A fire pre-blocking method for underground pipe gallery

Abstract

This application discloses a method for proactive fire prevention in underground utility tunnels, relating to the field of cable fire monitoring technology. The method includes: acquiring composite physical field characteristic parameters of the target area, including temperature change rate, acoustic vibration signal characteristic values, and gas concentration values; generating a trigger signal when the temperature change rate, acoustic vibration signal characteristic values, or gas concentration values ​​reach preset critical values; responding to the trigger signal with an intervention action; acquiring feedback parameters after the intervention action, and determining whether to repeat the intervention action based on the feedback parameters. This application achieves accurate identification of the imminent thermal-electrical runaway critical state of cable joints through the above method, advancing the intervention time from post-disaster to pre-disaster, fundamentally blocking the path of fault evolution into fire.

Description

Technical Field

[0001] This application relates to the field of cable fire monitoring technology, and in particular to a method for proactive fire prevention for underground utility tunnels. Background Technology

[0002] Currently, fire protection for power cables in underground utility tunnels mainly relies on two types of technologies. One type is fixed fire extinguishing devices triggered by a single physical quantity threshold, such as thermal aerosol fire extinguishers or ultra-fine dry powder fire extinguishing devices. The other type is a manual inspection and post-disaster response system based on video surveillance. The former is only activated after an open flame is generated, and cannot intervene during the critical window period when faults such as cable joint overheating or partial discharge occur but before a fire has formed, thus belonging to passive fire extinguishing after a disaster. The latter relies on manual judgment, has a large response delay, and is difficult to deal with sudden thermoelectric coupling faults.

[0003] The limitations of existing technologies lie in their lack of ability to detect the early sub-healthy state of cable faults and their inability to identify the coupling evolution law between thermal runaway and arc discharge; their intervention methods are singular, targeting only the open flame itself, and failing to simultaneously break the fault chain from multiple physical dimensions such as thermal suppression, insulation enhancement, and chemical inertization; moreover, they are all one-time actions, and cannot be adjusted in a closed-loop feedback manner based on the intervention effect. Once the first intervention fails, it will directly lead to a fire.

[0004] Based on the above analysis, the problems and shortcomings of the existing technology are as follows: Existing underground utility tunnel cable protection technologies cannot provide multi-dimensional coordinated intervention in cases of thermoelectric runaway or fire critical states, and lack feedback and adjustment capabilities. Summary of the Invention

[0005] This application provides a method for proactive fire prevention in underground utility tunnels, which can solve the problems of existing underground utility tunnel cable protection technologies being unable to provide multi-dimensional coordinated intervention in the event of thermoelectric runaway or fire critical state, and lacking feedback adjustment capabilities.

[0006] In a first aspect, embodiments of this application provide a method for proactive fire prevention in underground utility tunnels. The method includes: acquiring composite physical field characteristic parameters of a target area, including temperature change rate, acoustic vibration signal characteristic value, and gas concentration value; generating a trigger signal when the temperature change rate, acoustic vibration signal characteristic value, or gas concentration value reaches a preset critical value; responding to an intervention action based on the trigger signal; and acquiring feedback parameters after the intervention action is performed, and determining whether to repeat the intervention action based on the feedback parameters.

[0007] In one implementation of this application, a trigger signal is generated when the temperature change rate, the acoustic vibration signal characteristic value, or the gas concentration value reaches a preset critical value. Specifically, the temperature change rate includes the time derivative of the temperature, the acoustic vibration signal characteristic value includes the acoustic vibration signal intensity of a preset frequency band, and the gas concentration value is the characteristic gas concentration generated by the pyrolysis of the cable insulation material. When the time derivative, the acoustic vibration signal intensity, and the characteristic gas concentration are all detected to exceed the critical value, it is determined that the fire critical state has been entered, and an emergency trigger signal is generated.

[0008] In one implementation of this application, the method further includes: when one or both of the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration exceed a preset threshold, inputting the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration into a pre-trained probability model; obtaining the probability value representing the occurrence of a fire output by the probability model; and generating a probability trigger signal when the probability value exceeds a preset action probability threshold.

