Power cable detection method based on pyroelectric particle detection

Abstract

The invention relates to the technical field of power cables, in particular to a power cable detection method based on pyroelectric particle detection. Comprising the following steps: determining an optimal detection distance for measuring the pyroelectric particle content of the cable; dividing a plurality of temperature intervals; all types of heat release particles in each temperature interval are endowed with influence weights; constructing and training a fault diagnosis model; during detection, in the optimal detection distance of a to-be-detected cable, the sensor array is used for obtaining the contents of various heat release particles and the temperature of the to-be-detected cable; according to the temperature interval of the temperature of the to-be-measured cable, obtaining influence weights of various pyroelectric particles of the to-be-measured cable; weighting the pyroelectric particle content of the to-be-detected cable and the corresponding influence weight, and calculating an influence score of the to-be-detected cable; and inputting the influence score of the to-be-detected cable into the fault diagnosis model to obtain a diagnosis result. The method takes the pyroelectric particles as a detection target, analyzes the content of the pyroelectric particles through a deep learning model, and completes the heating fault diagnosis. Compared with a traditional cable detection method, the cable fault detection precision is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of power cable technology, and in particular to a power cable testing method based on pyroelectric particle detection. Background Technology

[0002] With the advancement of urbanization, the construction of power distribution networks has accelerated, leading to a significant increase in the number of substations. Cables, as a crucial component of the power grid system's normal operation, are vital carriers of power and signal transmission in substations, and are found almost throughout the entire substation network. However, the rapid development of the power grid has also resulted in frequent cable fires, making research into cable fire detection technologies a critical issue for current power distribution systems. Because cable trenches are located underground, it is often difficult to detect and address fires caused by cable faults in advance, leading to serious fire safety incidents. Furthermore, under high-temperature heat radiation and fire conditions, the polymers in cables can produce toxic and harmful gases such as carbon dioxide, carbon monoxide, and hydrogen chloride, causing significant damage to the natural environment and human health.

[0003] In the early stages of a cable fire, only low concentrations of flammable and non-flammable gases are present. The outer layer of the conductor is typically covered with insulation material, currently mostly polyvinyl chloride (PVC) and cross-linked polyethylene (XLPE). When the insulation material is heated to 150°C, it emits an unpleasant rubbery smell, primarily due to the gas released from the melting and volatilization of the outer insulation caused by overheating of the conductor. This gas is released from the decomposition of the cable's outer insulation material when heated, and it decomposes into pyrolytic particles. If the composition and concentration of these pyrolytic particles can be determined under specific environmental conditions (such as temperature and humidity) and at different measurement distances, precise detection can be achieved. Using this concentration as an early detection indicator for cable fires can advance the fire alarm time and avoid interference from dust and gaseous particles.

[0004] The current power cable fire detection technology system mainly includes three major monitoring methods: fiber optic temperature measurement, smoke detection, and thermal imaging monitoring. All three have significant technical bottlenecks in practical applications. Fiber optic temperature measurement systems, based on the Raman scattering principle, can achieve continuous temperature monitoring along the entire cable. However, they require laying dedicated fiber optic cables along the cable, making construction difficult in complex pipe networks and resulting in overall cabling costs as high as 80-120 yuan / meter. Furthermore, system debugging requires specialized optical instruments, making rapid deployment in existing cable networks difficult. Smoke detectors, as traditional fire-fighting equipment, have the advantage of easy installation, but their working principle relies on smoke particles triggering photoelectric sensing. They are slow to respond to the trace amounts of pyrolysis aerosols generated during the initial smoldering stage of the cable; experimental data shows an average alarm delay of 8-12 minutes in cable fires. Thermal imaging monitoring systems, through infrared thermal radiation detection, can achieve non-contact temperature measurement, but high-precision equipment costs over 50,000 yuan per unit and is prone to false alarms due to environmental water mist and dust. Actual measurements at a substation showed a false alarm rate as high as 23%. Summary of the Invention

[0005] In a first aspect, the present invention discloses a method for detecting power cables based on pyroelectric particle detection, the specific method of which is as follows:

[0006] Determine the optimal detection distance for measuring the pyroelectric particle content in the cable;

[0007] Divide the temperature range into several intervals, using the release content of various pyrolytic particles as the basis for dividing the temperature range;

[0008] Assign influence weights to all types of pyroelectric particles in each temperature range;

[0009] Build and train a fault diagnosis model;

[0010] During testing, at the optimal detection distance of the cable under test, the content of various pyroelectric particles and the temperature of the cable under test are obtained by a sensor array.

