Zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment

Abstract

The present invention relates to the technical field of fire alarm and fire extinguishing, and discloses a zero-false-alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment, comprising: a data acquisition module for acquiring environmental monitoring data and image data; an image analysis module for outputting image fire confidence using a pre-trained model; an information fusion judgment module for calculating the single fire confidence of each monitoring parameter and fusing them to obtain a comprehensive fire confidence; a dynamic test spray verification module for controlling the fire extinguishing agent to conduct a test spray, establishing a sensor prediction model to output predicted environmental monitoring data, and outputting a verification result; an adaptive dose control module for calculating the heat release amount and then the main fire extinguishing dose using the temperature rise-volume energy balance method; and a fire extinguishing execution module for implementing the main spray fire extinguishing and calculating the residual fire index, triggering supplementary spraying when a threshold is reached. The present invention improves the intelligence and reliability of the fire alarm and fire extinguishing system.

Description

Technical Field

[0001] The present invention belongs to the technical field of fire alarm and fire extinguishing, and in particular relates to a zero-false-alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment. Background Art

[0002] Existing automatic fire alarm and extinguishing systems often rely on single sensor signals or simple multi-sensor logic combinations for fire detection and extinguishing control. These systems suffer from high false alarm rates, delayed response times, wasteful use of extinguishing agent, and poor adaptability to complex fire scenarios. This is particularly true in key protected areas such as industrial plants, data centers, and power distribution rooms, where environmental factors are complex and variable. Single signals are susceptible to interference, leading to false or missed alarms and severely impacting system stability and safety. While some systems have attempted to incorporate multi-sensor information fusion, most remain at the level of simple weighting or logical judgment, lacking reliable mechanisms for deep multi-information fusion and fire authenticity verification. This makes it difficult to ensure timely response while reducing the risk of false alarms and misfires. Furthermore, the extinguishing dose in existing systems is often fixed and cannot be dynamically adjusted based on the actual fire scale. This results in low fire extinguishing efficiency and poor dose utilization. There are even cases where insufficient dose fails to extinguish the fire source or excessive dose causes secondary damage. Summary of the Invention

[0003] The present invention provides a zero-false-alarm automatic fire alarm and extinguishing system based on multi-condition judgment, which solves the technical problems in related technologies such as reliance on a single information source for fire determination, high false alarm rate, lack of effective authenticity verification, and inability to dynamically adjust the fire extinguishing dosage according to the actual fire situation.

[0004] The present invention provides a zero-false-alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment, comprising:

[0005] The data acquisition module is used to obtain environmental monitoring data and image data within the protection zone according to a preset sampling period, wherein the environmental monitoring data includes: smoke concentration, temperature, flame signal and carbon monoxide concentration;

[0006] An image analysis module, configured to output image fire confidence based on image data using a pre-trained model;

[0007] The information fusion judgment module is used to calculate the single fire confidence level of each monitoring parameter in the environmental monitoring data, and fuse each single fire confidence level with the image fire confidence level based on the DS evidence theory to obtain a comprehensive fire confidence level. When the comprehensive fire confidence level reaches the preset fire confirmation threshold, a fire trigger signal is output.

[0008] A test spray dynamic verification module is used to control the fire extinguishing agent to perform a small-dose test spray after receiving a fire trigger signal, establish a sensor prediction model based on the environmental monitoring data in a first preset time period before the test spray, output the predicted environmental monitoring data for a second preset time period, calculate the difference between the predicted environmental monitoring data and the environmental monitoring data in the second preset time period, and output a verification pass result when the difference meets a preset verification condition;

[0009] The adaptive dose control module is used to calculate the heat release of the protection zone according to the temperature rise-volume energy balance method based on the test spray phase and the current environmental monitoring data when receiving the verification result, and to calculate the main fire extinguishing dose in real time using the proportional-integral control logic and output the main fire extinguishing dose instruction;

[0010] The fire extinguishing execution module is used to implement the main spraying fire extinguishing after receiving the main fire extinguishing dosage instruction, and calculate the residual fire index according to the comprehensive fire confidence value within the set monitoring period. When the residual fire index reaches the preset supplementary spraying threshold, it triggers the supplementary spraying.