[0009] In one implementation of this application, the intervention action based on the trigger signal specifically includes: sequentially activating a first-level intervention action, a second-level intervention action, and a third-level intervention action based on an emergency trigger signal. The intervention actions include spraying an ultra-high-speed gas-liquid two-phase medium, spraying a high-pressure inert gas, and releasing an expandable flame-retardant material; activating the first-level intervention action based on a probability trigger signal, and continuously monitoring the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration after the first-level intervention action; if, within a preset confirmation time window, at least one of the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration continuously rises or exceeds the corresponding auxiliary threshold, activating the second-level and third-level intervention actions; if, within the confirmation time window, the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration all decrease or return to the normal range, terminating the intervention action and recording it as an alarm event.

[0010] In one implementation of this application, the method further includes: dynamically adjusting the duration of the injected gas-liquid two-phase medium according to the amplitude of the time differential value; dynamically adjusting the injection pressure and duration of the high-pressure inert gas according to the amplitude of the acoustic vibration signal intensity; and projecting the encapsulated expandable flame-retardant material onto the surface of the target area, wherein the expandable flame-retardant material forms a fireproof isolation layer after contacting the high-temperature surface.

[0011] In one implementation of this application, the method further includes: simultaneously initiating the first-level intervention action, transmitting a detection beam to the target area via an ultrasonic sensor array and receiving the reflected echo; based on the reflected echo, reconstructing in real time a three-dimensional point cloud model of the target area and the surrounding preset distance space, the three-dimensional point cloud model including the surface morphology of the monitored object, the spatial distribution of the sprayed medium, and the coverage boundary of the expanded flame-retardant material; comparing the three-dimensional point cloud model with a pre-stored normal state three-dimensional reference model to calculate the actual coverage and effectiveness of the current intervention action; if the actual coverage and effectiveness do not reach the preset fire intervention effect target value, adjusting the spray direction, spray angle, or spray pressure of the second-level or third-level intervention action in real time according to the deviation, until the actual coverage and effectiveness reach the fire intervention effect target value.

[0012] In one implementation of this application, the encapsulated expandable flame-retardant material is projected onto the surface of a target area. Specifically, this includes: uniformly doping the expandable flame-retardant material with magnetic particles; while projecting the expandable flame-retardant material, adjusting the field strength and direction of a magnetic field generator deployed around the nozzle based on the target area location and obstacle distribution identified by the three-dimensional point cloud model; and the expandable flame-retardant material moving along a preset obstacle avoidance path under the action of the magnetic field force to adhere to the surface of the target area.

[0013] In one implementation of this application, the target area includes multiple adjacent monitoring points, each monitoring point corresponds to an independent blocking device, and the blocking devices are connected through a wireless network to form a multi-agent system; the method further includes: broadcasting the monitored time derivative value, acoustic vibration signal intensity, characteristic gas concentration and remaining medium storage to adjacent devices through the blocking device via a wireless network; based on the received neighbor information, distributed reinforcement learning is used to make collaborative decisions to determine the start-up sequence, injection direction and injection intensity of the device.

[0014] In one implementation of this application, after an intervention action is performed, feedback parameters are obtained, and it is determined whether to repeat the intervention action based on the feedback parameters. Specifically, this includes: monitoring the local temperature, gas concentration, and insulation status after the intervention action using a temperature sensor, a gas sensor, and an ultrasonic sensor array; if the feedback parameters determine that the fault chain has not been completely blocked, triggering the repeated execution of at least one intervention action; and triggering the repeated execution of at least one fire-fighting intervention action based on the deviation between the feedback parameters and the preset target state.

[0015] In one implementation of this application, before generating a trigger signal when the temperature change rate, acoustic vibration signal characteristic value, or gas concentration value reaches a preset critical value, the method further includes: acquiring a historical monitoring data sequence of the target area; inputting the historical monitoring data sequence into a time-series prediction model, and outputting a predicted trajectory for the temperature change rate, acoustic vibration signal characteristic value, and gas concentration value within a future preset time window; performing similarity matching between the predicted trajectory for the temperature change rate, acoustic vibration signal characteristic value, and gas concentration value and multiple fire paths in a corresponding preset fire evolution path library; selecting the fire path with the highest similarity, identifying the current fire evolution mode type, and predicting the remaining available fire intervention time window; and selecting a corresponding fire intervention strategy from multiple preset fire intervention strategies based on the evolution mode type and the remaining available fire intervention time window, wherein the fire intervention strategy includes the activation sequence relationship, intensity, and duration of the intervention action.

