A high-rise building indoor fire source positioning method based on glass surface temperature

Abstract

The application discloses a high-rise building indoor fire source positioning method based on a glass surface temperature, and steps include: arranging m wireless thermocouples on the outer surface of the glass; measuring the horizontal distance from the wireless thermocouple at the center point position in the wireless thermocouple array to the walls on both sides of the fire room; saving and sending the collected temperature data signals every set time period t to a data processing end for average processing; taking the average real-time temperature glass data and the horizontal distance between the center wireless thermocouple and the walls on both sides as input data to obtain a data set; inputting the data set into a prediction program combined with a trained multilayer perception and KNN algorithm; and obtaining the predicted position block of the fire source through back propagation and category voting. The application realizes accurate identification of the position of the fire source, is not limited by a fire-fighting system and installation, has high deployment flexibility, high real-time reliability, meets the application of various practical scenes, and has strong universality and real-time performance.

Description

TECHNICAL FIELD

[0001] The application relates to a high-rise building indoor fire source positioning method based on a glass surface temperature and belongs to the technical field of fire source positioning. BACKGROUND

[0002] In the process of accelerating urbanization and urbanization construction, the original building form cannot meet the living and working environment requirements of the urban population, and the building floors are gradually increased. A large number of pipe shafts, cable shafts, smoke exhaust ducts and various vertical shafts are built in the high-rise building. Once a fire occurs, these shafts increase the vertical spreading way of the fire in the high-rise building and accelerate the spreading speed of the smoke. When carrying out fire extinguishing and rescue in the high-rise building, external extinguishing action is often needed. However, the water spraying height of the existing fire truck cannot meet the needs of fire extinguishing and rescue in the high-rise building, and the fire rescue personnel cannot complete the fire extinguishing and rescue of the high-rise building in time. The positioning of the indoor fire source of the high-rise building can greatly improve the efficiency of fire extinguishing and rescue and reduce the casualties and property losses caused by the fire.

[0003] At present, the room fire source positioning technology mainly includes visual fire positioning technology, wireless sensor network positioning technology and optical fiber sensor positioning technology. These positioning technologies all need fixed position sensors, which need to be installed during building construction. The fixed deployment flexibility is poor and there is a high probability of failure in the fire process. In addition, the time cost of these positioning technologies is high and cannot meet the real-time requirements in actual application scenarios. SUMMARY

[0004] The application provides a high-rise building indoor fire source positioning method based on a glass surface temperature. The positioning method has high deployment flexibility, strong real-time reliability and can meet the application of various actual scenes and has a wide range of applications.

[0005] In order to achieve the above purpose, the application provides a high-rise building indoor fire source positioning method based on a glass surface temperature, which comprises the following steps:

[0006] Step one, m wireless thermocouples are arranged on the outer surface of the fire room glass, and a square matrix is arranged on the largest glass in the fire room; the wireless thermocouples are connected with a temperature data signal acquisition module;

[0007] Step two, the horizontal distances of the wireless thermocouple at the center point position in the wireless thermocouple arrangement square matrix to the two side walls of the fire room are measured and recorded as D1 and D2;

[0008] Step three, the fire room is divided into r area-equivalent blocks, the block number is the same as the number of the arranged wireless thermocouples, the blocks are arranged in a square matrix and numbered as the fire source position category of the program output;

[0009] Step four, adjust the temperature data signal acquisition module, so that it receives the temperature data signal of each thermocouple at a frequency q;

[0010] Step five, open the fire source and temperature data signal acquisition module of the fire room, collect and read the temperature data of each thermocouple in real time, and save and send the collected temperature data signal every set time period t to the data processing end;

[0011] Step six, after receiving the temperature data, the data processing end averages all the temperature data obtained by each wireless thermocouple every set time period t to obtain m average temperature data;

[0012] Step seven, the data processing end combines the m average temperature data obtained into a 1*m matrix and adds D1 and D2 distance data to the 1*m matrix to obtain a 1*(m+2) matrix;

[0013] Step eight, input the 1*(m+2) matrix obtained in step seven into the trained prediction program combined with multilayer perception and KNN algorithm, and get the predicted location block of the fire source through back propagation and category voting.

[0014] Further, the specific process of obtaining the temperature data matrix in step six is:

[0015] (1) The data processing end receives the temperature data and obtains:

[0016]

[0017] T m(tq) is the tqth data collected by wireless thermocouple m in time period t; the temperature data collected by a single wireless thermocouple in time period t is tq;

[0018] (2) Take the average value of each row of temperature data in the matrix in step (1) to form a row matrix:

[0019]

[0020] T m is the average temperature data value of wireless thermocouple m in time period t.

