fault prediction method based on machine learning provided in an embodiment of the present invention;

[0060] Figure 3 A structural block diagram of a core board fault prediction device based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way, and the execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0062] In the description of the present invention, the meaning of "a plurality" is two or more than two. If there is a description of "a first" or "a second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features. In addition, unless otherwise defined, all technical and scientific terms used in this document have the same meaning as those commonly understood by those skilled in the art.

[0063] Reference Figure 1 The embodiment of the present invention provides a core board fault prediction system based on machine learning, including a data acquisition module, a data preprocessing module, a model training module, a data fusion module, a fault prediction module and an alarm and feedback module, wherein:

[0064] The data acquisition module is used to obtain multimodal data generated during the operation of the core board in real time. The multimodal data includes temperature data, power consumption data, voltage data, clock signal data, and onboard sensor data;

[0065] The data preprocessing module is used to perform denoising, normalization and feature enhancement on multimodal data to obtain feature-enhanced data;

[0066] The model training module is used to build and optimize the machine learning model based on the deep neural network and support vector machine algorithm according to the historical fault data, use the recursive neural network to capture the time series characteristics of the time series data of different fault modes, classify and label the fault modes through the supervised learning method, and obtain a trained fault prediction model;

[0067] The data fusion module is used to perform multi-level fusion of feature enhancement data to obtain fused feature data;

[0068] The fault prediction module is used to predict the core board fault based on the fusion feature data and the fault prediction model to obtain the fault prediction result;

[0069] The alarm and feedback module is used to issue alarm information and generate fault reports based on the fault prediction results.

[0070] Specifically, the data acquisition module is used to acquire multimodal data generated during the operation of the core board in real time. The multimodal data includes temperature, power consumption, voltage, clock signal and data of onboard sensors. The data acquisition module interacts with the core board through the embedded sensor interface and realizes high-frequency sampling and dynamic adjustment of data.

[0071] The data preprocessing module performs denoising, normalization and feature enhancement on the collected multimodal data. By applying filtering algorithms, outlier detection algorithms and signal repair algorithms, it ensures the integrity of the data input to the machine learning model. The data preprocessing module also includes a fault feature extraction function to automatically extract key feature parameters, including current fluctuations and power consumption anomalies.

[0072] The model training module builds and optimizes the machine learning model based on historical fault data and collected multimodal data. The model training module uses deep neural network and support vector machine algorithms to classify and label fault modes through supervised learning methods. At the same time, the model training module uses recursive neural networks to model time series data to capture the temporal characteristics of fault modes.

[0073] The data fusion module integrates the data collected by different sensors at multiple levels, uses Bayesian reasoning and Kalman filtering algorithms to improve the robustness of the data, and combines the real-time status of the core board to generate a unified feature vector to enhance the model's ability to identify complex faults.

[0074] The fault prediction module analyzes the multimodal data of the core board in real time based on the trained machine learning model, predicts the potential time and type of faults, and dynamically adjusts the prediction threshold of the model according to the possibility of faults, and uses the decision tree algorithm to warn of various fault modes.

[0075] The alarm and feedback module is used to send out an alarm signal and generate a fault report when the fault prediction module identifies a potential fault.

[0076] As an optional implementation, the data acquisition module specifically includes:

[0077] Embedded sensor interface management unit, used to interact with various sensors on the core board;

[0078] High-frequency sampling and dynamic adjustment module, used to dynamically adjust the sampling frequency according to the system load and sensor status;

[0079] The multimodal data acquisition unit is used to acquire multimodal data in real time according to the sampling frequency.

[0080] Specifically, the multimodal data acquisition unit is used to collect multimodal data generated during the operation of the core board in real time, including temperature data, power consumption data, voltage data, clock signal data and onboard sensor data, wherein:

[0081] Temperature data: The integrated temperature sensor monitors the temperature changes in different areas of the core board in real time;

[0082] Power consumption data: The total power consumption of the core board and the local power consumption of key components are monitored in real time through the power monitoring chip;

[0083] Voltage data: Monitor the voltage fluctuation of each power supply unit of the core board through the voltage detection module;

[0084] Clock signal data: Through the embedded clock module, the clock frequency and stability of the core board are captured in real time;

[0085] Onboard sensor data: including data from vibration, humidity, and other specific function sensors to ensure comprehensive awareness of the core board's working environment and physical state.