[0016] This application provides a method for proactive fire prevention in underground utility tunnels. Through the fusion monitoring of composite physical field characteristic parameters, it achieves accurate identification of the imminent thermal-electric runaway critical state of cable joints, advancing intervention from post-disaster to pre-disaster, fundamentally blocking the path from fault to fire evolution. A three-tiered progressive intervention system—thermal shock suppression, insulation enhancement, and chemical inerting—is constructed to simultaneously break the fault chain from three dimensions: temperature, electric field, and oxygen, significantly improving the success rate of prevention. A two-tiered early warning mechanism—emergency triggering and probabilistic triggering—is established, employing differentiated intervention strategies for different risk levels: full intervention in emergency situations and trial-confirmation-escalation in probabilistic situations, ensuring safety while avoiding resource waste and equipment damage caused by excessive intervention. Ultrasonic real-time imaging and 3D point cloud reconstruction technology are introduced to quantitatively evaluate the actual effect of intervention actions at the millisecond level, and subsequent intervention parameters are dynamically adjusted based on feedback deviations, significantly improving adaptability and reliability under complex operating conditions. Based on historical data sequence temporal prediction and fire path matching, fault evolution patterns can be predicted in advance, reserving sufficient intervention time windows, providing a scientific basis for the formulation of refined and customized intervention strategies. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a fire prevention method for underground utility tunnels, provided as an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] This application provides a method for proactive fire prevention in underground utility tunnels, which solves the problems of existing underground utility tunnel cable protection technologies being unable to provide multi-dimensional collaborative intervention in the event of thermoelectric runaway or fire critical state, and lacking feedback adjustment capabilities.

[0020] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0021] Figure 1 This is a flowchart illustrating a fire prevention method for underground utility tunnels, as provided in an embodiment of this application. Figure 1 As shown in the figure, the fire prevention method for underground utility tunnels provided in this application embodiment specifically includes the following steps: Step 10: Obtain the composite physical field characteristic parameters of the target area, including the temperature change rate, acoustic vibration signal characteristic value, and gas concentration value.

[0022] In this step, a high-response-speed temperature sensor, preferably a thin-film thermocouple or infrared temperature sensor, is installed on the surface or core of the cable joint to continuously collect temperature data. The collected temperature sequence is then differentiated in real time to obtain the rate of temperature change over time. Compared to absolute temperature values, the rate of temperature change can detect abnormal temperature rise trends earlier, effectively avoiding interference caused by ambient temperature fluctuations. An acoustic emission sensor is coupled and installed on the surface of the metal conductor or insulation layer of the cable joint, or a non-contact laser Doppler vibrometer is used to collect broadband acoustic vibration signals. Through digital filters or wavelet transform algorithms, the signal intensity of a preset frequency band corresponding to the characteristic frequency of the cable fault is extracted. The preset frequency band is determined based on numerous fault simulation experiments, including but not limited to ultrasonic signals generated by partial discharge, low-frequency vibration signals generated by mechanical loosening, and mid-frequency acoustic emission signals generated by the pyrolysis of insulation materials. A miniature gas sensor array is deployed in the airflow channel or enclosed space near the cable joint, including at least sensors targeting the characteristic gases of the cable insulation material's pyrolysis. For cross-linked polyethylene insulation, the characteristic gases include methane and ethylene; for ethylene propylene rubber insulation, the characteristic gases include carbon monoxide and hydrogen. The concentration value output by the sensor, after temperature and humidity compensation, is used as a characteristic parameter of the gas concentration.

[0023] Step 20: When the temperature change rate, acoustic vibration signal characteristic value, or gas concentration value reaches a preset critical value, a trigger signal is generated.

[0024] In this step, multiple parameters are judged in real time to identify whether the critical state requiring intervention has been entered. Traditional threshold judgment usually adopts fixed logic, but the evolution of cable faults is complex and diverse. A single judgment mechanism is difficult to balance sensitivity and accuracy. Therefore, two optional judgment mechanisms are provided, which are suitable for different scenarios. One of them can be selected or the two can be deployed in parallel according to actual needs.