[0021] Further, the specific process of step seven is to add D1 and D2 to the row matrix obtained in step six to form a new matrix:

[0022] [T1,T2,T3,...T m ,D1,D2].

[0023] Further, the training process of the prediction program combined with multilayer perception and KNN algorithm in step eight is:

[0024] (1) Simulate fire data using fire simulation software, build a fire data set, the specific process is:

[0025] a. Build a fire burning model, set parameters, including room geometry, fire location, fire heat release rate and fire area;

[0026] b. Set up temperature measuring points m on the glass of the fire room in an equal row and column square, record the measuring point position temperature, and record the distance D1, D2 from the center of the wireless thermocouple to the two ends of the wall;

[0027] c. Set the simulation time to get the simulated measuring point temperature data set {T mt}, T mt represents the temperature data of the mth measuring point at the tth second;

[0028] d. Pack the temperature measuring point data into a data set, where the vertical direction represents different times, and the horizontal direction represents different wireless thermocouple temperature parameters and distance parameters:

[0029]

[0030] (2) Use train_test_split to divide the data set into training set and test set;

[0031] (3) Create a multi-layer perception model (MLP), set the structure of the hidden layer, the activation function ReLU, and set the maximum number of iterations; Specifically: according to the obtained experimental data parameters, set the input feature dimension, that is, the number of temperature measuring points, the number of distance parameters; According to the input feature dimension, set the structure of the hidden layer, set the activation function ReLU and the maximum number of iterations;

[0032] (4) Call the fit method to train the multi-layer perception model, optimize the weights and biases of the network to minimize the loss function, the steps are:

[0033] a. Input data training set, initialize parameters, including weights and biases;

[0034] b. Pass the input data to each neuron in the input layer;

[0035] c. In the hidden layer, multiply the input data by the weight matrix of the current layer and add the bias term;

[0036] d. Perform nonlinear transformation on the weighted sum through the activation function Relu to get the output of the hidden layer:

[0037] H=σ(XW h ++b h )

[0038]

[0039] where H is the hidden layer, σ is the activation function, X is the input layer, W h is the weight, b h is the bias term, is the probability result obtained by the multi-layer perception, and the softmax function is used to convert the output into a probability distribution;

[0040] e. In the output layer, the output of the last hidden layer is multiplied by the weight matrix of the output layer and the bias term is added;

[0041] f. The final prediction output is calculated using the Softmax activation function;

[0042] g. Calculate the error, i.e. the loss value, between the model prediction value and the actual label;

[0043] h. Calculate the gradient of the loss function with respect to the weight and bias using the chain rule;

[0044] i. Update the parameters of the network using the calculated gradient;

[0045] j. Input the validation set to get the loss value;

[0046] k. Repeat steps b to j until the loss value of the validation data set no longer decreases or reaches the data training times;

[0047] (5) Read the output probability of the MLP model using the predict_proba method, and the output probability represents the prediction probability of each fire source location category;

[0048] (6) Create a KNN classifier and set the number of neighbors k;

[0049] (7) Train the KNN classifier using the features extracted from the multi-layer perception model;

[0050] (8) Call the fit method to train the KNN model, which stores the training data and training labels; During the fit process, the KNN model does not modify or learn the data. It simply saves the data inside the model;

[0051] (9) Use the predict method to make predictions on the test data, which is:

[0052] a. For each test sample input, calculate its Euclidean distance with all training samples:

[0053]

[0054] where n is the dimension; x i is the training sample, and y i is the test sample;

[0055] b. Select k nearest training samples according to the distance obtained in step a;

[0056] c. Vote or weighted vote the labels of the selected k neighbors to determine the final predicted fire source location category of the test sample, and the formula is:

[0057]

[0058] In the formula, is the prediction result; Π(y i = c) is an indicator function, which is equal to 1 if the category y i of the i-th neighbor is c, and is equal to 0 otherwise;

[0059] (10) Calculate the classification accuracy using the classification_report function, and the formula is:

[0060]

[0061] In the formula, Accuracy is the accuracy; n 准确 is the number of groups of the test set correctly predicted, and n 全部 is the total number of groups of the test set;

[0062] (11) Continuously increase the value of k of the classifier until the accuracy no longer increases;

[0063] (12) Complete the training and save the parameters of the multilayer perception and KNN model.