[0086] The embedded sensor interface management unit is used to exchange data with various sensors on the core board. It adopts a hybrid architecture of SPI (serial peripheral interface) and I2C (inter-integrated circuit bus) to ensure the efficiency and stability of data transmission, including:

[0087] (1) SPI bus: used for sensors with high-frequency data acquisition, providing high-speed real-time data transmission, suitable for fast-changing parameters such as temperature and voltage;

[0088] (2) I2C bus: used for lower-rate data sensors, such as humidity and vibration, to reduce system power consumption while ensuring data acquisition stability;

[0089] (3) Dynamic interface management: Automatically allocate bandwidth according to the working status of the sensor, give priority to the collection of important data (such as voltage and temperature), and avoid data transmission conflicts.

[0090] The high-frequency sampling and dynamic adjustment mechanism is used to achieve high-frequency sampling of multimodal data, and dynamically adjust the sampling frequency according to the system load and sensor status, as follows:

[0091] (1) Adaptive adjustment of sampling frequency: The sampling frequency is automatically adjusted according to the real-time status of the core board, such as voltage fluctuation, temperature rise, etc. The sampling frequency adjustment formula is:

[0092]

[0093] Among them, f new is the new sampling frequency, f base is the basic sampling frequency, V current is the current voltage, V ref is the reference voltage, β is the adjustment coefficient;

[0094] (2) Multimodal data synchronization mechanism: For sensor data with different sampling frequencies, a timestamp synchronization mechanism is used to ensure the temporal consistency of the data and avoid data invalidation or incorrect association;

[0095] The data acquisition module ensures efficient and accurate collection and real-time dynamic management of multimodal data through optimized multimodal data acquisition units, efficient embedded sensor interface management, and flexible high-frequency sampling and dynamic adjustment mechanisms, which helps to improve the system's comprehensive monitoring of the core board status and fault detection capabilities.

[0096] As an optional implementation, the data preprocessing module specifically includes:

[0097] A multimodal data denoising unit, used for denoising multimodal data;

[0098] An outlier detection unit, used to identify and remove abnormal or outlier data;

[0099] A signal repair unit, used to repair incomplete data caused by data loss or sensor failure;

[0100] A normalization processing unit is used to normalize multimodal data and eliminate scale differences between different data types;

[0101] The feature enhancement unit is used to optimize the multimodal data through feature enhancement technology to make the feature information more prominent.

[0102] Specifically, the multimodal data denoising unit is used to denoise the collected multimodal data to ensure the integrity and accuracy of the input data. The specific scheme is as follows:

[0103] (1) Filtering algorithm: Wavelet transform is used to denoise continuous signals such as temperature, voltage, and power consumption, removing high-frequency noise while retaining key signals. The denoising formula is expressed as:

[0104] S clean (t)=∑ k c k ψ k (t);

[0105] Among them, S clean (t) is the denoised signal, c k is the wavelet coefficient, ψ k (t) is the wavelet basis function;

[0106] (2) Low-pass filter: used to filter out high-frequency interference in the clock signal to ensure the smoothness of the signal;

[0107] (3) Outlier elimination: Use the preset threshold to perform preliminary screening on the power consumption and current fluctuation data, and eliminate abnormal data that exceeds the reasonable range.

[0108] The normalization processing unit is used to normalize the multimodal data, eliminate the scale differences between different data types, and ensure data consistency. The normalization method used is maximum and minimum normalization. The specific formula is:

[0109]

[0110] Among them, x′ is the normalized data, x is the original data, and x max and x min are the minimum and maximum values ​​in the data set, respectively. Temperature, power consumption, voltage and other data are processed uniformly through this normalization formula to ensure that different data dimensions have the same value range when inputting the model.

[0111] The feature enhancement unit optimizes multimodal data through feature enhancement technology to make key feature information more prominent. The specific solutions include:

[0112] (1) Principal component analysis: While reducing the data dimension, the most influential features for fault prediction are retained, which are calculated as:

[0113] Z = XW;

[0114] Among them, Z is the data after dimensionality reduction, X is the input data matrix, and W is the feature vector matrix;

[0115] (2) Feature standardization: Standardize the enhanced features to ensure that the feature values ​​are within the same mean and variance range, expressed as:

[0116]

[0117] Among them, o is the standardized data, μ is the mean, and σ is the standard deviation.

[0118] The outlier detection unit is used to identify and remove abnormal or outlier data to ensure that the data input into the model does not contain errors. It uses an outlier detection algorithm based on the Z score, which is expressed as:

[0119]

[0120] Among them, Z i is the Z score of the ith data point, x i is the i-th data, μ is the mean of the data, σ is the standard deviation of the data, if |Z i |>∈(preset threshold), the data is considered as an outlier.