[0011] Furthermore, a pre-trained model is used to output image fire confidence based on the image data, including:

[0012] S201, normalizing and cropping the current original frame image in the image data to output a standard size image;

[0013] S202, inputting the standard size image into the pre-trained convolutional neural network model to extract the image feature vector;

[0014] S203, inputting the image feature vector into the fully connected layer to output the single-frame fire confidence;

[0015] S204: Calculate a gradient-weighted class activation map heat map of the standard-sized image based on the single-frame fire confidence score, and calculate the ratio of the number of high-heat pixels in the heat map to the total number of pixels in the image to obtain a high-heat area ratio.

[0016] S205 , weighting the fire confidence of a single frame according to the area ratio of the high-heat region to obtain the fire confidence of the image.

[0017] Furthermore, the single fire reliability of each monitoring parameter in the environmental monitoring data is calculated, including:

[0018] S301, normalizing each monitoring parameter according to its own preset physical range to obtain the corresponding normalized monitoring parameter;

[0019] S302, calculating the difference between each normalized monitoring parameter and the corresponding preset alarm threshold;

[0020] S303 , performing Sigmoid mapping on each difference and performing first-order exponential smoothing to obtain each single fire confidence level.

[0021] Furthermore, based on the DS evidence theory, the single fire confidence and the image fire confidence are fused to obtain the comprehensive fire confidence, which includes:

[0022] S401, mapping each single fire confidence level and image fire confidence level to a corresponding basic probability distribution based on a preset domain, wherein the preset domain includes two propositions: the existence of a fire and the non-existence of a fire;

[0023] S402: Calculate a conflict coefficient for any two basic probability distributions based on the DS evidence theory. The conflict coefficient is the sum of the products of the probabilities supporting the proposition that a fire exists and the proposition that a fire does not exist, respectively, in the two basic probability distributions. The difference between one and the conflict coefficient is used as a normalization factor. The two basic probability distributions are fused according to the fusion rules of the DS evidence theory to obtain an updated fused distribution.

[0024] In step S403, the updated fused distribution is sequentially fused with the remaining basic probability distributions according to step S402 until all fusions are completed, and finally the probability of supporting the proposition that there is a fire in the fused distribution is output as the comprehensive fire confidence.

[0025] Furthermore, the small-dose test spraying includes: spraying a fire extinguishing agent of a preset proportion of the maximum design dose at the current moment, and the spraying duration is a first duration;

[0026] The sensor prediction model is constructed by adopting a first-order autoregressive structure in combination with a Kalman filter algorithm.

[0027] Further, calculating the difference between the predicted environmental monitoring data and the environmental monitoring data for the second preset time period, and outputting a verification pass result when the difference meets a preset verification condition, including:

[0028] S501, respectively calculating the difference between the predicted environmental monitoring data and the environmental monitoring data for a second preset time period to obtain residual data;

[0029] S502, calculating the overall root mean square error based on the residual data as a quantitative indicator of the test spraying effect;

[0030] S503 : When the overall root mean square error is less than the preset error threshold, a verification pass signal is output; otherwise, a verification fail signal is output.

[0031] Furthermore, upon receiving the verification result, the heat release of the protection zone is calculated according to the temperature rise-volume energy balance method based on the test spray phase and the current environmental monitoring data, including:

[0032] S601, obtaining the temperature in the current environmental monitoring data and the reference temperature before the test spraying, and calculating the difference between the two as the temperature rise;

[0033] S602: Multiply the temperature rise by the air density, specific heat capacity, and volume of the protection zone, and divide the result by a preset time parameter to obtain a preliminary heat release amount of the protection zone.

[0034] S603 , obtaining smoke concentration in current environmental monitoring data, calculating a ratio of the smoke concentration to a preset reference smoke concentration, and performing weighted correction on the preliminary heat release amount based on the ratio to obtain the heat release amount.