[0064] Further, the value range of the time period t is 10-60s, and the value range of the frequency q is 1HZ or more. The specific frequency is set according to the time period t, so that the temperature data collected in the time period t is more than one.

[0065] The application obtains the glass temperature data of the fire room by wireless thermocouples, calculates the temperature difference of multiple wireless thermocouples, and uses the classification model of the multilayer sensor combined with the KNN algorithm to better process the nonlinear relationship, so as to realize accurate identification of the fire source position. At the same time, the application can be applied to any form of building fire, and is not limited by the fire fighting system and installation, and is used for fire source positioning at any time and anywhere, has high deployment flexibility, strong real-time reliability, meets the application of various actual scenes, has strong universality and real-time performance. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 is the workflow diagram of the application;

[0067] Figure 2 is a schematic view of the horizontal distance between the central wireless thermocouple and the two end walls.

[0068] Figure 3 This is a flowchart of the prediction program combining the multilayer perceptron and KNN algorithm of this invention.

[0069] Figure 4 This is a comparison chart of the predicted results and actual results of an embodiment of the present invention. Detailed Implementation

[0070] The invention will now be further described with reference to the accompanying drawings.

[0071] like Figure 1 As shown, a method for locating indoor fire sources in high-rise buildings based on glass surface temperature includes the following steps:

[0072] Step 1: Install m wireless thermocouples on the outer surface of the glass of the fire-prone room, and arrange them in an array on the largest piece of glass in the fire-prone room; each wireless thermocouple is connected to a temperature data signal acquisition module.

[0073] Step Two, as follows Figure 2 As shown, measure the horizontal distances from the wireless thermocouple at the center point in the wireless thermocouple array to the two side walls of the room where the fire is taking place, and record them as D1 and D2.

[0074] Step 3: Divide the fire room into r blocks of equal area. The number of blocks is the same as the number of wireless thermocouples, and they are arranged in a square array. Number them as the fire source location category output by the program.

[0075] Step 4: Adjust the temperature data signal acquisition module to receive temperature data signals from each thermocouple at a frequency q.

[0076] Step 5: Activate the fire source and temperature data signal acquisition module in the fire room, collect and read the temperature data of each thermocouple in real time, and save and send the collected temperature data signals to the data processing terminal every set time period t.

[0077] Step 6: After receiving the temperature data, the data processing terminal averages all the temperature data obtained by each wireless thermocouple within a set time period t to obtain m average temperature data.

[0078] Step 7: The data processing unit combines the m average temperature data into a 1*m matrix and adds the distance data of D1 and D2 to the 1*m matrix to obtain a 1*(m+2) matrix.

[0079] Step 8: Input the 1*(m+2) matrix obtained in Step 7 into the prediction program that combines the trained multilayer perceptron and KNN algorithm. After backpropagation and class voting, the predicted location block of the fire source is obtained.

[0080] Example: (1) A fire room model was established according to the actual room size. The room model has a height of 2.8 m, a length and a width of 5 m, a starting height of 1 m, a width of 1.8 m, and a height of 1.5 m for the room window, which remains closed during the fire process, and a width of 0.8 m and a height of 2 m for the door, which is set opposite the room window and remains open during the fire process. The room floor is set to be a ceramic material, and the wall is set to be a concrete material. The combustible material has a foam thickness of 0.05 m on the upper surface, and the remaining surfaces and the interior are all yellow pine materials. The combustible material is arranged at nine positions in the room. Each group of experiments sets the fire source on the upper surface of a different single combustible material in the form of vent, sets the maximum heat release rate of each fire source to be between 300 and 400 kW, sets the simulation time to be 600 s, and changes the position of the window to perform multiple experiments to obtain sufficient data;

[0081] (2) Three thousand groups of data were randomly selected from the obtained data as a training data set, and 80% of them were used as the training set and 20% of them were used as the test set:

[0082] (3) As shown in Figure 3 , an MLPClassifier instance is created, the structure of the hidden layer is set to 64 neurons in the first layer and 32 neurons in the second layer, the activation function uses Relu, and the maximum number of iterations is set to 1000; the training set is used to train the multi-layer perceptron model (MLP model): the weights and biases are adjusted through the forward propagation and back propagation algorithms to minimize the prediction error, which is usually the cross-entropy loss;

