WO2022012295A1 - Fire detection method and apparatus - Google Patents

Abstract

A fire detection method and apparatus. The method comprises: determining a plurality of fire influence factors of nodes in a cabin (S101); determining a fire influence weight value of each fire influence factor in the plurality of fire influence factors of the nodes (S102); according to the fire influence weight value of each fire influence factor, determining an evaluation value of each node for the fire occurrence probability in the cabin (S103); determining the probability of actual fire in the cabin according to the evaluation value of each node for the fire occurrence probability in the cabin (S104); and determining, according to the probability of actual fire in the cabin, whether a fire alarm is given (S105). The fire detection method and apparatus can accurately detect a fire, and accurately determine whether a fire alarm is given, so that the timeliness and the accuracy of a fire alarm are ensured.

Description

Fire detection method and device technical field

The invention relates to the technical field of fire detection, in particular to a fire detection method and device.

Background technique

At present, in the detection of fire, the gray fuzzy neural network information fusion algorithm is usually used to fuse the measurement data obtained by the smoke detection module, the CO gas detection module and the temperature detection module, or the fuzzy logic control system with feedback (that is, the fuzzy logic control system with feedback is used). Logic fusion algorithm) fuses the measurement data obtained by different sensors or modules to determine whether a fire has occurred, but these methods have low detection accuracy, are prone to misjudgment of fire, and have poor detection timeliness, requiring too long to detect time.

SUMMARY OF THE INVENTION

The implementation of the present invention provides a fire detection method and device. The technical solution is as follows:

According to a first aspect of the embodiments of the present invention, a fire detection method is provided, including:

Identify multiple fire-influencing factors at each node in the engine room;

determining the fire influence weight value of each fire influence factor among the multiple fire influence factors of each node;

According to the fire influence weight value of each fire influence factor, the evaluation value of each node to the fire occurrence probability in the cabin is determined;

According to the evaluation value of the probability of fire occurrence in the cabin by each node, determine the probability of the actual occurrence of fire in the cabin;

According to the probability of a fire actually occurring in the cabin, it is judged whether to issue a fire alarm.

In one embodiment, the determining the fire influence weight value of each fire influence factor among the multiple fire influence factors of each node includes:

determining the judgment matrix for each of the fire influencing factors;

Consistency test is carried out on the judgment matrix of each fire influencing factor;

According to the judgment matrix of each fire influence factor that has passed the consistency check, the fire influence weight value of each fire influence factor is calculated.

In one embodiment, when there are multiple judgment matrices of each fire influencing factor, the fire influence weight of each fire influencing factor is calculated according to the judgment matrix of each fire influencing factor that has passed the consistency check values, including:

According to the judgment matrix of each of the fire-influencing factors that pass the consistency check, a preset number of fire-influencing weight values for the fire-influencing factors are calculated, wherein the preset number is equal to all the fire-influencing factors that pass the consistency check. the number of judgment matrices for each fire influencing factor; the preset number is a positive integer greater than or equal to 2;

The fire influence weight interval of each fire influence factor is determined according to the preset number of fire influence weight values of each fire influence factor.

In one embodiment, determining the evaluation value of the fire occurrence probability in the cabin by each node according to the fire influence weight value of each fire influence factor, including:

According to the fire influence weight value of each fire influence factor, determine the fire influence weight interval of each fire influence factor;

Determine the target influence function of each fire influence factor and the constraint condition of the target influence function according to the fire influence weight interval of each fire influence factor;

According to the target influence function of each fire influence factor and the constraint condition of the target influence function, calculate the optimal fire influence weight value of each fire influence factor of each node;

According to the optimal fire influence weight value of each fire influence factor of each node and the actual factor value of each fire influence factor of each node, the evaluation value of each node to the fire occurrence probability in the cabin is determined.

In one embodiment, determining the probability of a fire actually occurring in the cabin according to the evaluation value of the probability of occurrence of fire in the cabin by each node includes:

According to the evaluation value of each node on the probability of fire occurrence in the cabin, establish a support function between any two nodes in each node with respect to the evaluation value;

According to the support function between the any two nodes with respect to the evaluation value, establish an initial support degree matrix between the nodes with respect to the evaluation value;

Constructing an augmented support matrix for the initial support matrix between the evaluation values of each node; wherein, the augmented support matrix has one more row and one column compared to the initial support matrix;

According to the augmented support matrix, the probability that a fire actually occurs in the cabin is determined.

In one embodiment, the determining the probability of a fire actually occurring in the cabin according to the augmented support matrix includes:

determining, according to the augmented support matrix, an evaluation credibility coefficient of each node for the evaluation value of the probability of occurrence of fire in the cabin;

The probability of a fire actually occurring in the cabin is determined according to the evaluation value of the probability of occurrence of fire in the cabin by each node and the evaluation credibility coefficient of the evaluation value of the probability of occurrence of fire in the cabin by each node.

In one embodiment, the judging whether to issue a fire alarm according to the probability that a fire actually occurs in the cabin includes:

Calculate the probability threshold of fire occurrence in the cabin according to the fire occurrence critical factor value of each fire influencing factor of each node;

According to the actual probability of fire in the cabin and the probability threshold of fire in the cabin, it is judged whether to issue a fire alarm.

According to a second aspect of the embodiments of the present invention, a fire detection device is provided, comprising:

a first determination module, used for determining multiple fire-influencing factors of each node in the engine room;

a second determination module, configured to determine the fire influence weight value of each fire influence factor among the plurality of fire influence factors of each node;

a third determination module, configured to determine the evaluation value of each node on the probability of fire occurrence in the cabin according to the fire influence weight value of each fire influence factor;

a fourth determination module, configured to determine the probability of a fire actually occurring in the cabin according to the evaluation value of the probability of fire occurrence in the cabin by each node;

The judging module is used for judging whether to issue a fire alarm according to the probability of a fire actually occurring in the cabin.

In one embodiment, the second determining module includes:

a determination sub-module for determining the judgment matrix of each of the fire-influencing factors;

a check sub-module, used to check the consistency of the judgment matrix of each fire influencing factor;

The calculation sub-module is configured to calculate the fire influence weight value of each fire influence factor according to the judgment matrix of each fire influence factor that has passed the consistency check.

In one embodiment, the calculation submodule is specifically used for:

When there are multiple judgment matrices for each fire influencing factor, according to each judgment matrix of each fire influencing factor that passes the consistency check, a preset number of fire influence weight values for each fire influencing factor are calculated. , wherein the preset number is equal to the number of judgment matrices of the respective fire influencing factors that have passed the consistency check; the preset number is a positive integer greater than or equal to 2;

The fire influence weight interval of each fire influence factor is determined according to the preset number of fire influence weight values of each fire influence factor.