[0035] Furthermore, the heat release amount is multiplied by the preset fire suppression duration to obtain the target fire suppression energy; based on the difference between the target fire suppression energy and the energy sprayed during the trial spraying phase, the proportional integral algorithm is used to calculate the main fire extinguishing dose without limit, and after executing upper and lower limit constraints on it, the main fire extinguishing dose is finally obtained.

[0036] Furthermore, the residual fire index is calculated based on the comprehensive fire confidence value during the set monitoring period. When the residual fire index reaches the preset supplementary spraying threshold, supplementary spraying is triggered, including:

[0037] S701, reading the comprehensive fire confidence at the current moment and the residual fire index at the previous moment within the set monitoring period;

[0038] S702: The residual fire index at the previous moment is weighted and superimposed with the comprehensive fire confidence at the current moment according to a preset smoothing coefficient to update and generate the residual fire index at the current moment;

[0039] S703, comparing the residual fire index at the current moment with a preset supplementary injection threshold, outputting a supplementary injection trigger signal if the residual fire index reaches the preset supplementary injection threshold, otherwise keeping in standby mode.

[0040] The beneficial effects of the present invention are as follows: the present invention integrates environmental parameters and fire confidence of images through DS evidence theory, combined with test spraying verification, greatly reducing the false alarm rate; calculates heat release based on the temperature rise-volume energy balance method, dynamically determines the main fire extinguishing dose, and avoids waste and insufficiency; updates the residual fire index through weighted smoothing coefficients, accurately triggers supplementary spraying, and prevents re-ignition; and can adjust parameters according to the characteristics of the protection zone, taking into account different scenarios, thereby improving the intelligence and reliability of the system as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a module schematic diagram of the zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment of the present invention. DETAILED DESCRIPTION

[0042] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. In addition, features described with respect to some examples may also be combined in other examples.

[0043] It should be noted that, unless otherwise defined, the technical or scientific terms used in one or more embodiments of the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprising" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, but do not exclude other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0044] like Figure 1 As shown, the zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment includes:

[0045] The data acquisition module 101 is used to obtain environmental monitoring data and image data within the protection zone according to a preset sampling period, wherein the environmental monitoring data includes: smoke concentration, temperature, flame signal and carbon monoxide concentration;

[0046] The image analysis module 102 is configured to output image fire confidence levels using a pre-trained model based on image data.

[0047] The information fusion judgment module 103 is used to calculate the single fire confidence level of each monitoring parameter in the environmental monitoring data, and fuse each single fire confidence level with the image fire confidence level based on the DS evidence theory to obtain a comprehensive fire confidence level. When the comprehensive fire confidence level reaches a preset fire confirmation threshold, a fire trigger signal is output.

[0048] The test spray dynamic verification module 104 is used to control the fire extinguishing agent to perform a small-dose test spray after receiving the fire trigger signal, establish a sensor prediction model based on the environmental monitoring data in a first preset time period before the test spray, output the predicted environmental monitoring data for a second preset time period, calculate the difference between the predicted environmental monitoring data and the environmental monitoring data in the second preset time period, and output a verification pass result when the difference meets a preset verification condition;

[0049] The adaptive dose control module 105 is used to calculate the heat release of the protection zone according to the temperature rise-volume energy balance method based on the test spray phase and the current environmental monitoring data upon receiving the verification result, and to calculate the main fire extinguishing dose in real time using the proportional-integral control logic and output the main fire extinguishing dose instruction;

[0050] The fire extinguishing execution module 106 is used to implement the main spraying fire extinguishing after receiving the main fire extinguishing dosage instruction, and calculate the residual fire index according to the comprehensive fire confidence value within the set monitoring period, and trigger the supplementary spraying when the residual fire index reaches the preset supplementary spraying threshold.

[0051] In one embodiment of the present invention, the data acquisition module acquires environmental monitoring data and image data within the protected area according to a preset sampling period to achieve real-time perception of fire risk in the protected area. The preset sampling period can be set based on the fire hazard level and environmental complexity of the protected area. For example, in areas prone to fire and with rapidly changing environments, the sampling period can be set to 1 second to ensure data timeliness. In areas with lower fire risk and stable environments, the sampling period can be appropriately extended to 3-5 seconds to reduce system energy consumption.

