Integrated Utility Tunnel Cable Fire Intelligent Identification and Prevention and Control Linkage System

Abstract

This application relates to the technical field of utility tunnels, and provides an intelligent identification and prevention and control linkage system for cable fires in utility tunnels. In this system, a cable fire simulation device is used to simulate cable fires in the utility tunnel; an image acquisition unit is used to monitor the cable fire behavior in the tunnel in real time, and the collected simulated fire image information is transmitted to the deployed fire identification model in real time for fire identification, so as to judge the fire identification results of the fire identification model through the deployed fire warning model, and issue a fire extinguishing command or a warning command according to the judgment results; furthermore, after receiving the fire extinguishing command, the fire extinguishing linkage device responds, and the self-made fireproof rolling shutter with flexible steel brushes is first activated to block the fire point area; subsequently, the telescopic end of the electric push rod installed on the top of the tunnel quickly extends to break the heat-sensitive glass ball of the suspended fire extinguisher, realizing the rapid linkage prevention and control of cable fires in the tunnel.

Description

Technical Field

[0001] This application relates to the technical field of utility tunnels, and particularly to an intelligent identification and prevention and control linkage system for cable fires in utility tunnels. Background Art

[0002] With the acceleration of the urbanization process, the construction of urban underground utility tunnels is increasing day by day. As an underground space for accommodating various municipal pipelines, the power compartment of the utility tunnel is faced with a severe fire risk because a large number of cables are laid in it. Since the cables are usually laid continuously and the materials such as cable insulation layers are mostly flammable.

[0003] The high temperature and thick smoke generated by the fire will not only cause serious damage to the cables themselves, resulting in power outages and affecting the city's power supply, but may also endanger the structural safety of the entire utility tunnel and threaten the normal operation of other pipelines in the tunnel. At the same time, the toxic gases generated by combustion also pose a major threat to the lives of the personnel entering the tunnel for maintenance and other operations. Traditional fire detection methods have limitations in such a special environment as the utility tunnel. For example, smoke detectors may also be affected by the special smoke components and concentration change laws generated by cable fires in cable fires, resulting in false alarms or missed alarms, and at the same time unable to give early warnings for abnormal detections, which will pose a major threat to the lives of the personnel entering the tunnel for maintenance and other operations.

[0004] In terms of fire extinguishing, due to the long and narrow nature of the tunnel space, after the fire extinguishing agent is launched, it will diffuse along both ends of the tunnel under the initial kinetic energy of the fire extinguisher, weakening the fire extinguishing effect on the fire source. In addition, although the traditional suspended dry powder fire extinguishing system has good fire extinguishing efficiency, since the glass ball in the fire extinguisher can only act and respond after reaching 68 o °C, it is easy to cause a delay in fire extinguishing, and thus may miss the best time for fire extinguishing.

[0005] Therefore, there is an urgent need to provide a technical solution to address the above deficiencies in the prior art. Summary of the Invention

[0006] The purpose of this application is to provide an intelligent identification and prevention and control linkage system for cable fires in utility tunnels to solve or alleviate the problems existing in the above prior art.

[0007] To achieve the above purpose, this application provides the following technical solutions:

[0008] This application provides an intelligent identification and prevention and control linkage system for cable fires in utility tunnels, including: a utility tunnel cable fire simulation device, which is of a shell structure and is used for simulating fires of the utility tunnel cables;

[0009] The identification device monitors the fire simulation carried out in the cable tunnel fire simulation device in real time through the image acquisition unit, and transmits the acquired simulated fire image information to the deployed fire identification model for fire identification in real time, so as to judge the fire identification result of the fire identification model through the deployed fire warning model, and issue a fire extinguishing instruction or a warning instruction according to the judgment result;

[0010] The fire extinguishing linkage device is installed in the cable tunnel fire simulation device and is used to respond after receiving the fire extinguishing instruction issued by the identification device, start the fireproof rolling shutter with flexible steel brushes, block the fire point area in the cable tunnel fire simulation device, and then the telescopic end of the electric push rod installed on the top of the cable tunnel fire simulation device extends out to break the heat-sensitive glass ball of the hanging fire extinguisher installed on the top of the cable tunnel fire simulation device, so as to realize the fire extinguishing operation for the fire simulation in the cable tunnel fire simulation device.

[0011] Preferably, the cable tunnel fire simulation device includes: a tunnel platform, which is a rectangular shell structure with an opening on one side;

[0012] The size adjustment unit is used to block the opening of the tunnel platform to form a tunnel cavity with the tunnel platform; and it can move along the opening direction of the tunnel platform to adjust the spatial size of the tunnel cavity.

[0013] Preferably, fixed sliding rails extending in the opening direction are arranged on the inner side walls at both ends of the tunnel platform; correspondingly,

[0014] Both ends of the size adjustment unit are provided with fixed sliders slidably installed on the fixed sliding rails, so that the size adjustment unit can move along the opening direction of the tunnel platform.

[0015] Preferably, the cable tunnel fire simulation device further includes: an adjustable experimental oil pan, which is slidably installed on the inner side wall of the tunnel platform opposite to the opening and can be adjusted in position along the height direction and the horizontal direction of the tunnel platform.