In one embodiment, the third determining module includes:

a first determination sub-module, configured to determine the fire influence weight interval of each fire influence factor according to the fire influence weight value of each fire influence factor;

The second determination submodule is configured to determine the target influence function of each fire influence factor and the constraint condition of the target influence function according to the fire influence weight interval of each fire influence factor;

a calculation submodule, configured to calculate the optimal fire influence weight value of each fire influence factor of each node according to the target influence function of each fire influence factor and the constraint condition of the target influence function;

The third determination sub-module is configured to determine, according to the optimal fire influence weight value of each fire influence factor of each node and the actual factor value of each fire influence factor of each node, the effect of each node on the interior of the cabin. An estimate of the probability of fire occurrence.

In one embodiment, the fourth determining module is specifically configured to:

According to the evaluation value of each node on the probability of fire occurrence in the cabin, establish a support function between any two nodes in each node with respect to the evaluation value;

According to the support function between the any two nodes with respect to the evaluation value, establish an initial support degree matrix between the nodes with respect to the evaluation value;

Constructing an augmented support matrix for the initial support matrix between the evaluation values of each node; wherein, the augmented support matrix has one more row and one column compared to the initial support matrix;

According to the augmented support matrix, the probability that a fire actually occurs in the cabin is determined.

In one embodiment, the fourth determining module is further configured to:

determining, according to the augmented support matrix, an evaluation credibility coefficient of each node for the evaluation value of the probability of occurrence of fire in the cabin;

The probability of a fire actually occurring in the cabin is determined according to the evaluation value of the probability of occurrence of fire in the cabin by each node and the evaluation credibility coefficient of the evaluation value of the probability of occurrence of fire in the cabin by each node.

In one embodiment, the judging module is specifically used for:

Calculate the probability threshold of fire occurrence in the cabin according to the fire occurrence critical factor value of each fire influencing factor of each node;

Whether to issue a fire alarm is determined according to the actual probability of fire in the cabin and the probability threshold of fire in the cabin.

The following technical effects can be achieved through the technical solutions of the present invention:

By calculating the fire influence weight value of each fire influencing factor at each node, the estimated value of each node to the probability of fire occurrence in the cabin can be preliminarily calculated. However, the estimated value may not be accurate and needs to be adjusted. Therefore, the estimated value can be used to calculate The probability of a fire actually occurring in the cabin, so as to accurately detect the fire, and accurately judge whether to issue a fire alarm, so as to ensure the timeliness and accuracy of the fire alarm.

Other features and advantages of the present invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description, claims, and drawings.

Description of drawings

The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention.

FIG. 1 is a schematic diagram of the deployment of a fire detection sensor node provided by the present invention.

FIG. 2 is a fire probability error curve provided by the present invention.

FIG. 3 is a comparison diagram of detection time between a fire detection method provided by the present invention and two fire detection algorithms in the related art.

FIG. 4 is a relationship curve between the number k of judgment matrices and the false alarm rate of fire detection provided by the present invention.

FIG. 5 is a comparison diagram of detection accuracy of a fire detection method provided by the present invention compared with two fire detection algorithms in the related art.

FIG. 6 is a flowchart of a fire detection method provided by the present invention.

detailed description

Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. Where the following description refers to the drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the illustrative examples below are not intended to represent all implementations consistent with this disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as recited in the appended claims.

In order to solve the above technical problems, an embodiment of the present invention provides a fire detection method, which can be used in a fire detection program, system or device. As shown in FIG. 6 , the method includes steps S101 to S105:

In step S101, a plurality of fire-influencing factors of each node in the engine room are determined; the fire-influencing factors include but are not limited to: temperature, smoke concentration, CO concentration, and infrared light intensity. The cabin can be an aircraft cabin, an aviation cabin, etc., and each node is each sensor node. Of course, each sensor node can have one or more sensors of different types, which are used to detect the actual factor value of the fire influencing factor.

In step S102, the fire influence weight value of each fire influence factor among the multiple fire influence factors of each node is determined; the fire influence weight value is used to represent the weight of each fire influence factor in the occurrence of the fire. In addition, in order to ensure the accuracy of fire detection, each node adopts the same fire influence factor.

In step S103, according to the fire influence weight value of each fire influence factor, the evaluation value of each node on the probability of fire occurrence in the cabin is determined; the evaluation value is used to characterize the probability of fire occurrence at each node, The greater the evaluation value of a node, the greater the probability of fire at the node, and vice versa.

In step S104, the probability of a fire actually occurring in the cabin is determined according to the evaluation value of the probability of occurrence of fire in the cabin by each node;

In order to ensure the accuracy of the fire alarm, it is necessary to fuse the evaluation values of the fire probability of each node. At the same time, in order to avoid the influence on the accuracy of the fusion result when the sensor nodes deployed in the cabin fail, the present invention can be based on the support matrix. Multi-sensor node data is fused (that is, the evaluation value of fire occurrence probability of each node is fused).

In step S105, it is judged whether to issue a fire alarm according to the probability of a fire actually occurring in the cabin.

By calculating the fire influence weight value of each fire influencing factor at each node, the estimated value of each node to the probability of fire occurrence in the cabin can be preliminarily calculated. However, the estimated value may not be accurate and needs to be adjusted. Therefore, the estimated value can be used to calculate The probability of a fire actually occurring in the cabin, so as to accurately detect the fire, and accurately judge whether to issue a fire alarm, so as to ensure the timeliness and accuracy of the fire alarm.

In one embodiment, the determining the fire influence weight value of each fire influence factor among the multiple fire influence factors of each node includes:

determining the judgment matrix for each of the fire influencing factors;

Consistency test is carried out on the judgment matrix of each fire influencing factor;

According to the judgment matrix of each fire influence factor that has passed the consistency check, the fire influence weight value of each fire influence factor is calculated.

By determining the judgment matrix of each fire influencing factor, the judgment matrix can be checked for consistency. If the consistency check is passed, the judgment matrix is determined to be a reasonable matrix; otherwise, it should be discarded. Through the judgment matrix of each fire-influencing factor through the consistency check, the fire-influencing weight value of each fire-influencing factor is calculated, so as to determine the weight of each fire-influencing factor in the occurrence of fire.

For example, the relationship of the degree of influence between variables is quantitatively represented by a judgment matrix. Let the judgment matrix be C∈R ^n×n , where n represents the number of variables.

The judgment matrix C has the form:

matrix element

Indicates the importance ratio of the influence degree of the i-th and j-th variables on the fire situation.