[0025] As an optional embodiment, a trigger signal is generated when the temperature change rate, the characteristic value of the acoustic vibration signal, or the gas concentration value reaches a preset critical value. Specifically, it may include: Step 201: The temperature change rate includes the time derivative of the temperature, the characteristic value of the acoustic vibration signal includes the intensity of the acoustic vibration signal in a preset frequency band, and the gas concentration value is the characteristic gas concentration generated by the pyrolysis of the cable insulation material.

[0026] In this step, the temperature change rate is specified as the time derivative of temperature, the acoustic vibration signal characteristic value is specified as the acoustic vibration signal intensity of the preset frequency band, and the gas concentration value is specified as the characteristic gas concentration generated by the pyrolysis of cable insulation material. The preset frequency band is determined according to the fault type. If the focus is on partial discharge, the frequency band is set to the ultrasonic frequency band; if the focus is on mechanical loosening, the frequency band is set to the low frequency band.

[0027] Step 202: When the time derivative, acoustic vibration signal intensity, and characteristic gas concentration all exceed the critical values, it is determined that the fire has entered the critical state, and an emergency trigger signal is generated.

[0028] In this step, the time derivative, acoustic vibration signal intensity, and characteristic gas concentration are compared with their respective preset thresholds in real time. If and only if all three parameters are satisfied at the same time, it is determined that the target area has entered the critical state of thermo-electric runaway, and an emergency trigger signal is generated. This signal has the highest priority and will directly trigger all subsequent intervention actions.

[0029] As an optional embodiment, the method may further include: step 201': when one or both of the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration are detected to exceed a preset threshold, the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration are input into a pre-trained probability model.

[0030] In this step, when one or two of the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration are detected to exceed the preset threshold, but the condition that all three exceed the threshold is not yet met, the suspected state is entered. At this time, the three values ​​at the current moment are combined into a three-dimensional feature vector and input into the pre-trained probability model.

[0031] The probabilistic model is constructed by simulating various fault evolution processes of cable joints in a laboratory environment, including increased contact resistance, insulation layer scratches, localized dampness, and long-term overload. Simultaneously, time derivatives, acoustic vibration signal intensity, and characteristic gas concentration data are collected, and the actual state at each moment is labeled, including normal, sub-healthy, critical state, and fault. The raw time-series data undergoes sliding window processing to extract statistical features as model input. A binary classification model is trained using classification algorithms such as logistic regression, support vector machine, random forest, or deep neural network, outputting the probability value P that the current state belongs to an impending thermo-electric runaway. By learning from a large number of samples, the nonlinear coupling relationship between the three parameters can be captured, identifying subtle signs that the threshold method cannot detect.

[0032] Step 202': Obtain the probability value representing the occurrence of a fire as output by the probability model; Step 203': When the probability value exceeds the preset action probability threshold, generate a probability trigger signal.

[0033] In this step, the probability value P output by the model is obtained, with P ranging from [0,1]. P is then compared with a preset action probability threshold. If P is greater than the threshold, a probability trigger signal is generated, indicating that the system has a high degree of confidence in the imminent occurrence of thermal-electric runaway, but the urgency level is lower than that of an emergency trigger signal.

[0034] The emergency trigger channel uses a hardware comparator to capture sudden, multi-parameter simultaneous deterioration in extreme situations, ensuring immediate action in the most critical moments. The probability trigger channel uses MCU or DSP to run model inference, capturing gradual, single-parameter-first early faults, enabling proactive warnings and refined intervention. The outputs of the two channels are ORed to generate the final trigger signal; that is, as long as either channel generates a trigger signal, the system responds and intervenes. This collaborative mechanism ensures rapid response in extreme situations and enhances the ability to identify early faults.

[0035] Step 30: Respond to the intervention action based on the trigger signal.

[0036] As an optional embodiment, the intervention action in response to the trigger signal may specifically include: Step 301: According to the emergency trigger signal, the first-level intervention action, the second-level intervention action and the third-level intervention action are initiated in sequence. The intervention actions include spraying ultra-high-speed gas-liquid two-phase medium, spraying high-pressure inert gas and releasing expandable flame-retardant material.

