Intelligent fault detection and real-time hot plug system

Abstract

The invention relates to the technical field of automatic control, in particular to an intelligent fault detection and real-time hot plug system. The hot plug management unit is used for dynamically adding and removing the processing equipment based on a plug and play protocol; the data acquisition unit is used for acquiring state data of the equipment and receiving and storing data acquired from the equipment and the sensor in real time; the intelligent fault detection unit realizes monitoring, fault exploration and intelligent identification of the running state of the equipment through a feature extraction module, a fault detection module and an intelligent identification module; and when the signal feedback unit detects an abnormity or a fault, the signal feedback unit sends an alarm signal in time, so that the response speed and accuracy of the system to the equipment fault are improved, and the safe and stable operation of the equipment is ensured.

Description

Technical Field

[0001] The invention relates to the field of automatic control technology, in particular to an intelligent fault detection and real-time hot plug system. Background Art

[0002] The intelligent fault detection and real-time hot-swap system integrates fault detection and hot-swap, improving the reliability, flexibility and safety of industrial systems; through real-time monitoring and intelligent analysis, the system automatically identifies faults and reduces downtime; the hot-swap function supports dynamic equipment replacement without downtime, enhancing system availability; based on the neural network multi-threaded mode analysis of real-time data, it can not only predict the module type of the equipment, but also judge the abnormal situation of the equipment in advance; compared with traditional technologies, this intelligent recognition mechanism improves the ability to recognize complex fault modes and significantly improves the system's adaptability and intelligence.

[0003] The existing PLC system lacks an intelligent fault detection mechanism during the hot-swap process of devices, and is unable to promptly discover and handle problems such as device type mismatch, resulting in reduced system reliability. In addition, the existing hot-swap management unit mainly relies on static configuration and is unable to dynamically manage and optimize when devices are inserted or removed, affecting the flexibility and efficiency of the system. Therefore, an intelligent fault detection and real-time hot-swap system is designed. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent fault detection and real-time hot-swap system to solve the problems that the existing PLC system proposed in the above background technology lacks an effective fault detection mechanism during the hot-swap process of the device, and cannot timely discover and handle the problem of device type mismatch, resulting in reduced system reliability; and the existing hot-swap management unit mainly relies on static configuration, and cannot perform dynamic management and optimization when the device is inserted or removed, which affects the flexibility and efficiency of the system;

[0005] To achieve the above object, the present invention provides an intelligent fault detection and real-time hot swap system.

[0006] A hot-swap management unit is included, and the hot-swap management unit is used to handle the dynamic addition and removal of devices based on the plug-and-play protocol; the hot-swap management unit can identify newly connected devices in real time, automatically configure necessary resources, and ensure that the system does not interrupt operation during the device insertion or removal process; it can quickly perform hot-swap operations when a device failure is detected, optimize the device management and maintenance process, thereby improving the flexibility and reliability of the system and ensuring the continuous operation and stability of the device;

[0007] The data acquisition unit includes a data acquisition module, which is used to collect the status data of the equipment and adopts a fast storage mechanism to ensure that the data will not be lost, maintain the integrity and availability of the data, improve the data access efficiency, and provide reliable support for real-time analysis and fault detection;

[0008] The data collection unit also includes a high-buffer storage module, which is used to receive and store data collected from devices and sensors in real time, using a fast storage mechanism to ensure that data will not be lost, and maintain data integrity and availability;

[0009] An intelligent fault detection unit, which is used to monitor the operating status of the equipment, detect faults and intelligently identify them, monitor the operating status of the equipment, identify and judge faults, thereby improving the response speed and accuracy of the system;

[0010] The intelligent fault detection unit includes a feature extraction module, an intelligent identification module and a fault detection module;

[0011] The feature extraction module performs multi-scale analysis on the signal based on wavelet transform, processes instantaneous and non-stationary signals, and extracts instantaneous features and local features;

[0012] The intelligent identification module is based on a neural network and adopts a multi-threaded mode to analyze real-time data to predict its module type, thereby judging the abnormal situation of the equipment in real time;

[0013] The fault detection module receives data samples based on the high-cache storage module, and continuously monitors these data through anomaly detection algorithms, so as to quickly identify abnormal changes in the device status and determine whether there is a fault;

[0014] The signal feedback unit sends out an alarm signal in time when an abnormality or failure is detected.