[0016] Preferably, the shell structure of the cable tunnel fire simulation device is horizontally divided into multiple simulation areas; each of the simulation areas is respectively installed with the image acquisition unit and the hanging fire extinguisher.

[0017] Preferably, there are multiple fireproof rolling shutters, and the multiple fireproof rolling shutters are respectively arranged between adjacent two of the simulation areas; and after receiving the fire extinguishing instruction issued by the identification device, the two fireproof rolling shutters corresponding to the simulation area where the fire simulation is carried out move to block the simulation area where the fire simulation is carried out.

[0018] Preferably, one end of the fireproof rolling shutter is installed on the inner side wall of the pipe gallery cable fire simulation device, and can move in the front and back directions driven by the closing motor to block the simulation area where the fire simulation is carried out.

[0019] Preferably, a plurality of through grooves are arranged side by side in the height direction on the fireproof rolling shutter. Each through groove corresponds to the cables laid horizontally in the pipe gallery cable fire simulation device, and a flexible steel brush is arranged in each through groove; wherein, the through groove extends along the width direction of the pipe gallery platform.

[0020] Preferably, the fire recognition model deployed in the recognition device is trained based on the pre-acquired fire sample images, and the convolutional layer and the fully connected layer of the fire recognition model are updated based on the Adam method until the fire recognition error of the fire recognition model is less than or equal to a preset error threshold.

[0021] Preferably, the fire recognition model deployed in the recognition device extracts the fire features of the fire sample images through the convolutional layer according to the formula:

[0022]

[0023] Extract the fire features of the fire sample images; in the formula, is the th fire feature extracted by the th convolutional layer, is the activation function of the convolutional layer of the fire recognition model, is the th fire sample image; is the th convolutional layer of the fire recognition model, and is the weight between the th fire feature and the th fire feature, is the offset of the th fire feature in the th convolutional layer of the fire recognition model;

[0024] All are positive integers;

[0025] The fire recognition model reduces the dimension of the fire features of the extracted fire sample images through the pooling layer;

[0025] The fire recognition model classifies the fire features after dimensionality reduction processing through the fully connected layer according to the formula:

[0026]

[0027] In the formula, is the node of the th fully connected layer in the fire recognition model; is the node of the th fully connected layer in the fire recognition model; is the weight from the th fully connected layer to the th fully connected layer in the fire recognition model; is the bias term of the th fully connected layer; is the activation function of the fully connected layer of the fire recognition model;

[0028] The output layer of the fire recognition model calculates the fire occurrence probability according to the formula:

[0029]

[0030] ; where, is the th fire feature classification label output by the fully connected layer of the fire recognition model; among them, when, , it means a fire has occurred, when, , it means no fire has occurred.

[0031] Preferably, according to the formula:

[0032]

[0033] update the convolutional layer and the fully connected layer of the fire recognition model;

[0034] where, is the updated weight between the th fire feature and the th fire feature in the th convolutional layer of the fire recognition model; is the current weight between the th fire feature and the th fire feature in the th convolutional layer of the fire recognition model; is the self-learning rate of the weight corresponding to the convolutional layer of the fire recognition model;

[0035] is the updated weight from the th fully connected layer to the th fully connected layer of the fire recognition model; is the current weight from the th fully connected layer to the th fully connected layer in the fire recognition model; is the self - learning rate of the weights corresponding to the fully - connected layer of the fire identification model;

[0036] are respectively vectors in which the first - moment estimate and the second - moment estimate of the gradient are initialized to 0;

[0037] is the smoothing term.

[0038] Preferably, according to the formula:

[0039]

[0040] determine the th fire feature extracted from the fire sample image by the fire identification model The normalized predicted fire probability obtained by inputting into the fire warning model and the fire occurrence probability output by the fire identification model error , to judge the fire identification result of the fire identification model;

[0041] In the formula, is the activation function of the output layer of the fire identification model, is the weight matrix corresponding to the output layer of the fire warning model, is the hidden - layer value obtained by inputting the th fire feature extracted from the fire sample image by the fire identification model into the hidden layer of the fire warning model; is the bias term of the hidden layer; is the bias term of the output layer;

[0042] is the activation function of the hidden layer of the fire identification model, is the weight matrix of the th fire feature extracted from the fire sample image by the fire identification model ; is the hidden - layer value obtained by inputting the th fire feature extracted from the fire sample image by the fire identification model into the hidden layer of the fire warning model; is the hidden - layer value corresponding weight matrix.

[0043] Preferably, in response to the error being greater than the preset monitoring threshold , the identification device issues the warning instruction; in response to the error being less than or equal to the preset monitoring threshold , the identification device issues the fire extinguishing instruction.