[0083] (4) After training is completed, the MLP model is used to perform forward propagation on the training set and the test set to obtain the output of the hidden layer, i.e., the feature representation;

[0084] (5) The hidden layer of the MLP converts the input data into intermediate representations, which are used as feature extraction; the output of these hidden layers is considered as a new feature set for the subsequent KNN classifier, and the purpose of feature extraction is to convert the data into the abstract representation learned by the MLP;

[0085] (6) A KNeighborsClassifier instance is created, cross-validation is set, the k values of [1, 3, 5, 7, 9, 11] are selected, and the KNN classifier is trained using the features extracted from the MLP model; KNN does not require a learning process during the training phase, but stores the training data;

[0086] (7) The KNN classifier is used to predict the test data: KNN finds the nearest K neighbors based on distance measurement, and then decides the predicted category through majority voting; for a test sample, the Euclidean distance between it and all samples in the training set is calculated:

[0087]

[0088] According to the calculated distance, the training samples are sorted by distance from small to large, and the k training samples closest to the test sample are selected;

[0089] (8) Voting on the class labels of the selected k neighbors, each neighbor's class label will increase a vote, that is, there are 3 neighbors whose class labels are 1, and 2 neighbors whose class labels are 3, then the class of the sample is 1; Calculate the number of votes for each class label, and then select the class with the most votes as the predicted class of the test sample:

[0090]

[0091] The experimental data table is as follows:

[0092]

[0093]

[0094] Through the above examples, the following conclusions are obtained: when the fire source is located on one side of the glass, the wireless thermocouple temperature on the horizontal line of the glass will always show a trend of decreasing from the fire source side to the non-fire source side; when the fire source is located at the center vertical line position of the glass surface, the temperature along the horizontal line of the glass will not show any increasing or decreasing form; this change trend can be used as a basis for judging the horizontal position of the fire source;

[0095] Under the combined action of smoke and heat radiation, when the fire source is far away from the glass, the effect of smoke on the temperature of the glass is greater; when the fire source is close to the glass position, the effect of heat radiation and fire plume on the temperature of the glass is greater; when the fire source is far away from the glass, the temperature difference between the highest horizontal line thermocouple and the center horizontal line thermocouple is less than that between the center horizontal line thermocouple and the lowest horizontal line thermocouple, and when the fire source is close to the glass, the situation is reversed. When the fire source is located at the center plane of the room perpendicular to the glass, the difference is basically the same; this feature can be used as a basis for judging the longitudinal position of the fire source.

[0096] As shown in the experimental data table and Figure 4 The present application can exhibit high accuracy with a small number of training sets, and has strong universality, and can have high accuracy for different building structures. The present application is suitable for high-rise building fire fighting and rescue, realizes present installation, and has higher universality, timeliness and efficiency compared with other fire source positioning methods, and has high level of accuracy.

Claims

1. A method for locating a fire source in a high-rise building based on the temperature of a glass surface, characterized in that, Comprise the following steps: Step one, set m wireless thermocouples on the outer surface of the glass of the fire room, the square matrix is arranged on the largest piece of glass of the fire room; each wireless thermocouple is connected with a temperature data signal acquisition module; Step two, measure the horizontal distance from the wireless thermocouple at the center point position in the square matrix to the walls on both sides of the fire room, and record it as D1 and D2; Step three, divide the fire room into r blocks with the same area, the number of blocks is the same as the number of wireless thermocouples arranged, and the blocks are arranged in a square matrix and numbered as the fire source position category output by the program; Step four, adjust the temperature data signal acquisition module so that it receives the temperature data signal of each thermocouple at a frequency q; Step five, start the fire source in the fire room and the temperature data signal acquisition module, real-time collect and read the temperature data of each thermocouple, and save and send the collected temperature data signal to the data processing end every set time period t; Step six, after receiving the temperature data, the data processing end averages all the temperature data obtained by each wireless thermocouple every set time period t to obtain m average temperature data; Step seven, the data processing end combines the m average temperature data obtained into a 1*m matrix and adds the D1 and D2 distance data to the 1*m matrix to obtain a 1*(m+2) matrix; Step eight, input the 1*(m+2) matrix obtained in step seven into the prediction program combined with the trained multilayer perception and KNN algorithm, and get the predicted position block of the fire source through back propagation and category voting.