[0015] As a further improvement of the present technical solution, the status data includes voltage, current, temperature, vibration, and physical identification parameters, which are used to obtain data on the system operating status and provide basic data for subsequent analysis and fault detection.

[0016] As a further improvement of the technical solution, in the intelligent fault detection unit, the neural network adopts a multi-threaded mode to analyze the operation data of the equipment in real time to predict its module type and judge the abnormal situation of the equipment. The specific detection steps involved are:

[0017] S1.1. Build the network structure of the neural network;

[0018] S1.2, collect the sampled data received by the high-buffer storage module, including voltage, current, temperature, vibration, and physical identification parameters;

[0019] S1.3. Clean, transform and standardize the raw collected data through preprocessing algorithms to ensure data quality;

[0020] S1.4, based on the feature extraction module, wavelet transform is used to extract instantaneous features and local features;

[0021] S1.5. Improve the effectiveness and reliability of intelligent fault detection through prediction by neural network prediction algorithm;

[0022] S1.6. Use the cross entropy loss function to compare the predicted data with the inserted module. If the monitored data differs from the normal pattern, the system will issue an alarm.

[0023] As a further improvement of the technical solution, in S1.1, the specific steps of building the neural network structure are:

[0024] Determine the shape of the input layer to define the input layer to match the number of features; set the number of hidden layers and the number of neurons in each layer, and select the Sigmoid activation function to construct the hidden layer; set the output layer by selecting the number of neurons in the output layer according to the task;

[0025] As a further improvement of the present technical solution, in S1.3, based on a preprocessing algorithm, the preprocessing algorithm can quickly and easily fill in missing data by assuming that the change of data between known points is linear. The steps involved in the preprocessing algorithm are:

[0026] First, use the linear interpolation algorithm:

[0027]

[0028] In the formula, x 0 ,y 0 Indicates the horizontal and vertical coordinates of the first known data point; x 1 ,y 1 Indicates the horizontal and vertical coordinates of the second known data point; x indicates the interpolation position; y indicates the vertical coordinate obtained by interpolation calculation;

[0029] The ordinate obtained by interpolation calculation is used for data conversion. The label encoding algorithm used for data conversion is:

[0030] label(ci)=i;

[0031] In the formula, C represents a list of all categories; ci represents the integer to which each category code is mapped;

[0032] Finally, the result obtained by the label encoding algorithm is used for data standardization. The Min-Max standardization algorithm used for data standardization is:

[0033]

[0034] In the formula, x represents the original data value; xmin represents the minimum value of the feature; xmax represents the maximum value of the feature; and x′ represents the value after Min-Max standardization.

[0035] As a further improvement of the technical solution, in S1.4, the input signal is decomposed layer by layer by wavelet transform to extract instantaneous features and local features, specifically:

[0036] The low-frequency components of the signal are extracted through a low-pass filter to obtain the approximation coefficient;

[0037] A[n]=∑ k x[k]h[2n-k];

[0038] In the formula, x[k] represents the input signal; h[n] represents the filter coefficient of the low-pass filter; 2n represents downsampling; A[n] represents the approximation coefficient; k represents the index variable of each sample point of the input signal;

[0039] The high-frequency components of the signal are extracted through a high-pass filter to obtain detail coefficients;

[0040] D[n]=∑ k x[k]g[2n-k];

[0041] Wherein, x[k] represents the input signal; g[n] represents the filter coefficient of the high-pass filter; and D[n] represents the detail coefficient.

[0042] Through multi-scale analysis of wavelet transform, the instantaneous features of the extracted signal are high-pass filtered, and the local features are low-pass filtered.

[0043] As a further improvement of the present technical solution, in S1.5, module type prediction is performed based on a neural network prediction algorithm, so that the system can accurately identify the working status of the equipment and promptly determine whether there is an abnormality, thereby improving the effectiveness and reliability of intelligent fault detection; the prediction steps of the neural network prediction algorithm are:

[0044] Input layer to hidden layer:

[0045] z (1) =W (1) ·X+b (1) ;

[0046] a (1) =σ(z (1) );

[0047] Where W (1) represents the weight matrix of the first layer; X represents the input feature vector; b (1) represents the bias vector of the first layer; z (1) represents the linear combination result of the first layer; a (1) represents the activation output of the first layer; σ represents the Sigmoid activation function;

[0048] Hidden layer to output layer:

[0049] z (2) =W (2) ·a (1) +b (2) ;

[0050] In the formula, z (2) Represents the linear combination result of the output layer; W (2) represents the weight matrix from the hidden layer to the output layer; a (1) represents the activation output of the hidden layer; b (2) represents the bias vector of the second layer;

[0051] Apply the Sigmoid activation function:

[0052]

[0053] In the formula, a (2) represents the activation output of the output layer, expressed as a probability value; e represents the natural logarithm.