[0052] The smoke concentration is obtained through an optical scattering smoke sensor; the temperature is collected through a thermistor sensor; the flame signal is obtained through an infrared flame detector; the carbon monoxide concentration is collected through an electrochemical carbon monoxide sensor; and the image data is collected through high-definition cameras installed at key monitoring points in the protection area. The camera installation position must cover the main area of ​​the protection area to avoid monitoring blind spots.

[0053] In one embodiment of the present invention, outputting an image fire confidence score using a pre-trained model based on image data includes:

[0054] S201, normalize and resize the current original frame image in the image data to output a standard-sized image; wherein, normalization is to map the pixel values ​​of the original frame image to the range of 0 to 1 to eliminate the impact of pixel value differences under different lighting conditions on model analysis; resize cropping is to crop or scale the original frame image according to the input specification of 224×224 pixels required by the pre-trained model, and finally output a standard-sized image.

[0055] S202: Input the standard-sized image into a pre-trained convolutional neural network model to extract the image feature vector. The convolutional neural network model uses a ResNet model, which extracts deep features of the standard-sized image layer by layer through the convolutional layers, pooling layers, and other structures of the ResNet model. These features include key visual features related to fire, such as edges, textures, and color distribution, and ultimately outputs an image feature vector of a fixed dimension.

[0056] S203: Input the image feature vector into a fully connected layer to output a single-frame fire confidence score. The single-frame fire confidence score ranges from 0 to 1, with a larger value indicating a higher probability that the original frame image contains a fire.

[0057] S204, calculate the gradient weighted class activation mapping heat map of the standard size image according to the single-frame fire confidence, and calculate the ratio of the number of high-heat pixels in the heat map to the total number of pixels in the image to obtain the area ratio of the high-heat region; specifically, take the last convolutional layer feature map of the convolutional neural network model, obtain the gradient of the single-frame fire confidence to the feature map through back propagation, obtain the weight of each feature map through global average pooling, activate it with the ReLU activation function after weighted summation, and then interpolate to the standard size to obtain a heat map reflecting the contribution to fire judgment; the high-heat pixels represent pixels in the heat map whose pixel values ​​exceed the preset heat threshold, and the preset heat threshold is preferably set to 0.7.

[0058] S205. Weight the single-frame fire confidence according to the proportion of the high-heat area to obtain the image fire confidence. Specifically, when the proportion of the high-heat area is greater than a preset proportion of 30%, the single-frame fire confidence is positively weighted, and vice versa. The image fire confidence reflects the intensity and spatial distribution significance of the fire characteristics in the original frame image.

[0059] In one embodiment of the present invention, calculating a single fire reliability of each monitoring parameter in the environmental monitoring data includes:

[0060] S301 normalizes each monitoring parameter according to its preset physical range to obtain the corresponding normalized monitoring parameter. The preset physical range for smoke concentration is 0-1000 ppm, the preset physical range for temperature is 0-100°C, the preset physical range for flame signal is 0-500 mV, and the preset physical range for carbon monoxide concentration is 0-500 ppm. This normalization process uses a linear mapping method to map the actual measured value of each monitoring parameter to the range of 0 to 1, obtaining the corresponding normalized monitoring parameter.

[0061] S302, calculating the difference between each normalized monitoring parameter and the corresponding preset alarm threshold; preferably, the preset alarm threshold for smoke concentration is set to 0.3, corresponding to an actual value of 150 ppm, the preset alarm threshold for temperature is set to 0.6, corresponding to an actual value of 60°C, the preset alarm threshold for flame signal is set to 0.4, corresponding to an actual value of 200 mV, and the preset alarm threshold for carbon monoxide concentration is set to 0.2, corresponding to an actual value of 100 ppm;

[0062] S303 , performing Sigmoid mapping on each difference and performing first-order exponential smoothing to obtain each single fire confidence level; specifically, mapping each difference to a range of 0 to 1 using a Sigmoid function and performing first-order exponential smoothing.