[0044] Beneficial effects:

[0045] In the intelligent identification and prevention and control linkage system for cable fires in an integrated utility tunnel provided by the embodiments of the present application, a fire simulation of the cables in the integrated utility tunnel is carried out by a fire simulation device for utility tunnel cables with a housing structure; in the identification device, the fire simulation carried out in the fire simulation device for utility tunnel cables is monitored in real time by an image acquisition unit, and the collected simulated fire image information is transmitted in real time to a deployed fire identification model for fire identification, so as to judge the fire identification result of the fire identification model through a deployed fire warning model, and issue a fire extinguishing instruction or a warning instruction according to the judgment result; after the fire extinguishing linkage device installed in the fire simulation device for utility tunnel cables receives the fire extinguishing instruction issued by the identification device, it responds, and the self-made fireproof rolling shutter with flexible steel brushes is started first to block the fire point area; subsequently, the telescopic end of the electric push rod installed on the top of the tunnel quickly extends to break the heat-sensitive glass ball of the suspended fire extinguisher, realizing the rapid linkage prevention and control of cable fires in the utility tunnel.

[0046] Thereby, the video (image acquisition unit) collects multiple features in the utility tunnel, and effectively comprehensively detects the occurrence of a fire through the established dual identification and warning model for cable fires in the utility tunnel (fire identification model, fire warning model), improving the accuracy of fire identification, reducing the false alarm rate of cable fires in the utility tunnel, and improving the sensitivity and accuracy of judgment; and after a fire is identified, the fire extinguishing linkage device located in the fire simulation device for utility tunnel cables can be opened in time to achieve local blocking of the fire point in the utility tunnel and fire extinguishing agent spraying, improving the timeliness and automation level of extinguishing cable fires in the utility tunnel. Description of the drawings

[0047] The specification drawings forming a part of the present application are used to provide a further understanding of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. Among them:

[0048] Figure 1 It is a schematic diagram of the principle of an intelligent identification and prevention and control linkage system for cable fires in an integrated utility tunnel provided by some embodiments of the present application;

[0049] Figure 2 It is a schematic structural diagram of an intelligent identification and prevention and control linkage system for cable fires in an integrated utility tunnel provided by some embodiments of the present application;

[0050] Figure 3 It is a schematic structural diagram of a fire simulation device for utility tunnel cables provided by some embodiments of the present application;

[0051] Figure 4Explosion schematic diagram of a fixed slide rail and a fixed slider provided according to some embodiments of the present application;

[0052] Figure 5 Installation schematic diagram of an adjustable experimental oil pan provided according to some embodiments of the present application;

[0053] Figure 6 For Figure 5 Installation explosion schematic diagram of the adjustable experimental oil pan shown;

[0054] Figure 7 Structure schematic diagram of a fireproof rolling curtain provided according to some embodiments of the present application;

[0055] Figure 8 Installation schematic diagram of a closing motor provided according to some embodiments of the present application;

[0056] Figure 9 Working principle schematic diagram of a flexible steel brush provided according to some embodiments of the present application;

[0057] Figure 10 Assembly schematic diagram of a suspended fire extinguisher provided according to some embodiments of the present application;

[0058] Figure 11 For Figure 10 Installation schematic diagram of the dry powder fire extinguisher in;

[0059] Figure 12 Principle schematic diagram of an identification device provided according to some embodiments of the present application;

[0060] Figure 13 Structure schematic diagram of a dual identification and warning model provided according to some embodiments of the present application.

[0061] Explanation of reference numerals:

[0062] 1, Pipe gallery cable fire simulation device;

[0063] 101, Pipe gallery platform; 102, Dimension adjustment unit; 103, Adjustable experimental oil pan;

[0064] 111, Fixed slide rail; 112, Fixed slider;

[0065] 301, Suspended fire extinguisher; 302, Fireproof rolling curtain; 312, Through groove; 322, Flexible steel brush. Detailed implementation manners

[0066] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments. Each example is provided by way of explanation of the present application rather than a limitation thereof. In fact, those skilled in the art will appreciate that modifications and variations can be made to the present application without departing from the scope or spirit thereof. For example, features shown or described as part of one embodiment can be used in another embodiment to yield yet another embodiment. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the embodiments of the present invention should fall within the scope of protection of the embodiments of the present invention.

[0067] In the existing identification technology for judging cable fires based on the characteristics of composite gases, although the disadvantages of single-gas fire detectors are avoided to a certain extent, there are still problems of low identification accuracy and weak fire extinguishing timeliness in the identification process; at the same time, this technology still detects the gases generated by fires and has low identification accuracy for other fire characteristics. In addition, when extinguishing cable fires in confined spaces at present, traditional fire extinguishing methods still have problems such as poor fire extinguishing effect (the fire extinguishing agent diffuses around the fire source) and long response time (the temperature-sensitive glass bulb of the fire extinguisher responds at 68 o °C).

[0068] Based on this, an intelligent identification and prevention and control linkage system for cable fires in an integrated pipe gallery is provided in an embodiment of the present application. As Figures 1 to 13 shown, in this system, a pipe gallery cable fire simulation device 1 with a housing structure is used to simulate cable fires in the integrated pipe gallery. Specifically, the pipe gallery cable fire simulation device 1 includes: a pipe gallery platform 101 and a size adjustment unit 102; wherein, the pipe gallery platform 101 is a rectangular housing structure with an opening on one side, and the size adjustment unit 102 is used to block the opening of the pipe gallery platform 101, so that the size adjustment unit 102 and the pipe gallery platform 101 form a pipe gallery cavity; and the size adjustment unit 102 can move along the opening direction of the pipe gallery platform 101 to adjust the spatial size of the pipe gallery cavity to simulate integrated pipe galleries of different sizes and widths.