The rules for filling in element c _{ij are:}

If x _i and x _{j are} equally important, then take c _ij =1, c _ji =1;

If x _i is slightly more important than x _j , then take c _ij =3, c _ji =1/3;

If x _i is obviously more important than x _j , then take c _ij =5, c _ji =1/5;

If x _i is much more important than x _j , then take c _ij =7, c _ji =1/7;

If x _i is absolutely more important than x _j , then take c _ij =9, c _ji =1/9;

If _{the importance of x i} and x _j is between the above relationships, then c _ij can take 2, 4, 6, and 8; c _ji can take the values of 1/2, 1/4, 1/6, and 1/8. Therefore, the element c _ij can take any integer from 1 to 9 according to the above rules.

In one embodiment, when there are multiple judgment matrices of each fire influencing factor, the fire influence weight of each fire influencing factor is calculated according to the judgment matrix of each fire influencing factor that has passed the consistency check values, including:

According to the judgment matrix of each of the fire-influencing factors that pass the consistency check, a preset number of fire-influencing weight values for the fire-influencing factors are calculated, wherein the preset number is equal to all the fire-influencing factors that pass the consistency check. the number of judgment matrices for each fire influencing factor; the preset number is a positive integer greater than or equal to 2;

The fire influence weight interval of each fire influence factor is determined according to the preset number of fire influence weight values of each fire influence factor.

There can also be multiple judgment matrices. When there are multiple judgment matrices, a fire impact weight value can be calculated based on each judgment matrix that passes the consistency check. Therefore, when there are multiple judgment matrices that pass the consistency check, There will be multiple calculated fire influence weight values for each fire influence factor, and thus, the fire influence weight interval of each fire influence factor can be formed.

In one embodiment, determining the evaluation value of the fire occurrence probability in the cabin by each node according to the fire influence weight value of each fire influence factor, including:

According to the fire influence weight value of each fire influence factor, determine the fire influence weight interval of each fire influence factor;

According to the fire influence weight interval of each fire influence factor, determine the target influence function of each fire influence factor (for finding the smallest weight deviation in the interval) and the constraint condition of the target influence function (the constraint condition may be each The sum of the optimal fire influence weights of each fire influence factor of the node is 1);

According to the target influence function of each fire influence factor and the constraint condition of the target influence function, calculate the optimal fire influence weight value of each fire influence factor of each node;

According to the optimal fire influence weight value of each fire influence factor of each node and the actual factor value of each fire influence factor of each node, the evaluation value of each node to the fire occurrence probability in the cabin is determined. The actual factor value of each fire-influencing factor is the actual value of each fire-influencing factor. For example, the actual value range of the fire-influencing factor temperature (unit ℃) is [0, 100], and the actual value of the smoke concentration (unit ppm) is The interval is [100, 1000], the actual value interval where the CO concentration (unit ppm) is located is [10, 100], and the actual value interval where the infrared light intensity (unit Lux) is located is [100, 1000]. Different classes may affect the evaluation value. Before calculating the evaluation value, the actual factor value of each fire-influencing factor can be normalized.

After the fire influence weight interval of each fire influence factor is determined, the target influence function of each fire influence factor and the constraint conditions of the target influence function can be determined, so that according to the target influence function and the constraint conditions, the fire influence of each fire influence factor can be obtained from the fire influence factor. The optimal fire influence weight value is automatically selected from the weight interval, and the optimal fire influence weight value should be able to take into account all the values in the weight interval to the greatest extent. Therefore, the optimal fire influence weight value of each fire influence factor and all The actual factor value of each fire influencing factor of each node is described, and the determined evaluation value of each node for the probability of fire occurrence in the cabin can ensure that the final evaluation value can better reflect the actual situation; at the same time, the optimal weight is used for weighting Fusion can effectively avoid the tedious calculation process during fusion, and improve the response speed to fire.

In one embodiment, determining the probability of a fire actually occurring in the cabin according to the evaluation value of the probability of occurrence of fire in the cabin by each node includes:

According to the evaluation value of each node on the probability of fire occurrence in the cabin, establish a support function between any two nodes in each node with respect to the evaluation value;

make

and

^{Represents the evaluation value (i*} ,j ^* ∈1,2,…,N) of sensor nodes i ^* and j ^* on the probability of fire occurrence in the cabin at time k, respectively. if

and

If the difference is large, it means that the two sensor nodes do not support each other at time k, and if the difference is small, it means that the two nodes support each other. If a node is simultaneously supported by a majority of nodes, it is considered a valid fire probability estimate. Otherwise, the probability value evaluated by this node will be given a lower weight in the fusion process;

in order to express

and

The mutual support between , and the support function is defined as a decaying exponential function in the fuzzy set:

The parameter α is adjustable and is used to adjust the fusion accuracy, which is generally set to 0.8 [16].

while the support function

Represents the mutual support of node i ^* and node j ^* at the kth moment.

According to the support function between the any two nodes with respect to the evaluation value, establish an initial support degree matrix between the nodes with respect to the evaluation value;

Constructing an augmented support matrix for the initial support matrix between the evaluation values of each node; wherein, the augmented support matrix has one more row and one column compared to the initial support matrix;

According to the augmented support matrix, the probability that a fire actually occurs in the cabin is determined.

After establishing a support function between any two nodes of the nodes with respect to the evaluation value according to the evaluation value of the fire probability in the cabin by the nodes, since there is a support function between every two nodes Degree function, that is, there are many support degree functions between all nodes. In order to facilitate processing, the support degree function can be converted into an initial support degree matrix about the evaluation value, and in order to facilitate the adaptive allocation of weight coefficients for each node (ie, The evaluation credibility coefficient of the evaluation value of each node) to fuse and adjust the evaluation value of the probability of fire occurrence in the cabin to achieve the purpose of improving the accuracy of the probability of actual fire occurrence in the cabin, which can be constructed based on the initial support matrix The support matrix is extended to improve the accuracy of the calculated probability of a fire actually occurring in the cabin, thereby enabling accurate prediction of the fire.

In one embodiment, the determining the probability of a fire actually occurring in the cabin according to the augmented support matrix includes:

determining, according to the augmented support matrix, an evaluation credibility coefficient of each node for the evaluation value of the probability of occurrence of fire in the cabin;

The probability of a fire actually occurring in the cabin is determined according to the evaluation value of the probability of occurrence of fire in the cabin by each node and the evaluation credibility coefficient of the evaluation value of the probability of occurrence of fire in the cabin by each node.