[0121] The signal repair unit is used to repair incomplete data caused by data loss or sensor failure. The linear interpolation method is used for data repair, which is expressed as:

[0122]

[0123] Among them, x(t) is the time point data to be repaired, x(t1) and x(t2) are the data values ​​of adjacent known time points, and t1 and t2 are corresponding time points.

[0124] The fault feature extraction unit is used to automatically extract key feature parameters that affect fault prediction, including:

[0125] (1) Current fluctuation feature extraction: The amplitude of current fluctuation is captured by calculating the standard deviation of the current signal, which is expressed as:

[0126]

[0127] Among them, σ l is the standard deviation of current fluctuation, I i is the i-th current data point, μ I is the mean value of the current;

[0128] (2) Power consumption anomaly feature extraction: By detecting the difference between the peak value and the average value of power consumption, the occurrence of power consumption anomaly is determined, which is expressed as:

[0129] Pdiff =P peak -P avg ;

[0130] Among them, P peak is the peak power consumption, P avg is the average power consumption, if P diff If the preset threshold is exceeded, it is considered as abnormal power consumption;

[0131] The data preprocessing module ensures the integrity and validity of the data input into the machine learning model through efficient denoising, normalization, feature enhancement, outlier detection and signal repair mechanisms. The fault feature extraction unit effectively improves the system's ability to identify potential faults by accurately extracting key features.

[0132] As an optional implementation, the model training module specifically includes:

[0133] A historical data processing unit, used for data preprocessing and data fusion of historical fault data;

[0134] A deep neural network model building unit, used to build a deep neural network model based on a multi-layer perceptron;

[0135] A support vector machine classifier unit, which is used as an auxiliary classifier of the deep neural network model to classify the failure modes of historical failure data;

[0136] Recurrent neural network model unit, used to process time series data in fault prediction and capture the time series characteristics of different fault modes;

[0137] Model optimization and verification unit, used for model optimization and verification.

[0138] Specifically, the historical data processing unit is used to receive historical fault data and collected multi-modal data, and perform preliminary processing to ensure the validity and consistency of the data. The preliminary processing includes:

[0139] Data screening: Select the fault type data most relevant to the current system status from historical data;

[0140] Data balancing: Use oversampling and undersampling techniques to solve the problem of imbalanced data sets and ensure the training effect of machine learning models;

[0141] Data segmentation: Divide the dataset into training set, validation set, and test set to ensure the generalization ability of the model.

[0142] The deep neural network model building unit is used to build a deep neural network model based on a multi-layer perceptron. The model structure includes an input layer, several hidden layers and an output layer, as follows:

[0143] Input layer: The input layer receives the feature vector of multimodal data, and the vector length is equal to the number of features after preprocessing;

[0144] Hidden layer: Several hidden layers are used, each layer contains multiple neurons, and the ReLU activation function is used; the number of neurons and the number of layers in the hidden layer are determined by cross-validation;

[0145] Output layer: The output layer is used to classify the fault mode. The softmax function is used to calculate the probability of each fault category. The expression of the softmax function is:

[0146]

[0147] Among them, P(y i |x) is, z i is the weighted sum of the input to the output layer, and n is the number of fault categories.

[0148] The support vector machine classifier unit is used to classify the fault mode. As an auxiliary classifier of the deep neural network model, it enhances the classification effect and is particularly suitable for processing small sample data. The support vector machine uses the RBF kernel function, and the kernel function expression is:

[0149] K(x i ,x j )=exp(-γ∥x i -x j ∥ 2 );

[0150] Among them, K(x i ,x j ) is the sample x i With sample x j The kernel function value between , γ is the parameter of the kernel function;

[0151] Optimization objective: The goal of the support vector machine is to minimize the classification error and margin. The objective function is expressed as:

[0152]

[0153] Among them, w is the weight vector, C is the regularization parameter, ξ i is the slack variable.

[0154] The recursive neural network model unit is used to process time series data in fault prediction and capture the time series characteristics of the fault mode. The specific model is a long short-term memory network model, which is used to model time dependency. The long short-term memory network specifically includes:

[0155] The forget gate is represented by: f t =σ(W f ·[ht-1 ,x t ]+b f );

[0156] Among them, f t is the output of the forget gate, h t-1 is the hidden state of the previous time step, x t is the input of the current time step, W f is the forget gate weight matrix, b f is the bias term;

[0157] The input gate is represented as: i t =σ(W i ·[h t-1 ,x t ]+b i );

[0158] Among them, i t is the input gate output;

[0159] The cell state update is expressed as: C t =f t ·C t-1 +i t tanh(W C ·[h t-1 ,x t ]+b C );

[0160] Among them, C t is the cell state at the current time step, C t-1 is the cell state at the previous time step;

[0161] The output gate is represented as: o t =σ(W o ·[h t-1 ,x t ]+b o );

[0162] The final hidden state is represented as: h t =o t tanh(C t ).