[0069] In a specific example, the pipe gallery platform 101 mainly includes: a platform ceiling, platform side walls, a platform rear wall, a platform support structure, and a cable module. Among them, the platform ceiling, platform side walls, and platform rear wall are all made of steel structures, and the platform ceiling, platform side walls, and platform rear wall are welded to each other and together with the platform floor form a rectangular shell structure with a lateral opening; the platform support structure is located at the bottom of the rectangular shell structure and is used to support the rectangular shell structure; inside the platform rear wall (inside the rectangular shell structure), a cable module is arranged along the length direction of the pipe gallery platform 101. Specifically, the cable support components in the cable module are fixedly connected to the platform rear wall along the length direction of the pipe gallery platform 101, and the cable wires in the cable module are arranged on the cable support device. Here, multiple cable modules can be arranged simultaneously, and the multiple cable modules are evenly distributed side by side along the height direction of the pipe gallery platform 101 on the platform rear wall to more realistically simulate the cable tunnel scenario.

[0070] On the side of the pipe gallery rear wall facing the opening side of the pipe gallery platform 101, an oil pan chute is also arranged along the length direction of the pipe gallery platform 101, that is, an oil pan chute is arranged horizontally on the inner side wall opposite to the lateral opening of the pipe gallery, and an oil pan support plate is installed in the oil pan chute. Among them, the oil pan support plate is provided with a dovetail-shaped slider that is slidably sleeved with the oil pan chute. That is to say, the oil pan chute is a chute that extends along the horizontal direction of the pipe gallery platform 101 and has a dovetail-shaped cross-section, and the cross-section of the slider on the oil pan support plate that is adapted to the oil pan chute is also dovetail-shaped, whereby the oil pan support plate can move along the horizontal direction (length direction) of the pipe gallery platform 101 to adjust the position of the adjustable experimental oil pan 103 placed on the oil pan support plate along the length direction of the pipe gallery platform 101.

[0071] At the same time, an oil pan platform is slidably installed on the oil pan support plate, and the oil pan platform can adjust its position along the height direction of the pipe gallery platform 101, and further, the height of the adjustable experimental oil pan 103 placed on the oil pan platform can be adjusted. Specifically, the oil pan support plate is provided with a vertical chute that extends along the height direction of the pipe gallery platform 101, and the oil pan platform is slidably sleeved in the vertical chute, whereby the oil pan platform can move along the height direction on the oil pan support plate. Here, it should be noted that the vertical chute has a T-shaped cross-section or a dovetail-shaped cross-section; a threaded hole perpendicular to the bottom surface of the vertical chute is also provided on the oil pan platform, and a locking stud or a locking screw is installed in the threaded hole. When the oil pan platform is adjusted to the set height, by tightening the locking stud or the locking screw, the end face of the locking stud or the locking screw abuts against the bottom surface of the vertical chute, and further, the oil pan platform is locked on the oil pan support plate.

[0072] A size adjustment unit 102 is slidably installed on the opening side of the pipe gallery platform 101. By moving the size adjustment unit 102 in the width direction of the pipe gallery platform 101, the spatial size of the pipe gallery cavity is adjusted. Specifically, on the inner side walls at both ends of the pipe gallery platform 101, there are fixed slide rails 111 extending towards the opening direction. At both ends of the size adjustment unit 102, there are fixed sliders 112 slidably installed with the fixed slide rails 111. Through the cooperation of the fixed sliders 112 and the fixed slide rails 111, the size adjustment unit 102 can move along the opening direction of the pipe gallery platform 101.

[0073] In a specific example, at both ends of the detachable fireproof glass of the size adjustment unit 102, there are steel card slots respectively. On the steel card slots, there are fixed sliders 112 adapted to the fixed slide rails 111 respectively. By slidably inserting the fixed sliders 112 into the fixed sliders 112, the detachable fireproof glass can move along the width direction of the pipe gallery platform 101. Here, it should be noted that the detachable fireproof glass is adapted to the lateral opening of the pipe gallery platform 101 to close the lateral opening of the rectangular shell structure of the pipe gallery platform 101, forming a pipe gallery cavity. Thus, by moving the detachable fireproof glass along the width direction of the pipe gallery platform 101, the size of the pipe gallery cavity is adjusted to simulate cable pipe galleries of different sizes.

[0074] In the shell structure of the pipe gallery cable fire simulation device 1, there is also a ventilation adjustment module arranged. Through the ventilation adjustment module, the air volume and air speed in the pipe gallery cavity are simulated to further simulate different types of cable pipe gallery scenarios. Among them, the air inlet fan and the air outlet fan of the ventilation adjustment module are respectively installed at positions close to the platform side wall on the platform ceiling.

[0075] In this application, multiple simulation areas are divided horizontally in the shell structure of the pipe gallery cable fire simulation device 1. Each simulation area is respectively provided with an adjustable experimental oil pan 103, an image acquisition unit, and a suspended fire extinguisher 301. Between two adjacent simulation areas, they are isolated or connected through a fireproof rolling curtain 302 arranged. Here, it should be noted that the oil pan chute for installing the adjustable experimental oil pan 103 is discontinuous at the fireproof rolling curtain 302 installed between two adjacent simulation areas. Furthermore, the independent adjustment of the adjustable experimental oil pan 103 in each simulation area is realized. By simulating the position change of the adjustable experimental oil pan 103 in the simulation area, different ignition positions of the cable are simulated, and fire simulations at different positions are independently carried out in each simulation area to effectively simulate the uncertain ignition positions of the cables in the pipe gallery cable.