Based on the augmented support matrix, a higher evaluation credibility coefficient can be assigned to the evaluation value of each node with high reliability, and a lower evaluation credibility coefficient can be assigned to the evaluation value of each node with low credibility, so that Avoid failure or inaccurate detection of sensor nodes deployed in the cabin, which will adversely affect the accuracy of the fusion results of the evaluation values of each node, making the fusion of evaluation values fault-tolerant, thereby greatly improving the probability of actual fire in the cabin. Specifically, the technical scheme of the present invention conducts experiments with the gray-fuzzy neural network fusion algorithm and the fuzzy logic fusion algorithm in the related art from the perspectives of the time required to detect a fire and issue an alarm signal and the system false alarm rate. After comparison, it is found that the fire detection scheme of the present invention can detect fire within 10s, while reducing the false alarm rate to less than 0.5%, greatly reducing the fire false alarm rate and improving the accuracy and timeliness of fire detection.

In one embodiment, the judging whether to issue a fire alarm according to the probability that a fire actually occurs in the cabin includes:

Calculate the probability threshold of fire occurrence in the cabin according to the fire occurrence critical factor value of each fire influencing factor of each node (that is, the value of each fire influencing factor when the critical state of fire occurrence is reached);

Whether to issue a fire alarm is determined according to the actual probability of fire in the cabin and the probability threshold of fire in the cabin.

According to the fire occurrence critical factor value of each fire influencing factor of each node, the probability threshold of fire occurrence in the cabin can be accurately calculated, and then the currently calculated probability of actual fire occurrence in the cabin and the fire occurrence in the cabin can be calculated. By comparing the probability threshold of the cabin, it can be accurately judged whether a fire alarm is issued. Specifically, if the currently calculated probability of actual fire in the cabin is greater than or equal to the probability threshold of fire in the cabin, it means that a fire is about to occur, then Fire alarm is carried out, otherwise, it means that fire has not yet occurred, and there is no need to carry out fire alarm.

Finally, it should be clarified that those skilled in the art can freely combine the above-mentioned multiple embodiments according to actual needs.

The technical scheme of the present invention will be described in further detail below:

Aiming at the high false alarm rate when using a single sensor to detect fire in the aircraft cabin, a multi-sensor data fusion method is proposed to detect fire. First, the weights of temperature, smoke concentration, CO concentration and infrared light intensity in the occurrence of fire are calculated by the improved Analytic Hierarchy Process (AHP) on each sensor node of the wireless sensor network (WSN), and the variable weighted fusion method is used. Evaluate the probability of fire occurrence in the cabin; then based on the mutual support between the evaluation data of fire occurrence probability of each node, adaptively assign weight coefficients to each evaluation value, and weight and fuse the evaluation values of all nodes, and finally obtain the fire Occurrence probability; finally, it is judged whether there is a fire by comparing with the threshold probability. From the perspective of the time required for the system to detect a fire and issue an alarm signal and the system's false alarm rate, it is compared with the gray-fuzzy neural network fusion algorithm and the fuzzy logic fusion algorithm respectively. The experiments show that the fire detection algorithm can detect fire within 10s. At the same time, the false alarm rate is reduced to less than 0.5%, which verifies the superiority of the algorithm in terms of timeliness and accuracy. And by setting faulty sensor nodes, it is proved that the algorithm has a certain fault tolerance.

Based on the research of the above methods, in view of the characteristics of small and closed aircraft environment, the data fusion method under the action of various factors is considered in the study of the fire detection and alarm problem based on WSN. In order that the detection algorithm can still detect the fire accurately and timely in the case of many fusion variables, the present invention uses the improved AHP at each sensor node in the WSN to calculate the weight of each variable affecting the probability of fire occurrence, and calculates the weight of each variable affecting the probability of fire occurrence according to the multiple The variable weighted fusion evaluates the probability of fire occurrence in the cabin; at the same time, in order to avoid the influence of the faulty sensor on the detection accuracy, the present invention uses the adaptive weighted fusion method to fuse the evaluation data of each node. Among them, in the case of not knowing the detection ability of each node, the credibility of the fire assessment results of each node is calculated by constructing a support matrix, and a higher weight is assigned to a node with high credibility, and a lower weight is assigned otherwise. This minimizes the impact of faulty sensors on fusion results.

The main contribution of the present invention is to avoid the problems of reduced timeliness when the detection system fuses multi-variable data, and large deviations in fusion results when some sensors are faulty. First, the present invention considers the weight of each variable that affects the occurrence of fire, and designs a new "multi-variable weighted fusion" fire evaluation algorithm based on the improved AHP variable weight calculation method, which can still detect quickly when there are many fusion variables. In addition, the present invention fuses more related variables, which greatly reduces the false alarm rate of the system; secondly, the present invention uses adaptive weighted fusion to fuse the evaluation values of the fire occurrence probability of each sensor node. , by constructing a support matrix to assign weight coefficients to each node adaptively, to a certain extent, it avoids the problem of decreased fusion accuracy caused by the measurement deviation of some faulty sensors.

The structure of the invention is as follows: the first section introduces a method for evaluating the probability of fire occurrence by using the improved AHP and multivariate weighted fusion algorithm at each sensor node; the second section introduces a method by constructing an augmented support matrix , a method of adaptively weighted fusion of the estimated value of fire occurrence probability by each node. The third section is the experimental part, which verifies the timeliness, accuracy and fault tolerance of the fire detection algorithm proposed by the present invention. The fourth section is the conclusion of the present invention.

1. Fire assessment algorithm based on multivariate weighted fusion

In this section, an improved AHP method is studied, and a multivariate weight calculation method is proposed to determine the weights of temperature, smoke concentration, CO concentration and infrared light intensity in fire occurrence. A new "multi-variable weighted fusion" algorithm is proposed at each sensor node of WSN to obtain the evaluation value of each node's probability of fire occurrence in the cabin.

1.1 Improved AHP multivariate weighting algorithm

To evaluate and judge the occurrence of fire in the cabin, it is necessary to determine the relevant influencing factor variables, and the weight of the variables indicates their position in the evaluation process. The size of a variable weight will have an important impact on the fire assessment results. Therefore, it is a problem worth noting to determine the variable weight scientifically and reasonably. The present invention mainly adopts the improved AHP method to determine the variable weight [11]. The traditional AHP calculates the relative weight of the lowest layer to the highest layer by selecting a judgment matrix, and sorts the various schemes and measures in the lowest layer according to this weight. However, due to the single number of judgment matrices selected, the accuracy of the required weights is not high. In order to ensure the accuracy of the required weights, the present invention uses the improved AHP to calculate the weights occupied by each influencing factor variable in the process of fire occurrence, and by selecting multiple judgment matrices and constructing the objective function in the weight interval, the method is Optimal weights for variables.