[0063] This embodiment eliminates the physical dimension differences of different monitoring parameters through normalization, making the monitoring parameters comparable. The difference calculation with the preset alarm threshold accurately quantifies the extent to which the parameter exceeds the safety range. The Sigmoid mapping converts the difference into a confidence level in the range of 0 to 1, and combines it with first-order exponential smoothing to filter out instantaneous fluctuation interference. The resulting single fire confidence level can stably and accurately reflect the degree to which each parameter supports the existence of a fire.

[0064] In one embodiment of the present invention, the DS evidence theory establishes a basic probability distribution to represent the support of evidence for a proposition, calculates a conflict coefficient to measure the contradiction of evidence, and then synthesizes multi-source evidence according to fusion rules to output a comprehensive support degree, thereby achieving effective integration of multiple information. Based on the DS evidence theory, each single fire confidence level and the image fire confidence level are fused to obtain a comprehensive fire confidence level, including:

[0065] S401: Based on a preset universe, each single fire confidence level and image fire confidence level is mapped to a corresponding basic probability distribution. The preset universe includes two propositions: the existence of a fire and the non-existence of a fire. Specifically, for the single fire confidence level and image fire confidence level of an environmental monitoring parameter, the basic probability distribution is defined as: , ,in, represents the basic probability distribution of the i-th evidence of the fire proposition, H represents the fire proposition, means there is no fire proposition, represents the corresponding fire confidence, The basic probability distribution of the i-th piece of evidence for the proposition that there is no fire.

[0066] S402: Calculate a conflict coefficient for any two basic probability distributions based on the DS evidence theory. The conflict coefficient is the sum of the products of the probabilities supporting the proposition that a fire exists and the proposition that a fire does not exist, respectively, in the two basic probability distributions. The difference between one and the conflict coefficient is used as a normalization factor. The two basic probability distributions are fused according to the fusion rules of the DS evidence theory to obtain an updated fused distribution.

[0067] Specifically, the calculation formula of the conflict coefficient is: , k represents the conflict coefficient, which is used to reflect the degree of conflict between the two basic probability distributions, that is, the sum of the products of the probabilities that one side supports the existence of fire and the other side supports the non-existence of fire; and They represent the basic probability distribution of the ath and bth evidence of the fire proposition respectively, and They represent the basic probability distribution of the a-th and b-th evidence of the proposition that there is no fire respectively; the core logic of the DS fusion rule is: first calculate the conflict coefficient, then use 1-conflict coefficient as the normalization factor, sum the probabilities of each piece of evidence supporting the same proposition, and obtain the probability distribution of the proposition after fusion. At the same time, process the uncertainty part and finally output the comprehensive evidence support to achieve effective fusion of multi-source information.

[0068] In step S403, the updated fused distribution is sequentially fused with the remaining basic probability distributions according to step S402 until all fusions are completed, and finally the probability of supporting the proposition that there is a fire in the fused distribution is output as the comprehensive fire confidence.

[0069] When the comprehensive fire confidence reaches the preset fire confirmation threshold, a fire trigger signal is output to start the subsequent test spray dynamic verification process, realizing the connection from fire monitoring to fire extinguishing response.

[0070] This embodiment automatically identifies and processes conflicting information between different sensors through the calculation and normalization of conflict coefficients, avoiding a single false alarm sensor dominating the decision-making; during the fusion process, the weights of evidence supporting the same proposition are superimposed and enhanced, making the comprehensive fire confidence closer to the actual state.

[0071] In one embodiment of the present invention, the small-dose test spray includes: spraying a preset proportion of the maximum design dose of fire extinguishing agent at the current moment, and the spraying duration is a first duration; wherein the preset proportion is preferably set to 3%, and the first duration is preferably set to 1 second;

[0072] The sensor prediction model adopts a first-order autoregressive structure and is constructed in combination with a Kalman filter algorithm. Specifically, the sensor prediction model is established based on the environmental monitoring data of the previous moment, and the observation noise and process noise in the prediction process are dynamically corrected in combination with the Kalman filter algorithm, and the predicted environmental monitoring data for the second preset time period after the test spraying is output.