[0076] When a fire simulation is carried out in the simulation area of the cable tunnel fire simulation device 1, the real-time state of the fire simulation is monitored in real time by the image acquisition unit installed on the platform ceiling, and the collected simulated fire image information is transmitted to the deployed fire identification model in real time for fire identification; the fire warning model is used to judge the fire identification result of the fire identification model, and a fire extinguishing instruction or a warning instruction is issued according to the judgment result. Thus, the fire identification model identifies whether a fire occurs through the real-time simulated fire image information collected by the image acquisition unit, and the fire warning model judges the identification result; if the error between the identification result of the fire identification model and the judgment result of the fire warning model is greater than the preset monitoring threshold, the identification device issues a warning instruction to notify the staff to conduct on-site inspection and processing; if the error between the identification result of the fire identification model and the judgment result of the fire warning model is less than or equal to the preset monitoring threshold, the identification device issues a fire extinguishing instruction to turn on the fire extinguishing linkage device for fire extinguishing. Furthermore, through the double identification and warning of the fire, the abnormal situation of the single model prediction error is effectively avoided, the untimely fire extinguishing is reduced, and the loss is reduced. Through the double identification and warning, the fire situation in the cable tunnel can be accurately identified and monitored in a timely manner.

[0077] In a specific example, by establishing a fire identification model for the cable tunnel, the deep features of the real-time monitored images are automatically extracted, including but not limited to extracting flame shape features, flame color features, smoke color features, flame motion features, smoke motion features, etc. Among them, the fire identification model deployed in the identification device is trained based on the pre-obtained fire sample images, and the convolutional layer and the fully connected layer of the fire identification model are updated based on the Adam method until the fire identification error of the fire identification model is less than or equal to the preset error threshold.

[0078] In this embodiment, after collecting the fire sample images and manually annotating the flames and smoke in them, they are input into the fire identification model to train the fire identification model. The hidden layer of the fire identification model contains several convolutional layers, pooling layers and fully connected layers. Among them, the convolutional layer of the fire identification model follows the formula:

[0079]

[0080] Extract the fire features of the fire sample images; in the formula, is the th fire feature extracted by the th convolutional layer, is the activation function of the convolutional layer of the fire identification model, is the th fire sample image; is the th convolutional layer in the fire identification model and the the weight between the first fire feature and the nth fire feature, is the offset of the nth fire feature in the nth convolutional layer of the fire recognition model; Both are positive integers. Both are positive integers.

[0081] The fire feature image obtained through the convolutional layer of the fire recognition model is too long. The fire recognition model uses the pooling layer to reduce the dimension of the fire features of the extracted fire sample images. The operation speed is effectively improved through the image features after dimension reduction, and important features such as the shape, color, and movement of the flame and smoke in the fire are effectively extracted. Specifically, when the convolutional layer of the fire recognition model extracts the feature matrix of the fire features, the fire uses the max pooling method to reduce the dimension of the fire features and extract the maximum data features in each sub-matrix. Among them, the flame shape features mainly include the edge features of the flame shape and the flame height; the flame color features mainly include the pixel distributions of red, white, and yellow in the flame; the smoke color features include the pixel distributions of black and gray in the smoke; the flame movement features include the speed, range, and direction of flame spread; the smoke movement features include the speed, range, and direction of smoke spread.

[0082] Then, the feature image obtained by the pooling layer is converted into a one-dimensional vector through the fully connected layer of the fire recognition model, and it acts as a classifier to classify the fire features. For example, continuous flames, discontinuous flames, etc. in the edge features of the flame shape are classified. Specifically, the fully connected layer of the fire recognition model follows the formula:

[0083]

[0084] classifies the fire features after dimension reduction processing; in the formula, is the node of the nth fully connected layer in the fire recognition model; is the node of the mth fully connected layer in the fire recognition model, is the weight from the mth fully connected layer to the nth fully connected layer in the fire recognition model, is the bias term of the mth fully connected layer, is the activation function of the fully connected layer of the fire recognition model. is the weight from the mth fully connected layer to the nth fully connected layer in the fire recognition model, is the bias term of the mth fully connected layer, is the bias term of the nth fully connected layer, is the activation function of the fully connected layer of the fire recognition model.

[0085] The fully connected layer of the fire recognition model classifies fire features into two categories: with fire and without fire. Then, through the output layer of the fire recognition model, the image classification labels of the fully connected layer are output. Specifically, the output layer of the fire recognition model uses the Softmax function according to the formula:

[0086]

[0087] to calculate the probability of fire occurrence ; where is the th fire feature classification label output by the fully connected layer of the fire recognition model; among them, when , it means a fire has occurred, when , it means no fire has occurred.