[0011] Based on the temperature range of the cable under test, obtain the influence weights of various pyroelectric particles in the cable under test.

[0012] The influence score of the cable under test is calculated by weighting the pyroelectric particle content of the cable under test and the corresponding influence weights.

[0013] Input the influence score of the cable under test into the fault diagnosis model to obtain the diagnosis results.

[0014] Furthermore, the optimal detection distance for measuring the pyroelectric particle content of the cable was determined using the following method:

[0015] By maintaining the same temperature conditions and adjusting the sensor distance, the content of pyroelectric particles can be analyzed and obtained.

[0016] The location with the highest total content of pyroelectric particles is selected as the optimal detection distance.

[0017] Furthermore, several temperature ranges are divided, as follows:

[0018] Cut a section of cable sheath and place it into the heating test apparatus;

[0019] Control the heating device to reach different temperature points;

[0020] At different temperature points, the gas is introduced into the gas chromatograph through the gas outlet valve for pyroelectric particle analysis to obtain the composition and content of each pyroelectric particle;

[0021] Several analytical arrays were constructed based on temperature nodes, pyroelectric particle composition, and pyroelectric particle content.

[0022] Cluster the analysis arrays using the K-Means algorithm;

[0023] Temperature nodes corresponding to the same type of analysis array are divided into the same temperature range.

[0024] Furthermore, the heating test device used to divide several temperature ranges includes a heating stage, which is connected to an inlet valve and an outlet valve respectively. The outlet valve is connected to a gas chromatograph through a cooling pipe and a gas valve. The heating stage temperature is controlled by an XMT614 intelligent temperature controller.

[0025] Furthermore, influence weights are assigned to all types of pyroelectric particles within each temperature range, as follows:

[0026] Obtain the content of pyrolytic particles in a single temperature range;

[0027] The influence weight of a single pyroelectric particle on the overall pyroelectric particles at a single temperature is calculated by dividing the content of a single pyroelectric particle by the total content of pyroelectric particles.

[0028] Furthermore, the K-Means algorithm is used to cluster the analysis arrays, as follows:

[0029] Randomly select K initial cluster centers Ci from the dataset, 1≤i≤k, calculate the Euclidean distance between the remaining data objects and the cluster centers Ci, find the cluster center Ci that is closest to the target data object, and assign the data object to the cluster corresponding to the cluster center Ci;

[0030] Calculate the average value of the data objects in each cluster, and use this value as the new cluster center for the next iteration. Continue this process until the cluster centers no longer change or the maximum number of iterations is reached. The Euclidean distance between data objects and cluster centers in space is calculated using the following formula:

[0031]

[0032] Where X is the data object; Ci is the i-th cluster center; m is the dimension of the data object; Xj and Cij are the j-th attribute values ​​of X and Ci, respectively; the SSE calculation formula for the entire dataset is:

[0033]

[0034] Wherein, the size of SSE indicates the quality of the clustering result; k is the number of clusters.

[0035] Furthermore, the fault diagnosis model is a multilayer perceptron, and the specific construction method is as follows:

[0036] Initialize the network: Assume the input layer has n nodes, the output layer has m nodes, and the hidden layers have l layers; the weights of each connection from the input layer to the hidden layers are... The weights of each connection from the hidden layer to the output layer are: The bias from the input layer to the hidden layer is a. j The bias from the hidden layer to the output layer is b. k The learning rate is The activation function is g(x), which is set as the Sigmoid function, and the specific formula is as follows:

[0037]

[0038] Calculating the output of the hidden layer: The output of the hidden layer consists of the input signal and weights. and bias a j Composition, output H j for:

[0039]

[0040] The output of the output layer is calculated from the output of the hidden layer. Weight and bias b k Composition, output O k for:

[0041]

[0042] Calculation error: The error formula is as follows: ,

[0043] Where Y k Let Y be the desired output. k -O k =e k ,

[0044] Then the error E can be expressed as:

[0045]

[0046]

[0047] In the formula, i=1,...n, j=1,...l, k=1,...,m;

[0048] Updating weights: The weights ω and ω… as well as the biases a and b are updated through backpropagation. The specific formulas are as follows:

[0049]

[0050]

[0051]

[0052] .