[0054] As a further improvement of the technical solution, in S1.6, the predicted data is compared with the inserted module based on a cross entropy loss function, and the expression of the function is:

[0055] L=-(ylog(a (2) )+(1-y)log(1-a (2) ));

[0056] Where L represents the difference between the model prediction and the true label; y represents the true label; a represents the probability that the sample belongs to the positive class; log represents the natural logarithm.

[0057] As a further improvement of the technical solution, the fault detection module identifies abnormal changes and abnormal behaviors of the device state based on an abnormal detection algorithm, and is used to identify abnormal changes in the device state and determine whether there is a fault. The abnormal detection algorithm is specifically:

[0058]

[0059] In the formula, Z represents Z-score; X represents the sample value to be tested; μ represents the mean of the sample value; σ represents the standard deviation of the sample value;

[0060] The significance of Z-score is that it converts the original data into unitless standardized values, which can be used to compare data in different units or scales. In the standard normal distribution, about 68% of the data will have a Z-score between -1 and 1, and about 95% of the data will have a Z-score between -2 and 2. Observations with an absolute value of Z-score greater than 2 may be considered outliers.

[0061] In the intelligent fault detection unit, the Z-score of the real-time collected data can be calculated and compared with a preset threshold to determine whether the data is abnormal; if the absolute value of the Z-score exceeds the threshold, the system considers that an abnormality has been detected and triggers a further signal feedback mechanism.

[0062] As a further improvement of the present technical solution, the signal feedback unit includes an alarm and indication module, which sends out sound and light signals to remind the user when the system detects that the feature prediction results of the neural network are inconsistent with those of the original module.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1. This intelligent fault detection and real-time hot-swap system integrates advanced fault detection technology to monitor device status in real time and automatically identify faults, thereby reducing manual intervention and potential downtime and improving the overall reliability of the system. The hot-swap function design allows devices to be dynamically added or removed without turning off the power, achieving zero-downtime maintenance and upgrades, and enhancing the availability and flexibility of the system.

[0065] 2. In this intelligent fault detection and real-time hot-swap system, the modular design allows the system to be easily expanded and upgraded as needed to adapt to different industrial application scenarios, improve the system's maintainability and adaptability to future technological developments. The system can collect and analyze large amounts of data, provide in-depth business insights, and support more accurate data-driven decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is the overall flow chart of the present invention;

[0067] The meaning of each number in the figure is:

[0068] 1. Hot-swap management unit; 2. Data acquisition unit; 3. Intelligent fault detection unit; 4. Signal feedback unit. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0070] Example

[0071] See also Figure 1 As shown, an intelligent fault detection and real-time hot swap system is provided, including

[0072] A hot-swap management unit 1, wherein the hot-swap management unit 1 is used to process the dynamic addition and removal of devices based on the plug-and-play protocol; the hot-swap management unit 1 can identify newly connected devices in real time, automatically configure necessary resources, and ensure that the system does not interrupt operation during the device insertion or removal process; it can quickly perform hot-swap operations when a device failure is detected, optimize the device management and maintenance process, thereby improving the flexibility and reliability of the system and ensuring the continuous operation and stability of the device;

[0073] Data acquisition unit 2, the data acquisition unit 2 includes a data acquisition module, the data acquisition module is used to collect status data of the device, the status data includes voltage, current, temperature, vibration, physical identification parameters, and provide basic data for subsequent analysis and fault detection;

[0074] The data collection unit 2 also includes a high-buffer storage module, which is used to receive and store sampled data collected from devices and sensors in real time, and uses a fast storage mechanism to ensure that data will not be lost, thereby maintaining data integrity and availability.