[0163] The model optimization and verification unit is used to optimize and verify the model, including:

[0164] (1) Loss function: The loss function of deep neural network and recurrent neural network adopts cross entropy loss function, which is expressed as:

[0165]

[0166] Among them, L is the loss function, y iis the true label, p i is the predicted probability;

[0167] (2) Optimization algorithm: The Adam optimization algorithm is used to train and optimize the model, and the learning rate is determined by cross-validation;

[0168] (3) Model validation: The model is evaluated through K-fold cross validation to ensure the generalization ability and reliability of the model;

[0169] The model training module combines deep neural networks, support vector machines and recursive neural networks to achieve accurate classification, labeling and time series modeling of fault modes. This module is not only suitable for large-scale data sets, but also can capture the timing characteristics of complex fault modes, effectively improving the accuracy and robustness of fault prediction.

[0170] As an optional implementation, the data fusion module specifically includes:

[0171] A Bayesian reasoning fusion unit, used for performing probability fusion on feature enhancement data from different sources by using a Bayesian reasoning method to obtain first fused data;

[0172] A Kalman filter fusion unit, used for real-time fusion and optimization of dynamically changing feature enhancement data to obtain second fused data;

[0173] A unified feature vector generating unit, used for generating fused feature data according to the first fused data and the second fused data;

[0174] The real-time status combining unit is used to dynamically adjust the weight of data fusion according to the real-time running status of the core board.

[0175] Specifically, the multimodal data multi-level fusion unit is used to receive multimodal data (including temperature, power consumption, voltage, clock signal and onboard sensor data) from different sensors, and integrate them into a unified feature vector through a multi-level fusion strategy. The specific implementation scheme is as follows:

[0176] (1) Time synchronization mechanism: Data from different sensors are synchronized according to timestamps to ensure that the data are fused within the same time window to avoid inconsistent data timing;

[0177] (2) Data layer processing:

[0178] First-layer fusion: weighted average fusion of data from sensors of the same type (such as multiple temperature sensors) to reduce the impact of fluctuations in single sensor data on the system;

[0179] Second layer fusion: feature-level fusion of data from different types of sensors (such as temperature, power consumption, and voltage) to generate preliminary multimodal data vectors.

[0180] The Bayesian reasoning fusion unit uses the Bayesian reasoning method to perform probability fusion on data from different sources, eliminate the influence of noise, and improve the robustness of data fusion. The Bayesian reasoning formula is expressed as:

[0181]

[0182] Where P(A|B) represents the probability of event A (such as a certain failure mode) occurring under the condition that sensor data B is observed, P(B|A) is the probability of sensor data B occurring under failure mode A, P(A) is the prior probability of the failure mode, and P(B) is the observed probability of the sensor data;

[0183] Data fusion process: Utilize the prior and observed data of different sensor data to dynamically adjust the credibility of each sensor, realize the fusion of multi-sensor data through Bayesian reasoning, and improve the fault tolerance and robustness of the system.

[0184] The Kalman filter fusion unit is used to perform real-time fusion and optimization of dynamically changing data to improve the accuracy of data fusion. It is especially suitable for processing real-time changing data such as voltage and power consumption. The specific steps are as follows:

[0185] State prediction: Kalman filtering predicts the sensor data at the next moment through the system model. The prediction formula is expressed as:

[0186]

[0187] in, is the predicted state at the kth moment, A is the state transfer matrix, B is the control matrix, u k is the control input;

[0188] Update the fusion of predictions and observations: through the Kalman gain K k The predicted value and the actual observed value are integrated and the state is updated, which is expressed as:

[0189]

[0190] in, is the updated state estimate, z k is the actual observation value at the kth moment, H is the observation matrix, K k is the Kalman gain, expressed as:

[0191] K k =P k|k-1 H T (HP k|k-1 H T +R) -1 ;

[0192] Among them, P k|k-1 is the prediction error covariance matrix, and R is the observation noise covariance matrix.

[0193] The unified feature vector generation unit combines the fusion results of Kalman filtering and Bayesian reasoning to generate a unified feature vector for fault identification. The specific steps are as follows:

[0194] Feature normalization: Normalize different types of data to ensure that each data type is within the same numerical range to avoid large differences in feature values ​​that affect the learning effect of the model;

[0195] Feature integration: Integrate the results of different sensor data after Bayesian reasoning and Kalman filtering optimization to generate the final unified feature vector, which is expressed as:

[0196] F=[f1,f2,...,f n ];

[0197] Among them, F is a unified eigenvector, f1,f2,...,f n The key features extracted after fusion are input into the machine learning model for fault prediction.