[0073] In this embodiment, a small-dose trial spray can not only test the environmental response through actual intervention, but also avoid excessive waste of reagents due to misspraying; the prediction model combining the first-order autoregressive structure and Kalman filtering can accurately capture the changing trend of environmental parameters, effectively filter out instantaneous interference, and improve the matching degree between predicted data and actual data, providing a reliable model basis for subsequent verification of fire authenticity and reducing the probability of misjudgment due to environmental fluctuations.

[0074] In one embodiment of the present invention, calculating the difference between the predicted environmental monitoring data and the environmental monitoring data for the second preset time period, and outputting a verification pass result when the difference meets a preset verification condition, includes:

[0075] S501, respectively calculating the difference between the predicted environmental monitoring data and the environmental monitoring data for the second preset time period to obtain residual data; specifically, calculating the difference between each parameter in the predicted environmental monitoring data and the environmental monitoring data moment by moment to obtain residual data.

[0076] S502, calculating the overall root mean square error based on the residual data as a quantitative indicator of the test spraying effect; the overall root mean square error reflects the overall deviation between the prediction model and the actual environment, and the smaller the value, the smaller the deviation.

[0077] S503: When the overall root mean square error is less than the preset error threshold, a verification pass signal is output, allowing the system to enter the main spray fire extinguishing process; otherwise, a verification fail signal is output, only the sound and light alarm is triggered and the time is recorded, and the subsequent fire extinguishing operation is terminated.

[0078] This embodiment achieves secondary verification of fire authenticity through residual analysis and root mean square error quantification verification, effectively filtering out false fire signals caused by sensor false alarms and environmental interference.

[0079] In one embodiment of the present invention, upon receiving the verification result, the heat release amount of the protection zone is calculated according to the temperature rise-volume energy balance method based on the test spraying stage and the current environmental monitoring data, including:

[0080] S601, obtaining the temperature in the current environmental monitoring data and the reference temperature before the test spraying, and calculating the difference between the two as the temperature rise;

[0081] S602: Multiply the temperature rise by the air density, specific heat capacity, and volume of the protection zone, and divide the result by a preset time parameter to obtain a preliminary heat release amount for the protection zone. The preset time parameter represents the time interval from the start of the test spraying to the current moment. The air density and specific heat capacity are values ​​under standard conditions.

[0082] S603, obtain the smoke concentration in the current environmental monitoring data, calculate the ratio of the smoke concentration to the preset reference smoke concentration, and perform weighted correction on the preliminary heat release based on the ratio to obtain the heat release. The heat release calculation formula is: , where Q represents the heat release, represents the initial heat release, represents the correction factor, Indicates smoke density. Indicates the preset reference smoke concentration, taking the smoke concentration baseline value of 100ppm in a fire scene.

[0083] In one embodiment of the present invention, the heat release amount is multiplied by the preset fire suppression duration to obtain the target fire suppression energy; based on the difference between the target fire suppression energy and the energy sprayed during the trial spraying phase, the proportional integral algorithm is used to calculate the main fire extinguishing dose without limit, and after applying upper and lower limit constraints to it, the main fire extinguishing dose is finally obtained.

[0084] Specifically, the preset fire suppression duration is set according to the fire hazard level of the protection area; the calculation formula for the unlimited main fire extinguishing dose is: ,in, Indicates the unlimited primary extinguishing dose, represents the proportionality coefficient, Indicates the integral coefficient. The integral term accumulates the energy error since the start of the test spray. represents the difference between the target suppression energy and the energy injected during the trial injection phase, and t represents the time interval from the start of the trial injection to the completion of the calculation of the main fire extinguishing dose. These upper and lower bounds ensure that the main fire extinguishing dose is within a safe and effective range, preventing over-injection or under-injection caused by algorithm runaway.