[0075] Intelligent fault detection unit 3, the intelligent fault detection unit 3 is used for monitoring the operation status of the equipment, fault investigation and intelligent identification, monitoring the operation status of the equipment, identifying and judging the fault, thereby improving the response speed and accuracy of the system;

[0076] The intelligent fault detection unit 3 includes a feature extraction module, an intelligent identification module and a fault detection module;

[0077] The feature extraction module can provide time and frequency information simultaneously based on wavelet transform, process instantaneous and non-stationary signals, perform multi-scale analysis on the signal, and extract instantaneous features and local features;

[0078] The intelligent identification module is based on a neural network and adopts a multi-threaded mode to analyze real-time data to predict its module type, thereby judging the abnormal situation of the equipment in real time;

[0079] The fault detection module receives the sampled data based on the high-cache storage module, and continuously monitors the sampled data through the anomaly detection algorithm, so as to quickly identify the abnormal changes in the device status and determine whether there is a fault;

[0080] The signal feedback unit 4 sends out an alarm signal in time when the signal feedback unit 4 detects an abnormality or a fault.

[0081] In the intelligent fault detection unit 3, the neural network adopts a multi-threaded mode to analyze the operation data of the equipment in real time to predict its module type and judge the abnormal situation of the equipment. The specific detection steps involved are:

[0082] S1.1. Build the network structure of the neural network. The specific steps for building the network structure of the neural network are:

[0083] Determine the shape of the input layer to define the input layer to match the number of features; set the number of hidden layers and the number of neurons in each layer, and select the Sigmoid activation function to construct the hidden layer; set the output layer by selecting the number of neurons in the output layer according to the task;

[0084] S1.2, collect the sampled data received by the high-buffer storage module, including voltage, current, temperature, vibration, and physical identification parameters;

[0085] S1.3. Clean, transform and standardize the raw collected data through preprocessing algorithms to ensure data quality. The steps involved in the preprocessing algorithm are:

[0086] First, use the linear interpolation algorithm:

[0087]

[0088] In the formula, x 0 ,y 0 Indicates the horizontal and vertical coordinates of the first known data point; x 1 ,y 1 Represents the horizontal and vertical coordinates of the second known data point; x represents the interpolation position; y represents the vertical coordinate obtained by interpolation calculation.

[0089] The ordinate obtained by interpolation calculation is used for data conversion. The label encoding algorithm used for data conversion is:

[0090] label(ci)=i;

[0091] In the formula, C represents a list of all categories; ci represents the integer to which each category code is mapped;

[0092] Finally, the result obtained by the label encoding algorithm is used for data standardization. The Min-Max standardization algorithm used for data standardization is:

[0093]

[0094] In the formula, x represents the original data value; xmin represents the minimum value of the feature; xmax represents the maximum value of the feature; x′ represents the value after Min-Max standardization;

[0095] S1.4, based on the feature extraction module, wavelet transform is used to extract instantaneous features and local features, wherein the wavelet transform decomposes the input signal layer by layer to extract instantaneous features and local features, specifically:

[0096] The low-frequency components of the signal are extracted through a low-pass filter to obtain the approximation coefficient;

[0097] A[n]=∑ k x[k]h[2n-k];

[0098] In the formula, x[k] represents the input signal; h[n] represents the filter coefficient of the low-pass filter; 2n represents downsampling; A[n] represents the approximation coefficient; k represents the index variable of each sample point of the input signal;

[0099] The high-frequency components of the signal are extracted through a high-pass filter to obtain detail coefficients;

[0100] D[n]=∑ k x[k]g[2n-k];

[0101] Wherein, x[k] represents the input signal; g[n] represents the filter coefficient of the high-pass filter; and D[n] represents the detail coefficient.

[0102] Through multi-scale analysis of wavelet transform, the instantaneous features of the extracted signal are high-pass filtered, and the local features are low-pass filtered.

[0103] S1.5. Improve the effectiveness and reliability of intelligent fault detection through prediction by a neural network prediction algorithm, and perform module type prediction. The prediction steps of the neural network prediction algorithm are as follows:

[0104] Input layer to hidden layer:

[0105] z (1) =W (1) ·X+b (1) ;

[0106] a (1) =σ(z (1) );

[0107] Where W (1)represents the weight matrix of the first layer; X represents the input feature vector; b (1) represents the bias vector of the first layer; z (1) represents the linear combination result of the first layer; a (1) represents the activation output of the first layer; σ represents the Sigmoid activation function;

[0108] Hidden layer to output layer:

[0109] z (2) =W (2) ·a (1) +b (2) ;

[0110] In the formula, z (2) Represents the linear combination result of the output layer; W (2) represents the weight matrix from the hidden layer to the output layer; a (1) represents the activation output of the hidden layer; b (2) represents the bias vector of the second layer;

[0111] Apply the Sigmoid activation function:

[0112]

[0113] In the formula, a (2) represents the activation output of the output layer, expressed as a probability value; e represents the natural logarithm.