[0198] The real-time status combination unit is used to dynamically adjust the weight of data fusion according to the real-time operating status of the core board (such as load, voltage fluctuation, etc.), and the steps include:

[0199] State monitoring: monitor the real-time operating status of the system. If the data of some sensors is more important in certain states, increase their weight. The weight adjustment formula is expressed as:

[0200]

[0201] Among them, W new is the adjusted weight, W old is the current weight, S current is the current state value, S ref is the reference state value, and λ is the adjustment coefficient.

[0202] The data fusion module uses Bayesian reasoning and Kalman filtering algorithms to perform multi-level fusion of data from different sensors, and generates a unified feature vector based on the real-time status of the core board, which greatly improves the system's ability to identify complex fault modes and enhances the robustness and dynamic adaptability of data processing.

[0203] Further as an optional implementation, the fault prediction module specifically includes:

[0204] A fault prediction unit, used to input the fused feature data into a fault prediction model to predict the potential fault occurrence time and the corresponding fault type;

[0205] The dynamic prediction threshold adjustment unit is used to dynamically adjust the prediction threshold of the fault prediction model according to the probability of the fault occurring.

[0206] Specifically, the real-time fault prediction unit analyzes the multimodal data (such as temperature, power consumption, voltage, clock signal, etc.) collected during the operation of the core board in real time based on the trained machine learning model (including deep neural network, support vector machine and recursive neural network), and predicts the potential fault occurrence time and fault type. The specific implementation plan is as follows:

[0207] Data input: Real-time acquisition of multimodal data processed by the data fusion module, and inputting it into the trained machine learning model for analysis;

[0208] Prediction output: The model outputs the fault probability and fault type label of the current time step, where the fault type may include voltage anomaly, power overload, clock signal failure, etc.

[0209] The fault time prediction unit predicts the occurrence time of potential faults based on the time series characteristics of the recurrent neural network. The specific implementation scheme is as follows:

[0210] Time series modeling: Use the model to process the time series data of the core board during operation to capture the timing characteristics of the fault. The output of the long short-term memory network model includes the time step t of the fault occurrence. fault , the prediction formula is expressed as:

[0211] t fault = argmax t (P fault (t));

[0212] Among them, P fault (t) is the probability of failure occurring at time step t. The model selects the time step with the highest probability of failure as the prediction time of potential failure.

[0213] The fault type prediction unit classifies real-time data through machine learning models (deep neural network and support vector machine) to predict potential fault types. The fault type classification task is based on the softmax function or support vector machine classifier output, including:

[0214] The deep neural network model outputs the probability of fault type:

[0215]

[0216] Among them, P(y i |x) represents the probability that the current fault mode belongs to category i, z i is the weighted input for this category, and n is the number of fault types;

[0217] The support vector machine fault classification decision function is expressed as:

[0218]

[0219] Among them, K(x i ,x j ) is the SVM kernel function, and b is the bias term.

[0220] The dynamic prediction threshold adjustment unit is used to dynamically adjust the prediction threshold of the model according to the probability of fault occurrence, ensuring accurate early warning at the early stage of fault occurrence. The specific adjustment plan is as follows:

[0221] The fault prediction threshold adjustment formula is:

[0222] T new =T old +α×(P fault -P threshold );

[0223] Among them, T new is the adjusted prediction threshold, T o ld is the current threshold, α is the adjustment coefficient, P fault is the predicted probability of the current fault, P threshold is the preset failure probability threshold;

[0224] Threshold adjustment mechanism: When P fault Above the preset threshold P threshold When P fault When it is below the threshold, raise the threshold to avoid false positives.

[0225] The multi-fault mode early warning unit combines the decision tree algorithm to provide real-time early warning for multiple fault modes. The decision tree is used to quickly classify different fault types and modes and trigger early warnings based on the fault modes. The specific scheme is as follows:

[0226] Decision tree algorithm: define fault classification rules according to different fault modes, and finally output fault types and warning signals through conditional judgment of decision tree nodes. The loss function of the decision tree is Gini impurity, and the calculation formula is:

[0227]

[0228] Among them, Gini(D) represents the Gini impurity of data set D, P i is the proportion of samples belonging to category i in the data set, and n is the number of fault categories;

[0229] Early warning trigger mechanism: When the fault prediction probability P faultExceeds the dynamically adjusted threshold T new And when the decision tree algorithm determines that the current mode meets a certain fault type, a warning signal is triggered.