[0085] In one embodiment of the present invention, a residual fire index is calculated based on a comprehensive fire confidence value within a set monitoring period, and supplementary spraying is triggered when the residual fire index reaches a preset supplementary spraying threshold, including:

[0086] S701, reading the comprehensive fire confidence at the current moment and the residual fire index at the previous moment within the set monitoring period;

[0087] S702: The residual fire index at the previous moment is weighted and superimposed with the comprehensive fire confidence at the current moment according to a preset smoothing coefficient to generate the residual fire index at the current moment. The calculation formula of the residual fire index is: , represents the residual fire index at the current time s, Indicates the residual fire index of the previous moment. The residual fire index is 0 at the initial moment. represents the smoothing coefficient, Represents the comprehensive fire confidence at the current moment.

[0088] S703, compare the residual fire index at the current moment with the preset supplementary spraying threshold. If the residual fire index reaches the preset supplementary spraying threshold, a supplementary spraying trigger signal is output; otherwise, the system remains in standby mode. The preset supplementary spraying threshold is preferably set to 0.3. If the residual fire index is ≥ 0.3, it indicates that there is still a significant residual fire risk after the main spraying. The system immediately outputs a supplementary spraying trigger signal, and the control device performs supplementary spraying at 25% of the main fire extinguishing dose. If the residual fire index is < 0.3, it is determined that the fire has been effectively controlled and the system remains in standby mode until the end of the monitoring period.

[0089] It should be noted that the intervals and thresholds are set for ease of comparison. The threshold size depends on the amount of sample data and the cardinality set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless numerical calculations. These formulas are derived from software simulations of the most recent real-world conditions using large amounts of data. The preset parameters in these formulas are set by those skilled in the art based on actual conditions.

[0090] The above describes the embodiments of the present invention, but the present invention is not limited to the above specific implementation methods. The above specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make many forms based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment, characterized by: include: The data acquisition module is used to obtain environmental monitoring data and image data within the protection zone according to a preset sampling period, wherein the environmental monitoring data includes: smoke concentration, temperature, flame signal and carbon monoxide concentration; An image analysis module is used to output image fire confidence based on image data using a pre-trained model; The information fusion judgment module is used to calculate the single fire confidence level of each monitoring parameter in the environmental monitoring data, and fuse each single fire confidence level with the image fire confidence level based on the DS evidence theory to obtain a comprehensive fire confidence level. When the comprehensive fire confidence level reaches the preset fire confirmation threshold, a fire trigger signal is output. A test spray dynamic verification module is used to control the fire extinguishing agent to perform a small-dose test spray after receiving a fire trigger signal, establish a sensor prediction model based on the environmental monitoring data in a first preset time period before the test spray, output the predicted environmental monitoring data for a second preset time period, calculate the difference between the predicted environmental monitoring data and the environmental monitoring data in the second preset time period, and output a verification pass result when the difference meets a preset verification condition; The adaptive dose control module is used to calculate the heat release of the protection zone using the temperature rise-volume energy balance method based on the test spray phase and current environmental monitoring data upon receiving the verification result, and to calculate the main fire extinguishing dose in real time using proportional-integral control logic and output the main fire extinguishing dose instruction; The calculation process of heat release includes: S601, obtaining the temperature in the current environmental monitoring data and the reference temperature before the test spraying, and calculating the difference between the two as the temperature rise; S602: Multiply the temperature rise by the air density, specific heat capacity, and volume of the protection zone, and divide the result by a preset time parameter to obtain a preliminary heat release amount of the protection zone. S603, obtaining smoke concentration in current environmental monitoring data, calculating a ratio of the smoke concentration to a preset reference smoke concentration, and performing a weighted correction on the preliminary heat release amount based on the ratio to obtain a heat release amount; The target fire suppression energy is obtained by multiplying the heat release amount by the preset fire suppression duration. Based on the difference between the target fire suppression energy and the energy sprayed during the trial spraying phase, the proportional integral algorithm is used to calculate the unlimited main fire extinguishing dose. After applying upper and lower limit constraints to it, the main fire extinguishing dose is finally obtained. The fire extinguishing execution module is used to implement the main spraying fire extinguishing after receiving the main fire extinguishing dosage instruction, and calculate the residual fire index according to the comprehensive fire confidence value within the set monitoring period. When the residual fire index reaches the preset supplementary spraying threshold, it triggers the supplementary spraying.