The basic process of the improved AHP method for evaluating fire problems is: 1) Determine the factors and variables that affect the occurrence of fire; 2) Establish a judgment matrix for each variable; 3) Consistency test on the judgment matrix; 4) Eigenvector method to calculate the variable weight 5) Set the objective function to obtain the optimal weight. The weight calculation steps are as follows:

Step 1: Set the factor variables that affect the occurrence of fire as temperature, smoke concentration, CO concentration and infrared light intensity. The variables are represented by x _i (i=1,2,3,,n), and i is the variable subscript. Let x ₁ represent temperature, x ₂ represent smoke concentration, x ₃ represent CO concentration, and x ₄ represent infrared light intensity;

Step 2: Establish a judgment matrix for each variable:

The influence degree relationship between variables is quantitatively represented by a judgment matrix. Let the judgment matrix be C∈R ^n×n , where n represents the number of variables.

The judgment matrix C has the form:

matrix element

Indicates the importance ratio of the influence degree of the i-th and j-th variables on the fire situation.

The rules for filling in element c _{ij are:}

If x _i and x _{j are} equally important, then take c _ij =1, c _ji =1;

If x _i is slightly more important than x _j , then take c _ij =3, c _ji =1/3;

If x _i is obviously more important than x _j , then take c _ij =5, c _ji =1/5;

If x _i is much more important than x _j , then take c _ij =7, c _ji =1/7;

If x _i is absolutely more important than x _j , then take c _ij =9, c _ji =1/9;

If _{the importance of x i} and x _j is between the above relationships, then c _ij can take 2, 4, 6, and 8; c _ji can take the values of 1/2, 1/4, 1/6, and 1/8. Therefore, the element c _ij can take any integer from 1 to 9 according to the above rules.

Step 3: Check the consistency of the judgment matrix:

Set Consistency Ratio

in,

(λ is the largest eigenvalue of C, and n is the order of the judgment matrix). RI is the average random consistency index, and its value is determined by the different orders of the judgment matrix (as shown in Table 1). When CR<0.1, the corresponding judgment matrix C conforms to the consistency test and is a reasonable matrix, otherwise it should be discarded.

Table 1 RI values under different orders

matrix order	1	2	3	4	5	6	7	8
RI	0	0	0.52	0.89	1.12	1.26	1.36	1.41

Step 4: Eigenvector method to find the variable weight interval:

Based on the judgment matrix constructed above, the weight of each variable is calculated by the eigenvector method, and the calculated weight is written into the form of an interval, which is recorded as the weight interval of the variable (this interval is composed of independent weight points).

Corresponding to an arbitrarily selected reasonable judgment matrix, a weight of the variable can be calculated according to the eigenvector method. The weight calculation formula is as follows:

In the formula, i is the variable subscript (i=1,2,3,,n), and k represents the number of reasonable judgment matrices selected (k=1,2,3,,m; is a positive integer);

is a weight of the i-th variable obtained under the k-th judgment matrix, so k weights can be obtained for each variable i.

The k weights of each variable obtained above are written in the form of an interval, and the interval is the weight interval of the variable. For any variable i, a weight interval of the variable i can be obtained based on formulas (2) to (4), and the interval includes k weights.

Step 5: Set the objective function to obtain the optimal weight:

The optimal weight should be able to take into account all the value information in the weight interval to the greatest extent, so as to ensure that the final evaluation result can better reflect the actual situation; at the same time, using the optimal weight for weighted fusion can effectively avoid the tedious calculation process during fusion, improve Speed of reaction to fire. Let w _i represent a weight of the i-th variable, and construct the objective function that minimizes the weight deviation in the interval (that is, the objective influence function above) as:

The constraints of the objective function are:

Based on formulas (5)-(6), the optimal weight of the i-th variable is obtained as

1.2 Fire probability evaluation algorithm based on variable weighted fusion

According to the actual value of each variable in different environments in the cabin, the temperature (unit °C) is set in the interval [0, 100], the smoke concentration (in ppm) in the interval [100, 1000], and the CO concentration (in ppm) in the interval [ 10, 100], the infrared light intensity (unit Lux) is in the interval [100, 1000]. A conversion function is constructed to map the real value of the variable to [0-1] to avoid affecting the calculation result due to different dimensions. The conversion function is:

Among them, max( _xi ) is the maximum value in the interval where the variable i is located, and min( _xi ) is the minimum value in the interval where the variable i is located.

Suppose N sensor nodes are deployed in the cabin, the evaluation value of each node for the probability of fire occurrence in the cabin is p, and the normalized variable value is recorded as

The optimal weight of the variable is

The variable weighted result is the probability of fire occurrence, and the calculation formula is:

in

It is the evaluation result of the probability of fire occurrence by node i ^*.

2. Fusion algorithm based on node data support

During fire detection, if the probability of fire in the cabin is greater than the threshold probability, a fire alarm signal needs to be issued. In order to ensure the accuracy of fire alarm, it is necessary to fuse the fire probability values evaluated by each node. At the same time, in order to avoid the influence on the accuracy of fusion results when the sensor nodes deployed in the cabin fail, this section proposes a multi-sensor node data fusion method based on support matrix. This method objectively reflects the support degree between the evaluation data of nodes without knowing the evaluation ability of each node to the probability of fire occurrence. By constructing an augmented matrix, the weight coefficient of each node's fire probability evaluation value in the fusion process is adaptively adjusted to achieve the best fusion effect. The feature of this method is that it can fuse a large amount of data online, and in the process of adaptive weight coefficient allocation, by assigning a higher weight to the fire evaluation value with high reliability, and otherwise assigning a lower weight, the algorithm has a certain fault tolerance. ability.

2.1 Constructing the support matrix

There are N sensor nodes (each node is composed of temperature, smoke concentration, CO concentration and infrared light intensity sensors) to measure environmental variables, and the probability of fire occurrence in the cabin at this moment can be obtained through variable weighted fusion calculation at each node. make

and

^{Represents the evaluation value (i*} ,j ^* ∈1,2,…,N) of sensor nodes i ^* and j ^* on the probability of fire occurrence in the cabin at time k, respectively. if

and

If the difference is large, it means that the two sensor nodes do not support each other at time k, and if the difference is small, it means that the two nodes support each other. If a node is simultaneously supported by a majority of nodes, it is considered a valid fire probability estimate. Otherwise, the probability value evaluated by this node will be given a lower weight in the fusion process.

in order to express

and

The mutual support between , and the support function is defined as a decaying exponential function in the fuzzy set:

The parameter α is adjustable and is used to adjust the fusion accuracy, which is generally set to 0.8 [16].