[0088] During the training process of the fire recognition model, by comparing the output value of the fire recognition model with the true value, the loss function value reflects the error between the output value of the fire recognition model and the true value (the fire recognition error of the fire recognition model). Among them, according to the formula:

[0089]

[0090] where is the true value corresponding to the rd fire sample image, that is, whether a fire has occurred, or ; is the output value of the fire recognition model corresponding to the th fire sample image, that is, the probability of fire occurrence , and 1;

[0091]

[0092] is the flame sample image input to the fire recognition model, is the current weight between the th and th fire features in the th convolutional layer of the fire recognition model, is the current weight from the th fully connected layer to the th fully connected layer in the fire recognition model. In this application, the weights between the fire features of the convolutional layer and the fully connected layer of the fire recognition model are updated by the Adam method to make the prediction error of the fire recognition model smaller. Specifically, according to the formula:

[0093]

[0094] Update the convolutional layer and the fully connected layer of the fire recognition model. Wherein, is the updated weight value between the th fire feature and the th fire feature in the th convolutional layer of the fire recognition model, is the current weight value between the th fire feature and the th fire feature in the th convolutional layer of the fire recognition model, is the self-learning rate of the weight value corresponding to the convolutional layer of the fire recognition model; is the updated weight value from the th fully connected layer to the th fully connected layer of the fire recognition model, is the current weight value from the th fully connected layer to the th fully connected layer in the fire recognition model, is the self-learning rate of the weight value corresponding to the fully connected layer of the fire recognition model; are vectors in which the first-order moment estimate and the second-order moment estimate of the gradient are respectively initialized to 0; is the smoothing term.

[0095] In this application, the image acquisition unit acquires the images in the cable fire simulation device 1 of the pipe gallery through the implementation, and analyzes the probability of fire occurrence through the fire recognition model. Judging whether a fire occurs solely through the fire recognition model may still lead to false alarms or missed alarms of fires. For this reason, in this application, the recognition result of the fire recognition model is judged by deploying a fire warning model at the same time to further improve the prediction accuracy of the cable fire in the pipe gallery.

[0096] Specifically, the fire warning model is calculated according to the formula:

[0097]

[0098] Determine the th fire feature extracted from the fire sample image by the fire recognition model The normalized predicted fire probability obtained by inputting into the fire warning model and the fire occurrence probability output by the fire recognition model of the error

[0099] Wherein, is the activation function of the output layer of the fire identification model, is the weight matrix corresponding to the output layer of the fire warning model, is the hidden layer value obtained by inputting the th fire feature extracted from the fire sample image by the fire identification model into the hidden layer of the fire warning model; is the bias term of the hidden layer; is the bias term of the output layer; is the activation function of the hidden layer of the fire identification model, is the th fire feature extracted from the fire sample image by the fire identification model weight matrix; is the hidden layer value obtained by inputting the th fire feature extracted from the fire sample image by the fire identification model into the hidden layer of the fire warning model; is the hidden layer value corresponding weight matrix.

[0100] In the fire warning model, when the th fire feature extracted from the fire sample image by the fire identification model is input into the fire warning model, and the normalized predicted fire probability and the fire occurrence probability output by the fire identification model error is greater than the preset monitoring threshold th fire feature extracted from the fire sample image by the fire identification model is input into the fire warning model, and the normalized predicted fire probability and the fire occurrence probability output by the fire identification model error is less than or equal to the preset monitoring threshold

[0101] When this occurs, the identification device issues a fire extinguishing instruction, and the fire extinguishing linkage device installed in the cable fire simulation device 1 in the pipe gallery performs a fire extinguishing operation on the fire simulation carried out in the cable fire simulation device 1 in the pipe gallery. Specifically, after receiving the fire extinguishing instruction issued by the identification device, the fire extinguishing linkage device is activated, and a suspended fire extinguisher 301 installed in the cable fire simulation device 1 in the pipe gallery performs a fire extinguishing operation on the fire simulation carried out in the cable fire simulation device 1 in the pipe gallery.Between two adjacent simulation areas in the pipe gallery cavity, they are isolated or connected by the fire shutter 302. On the one hand, it can realize the simulation of pipe galleries of different sizes. On the other hand, it can realize the fire isolation of different areas and reduce losses. Among them, after receiving the fire extinguishing instruction issued by the identification device, the two fire shutters 302 corresponding to the simulation area where the fire simulation is carried out move to block the simulation area where the fire simulation is carried out, so as to realize the closed fire extinguishing operation of the simulation area where the fire simulation is carried out.

[0102] In a specific application scenario, one end of the fire shutter 302 is installed on the inner side wall of the pipe gallery cable fire simulation device 1, and the closing motor is installed on the platform ceiling and drives the movement along the width direction (i.e., the front-back direction) of the pipe gallery platform 101 through gear transmission to block the simulation area where the fire simulation is carried out. A plurality of through grooves 312 are arranged side by side along the height direction on the fire shutter 302. Each through groove 312 extends along the width direction of the pipe gallery platform 101 and corresponds to the cables arranged horizontally in the pipe gallery cable fire simulation device 1, and each through groove 312 is provided with a flexible steel brush 322.