[0053] Furthermore, the fault diagnosis model is a support vector machine, and the specific construction method is as follows:

[0054] Given dataset In the sample space, the dividing hyperplane is represented as:

[0055]

[0056] Where w=(w1,...,w p ) is the normal vector of the hyperplane, and b is the offset;

[0057] Calculate the distance from any point in the sample space to the hyperplane:

[0058]

[0059] in, Assuming there is A linear equation is used to divide the data into two classes. When y = +1, we have... When y = -1, we have ;

[0060] Find the maximum boundary line, let:

[0061]

[0062] get This formula represents the functional margin, and the margin between two heterogeneous support vectors is:

[0063]

[0064] The goal is to find the hyperplane H with the largest spacing, i.e.:

[0065]

[0066]

[0067] Rewritten as:

[0068]

[0069]

[0070] Solving for w and b using the dual problem is relatively easy, and it can be optimized using the Lagrange multiplier method as follows:

[0071]

[0072]

[0073] Solving for the hyperplane H yields:

[0074]

[0075] Where k(x) i ,x j ) is a kernel function, responsible for mapping samples from a low-dimensional space to a high-dimensional space to make them linearly separable.

[0076] Secondly, the present invention discloses a power cable detection device for pyroelectric particle detection, comprising: a pyroelectric particle acquisition module, a distance control module, a temperature control module, a humidity sensing module, a gas ventilation module, and a fire extinguishing module.

[0077] Temperature control module controls the temperature of the heating cable sheath;

[0078] The pyroelectric particle acquisition module identifies the release of pyroelectric particles from the cable sheath at different temperatures and transmits the data to the pyroelectric particle recognition module after A / D conversion by a micro-sensor.

[0079] Distance control module, used to control the distance between the sensor array and the pyroelectric particles emitted by the burning cable;

[0080] The humidity sensing module collects the humidity in the air during combustion;

[0081] Gas ventilation module and fire extinguishing module ensure operational safety.

[0082] Due to the adoption of the above technical solutions, this application has the following beneficial effects:

[0083] 1. This application uses pyroelectric particles as the detection target and analyzes the content of pyroelectric particles through a deep learning model to complete the diagnosis of heating faults. Compared with traditional cable testing methods, it significantly improves the accuracy of cable fault detection.

[0084] 2. To address the issue that the content of pyroelectric particles may deviate due to changes in detection distance, this application provides a method for optimizing the detection distance, thereby further improving the accuracy of cable fault detection.

[0085] 3. The content of different pyrogenetic particles released by cable sheath varies greatly with different ambient temperatures. If we can identify the types of pyrogenetic particles with more obvious fault characteristics in each temperature range and assign them higher influence weights, we can amplify the impact of fault characteristics on the deep learning model and further improve the accuracy of cable fault detection.

[0086] 4. This application uses the influence score of self-constructed cables as the input information of the deep learning model. Compared with inputting the content of all pyroelectric particles into the deep learning model, it significantly reduces the number of model variables and reduces the parameter complexity of the deep learning model.

[0087] 5. The power cable detection device for pyroelectric particle detection provided in this application identifies pyroelectric particles through a multi-sensor array, and then uses temperature, humidity and distance to control the variables during measurement, taking into account the environment when multiple cables are heating up, to perform high-precision and high-sensitivity monitoring of pyroelectric particles in the very early stage of cable fire, providing strong technical support for the very early monitoring of subsequent cable fires.

[0088] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0089] The accompanying drawings of this invention are described below.

[0090] Figure 1 This is a schematic diagram of the process of the present invention.

[0091] Figure 2 This is a schematic diagram of the heating test apparatus.

[0092] Figure 3 This is a schematic diagram of a power cable testing device for pyroelectric particle detection. Detailed Implementation

[0093] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0094] Example 1:

[0095] A power cable testing method based on pyroelectric particle detection, such as Figure 1 As shown, the specific steps are as follows:

[0096] S1. Determine the optimal detection distance for measuring the pyroelectric particle content of the cable. The specific steps are as follows:

[0097] S11. Under the same fixed temperature conditions, adjust the sensor distance and analyze the content of pyroelectric particles.

[0098] S12. Select the location with the highest total content of pyroelectric particles as the optimal detection distance.

[0099] S2. Divide the temperature range into several intervals, using the release content of various pyrolytic particles as the basis for dividing the temperature range. The specific steps are as follows:

[0100] S21. Cut the cable sheath and place it into the heating test device.