Support function

Represents the mutual support of node i ^* and node j ^* at the kth moment, which can generally be expressed in matrix form:

After getting the support matrix

After that, the mutual support relationship between the nodes can be determined. for

the i ^* th column of ,

The larger the value, the higher the reliability of the fire probability evaluation value obtained at the node i ^{*, and the lower the reliability otherwise.}

2.2 Constructing the Augmented Support Matrix

In actual detection, a large number of fire probability evaluation values will be obtained. In order to integrate all evaluation values in each sampling time, an augmented support matrix is defined, which increases the dimension of the support matrix by one row and one column. The purpose of constructing this new dimension support matrix is to measure the mutual support between all current evaluation values and previous evaluation values, so as to adaptively assign weight coefficients to each node.

The specific steps of constructing the augmented matrix are as follows:

When k=1, use the average of the first N evaluation values

As an initial fire probability assessment value

Calculate the new rows and columns of the augmented support matrix at time k:

Represents the fusion result of the fire probability evaluation values of all nodes at time (k-1).

The augmented support matrix at time k can be defined as:

It reflects the comprehensive support of all nodes in each sampling time.

2.3 Weighted fusion node evaluation value

Assume

Express

The fusion weight coefficient (w _N+1 (k) is the fusion result at (k-1) time

weight coefficient),

satisfy:

The augmented support matrix

, through the

Integrate the i ^* th column to get

reliability, therefore,

is true

's points. set a vector

Each element is the result of integrating the i*th of sj*i*(k),

for:

Among them, i ^* ,j ^* =1,2,...,N+1.

According to formula (12), formula (14) becomes:

Wherein, W=[w ₁ (k),w ₂ (k),…w _N+1 (k)] ^T , A=[a ₁ (k),a ₂ (k),…a _N+1 (k )] ^T .

because

is a nonnegative symmetric matrix, by the Frobenius-Perron theorem,

Has the largest eigenvalue λ ^* (λ ^* > 0).

therefore,

Calculate the positive eigenvector A corresponding to ^{λ*, there are:}

W = λ ^* A (17)

Now the proportional relationship between the available weight coefficient and the eigenvector is:

According to formula (13), calculate the weight coefficient of each node:

Among them, i ^* ,j ^* =1,2,...,N+1.

Then, to get the final fusion expression:

in

It represents the probability of fire occurrence in the cabin after fusing the evaluation values of all sensor nodes, and compares it with the fire threshold probability. If it is greater than the threshold probability, the system sends out a fire alarm signal.

The critical value of fire occurrence is set as: the temperature is 55℃, the smoke concentration is 700ppm, the CO concentration is 20ppm, and the infrared light intensity is 760Lux. After the above variables are normalized and entered into formula (8), the result is the fire threshold probability.

When using the variable weighted fusion and the node fire probability evaluation data weighted fusion method to calculate the fire probability in the cabin, steps 1 to 5 in the first section are offline calculation processes, which aim to obtain the variables x _i (i=1,2,3,4 )Weights. When online detection of the fire condition in the cabin, the environmental data detected by the sensor is first normalized by formula (7) on each node, and then the fire probability is evaluated by formula (8); then based on the node data, mutual support The probability value of each node is estimated by self-adaptive weight coefficient distribution, and the actual fire probability value in the cabin is calculated after weighted fusion.

Ultimately, will

Compare with the threshold probability to judge whether to send out a fire alarm signal.

3. Simulation experiment verification

50 sensor nodes are randomly and uniformly deployed in the space. Each node is composed of temperature, smoke concentration, CO concentration and infrared light intensity sensors. The schematic diagram of node deployment is shown in Figure 1. The environmental parameters collected by each node are:

Variable units x1 is °C, x2 is ppm, x3 is ppm and x4 is Lux. Under the same experimental background, the variables that affect the probability of fire occurrence: temperature, smoke concentration, CO concentration and infrared light intensity are firstly calculated. The fusion algorithm is compared with the fuzzy logic fusion algorithm to verify the superiority of the online detection method (ie, the fire detection method of the present invention) in terms of timeliness, accuracy and fault tolerance of fire detection.

3.1 Calculation of the weight of each variable

According to the relevant introduction on the calculation of variable weights in Section 1, the weights of each variable are calculated as follows. Before calculating the weight of each variable, first make the following assumptions, the _{importance relationship of the degree of influence of variables x i} (i=1, 2, 3, 4) on the fire is as follows:

Assumption 1: x ₁ >x ₂ >x ₃ >x ₄ ;

Assumption 2: x ₁ >x ₂ >x ₄ >x ₃ ;

Assumption 3: x ₁ >x ₃ >x ₂ >x ₄ ;

Assumption 4: x ₁ >x ₃ >x ₄ >x ₂ ;

Assumption 5: x ₁ >x ₄ >x ₂ >x ₃ ;

Assumption 6: x ₁ >x ₄ >x ₃ >x ₂ ;

Assumption 24: x ₄ >x ₃ >x ₂ >x ₁

There are a total of 24 possible situations, and the above 24 hypotheses are recorded as A1, A2, …, A24.

Table 2 Calculation results of variable weight interval under assumption 13

According to the relationship between the influence degrees of the variables in the above assumptions, this experiment calculates the weight interval of each variable when k is 5 (where B1 to B5 are the judgment matrix), and uses formulas (5) and (6) to calculate each variable in the weight interval. Optimal weights for variables. Table 2 describes the relevant process of weight calculation in Section 1 (where Table 2 records the weight interval obtained under assumption 13), and Table 3 lists the corresponding output results:

Table 3 Output results of optimal weights for each variable under all assumptions

In this experiment, five sets of environmental parameters were selected: X1=[38,500,18,188]; X2=[89,580,30,200]; X3=[39,450,70,229]; X4=[51,750,38,199]; X5=[39,480,19,705]. Using the online detection method of the present invention under five sets of parameters, the probability of fire occurrence under the assumptions of A1, A2, . . . , A24 is obtained respectively. When the input parameters are the same, compare the fire occurrence probability calculated under the assumptions A1, A2,..., A24 with the fire occurrence probability obtained by the fuzzy fusion algorithm based on the "IF THEN" statement (through experiments on the car, it is proved that This method is feasible in fire detection), the curves shown in Figure 2 represent the error curves of A1, A2, ..., A24 relative to the fire probability obtained by the method proposed in the present invention. It can be seen from the simulation results that under the assumption condition A13 The lower error is the smallest. It can be seen that the _{relationship between the influence degree of variable x i} (i=1, 2, 3, 4) on the probability of fire occurrence is x ₃ > x ₁ > x ₂ > x ₄ , and the weight of each variable is [0.1474, 0.1439, 0.6492, 0.0595].