[0103] Among them, the fire shutter 302 is of steel structure, and the flexible steel brush 322 is installed at the position of the through groove 312, so that when the fire shutter 302 moves driven by the closing motor, the cable is located in the through groove 312 and is not affected by the movement of the fire shutter 302. Here, the flexible steel brush 322 is connected (such as welded) to the fire shutter 302 through a steel brush connecting plate. During the movement of the fire shutter 302, when the flexible steel brush 322 touches the cable, due to the resistance of the cable, the flexible steel brush 322 automatically lifts. Thus, not only can the fire shutter 302 isolate and prevent the spread of fire, but also the integrity of the cable is not damaged.

[0104] The closing motor drives the fire shutter 302 to move along the width direction of the pipe gallery platform 101 through gear transmission. When a fire is detected, the closing motor starts, drives the fire shutter 302 to move along the rolling door slide rails arranged on the platform ceiling and the platform floor and automatically closes, isolating the fire occurrence area separately, so as to isolate and prevent the spread of fire in time.

[0105] In a specific application scenario, after a fire is detected, an electric push rod installed on the platform ceiling of the pipe gallery platform 101 breaks the heat-sensitive glass bulb of the dry powder fire extinguisher, so that the dry powder extinguishing agent in the dry powder fire extinguisher extinguishes the fire (simulated) in the corresponding simulated area of the pipe gallery cable fire simulation device 1. Among them, the dry powder fire extinguisher and the electric push rod are both installed on the platform ceiling, and the telescopic end of the electric push rod is directly opposite to the heat-sensitive glass bulb of the dry powder fire extinguisher. Therefore, after a fire is detected, the telescopic end of the electric push rod quickly extends out, hits the heat-sensitive glass bulb, breaks the heat-sensitive glass bulb, and the dry powder fire extinguisher sprays out to extinguish the fire (simulated) in the corresponding simulated area.

[0106] Therefore, multiple fire characteristics in the pipe gallery are collected by the video (image acquisition unit), and the established cable fire identification and early warning model (fire identification model, fire early warning model) effectively comprehensively detects the occurrence of the fire, improves the accuracy of fire identification, reduces the false alarm rate of the pipe gallery cable fire, and improves the sensitivity and accuracy of judgment; and after the fire is detected, the fire extinguishing linkage device in the pipe gallery cable fire simulation device 1 can be timely opened to realize the rapid start of the comprehensive pipe gallery fire extinguisher, timely and effectively handle the fire situation in the fire occurrence area, improve the timeliness and automation of fire extinguishing, and reduce the passivity of manual fire extinguishing.

[0107] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0108] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0109] In the present invention, unless otherwise clearly defined or limited, terms such as "installed", "connected", "joined", "fixed", etc. shall be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or integrated; it may be a mechanical connection, an electrical connection, or capable of communicating with each other; it may be directly connected, or indirectly connected through an intermediate medium, and it may be the internal communication of two components or the interaction relationship between two components, unless otherwise clearly defined. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0110] In the present invention, unless otherwise clearly defined or limited, the first feature being "on" or "under" the second feature may be that the first and second features are in direct contact, or the first and second features are indirectly in contact through an intermediate medium. Moreover, the first feature being "above", "over" and "on top of" the second feature may be that the first feature is directly above or obliquely above the second feature, or merely indicates that the first feature has a higher horizontal height than the second feature. The first feature being "under", "below" and "beneath" the second feature may be that the first feature is directly below or obliquely below the second feature, or merely indicates that the first feature has a lower horizontal height than the second feature.

[0111] In the present invention, terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0112] The above are only the preferred embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. An intelligent identification and prevention and control linkage system for cable fire in an integrated pipe gallery, characterized in that: include: The pipe gallery cable fire simulation device is a shell structure, used for simulating the fire of the integrated pipe gallery cable; The identification device monitors the fire simulation in the cable fire simulation device in real time through an image acquisition unit, and transmits the collected simulated fire image information to the deployed fire identification model in real time, and the fire identification model identifies whether a fire occurs through the simulated fire image, and judges the fire identification result of the fire identification model through the deployed fire warning model. If the error between the fire identification result of the fire identification model and the judgment result of the fire warning model is greater than a preset monitoring threshold, a warning instruction is issued; If the error between the fire identification result of the fire identification model and the judgment result of the fire warning model is less than or equal to the preset monitoring threshold, a fire extinguishing instruction is issued; A fire extinguishing linkage device is installed in the pipe gallery cable fire simulation device, and is used to respond after receiving a fire extinguishing command from the identification device, so that the fireproof roller shutter with a flexible steel brush is activated, and after the fire point area in the pipe gallery cable fire simulation device is blocked, the telescopic end of the electric push rod installed on the top of the pipe gallery cable fire simulation device is extended to smash the temperature-sensitive glass ball of the suspended fire extinguisher installed on the top of the pipe gallery cable fire simulation device, thereby realizing the fire extinguishing operation of the fire simulation carried out in the pipe gallery cable fire simulation device.