2. The zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment according to claim 1 is characterized in that: The pre-trained model is used to output the image fire confidence based on the image data, including: S201, normalizing and resizing the current original frame image in the image data to output a standard size image; S202, inputting the standard size image into the pre-trained convolutional neural network model to extract the image feature vector; S203, inputting the image feature vector into the fully connected layer to output the single-frame fire confidence; S204: Calculate a gradient-weighted class activation map heat map of the standard-sized image based on the single-frame fire confidence score, and calculate the ratio of the number of high-heat pixels in the heat map to the total number of pixels in the image to obtain a high-heat area ratio. S205 , weighting the fire confidence of a single frame according to the area ratio of the high-heat region to obtain the fire confidence of the image.

3. The zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment according to claim 1 is characterized in that: Calculate the single fire reliability of each monitoring parameter in the environmental monitoring data, including: S301, normalizing each monitoring parameter according to its own preset physical range to obtain the corresponding normalized monitoring parameter; S302, calculating the difference between each normalized monitoring parameter and the corresponding preset alarm threshold; S303 , performing Sigmoid mapping on each difference and performing first-order exponential smoothing to obtain each single fire confidence level.

4. The zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment according to claim 2 is characterized in that: Based on the DS evidence theory, the single fire confidence and the image fire confidence are fused to obtain the comprehensive fire confidence, which includes: S401, mapping each single fire confidence level and image fire confidence level to a corresponding basic probability distribution based on a preset domain, wherein the preset domain includes two propositions: the existence of a fire and the non-existence of a fire; S402: Calculate a conflict coefficient for any two basic probability distributions based on the DS evidence theory. The conflict coefficient is the sum of the products of the probabilities supporting the proposition that a fire exists and the proposition that a fire does not exist, respectively, in the two basic probability distributions. The difference between one and the conflict coefficient is used as a normalization factor. The two basic probability distributions are fused according to the fusion rules of the DS evidence theory to obtain an updated fused distribution. In step S403, the updated fused distribution is sequentially fused with the remaining basic probability distributions according to step S402 until all fusions are completed, and finally the probability of supporting the proposition that there is a fire in the fused distribution is output as the comprehensive fire confidence.

5. The zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment according to claim 1 is characterized in that: The small-dose test spray includes: spraying a fire extinguishing agent at a preset ratio of the maximum design dose at the current moment, and the spraying duration is the first duration; The sensor prediction model is constructed by adopting a first-order autoregressive structure in combination with a Kalman filter algorithm.

6. The zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment according to claim 1 is characterized in that: Calculating the difference between the predicted environmental monitoring data and the environmental monitoring data for the second preset time period, and outputting a verification pass result when the difference meets a preset verification condition, including: S501, respectively calculating the difference between the predicted environmental monitoring data and the environmental monitoring data for a second preset time period to obtain residual data; S502, calculating the overall root mean square error based on the residual data as a quantitative indicator of the test spraying effect; S503 : When the overall root mean square error is less than the preset error threshold, a verification pass signal is output; otherwise, a verification fail signal is output.

7. The zero false alarm automatic fire alarm and fire extinguishing system based on multi-condition judgment according to claim 4 is characterized in that: The residual fire index is calculated based on the comprehensive fire confidence value during the set monitoring period. When the residual fire index reaches the preset supplementary spraying threshold, supplementary spraying is triggered, including: S701, reading the comprehensive fire confidence at the current moment and the residual fire index at the previous moment within the set monitoring period; S702: The residual fire index at the previous moment is weighted and superimposed with the comprehensive fire confidence at the current moment according to a preset smoothing coefficient to update and generate the residual fire index at the current moment; S703, comparing the residual fire index at the current moment with a preset supplementary injection threshold, outputting a supplementary injection trigger signal if the residual fire index reaches the preset supplementary injection threshold, otherwise keeping in standby mode.

Images