3.2 Analysis of timeliness and accuracy of online detection algorithm

From the perspective of fire detection and the time required to issue an alarm signal, Environment variable parameters to detect 50 nodes in the space

Substitute the above three fire detection algorithms respectively, carry out 100 independent repeated experiments, and record the respective detection time. where x ₁ takes a value _{in the interval (55, 100], x 2} takes a value in the interval (700, 1000], x ₃ takes a value _{in the interval (20, 100], and x 4} takes a value in the interval (760, 1000]. From It can be seen from the simulation curve in Fig. 3 that under the condition of the same input variables, the online detection algorithm of the present invention can complete the fire detection and alarm within 10s, while the gray-fuzzy neural network fusion algorithm in the related art completes the fire within 23s Detection and alarm, the fuzzy logic fusion algorithm in the related art completes fire detection and alarm within 20s. Experiments show that when the input variables are the same, the online detection algorithm proposed by the present invention is superior to the other two algorithms in terms of fire detection timeliness .

The core of the online detection algorithm is to calculate the weight occupied by each variable. The set of weights [0.6492, 0.1474, 0.1439, 0.0595] in the above experiment is obtained when the number of judgment matrices is 5. When k is 5, the fire false alarm rate is 3% under the background of this experiment, and taking this as the initial condition, the relationship between the number k selected for judgment matrix and the fire detection false alarm rate is analyzed. The simulation curve in Fig. 4 shows that the more reasonable judgment matrices are selected, the more accurate the obtained variable weights will be, and the more accurate the evaluation results of fire occurrence probability of each node will be. When k is greater than 20, the online detection algorithm reduces the false alarm rate to less than 0.5%, which effectively improves the accuracy of fire detection.

3.3 Fault Tolerance Analysis of Online Detection Algorithms

On the basis of 50 sensor nodes, when the number of faulty nodes is gradually increased, the deviation between the detected value and the true value of the probability of fire occurrence is observed. Assuming that the ambient temperature is 38°C, the smoke concentration is 650ppm, the CO concentration is 14ppm, and the infrared light intensity is 700lux and the assumption remains unchanged, it is assumed that the fire probability detected when all nodes are working normally is the true value, as shown in Table 4.

Table 4 The corresponding real values of fire probability under three detection algorithms

fire detection algorithm	fire probability
Grey Fuzzy Neural Network Fusion Method	30.56%
Fuzzy logic fusion method	29.48%
On-line detection method of the present invention	28.62%

The squared error (SE) is used as the standard to evaluate the detection accuracy of the above three algorithms, and SE is defined as:

Among them, j is the subscript of the detection algorithm (j=1, 2, 3); p ^j represents the true value of the probability of fire occurrence under the j algorithm;

It represents the estimated value of the probability of fire occurrence after fusing 50 nodes under the j algorithm.

It can be seen from the simulation curve in Fig. 5 that when the number of faulty nodes reaches 10, the SE of the gray-fuzzy neural network information fusion algorithm and the fuzzy logic control system with feedback in the related technology will increase sharply, and the detection accuracy of the algorithm will decrease, which cannot meet the practical application. System Requirements. However, the online detection algorithm proposed by the present invention performs self-adaptive distribution of weight coefficients based on the mutual support between data due to node data fusion. Therefore, when the number of faulty nodes increases, its SE value increases gently, and after the number of faulty nodes reaches 30, a large deviation occurs. Therefore, in an actual fire detection system, the fault tolerance capability of the online detection algorithm of the present invention is better.

4 Conclusion

The fire online detection algorithm of the present invention first calculates the weight of each variable by using the improved AHP at each node of the WSN, and proposes a new "variable weighted fusion algorithm" to evaluate the probability of fire occurrence, so that the system can still be used when there are many input variables. It can detect fire in time and accurately. Experiments show that the algorithm can complete the detection and alarm of fire within 10s. Compared with other fire detection algorithms, it greatly reduces the time required for fire detection and alarm, effectively avoids the spread of fire, and to a certain extent makes the follow-up fire. Firefighting has become smoother. When using WSN to collect environmental data, there is a problem that the detection accuracy is damaged or even the system is paralyzed due to sensor failure. The fusion is given a higher weight, and vice versa. By adaptively assigning weight coefficients to each node, the influence of deviation data collected by faulty sensors on fusion results is reduced, so the online detection algorithm has strong fault tolerance.

Corresponding to the above fire detection method provided by the embodiment of the present invention, the embodiment of the present invention further provides a fire detection device, the device comprising:

a first determination module, used for determining multiple fire-influencing factors of each node in the engine room;

a second determination module, configured to determine the fire influence weight value of each fire influence factor among the plurality of fire influence factors of each node;

a third determination module, configured to determine the evaluation value of each node on the probability of fire occurrence in the cabin according to the fire influence weight value of each fire influence factor;

a fourth determination module, configured to determine the probability of a fire actually occurring in the cabin according to the evaluation value of the probability of fire occurrence in the cabin by each node;

The judging module is used for judging whether to issue a fire alarm according to the probability of a fire actually occurring in the cabin.

In one embodiment, the second determining module includes:

a determination sub-module for determining the judgment matrix of each of the fire-influencing factors;

a check sub-module, used to check the consistency of the judgment matrix of each fire influencing factor;

The calculation sub-module is configured to calculate the fire influence weight value of each fire influence factor according to the judgment matrix of each fire influence factor that has passed the consistency check.

In one embodiment, the calculation submodule is specifically used for:

When there are multiple judgment matrices for each fire influencing factor, according to each judgment matrix of each fire influencing factor that passes the consistency check, a preset number of fire influence weight values for each fire influencing factor are calculated. , wherein the preset number is equal to the number of judgment matrices of the respective fire influencing factors that have passed the consistency check; the preset number is a positive integer greater than or equal to 2;

According to the preset number of fire influence weight values of each fire influence factor, the fire influence weight interval of each fire influence factor is determined.

In one embodiment, the third determining module includes:

a first determination sub-module, configured to determine the fire influence weight interval of each fire influence factor according to the fire influence weight value of each fire influence factor;

The second determination submodule is configured to determine the target influence function of each fire influence factor and the constraint condition of the target influence function according to the fire influence weight interval of each fire influence factor;

a calculation submodule, configured to calculate the optimal fire influence weight value of each fire influence factor of each node according to the target influence function of each fire influence factor and the constraint condition of the target influence function;

The third determination sub-module is configured to determine, according to the optimal fire influence weight value of each fire influence factor of each node and the actual factor value of each fire influence factor of each node, the effect of each node on the interior of the cabin. An estimate of the probability of fire occurrence.