[0101] S22. Control the heating device to reach different temperature nodes.

[0102] S23. At different temperature points, the gas is introduced into the gas chromatograph through the gas outlet valve for pyrogenetic particle analysis to obtain the composition and content of each pyrogenetic particle.

[0103] S24. Construct several analysis arrays based on temperature nodes, pyroelectric particle composition, and pyroelectric particle content.

[0104] S25. Cluster each analysis array using the K-Means algorithm.

[0105] S26. Temperature nodes corresponding to the same type of analysis array are divided into the same temperature range.

[0106] In step S25, the K-Means algorithm pre-specifies the initial number of clusters and initializes the cluster centers. Based on the distance between samples, it divides the sample set into clusters according to the similarity between the data objects and the cluster centers. It continuously updates the positions of the cluster centers, continuously reducing the sum of squared errors (SSE) of the clusters. Clustering ends when the SSE no longer changes or the objective function converges, yielding the final result. The specific method is as follows:

[0107] First, randomly select K initial cluster centers Ci (1≤i≤k) from the dataset. Calculate the Euclidean distance between the remaining data objects and the cluster centers Ci. Find the cluster center Ci that is closest to the target data object and assign the data object to the cluster corresponding to the cluster center Ci.

[0108] Next, the average value of the data objects in each cluster is calculated, and this value is used as the new cluster center for the next iteration. This process continues until the cluster centers no longer change or the maximum number of iterations is reached. The Euclidean distance between a data object and its cluster center in space is calculated using the following formula:

[0109]

[0110] Where X is the data object; Ci is the i-th cluster center; m is the dimension of the data object; Xj and Cij are the j-th attribute values ​​of X and Ci, respectively. The SSE calculation formula for the entire dataset is:

[0111]

[0112] Wherein, the size of SSE indicates the quality of the clustering result; k is the number of clusters.

[0113] In this embodiment, the heating test device used to divide the temperature range is, for example... Figure 2 As shown, it includes a heating stage, which is connected to an inlet valve and an outlet valve respectively. The outlet valve is connected to a gas chromatograph through a cooling pipe and a gas valve. The heating stage temperature is controlled by an XMT614 intelligent temperature controller.

[0114] S3. Assign influence weights to all types of pyroelectric particles in each temperature range. The specific steps are as follows:

[0115] S31. Obtain the content of each pyroelectric particle in a single temperature range.

[0116] S32. Divide the content of a single pyroelectric particle by the total content of pyroelectric particles to calculate the influence weight of a single pyroelectric particle on the overall pyroelectric particles at a single temperature.

[0117] S4. Construct and train the fault diagnosis model.

[0118] In one embodiment, the fault diagnosis model is a multilayer perceptron, and the specific construction method is as follows:

[0119] Initialize the network: Assume the input layer has n nodes, the output layer has m nodes, and the hidden layers have l layers; the weights of each connection from the input layer to the hidden layers are... The weights of each connection from the hidden layer to the output layer are: The bias from the input layer to the hidden layer is a. j The bias from the hidden layer to the output layer is b. k The learning rate is The activation function is g(x), which is set as the Sigmoid function, and the specific formula is as follows:

[0120]

[0121] Calculating the output of the hidden layer: The output of the hidden layer consists of the input signal and weights. and bias a j Composition, output H j for:

[0122]

[0123] The output of the output layer is calculated from the output of the hidden layer. Weight and bias b k Composition, output O k for:

[0124]

[0125] Calculation error: The error formula is as follows: ,

[0126] Where Y k Let Y be the desired output. k -O k =e k ,

[0127] Then the error E can be expressed as:

[0128]

[0129]

[0130] In the formula, i=1,...n, j=1,...l, k=1,...,m;

[0131] Updating weights: The weights ω and ω… as well as the biases a and b are updated through backpropagation. The specific formulas are as follows:

[0132]

[0133]

[0134]

[0135] .