2. The method for locating a fire in a high-rise building based on the temperature of a glass surface according to claim 1, wherein The specific process of obtaining the temperature data matrix in step six is: (1) the data processing end receives the temperature data and obtains: T m(tq) Tq is the data collected by the wireless thermocouple m in the time period t; the temperature data collected by a single wireless thermocouple in the time period t is tq; (2) take the average value of each row of temperature data in the matrix of step (1) to form a row matrix: T m Tavg is the average temperature data value for the wireless thermocouple m over the time period t.

3. The method for locating a fire in a high-rise building based on the temperature of a glass surface according to claim 2, wherein The specific process of step seven is to add D1 and D2 to the row matrix obtained in step six to form a new matrix: [T1, T2, T3,... T m , D1, D2].

4. The method for locating a fire in a high-rise building based on the temperature of a glass surface according to claim 3, wherein The training process of the prediction program combined with the multilayer perception and KNN algorithm in step eight is: (1) use fire simulation software to simulate fire data and build a fire data set, the specific process is: a. Build a fire burning model and set parameters, including room geometry, fire source position, fire source heat release rate and fire source area; b. Set up temperature measuring points m on the glass of the fire room in an equal row and column square matrix, record the temperature of the measuring point position, and record the distance D1 and D2 from the center of the wireless thermocouple to the walls on both ends; c. Set simulation time, get simulated temperature data set {T mt} of measuring point mt m represents the mth measuring point, t represents the tth second d. Pack the temperature measuring point data into a data set, where the vertical represents different times, and the horizontal represents different wireless thermocouple temperature parameters and distance parameters: (2) use train_test_split to divide the data set into training set and test set; (3) create a multilayer perception model, set the structure of the hidden layer, the activation function ReLU, and set the maximum number of iterations; Specifically: set the input feature dimension according to the obtained experimental data parameters, that is, the number of temperature measuring points and the number of distance parameters; set the structure of the hidden layer according to the input feature dimension, set the activation function ReLU and the maximum number of iterations; (4) Call the fit method to train the multi-layer perceptron model, optimize the weights and biases of the network to minimize the loss function, the steps are: a. Input the training data set, initialize the parameters, including weights and biases; b. Pass the input data to each neuron in the input layer; c. In the hidden layer, multiply the input data with the weight matrix of the current layer and add the bias term; d. Nonlinearly transform the weighted sum through the activation function Relu to get the output of the hidden layer: H = σ(XW h ++b h ) where H is the hidden layer, σ is the activation function, X is the input layer, W h is the weight, b h is the bias term, is the probability result obtained by the multi-layer perception, and the softmax function is used to convert the output into a probability distribution. e. In the output layer, multiply the output of the last hidden layer with the weight matrix of the output layer and add the bias term; f. Calculate the final predicted output using the Softmax activation function; g. Calculate the error between the model's predicted value and the actual label, i.e. the loss value; h. Calculate the gradient of the loss function with respect to the weights and biases using the chain rule; i. Update the network's parameters using the calculated gradients; j. Input the validation set to get the loss value; k. Repeat steps b to j until the loss value of the validation data set no longer decreases or reaches the data training times; (5) Use the predict_proba method to read the output probability of the multi-layer perceptron model, the output probability represents the predicted probability of each fire source location category; (6) Create a KNN classifier and set the number of neighbors k value; (7) Train the KNN classifier using the features extracted from the multi-layer perceptron model; (8) Call the fit method to train the KNN model, the steps are to store the training data and training labels; (9) Use the predict method to make predictions on the test data, the steps are: a. For each test sample input, calculate its Euclidean distance with all training samples: where n is the dimensionality; x i is the training sample, y i is the test sample; b. Select the k nearest training samples according to the distance obtained in step a; c. Vote or weighted vote on the labels of the selected k neighbors to determine the final predicted fire source location category of the test sample, the formula is: wherein is the prediction; Π(y i = c) is an indicator function that equals 1 if the class y i of the i-th neighbor is c and 0 otherwise. (10) Use the classification_report function to calculate the classification accuracy, the formula is: where Accuracy is the accuracy rate; n 准确 is the number of test set groups that are correctly predicted 全部 is the total number of test set groups (11) Keep increasing the k value of the classifier until the accuracy no longer increases; (12) Complete the training and save the parameters of the multi-layer perceptron and KNN models.

5. The method for locating an indoor fire source in a high-rise building based on a glass surface temperature according to claim 1, wherein, Set the value range of the time period t to be 10-60s, and the value range of the frequency q to be 1HZ or above.

Images