2. The integrated pipe gallery cable fire intelligent identification and prevention and control linkage system according to claim 1 is characterized in that: The pipe gallery cable fire simulation device comprises: The pipe gallery platform is a rectangular shell structure with an opening on one side, and fixed slide rails extending toward the opening are arranged on the inner side walls at both ends; A size adjustment unit is used to block the opening of the pipe gallery platform to form a pipe gallery cavity with the pipe gallery platform; and fixed sliders slidably mounted with the fixed slide rails are provided at both ends, so that the size adjustment unit can move along the opening direction of the pipe gallery platform to adjust the spatial size of the pipe gallery cavity; The adjustable experimental oil pan is slidably mounted on the inner wall of the pipe gallery platform opposite to the opening, and can be adjusted in position along the height direction and the horizontal direction of the pipe gallery platform.

3. The integrated pipe gallery cable fire intelligent identification and prevention and control linkage system according to claim 2 is characterized in that: The shell structure of the pipe gallery cable fire simulation device is divided into a plurality of simulation areas in the horizontal direction; The image acquisition unit and the suspended fire extinguisher are installed in each simulation area respectively.

4. The integrated pipe gallery cable fire intelligent identification and prevention and control linkage system according to claim 3 is characterized in that: There are multiple fireproof roller shutters, which are respectively arranged between two adjacent simulation areas; and after receiving the fire extinguishing command issued by the identification device, the two fireproof roller shutters corresponding to the simulation area for fire simulation move to block the simulation area for fire simulation.

5. The integrated pipe gallery cable fire intelligent identification and prevention and control linkage system according to claim 4 is characterized in that: A plurality of through grooves are arranged in parallel along the height direction on the fireproof roller shutter, each of the through grooves corresponds to the cables arranged in the horizontal direction in the pipe gallery cable fire simulation device, and each of the through grooves is provided with the flexible steel brush; wherein the through grooves extend along the width direction of the pipe gallery platform.

6. The integrated pipe gallery cable fire intelligent identification and prevention and control linkage system according to claim 1 is characterized in that: The fire identification model deployed in the identification device is trained based on the pre-acquired fire sample image, and the convolution layer and the fully connected layer of the fire identification model are updated based on the Adam method until the fire identification error of the fire identification model is less than or equal to a preset error threshold.

7. The integrated pipe gallery cable fire intelligent identification and prevention and control linkage system according to claim 6 is characterized in that: The fire identification model deployed in the identification device is constructed through a convolutional layer according to the formula: ； Extract the fire features of the fire sample image; where, For the The convolutional layer extracts Fire characteristics, is the activation function of the convolutional layer of the fire recognition model, For the a fire sample image; The fire identification model In the convolutional layer Fire characteristics and The weights between the fire characteristics, The fire identification model In the convolutional layer The offset of each fire feature; All are positive integers; The fire identification model performs dimensionality reduction processing on the fire features of the extracted fire sample images through a pooling layer; The fire identification model is implemented through a fully connected layer according to the formula: ； Classify the fire characteristics after dimensionality reduction; where, The fire identification model nodes of the fully connected layer; The fire identification model nodes in the fully connected layer, The fire identification model The fully connected layers to The weights of the fully connected layers, For the The bias term of the fully connected layer, is the activation function of the fully connected layer of the fire identification model; The output layer of the fire identification model is processed by the Softmax function according to the formula: ； Calculate the probability of fire ; In the formula, is the output of the fully connected layer of the fire identification model fire feature classification labels; among them, hour, , indicating a fire has occurred, hour, , indicating that no fire occurred.

8. The integrated pipe gallery cable fire intelligent identification and prevention and control linkage system according to claim 7 is characterized in that: According to the formula: ； Updating the convolutional layer and the fully connected layer of the fire identification model; In the formula, The fire identification model In the convolutional layer Fire characteristics and The updated weights between the fire features are The fire identification model In the convolutional layer Fire characteristics and The current weights between the fire characteristics, is the self-learning rate of the weights corresponding to the convolutional layer of the fire identification model; is the fire identification model The fully connected layers to The updated weights of the fully connected layers are The fire identification model The fully connected layers to The current weight of the fully connected layer, is the self-learning rate of the weights corresponding to the fully connected layer of the fire identification model; are vectors of the first-order moment estimate and the second-order moment estimate of the gradient initialized to 0; is the smoothing term.

9. The integrated pipe gallery cable fire intelligent identification and prevention and control linkage system according to claim 6 is characterized in that: According to the formula: ； Determine the first Fire characteristics Input the normalized predicted fire probability obtained in the fire warning model The fire occurrence probability output by the fire identification model Error , to judge the fire identification result of the fire identification model; In the formula, is the activation function of the output layer of the fire identification model, is the weight matrix corresponding to the output layer of the fire warning model, The fire identification model extracts the first A hidden layer value obtained by inputting a fire feature into the hidden layer of the fire warning model; is the bias term of the hidden layer; is the bias term of the output layer; is the activation function of the hidden layer of the fire recognition model, The fire identification model extracts the first Fire characteristics The weight matrix of The fire identification model extracts the first A hidden layer value obtained by inputting a fire feature into the hidden layer of the fire warning model; is the hidden layer value The corresponding weight matrix.

10. The integrated pipe gallery cable fire intelligent identification and prevention linkage system according to claim 9 is characterized in that: Responding to Error Greater than the preset monitoring threshold , the identification device issues the warning instruction; Responding to Error Less than or equal to the preset monitoring threshold , the identification device issues the fire extinguishing instruction.

Images