[0136] In another embodiment, the fault diagnosis model is a support vector machine, and the specific construction method is as follows:

[0137] Given dataset In the sample space, the dividing hyperplane is represented as:

[0138]

[0139] Where w=(w1,...,w p ) is the normal vector of the hyperplane, and b is the offset;

[0140] Calculate the distance from any point in the sample space to the hyperplane:

[0141]

[0142] in, Assuming there is A linear equation is used to divide the data into two classes. When y = +1, we have... When y = -1, we have ;

[0143] Find the maximum boundary line, let:

[0144]

[0145] get This formula represents the functional margin, and the margin between two heterogeneous support vectors is:

[0146]

[0147] The goal is to find the hyperplane H with the largest spacing, i.e.:

[0148]

[0149]

[0150] Rewritten as:

[0151]

[0152]

[0153] Solving for w and b using the dual problem is relatively easy, and it can be optimized using the Lagrange multiplier method as follows:

[0154]

[0155]

[0156] Solving for the hyperplane H yields:

[0157]

[0158] Where k(x) i ,x j ) is a kernel function, responsible for mapping samples from a low-dimensional space to a high-dimensional space to make them linearly separable.

[0159] S5. During detection, at the optimal detection distance of the cable under test, the sensor array acquires the content of various pyroelectric particles and the temperature of the cable under test. Through the A / D conversion circuit in the microcontroller, the converted multi-channel sensor signals are transmitted to the pyroelectric particle recognition module via RS485.

[0160] S6. Based on the temperature range of the cable under test, obtain the influence weights of various pyroelectric particles in the cable under test.

[0161] S7. The influence score of the cable under test is calculated by weighting the pyrogenetic particle content of the cable under test and the corresponding influence weight.

[0162] S8. Input the influence score of the cable under test into the fault diagnosis model to obtain the diagnosis results.

[0163] Example 2:

[0164] A power cable testing device for pyroelectric particle detection, such as Figure 3 As shown, it includes: a pyroelectric particle acquisition module, a distance control module, a temperature control module, a humidity sensing module, a gas ventilation module, and a fire extinguishing module;

[0165] The temperature control module identifies the release of pyroelectric particles from the cable insulation material at different temperatures by heating it. The distance control module controls the distance between the sensor array and the pyroelectric particles emitted by the burning cable. The pyroelectric particle acquisition module identifies the composition and concentration of multiple gases through a MOS sensor array. A humidity sensor collects the humidity in the air during combustion. A gas ventilation and fire extinguishing module ensures the safety of the experiment. After A / D conversion by the microcontroller, data is transmitted via RS485 protocol to the pyroelectric particle identification module for real-time monitoring of the experiment. By identifying pyroelectric particles through a multi-sensor array and using temperature, humidity, and distance control variables during measurement, considering the environment during multiple cable heating events, high-precision and high-sensitivity monitoring of pyroelectric particles in the very early stages of cable fires is achieved, providing strong technical support for subsequent early detection of cable fires.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting power cables based on pyroelectric particle detection, characterized in that, The specific method is as follows: Determine the optimal detection distance for measuring the pyroelectric particle content in the cable; Divide the temperature range into several intervals, using the release content of various pyrolytic particles as the basis for dividing the temperature range; Assign influence weights to all types of pyroelectric particles in each temperature range; Build and train a fault diagnosis model; During testing, at the optimal detection distance of the cable under test, the content of various pyroelectric particles and the temperature of the cable under test are obtained by a sensor array. Based on the temperature range of the cable under test, obtain the influence weights of various pyroelectric particles in the cable under test. The influence score of the cable under test is calculated by weighting the pyroelectric particle content of the cable under test and the corresponding influence weights. Input the influence score of the cable under test into the fault diagnosis model to obtain the diagnosis results.

2. The power cable testing method based on pyroelectric particle detection as described in claim 1, characterized in that, The optimal detection distance for measuring the pyroelectric particle content of cables is determined using the following method: By maintaining the same temperature conditions and adjusting the sensor distance, the content of pyroelectric particles can be analyzed and obtained. The location with the highest total content of pyroelectric particles is selected as the optimal detection distance.

3. The power cable testing method based on pyroelectric particle detection as described in claim 1, characterized in that, Divide the temperature range into several intervals, as follows: Cut a section of cable sheath and place it into the heating test apparatus; Control the heating device to reach different temperature points; At different temperature points, the gas is introduced into the gas chromatograph through the gas outlet valve for pyroelectric particle analysis to obtain the composition and content of each pyroelectric particle; Several analytical arrays were constructed based on temperature nodes, pyroelectric particle composition, and pyroelectric particle content. Cluster the analysis arrays using the K-Means algorithm; Temperature nodes corresponding to the same type of analysis array are divided into the same temperature range.