[0230] The fault report generation unit is used to automatically generate a fault analysis report after fault prediction. The report content includes the fault type, predicted fault occurrence time, fault occurrence probability and corresponding repair suggestions. The report generation process is as follows:

[0231] Report content: Analyze the detailed information of each fault prediction, generate a report containing fault mode, probability of occurrence, potential occurrence time, key data characteristics, etc., and output repair suggestions;

[0232] Fault report output format: Standardized report format allows engineers to quickly understand the cause of the fault and how to handle it.

[0233] The fault prediction module analyzes the multimodal data of the core board in real time based on the trained machine learning model, combined with the time series modeling function of the recursive neural network, to effectively predict the time and type of potential faults. It dynamically adjusts the prediction threshold and uses the decision tree algorithm to warn of multiple fault modes, significantly improving the accuracy and timeliness of fault prediction, providing a complete fault analysis report, and enhancing the system's ability to prevent and respond to complex faults.

[0234] As an optional implementation, the alarm and feedback module specifically includes:

[0235] An alarm signal triggering unit is used to send out an alarm signal when the fault prediction module identifies a potential fault. The alarm signal includes a low-level alarm, a medium-level alarm and a high-level alarm;

[0236] An alarm signal sending unit, used to send the generated alarm signal to relevant personnel through different channels;

[0237] The fault report generation unit is used to automatically generate a fault report after the fault prediction module identifies a potential fault. The fault report includes the fault type, predicted fault probability, fault occurrence time, original multimodal data, and repair suggestions.

[0238] Specifically, the alarm signal trigger unit is used to send out an alarm signal when the fault prediction module identifies a potential fault. The alarm mechanism is designed as low-level alarm, medium-level alarm and high-level alarm to ensure differentiated responses to different fault severities. The specific implementation plan is as follows:

[0239] (1) Alarm condition: When the fault prediction module calculates the fault occurrence probability P fault Exceeds the dynamically adjusted threshold T new When , the alarm signal is triggered, and the alarm triggering formula is:

[0240]

[0241] Among them, Alarm=1 means the alarm is triggered, and Alarm=0 means no alarm;

[0242] (2) Multi-level alarm mechanism: According to the probability of failure, the alarm signals are divided into different levels, including:

[0243] Low-level alarm: When P fault Just exceeded the threshold T new , issues a preliminary warning to remind the operation and maintenance personnel to conduct inspections;

[0244] Medium Alert: When P fault If the threshold is significantly exceeded, a medium-level alarm is issued, indicating that the potential failure risk is high and a timely response is required;

[0245] Advanced Alert: When P fault When a set high-risk threshold is reached, an emergency alert is issued, indicating that a failure is imminent and immediate action is required.

[0246] The alarm signal sending unit is used to send the generated alarm signal to relevant personnel through different channels, including:

[0247] Sound alarm: sound prompts are issued through buzzers or sound systems, especially suitable for on-site monitoring;

[0248] Visual alarm: An alarm icon or warning light signal (such as flashing red) appears on the monitoring terminal display screen, and the fault details can be displayed at the same time;

[0249] Remote Notification: Alarm signals are sent to relevant personnel’s mobile devices via SMS, email or app notification to ensure remote monitoring and timely response.

[0250] The fault report generation unit is used to automatically generate a fault report after the fault prediction module identifies a potential fault. The fault report includes the fault type, predicted fault probability, fault occurrence time, related multimodal data, and repair suggestions.

[0251] The alarm and feedback module also integrates a feedback mechanism to record the progress and feedback information of the operation and maintenance personnel on the alarm, helping the system to optimize future fault prediction and alarm strategies. The specific implementation plan is as follows:

[0252] (1) Feedback record: After the operation and maintenance personnel handle the fault, the system automatically records the following information:

[0253] The actual troubleshooting time and process;

[0254] The results of the failure repair success or failure;

[0255] The steps and time required to fix the problem;

[0256] (2) System learning and optimization: Based on feedback information, the system will update the fault handling database and optimize the alarm strategy when predicting similar faults in the future. By continuously learning from historical feedback data, the accuracy of fault prediction and alarm will be gradually improved.

[0257] In some optional embodiments, the alarm and feedback module also includes an emergency handling suggestion unit, which is used to automatically generate emergency handling suggestions when a high-level alarm is triggered, to help operation and maintenance personnel take effective measures to reduce the impact of the failure. The emergency handling suggestions include: immediately disconnecting the power supply or core board at risk of failure; measures to limit the expansion of the scope of the failure; and providing quick replacement solutions for key components.

[0258] Optionally, the alarm and feedback module also integrates a feedback mechanism to record the progress and feedback information of the operation and maintenance personnel on the alarm.