In one embodiment, the fourth determining module is specifically configured to:

According to the evaluation value of each node on the probability of fire occurrence in the cabin, establish a support function between any two nodes in each node with respect to the evaluation value;

According to the support function between the any two nodes with respect to the evaluation value, establish an initial support degree matrix between the nodes with respect to the evaluation value;

Constructing an augmented support matrix for the initial support matrix between the evaluation values of each node; wherein, the augmented support matrix has one more row and one column compared to the initial support matrix;

According to the augmented support matrix, the probability that a fire actually occurs in the cabin is determined.

In one embodiment, the fourth determining module is further configured to:

determining, according to the augmented support matrix, an evaluation credibility coefficient of each node for the evaluation value of the probability of occurrence of fire in the cabin;

The probability of a fire actually occurring in the cabin is determined according to the evaluation value of the probability of occurrence of fire in the cabin by each node and the evaluation credibility coefficient of the evaluation value of the probability of occurrence of fire in the cabin by each node.

In one embodiment, the judging module is specifically used for:

Calculate the probability threshold of fire occurrence in the cabin according to the fire occurrence critical factor value of each fire influencing factor of each node;

Whether to issue a fire alarm is determined according to the actual probability of fire in the cabin and the probability threshold of fire in the cabin.

Other embodiments of the present disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or techniques in the technical field not disclosed by the present disclosure . The specification and examples are to be regarded as exemplary only, with the true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise structures described above and illustrated in the accompanying drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims (10)

1. A fire detection method is characterized in that, comprises the following steps:

   Identify multiple fire-influencing factors at each node in the engine room;

   determining the fire influence weight value of each fire influence factor among the multiple fire influence factors of each node;

   According to the fire influence weight value of each fire influence factor, the evaluation value of each node to the probability of fire occurrence in the cabin is determined;

   According to the evaluation value of the probability of fire occurrence in the cabin by each node, determine the probability of the actual occurrence of fire in the cabin;

   According to the probability of a fire actually occurring in the cabin, it is judged whether to issue a fire alarm.
2. The method of claim 1, wherein:

   The determining of the fire influence weight value of each fire influence factor among the multiple fire influence factors of each node includes:

   determining the judgment matrix for each of the fire influencing factors;

   Consistency test is carried out on the judgment matrix of each fire influencing factor;

   According to the judgment matrix of each fire influence factor that has passed the consistency check, the fire influence weight value of each fire influence factor is calculated.
3. The method according to claim 2, characterized in that:

   When there are multiple judgment matrices of each fire influencing factor, calculating the fire influence weight value of each fire influencing factor according to the judgment matrix of each fire influencing factor that has passed the consistency check, including:

   According to the judgment matrix of each fire influence factor that passes the consistency check, a preset number of fire influence weight values for each fire influence factor are calculated, wherein the preset number is equal to all the fire influence factors that pass the consistency check. the number of judgment matrices for each fire influencing factor; the preset number is a positive integer greater than or equal to 2;

   The fire influence weight interval of each fire influence factor is determined according to the preset number of fire influence weight values of each fire influence factor.
4. The method of claim 1, wherein:

   According to the fire influence weight value of each fire influence factor, the evaluation value of each node to the fire occurrence probability in the cabin is determined, including:

   According to the fire influence weight value of each fire influence factor, determine the fire influence weight interval of each fire influence factor;

   Determine the target influence function of each fire influence factor and the constraint condition of the target influence function according to the fire influence weight interval of each fire influence factor;

   According to the target influence function of each fire influence factor and the constraint condition of the target influence function, calculate the optimal fire influence weight value of each fire influence factor of each node;

   According to the optimal fire influence weight value of each fire influence factor of each node and the actual factor value of each fire influence factor of each node, the evaluation value of each node to the fire occurrence probability in the cabin is determined.
5. The method of claim 1, wherein:

   The determining the probability of a fire actually occurring in the cabin according to the evaluation value of the probability of occurrence of fire in the cabin by each node includes:

   According to the evaluation value of each node on the probability of fire occurrence in the cabin, establish a support function between any two nodes in each node with respect to the evaluation value;

   According to the support function between the any two nodes with respect to the evaluation value, establish an initial support degree matrix between the nodes with respect to the evaluation value;

   Constructing an augmented support matrix for the initial support matrix between the evaluation values of each node; wherein, the augmented support matrix has one more row and one column compared to the initial support matrix;

   According to the augmented support matrix, the probability that a fire actually occurs in the cabin is determined.
6. The method of claim 5, wherein:

   The determining, according to the augmented support matrix, the probability that a fire actually occurs in the cabin includes:

   determining, according to the augmented support matrix, the evaluation credibility coefficient of each node for the evaluation value of the probability of occurrence of fire in the cabin;

   The probability of a fire actually occurring in the cabin is determined according to the evaluation value of the probability of occurrence of fire in the cabin by each node and the evaluation credibility factor of the evaluation value of the probability of occurrence of fire in the cabin by each node.
7. The method according to any one of claims 1 to 6, wherein,

   The judging whether to issue a fire alarm according to the probability of a fire actually occurring in the cabin includes:

   Calculate the probability threshold of fire occurrence in the cabin according to the fire occurrence critical factor value of each fire influencing factor of each node;

   Whether to issue a fire alarm is determined according to the actual probability of fire in the cabin and the probability threshold of fire in the cabin.
8. A fire detection device, characterized in that the device comprises:

   a first determination module, used for determining multiple fire-influencing factors of each node in the engine room;

   a second determination module, configured to determine the fire influence weight value of each fire influence factor among the multiple fire influence factors of each node;

   a third determination module, configured to determine the evaluation value of each node on the probability of fire occurrence in the cabin according to the fire influence weight value of each fire influence factor;

   a fourth determination module, configured to determine the probability of a fire actually occurring in the cabin according to the evaluation value of the probability of occurrence of fire in the cabin by each node;

   The judging module is used for judging whether to issue a fire alarm according to the probability of a fire actually occurring in the cabin.
9. The device of claim 8, wherein:

   The second determining module includes:

   a determination sub-module for determining the judgment matrix of each of the fire-influencing factors;

   a check sub-module, used to check the consistency of the judgment matrix of each fire influencing factor;

   The calculation sub-module is configured to calculate the fire influence weight value of each fire influence factor according to the judgment matrix of each fire influence factor that has passed the consistency check.
10. The device according to claim 9, characterized in that:

    The calculation sub-module is specifically used for:

    When there are multiple judgment matrices for each fire influencing factor, according to each judgment matrix for each fire influencing factor that passes the consistency check, calculate a preset number of fire influence weight values for each fire influencing factor , wherein the preset number is equal to the number of judgment matrices of the respective fire influencing factors that have passed the consistency check; the preset number is a positive integer greater than or equal to 2;

    The fire influence weight interval of each fire influence factor is determined according to the preset number of fire influence weight values of each fire influence factor.

Images