4. The power cable testing method based on pyroelectric particle detection as described in claim 3, characterized in that, The heating test apparatus used to divide several temperature ranges includes a heating stage, which is connected to an inlet valve and an outlet valve respectively. The outlet valve is connected to a gas chromatograph through a cooling pipe and a gas valve. The heating stage temperature is controlled by an XMT614 intelligent temperature controller.

5. The power cable testing method based on pyroelectric particle detection as described in claim 3, characterized in that, Assign influence weights to all types of pyroelectric particles within each temperature range, as follows: Obtain the content of pyrolytic particles in a single temperature range; The influence weight of a single pyroelectric particle on the overall pyroelectric particles at a single temperature is calculated by dividing the content of a single pyroelectric particle by the total content of pyroelectric particles.

6. The power cable testing method based on pyroelectric particle detection as described in claim 3, characterized in that, The K-Means algorithm is used to cluster the analysis arrays. The specific method is as follows: Randomly select K initial cluster centers Ci from the dataset, 1≤i≤k, calculate the Euclidean distance between the remaining data objects and the cluster centers Ci, find the cluster center Ci that is closest to the target data object, and assign the data object to the cluster corresponding to the cluster center Ci; Calculate the average value of the data objects in each cluster, use this value as the new cluster center, and proceed with the next iteration until the cluster center no longer changes or the maximum number of iterations is reached. The formula for calculating the Euclidean distance between data objects and cluster centers in space is: Where X is the data object; Ci is the i-th cluster center; m is the dimension of the data object; Xj and Cij are the j-th attribute values ​​of X and Ci, respectively; the SSE calculation formula for the entire dataset is: Where SSE represents the quality of the clustering result; k is the number of clusters.

7. The power cable testing method based on pyroelectric particle detection as described in claim 1, characterized in that, The fault diagnosis model is a multilayer perceptron, and the specific construction method is as follows: Initialize the network: Assume the input layer has n nodes, the output layer has m nodes, and the hidden layers have l layers; the weights of each connection from the input layer to the hidden layers are... The weights of each connection from the hidden layer to the output layer are: The bias from the input layer to the hidden layer is a. j The bias from the hidden layer to the output layer is b. k The learning rate is The activation function is g(x), which is set as the Sigmoid function, and the specific formula is as follows: Calculating the output of the hidden layer: The output of the hidden layer consists of the input signal and weights. and bias a j Composition, output H j for: The output of the output layer is calculated from the output of the hidden layer. Weight and bias b k Composition, output O k for Calculation error: The error formula is as follows: , where Y k Let Y be the desired output. k -O k =e k Then the error E can be expressed as: In the formula, i=1,...n, j=1,...l, k=1,...,m; updating weights: weights ω and ω⋯, as well as biases a and b, are updated through backpropagation, as shown in the following formula: .

8. The power cable testing method based on pyroelectric particle detection as described in claim 1, characterized in that, The fault diagnosis model is a support vector machine, and the specific construction method is as follows: Given a dataset In the sample space, the dividing hyperplane is represented as: Where w=(w1,...,w p Let y be the normal vector of the hyperplane, and b be the offset; calculate the distance from any point in the sample space to the hyperplane: in, Assuming there is A linear equation is used to divide the data into two classes. When y = +1, we have... When y = -1, we have Find the maximum dividing line of the boundary, let: get This formula represents the functional margin, and the margin between two heterogeneous support vectors is: The goal is to find the hyperplane H with the largest spacing, i.e.: Rewritten as: Solving for w and b using the dual problem is relatively easy, and it can be optimized using the Lagrange multiplier method as follows: Solving for the hyperplane H yields: Where k(x) i ,x j ) is a kernel function, responsible for mapping samples from a low-dimensional space to a high-dimensional space to make them linearly separable.

9. A power cable testing device for pyroelectric particle detection, characterized in that, include: The system includes a pyroelectric particle acquisition module, a distance control module, a temperature control module, a humidity sensing module, a gas ventilation module, and a fire extinguishing module. Temperature control module controls the temperature of the heating cable sheath; The pyroelectric particle acquisition module identifies the release of pyroelectric particles from the cable sheath at different temperatures and transmits the data to the pyroelectric particle recognition module after A / D conversion by a micro-sensor. Distance control module, used to control the distance between the sensor array and the pyroelectric particles emitted by the burning cable; The humidity sensing module collects the humidity in the air during combustion; Gas ventilation module and fire extinguishing module ensure operational safety.

Images