[0259] Optionally, fault reports can be exported in multiple formats for easy transmission and sharing.

[0260] It can be recognized that the present invention ensures that the multimodal data (such as temperature, power consumption, voltage, clock signal, etc.) collected from the core board in real time has high accuracy and consistency before being input into the machine learning model through the multi-level optimization design of data collection, preprocessing and fusion. The data is fused by Bayesian reasoning and Kalman filtering algorithm, which effectively improves the robustness of the data and reduces the influence of noise and sensor errors. The feature extraction and enhancement algorithm ensures the accurate capture of key features (such as current fluctuations and power consumption anomalies), and provides a reliable data basis for subsequent model training and fault prediction, thereby greatly improving the system's ability to recognize complex fault modes, improving the accuracy of core board fault prediction and the response efficiency of core board faults; through the collaborative work of deep neural networks, support vector machines and recursive neural networks, the system can accurately model fault modes from multimodal historical data and real-time data, especially in capturing timing fault characteristics. It performs well in real time, so it can predict the time and type of potential failures in real time, and dynamically adjust the prediction threshold according to the probability of failure, to ensure accurate warning in the early stage of failure. This intelligent prediction mechanism effectively improves the efficiency of fault handling, helps operation and maintenance personnel take measures in advance, and reduces downtime and losses; the alarm and feedback module provides a multi-level alarm trigger mechanism and automatic fault report generation function, which can promptly notify operation and maintenance personnel and provide detailed fault analysis and repair suggestions, and can issue alarms through multiple channels such as sound, vision and remote notification to ensure timely response in different environments; in addition, the system integrates a feedback mechanism, through the processing feedback of operation and maintenance personnel, continuously optimizes the system's fault prediction model and alarm strategy, and enhances the system's adaptive ability. With the accumulation of feedback information, the system can more accurately adjust the model and response mechanism when dealing with complex and new faults, further improving the accuracy of fault prediction and system stability.

[0261] Reference Figure 2 The embodiment of the present invention provides a core board fault prediction method based on machine learning, comprising the following steps:

[0262] According to historical fault data, a machine learning model is built and optimized based on deep neural network and support vector machine algorithm. Recurrent neural network is used to capture the time series characteristics of time series data of different fault modes. The fault modes are classified and labeled through supervised learning method to obtain a trained fault prediction model.

[0263] Real-time acquisition of multi-modal data generated during the operation of the core board, including temperature data, power consumption data, voltage data, clock signal data, and onboard sensor data;

[0264] De-noising, normalizing and feature enhancing the multimodal data to obtain feature enhanced data;

[0265] Perform multi-level fusion on the feature enhancement data to obtain fused feature data;

[0266] According to the fusion feature data and the fault prediction model, the core board fault prediction is performed to obtain the fault prediction result;

[0267] Issue alarm information and generate fault reports based on fault prediction results.

[0268] The contents of the above system embodiments are all applicable to the present method embodiments. The functions specifically implemented by the present method embodiments are the same as those of the above system embodiments, and the beneficial effects achieved are also the same as those achieved by the above system embodiments.

[0269] Reference Figure 3 The embodiment of the present invention provides a core board fault prediction device based on machine learning, comprising:

[0270] at least one processor;

[0271] at least one memory for storing at least one program;

[0272] When the at least one program is executed by the at least one processor, the at least one processor implements the core board fault prediction method based on machine learning.

[0273] The contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0274] An embodiment of the present invention also provides a computer-readable storage medium, which stores a program executable by a processor. When the program executable by the processor is executed by the processor, it is used to execute the above-mentioned core board fault prediction method based on machine learning.

[0275] A computer-readable storage medium of an embodiment of the present invention can execute a core board fault prediction method based on machine learning provided by an embodiment of the method of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0276] The embodiment of the present invention also discloses a computer program product or a computer program, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 1 The method shown.

[0277] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the above-mentioned boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0278] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified to the contrary, one or more of the above-mentioned functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0279] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the above methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0280] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0281] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the above-mentioned program is printed, since the above-mentioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or processing in other suitable ways as necessary, and then stored in a computer memory.