[0114] S1.6. Use the cross entropy loss function to compare the predicted data with the inserted module. If the monitored data is different from the normal mode, the system will issue an alarm. The predicted data is compared with the inserted module based on the cross entropy loss function. The expression of this function is:

[0115] L=-(ylog(a (2) )+(1-y)log(1-a (2) ));

[0116] Where L represents the difference between the model prediction and the true label; y represents the true label; a represents the probability that the sample belongs to the positive class; log represents the natural logarithm.

[0117] The anomaly detection algorithm is specifically:

[0118]

[0119] In the formula, Z represents Z-score; X represents the sample value to be tested; μ represents the mean of the sample value; σ represents the standard deviation of the sample value.

[0120] The significance of Z-score is that it converts the original data into unitless standardized values, which can be used to compare data in different units or scales. In the standard normal distribution, about 68% of the data will have a Z-score between -1 and 1, and about 95% of the data will have a Z-score between -2 and 2. Observations with an absolute value of Z-score greater than 2 may be considered outliers.

[0121] In the intelligent fault detection unit 3, the Z-score of the real-time collected data can be calculated and compared with a preset threshold to determine whether the data is abnormal; if the absolute value of the Z-score exceeds the threshold, the system considers that an abnormality has been detected and triggers a further signal feedback mechanism.

[0122] The signal feedback unit 4 includes an alarm and indication module. When the system detects that the feature prediction results of the neural network and the original module are inconsistent, it sends out sound and light signals to remind the user.

[0123] 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 by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. An intelligent fault detection and real-time hot swap system, characterized in that: include A hot-swap management unit (1), the hot-swap management unit (1) being used to process dynamic addition and removal of devices based on a plug-and-play protocol; A data acquisition unit (2), the data acquisition unit (2) comprising a data acquisition module, the data acquisition module being used to acquire status data of the device; The data collection unit (2) also includes a high-buffer storage module, which is used to receive and store sampled data collected from the device and the sensor in real time, and uses a fast storage mechanism to ensure that the data will not be lost; An intelligent fault detection unit (3), the intelligent fault detection unit (3) is used to monitor the operating status of the equipment, detect faults and intelligently identify them, monitor the operating status of the equipment, identify and judge faults; The intelligent fault detection unit (3) comprises a feature extraction module, an intelligent identification module and a fault detection module; The feature extraction module can provide time and frequency information simultaneously based on wavelet transform, process instantaneous and non-stationary signals, perform multi-scale analysis on the signal, and extract instantaneous features and local features; The intelligent identification module is based on a neural network and adopts a multi-threaded mode to analyze real-time data to predict its module type, thereby judging the abnormal situation of the equipment in real time; The fault detection module receives the sampled data based on the high-cache storage module, and continuously monitors the sampled data through the anomaly detection algorithm, so as to quickly identify the abnormal changes in the device status and determine whether there is a fault; A signal feedback unit (4), wherein when the signal feedback unit (4) detects an abnormality or a fault, the signal feedback unit (4) promptly sends out an alarm signal.

2. The intelligent fault detection and real-time hot-swap system according to claim 1, characterized in that: The status data includes voltage, current, temperature, vibration, and physical identification parameters, which provide basic data for subsequent analysis and fault detection.

3. The intelligent fault detection and real-time hot-swap system according to claim 1, characterized in that: In the intelligent fault detection unit (3), the neural network adopts a multi-threaded mode to analyze the operation data of the device in real time to predict its module type and determine the abnormal situation of the device. The specific detection steps involved are: S1.

1. Build the network structure of the neural network; S1.2, collect the sampled data received by the high-buffer storage module, including voltage, current, temperature, vibration, and physical identification parameters; S1.

3. Clean, transform and standardize the raw collected data through preprocessing algorithms to ensure data quality; S1.4, based on the feature extraction module, wavelet transform is used to extract instantaneous features and local features; S1.