[0282] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0283] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0284] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0285] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A core board fault prediction system based on machine learning, characterized in that: It includes data acquisition module, data preprocessing module, model training module, data fusion module, fault prediction module and alarm and feedback module, among which: The data acquisition module is used to acquire multimodal data generated during the operation of the core board in real time, and the multimodal data includes temperature data, power consumption data, voltage data, clock signal data and onboard sensor data; The data preprocessing module is used to perform denoising, normalization and feature enhancement processing on the multimodal data to obtain feature enhanced data; The model training module is used to build and optimize the machine learning model based on the deep neural network and support vector machine algorithm according to the historical fault data, use the recursive neural network to capture the time series characteristics of the time series data of different fault modes, classify and label the fault modes through the supervised learning method, and obtain a trained fault prediction model; The data fusion module is used to perform multi-level fusion on the feature enhancement data to obtain fused feature data; The fault prediction module is used to predict the core board fault according to the fusion feature data and the fault prediction model to obtain a fault prediction result; The alarm and feedback module is used to issue alarm information and generate a fault report according to the fault prediction result.

2. The core board fault prediction system based on machine learning according to claim 1 is characterized in that: The data acquisition module specifically includes: Embedded sensor interface management unit, used to interact with various sensors on the core board; High-frequency sampling and dynamic adjustment module, used to dynamically adjust the sampling frequency according to the system load and sensor status; The multimodal data acquisition unit is used to acquire the multimodal data in real time according to the sampling frequency.

3. The core board fault prediction system based on machine learning according to claim 1 is characterized in that: The data preprocessing module specifically includes: A multimodal data denoising unit, used for performing denoising processing on the multimodal data; An outlier detection unit, used to identify and remove abnormal or outlier data; A signal repair unit, used to repair incomplete data caused by data loss or sensor failure; A normalization processing unit, used for normalizing the multimodal data to eliminate scale differences between different data types; The feature enhancement unit is used to optimize the multimodal data through feature enhancement technology to make the feature information more prominent.

4. The core board fault prediction system based on machine learning according to claim 1 is characterized in that: The model training module specifically includes: A historical data processing unit, used for performing data preprocessing and data fusion on the historical fault data; A deep neural network model building unit, used to build a deep neural network model based on a multi-layer perceptron; A support vector machine classifier unit, as an auxiliary classifier of the deep neural network model, is used to classify the failure mode of the historical failure data; Recurrent neural network model unit, used to process time series data in fault prediction and capture the time series characteristics of different fault modes; Model optimization and verification unit, used for model optimization and verification.

5. The core board fault prediction system based on machine learning according to claim 1 is characterized in that: The data fusion module specifically includes: A Bayesian reasoning fusion unit, used to perform probability fusion on the feature enhancement data from different sources by using a Bayesian reasoning method to obtain first fused data; A Kalman filter fusion unit, used for real-time fusion and optimization of the dynamically changing feature enhancement data to obtain second fused data; a unified feature vector generating unit, configured to generate the fused feature data according to the first fused data and the second fused data; The real-time status combining unit is used to dynamically adjust the weight of data fusion according to the real-time running status of the core board.

6. The core board fault prediction system based on machine learning according to claim 1 is characterized in that: The fault prediction module specifically includes: A fault prediction unit, used to input the fused feature data into the fault prediction model to predict a potential fault occurrence time and a corresponding fault type; The dynamic prediction threshold adjustment unit is used to dynamically adjust the prediction threshold of the fault prediction model according to the probability of fault occurrence.

7. A core board fault prediction system based on machine learning according to any one of claims 1 to 6, characterized in that: The alarm and feedback module specifically includes: An alarm signal triggering unit, used to send out an alarm signal when the fault prediction module identifies a potential fault, wherein the alarm signal includes a low-level alarm, a medium-level alarm and a high-level alarm; An alarm signal sending unit, used to send the generated alarm signal to relevant personnel through different channels; A fault report generating unit is used to automatically generate a fault report after the fault prediction module identifies a potential fault, wherein the fault report includes the fault type, predicted fault probability, fault occurrence time, original multimodal data, and repair suggestions.

8. A core board fault prediction method based on machine learning, characterized in that: The following steps are involved: According to historical fault data, a machine learning model is built and optimized based on deep neural network and support vector machine algorithm. Recursive neural network is used to capture the time series characteristics of time series data of different fault modes. The fault modes are classified and labeled through supervised learning method to obtain a trained fault prediction model. Acquire multimodal data generated during the operation of the core board in real time, wherein the multimodal data includes temperature data, power consumption data, voltage data, clock signal data, and onboard sensor data; Performing denoising, normalization and feature enhancement processing on the multimodal data to obtain feature enhanced data; Performing multi-level fusion on the feature enhancement data to obtain fused feature data; Perform core board fault prediction based on the fused feature data and the fault prediction model to obtain a fault prediction result; An alarm message is issued and a fault report is generated according to the fault prediction result.

9. A core board fault prediction device based on machine learning, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the core board fault prediction method based on machine learning as described in claim 8.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to execute a core board fault prediction method based on machine learning as described in claim 8 when executed by the processor.

Images