5. Improve the effectiveness and reliability of intelligent fault detection through prediction by neural network prediction algorithm; S1.

6. Use the cross entropy loss function to compare the predicted data with the inserted module. If the monitored data differs from the normal pattern, the system will issue an alarm.

4. The intelligent fault detection and real-time hot-swap system according to claim 6, characterized in that: In S1.1, the specific steps of building the neural network structure are: Determine the shape of the input layer to define the input layer to match the number of features; set the number of hidden layers and the number of neurons in each layer, and select the Sigmoid activation function to construct the hidden layer; set the output layer by selecting the number of neurons in the output layer according to the task.

5. The intelligent fault detection and real-time hot swap system according to claim 3 is characterized in that: In S1.3, the steps involved in the preprocessing algorithm are: First, use the linear interpolation algorithm: In the formula, x0, y0 represent the horizontal and vertical coordinates of the first known data point; x1, y1 represent the horizontal and vertical coordinates of the second known data point; x represents the interpolation position; y represents the vertical coordinate obtained by interpolation calculation; The ordinate obtained by interpolation calculation is used for data conversion. The label encoding algorithm used for data conversion is: label(ci)=i; In the formula, C represents a list of all categories; ci represents the integer to which each category code is mapped; Finally, the result obtained by the label encoding algorithm is used for data standardization. The Min-Max standardization algorithm used for data standardization is: In the formula, x represents the original data value; xmin represents the minimum value of the feature; xmax represents the maximum value of the feature; and x′ represents the value after Min-Max standardization.

6. The intelligent fault detection and real-time hot swap system according to claim 1, characterized in that: In S1.4, the input signal is decomposed layer by layer by wavelet transform to extract instantaneous features and local features, specifically: The low-frequency components of the signal are extracted through a low-pass filter to obtain the approximation coefficient; A[n]=Σ k x[k]h[2n-k]; In the formula, x[k] represents the input signal; h[n] represents the filter coefficient of the low-pass filter; 2n represents downsampling; A[n] represents the approximation coefficient; k represents the index variable of each sample point of the input signal; The high-frequency components of the signal are extracted through a high-pass filter to obtain detail coefficients; D[n]=∑ k x[k]g[2n-k]; Wherein, x[k] represents the input signal; g[n] represents the filter coefficient of the high-pass filter; and D[n] represents the detail coefficient.

7. The intelligent fault detection and real-time hot swap system according to claim 6, characterized in that: In S1.5, module type prediction is performed based on a neural network prediction algorithm, and the prediction steps of the neural network prediction algorithm are: Input layer to hidden layer: z (1) =W (1) ·X+b (1) ; a (1) =σ(z (1) ); Where W (1) represents the weight matrix of the first layer; X represents the input feature vector; b (1) represents the bias vector of the first layer; z (1) represents the linear combination result of the first layer; a (1) represents the activation output of the first layer; σ represents the Sigmoid activation function; Hidden layer to output layer: z (2) =W (2) ·a (1) +b (2) ; In the formula, z (2) Represents the linear combination result of the output layer; W (2) represents the weight matrix from the hidden layer to the output layer; a (1) represents the activation output of the hidden layer; b (2) represents the bias vector of the second layer; Apply the Sigmoid activation function: In the formula, a (2) represents the activation output of the output layer, expressed as a probability value; e represents the natural logarithm.

8. The intelligent fault detection and real-time hot swap system according to claim 1, characterized in that: In S1.6, the predicted data is compared with the inserted module based on the cross entropy loss function, and the expression of the function is: L6-(ylog(a). (2) )+(1-y)log(1-a (2) )) Where L represents the difference between the model prediction and the true label; y represents the true label; a represents the probability that the sample belongs to the positive class; log represents the natural logarithm.

9. The intelligent fault detection and real-time hot swap system according to claim 1, characterized in that: The anomaly detection algorithm is specifically: In the formula, Z represents Z-score; X represents the sample value to be tested; μ represents the mean of the sample value; σ represents the standard deviation of the sample value.

10. The intelligent fault detection and real-time hot swap system according to claim 1, characterized in that: The signal feedback unit (4) comprises an alarm and indication module, which emits an audible and visual signal to alert the user when the system detects that the neural network predicts the characteristics of the original module inconsistently.

Images