[0037] In this step, when an emergency trigger signal is generated, it is determined that the target area has entered a highly certain critical state, and immediate and all-out intervention is necessary without any delay or hesitation. Therefore, the three-level intervention actions are initiated in sequence according to a preset fixed time sequence, with no deliberate time interval between each level, in order to achieve the maximum blocking effect.

[0038] The system sprays an ultra-high-speed gas-liquid two-phase medium. An emergency trigger signal activates the first high-speed solenoid valve, which connects the liquid storage chamber and the gas storage chamber. The liquid storage chamber contains a highly effective liquid medium, preferably perfluorohexanone or a water-based extinguishing agent. The gas storage chamber contains a high-pressure driving gas, preferably nitrogen or compressed air. After the solenoid valve opens, the high-pressure gas flows at high speed through the venturi structure, creating a negative pressure in the mixing chamber and drawing the liquid from the liquid storage chamber. The two are then violently mixed in the mixing chamber to form an ultra-high-speed gas-liquid two-phase flow, which is then directed to the local overheated hot spot in a cone-shaped mist form through a specially designed nozzle.

[0039] The jet velocity of the ultra-high-speed gas-liquid two-phase flow can cover the hot spot area in a very short time. After the droplets come into contact with the high-temperature surface, they quickly vaporize and absorb a large amount of latent heat of vaporization, reducing the temperature of the hot spot to a safe range. At the same time, the generation of water vapor or perfluorohexanone vapor can also locally displace oxygen, forming a preliminary inertization effect.

[0040] Step 302: Initiate the first-level intervention action based on the probability trigger signal, and continuously monitor the time derivative, acoustic vibration signal intensity, and characteristic gas concentration after the first-level intervention action.

[0041] In this step, upon receiving a probability-triggered signal, it is determined that the target area has a high risk of failure, but has not yet reached the most urgent state. To avoid excessive intervention causing unnecessary resource consumption and equipment damage, the first-level intervention action is initiated by spraying ultra-high-speed gas-liquid two-phase medium for local cooling, while simultaneously entering the confirmation-escalation mode. While initiating the first-level intervention action, three parameters—time derivative, acoustic vibration signal intensity, and characteristic gas concentration—continue to be collected and monitored at a high sampling frequency, indicating an exploratory intervention phase. The response at the fault point is observed through the first-level cooling. If the parameters improve rapidly after cooling, it indicates that the risk is controllable or it is a false alarm; if the parameters continue to deteriorate after cooling, it indicates that the fault is escalating, requiring further intervention.

[0042] Step 303: If, within the preset confirmation time window, at least one of the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration continuously rises or exceeds the corresponding auxiliary threshold, the second-level intervention action and the third-level intervention action are initiated.

[0043] In this step, a preset confirmation time window is initiated. The duration of this window can be set according to the specific application scenario. During the window period, the dynamic changes of the three parameters are continuously monitored, and it is determined whether any of the following escalation conditions are met: Condition A: Any parameter continues to rise, and the rate of rise exceeds a preset rate threshold; Condition B: Any parameter exceeds an auxiliary threshold that is more lenient than the initial threshold. The auxiliary threshold is usually set to 1.2-1.5 times the initial threshold to determine whether the parameter still deteriorates to a more severe level after intervention. If any of the above conditions are met within the confirmation time window, it is determined that the fault is escalating, and a single first-level intervention is insufficient to completely block it. At this time, the second-level and third-level intervention actions are immediately initiated, spraying high-pressure inert gas and releasing expandable flame-retardant material in the same manner as in step 301 to achieve complete blockage.

[0044] Step 304: If, within the confirmation time window, the time derivative value, the intensity of the acoustic vibration signal, and the concentration of the characteristic gas all decrease or return to the normal range, terminate the intervention action and record it as an alarm event.

[0045] In this step, if none of the three parameters show a deteriorating trend within the confirmation time window, but instead tend to decrease or have returned to the normal range, it is determined that the initial probability trigger may have originated from transient interference, sensor noise, or false alarm. Subsequent intervention actions will be terminated, the second and third level interventions will not be initiated, and this event will be recorded as an alarm event.

[0046] Specifically, the recorded alarm event information includes: trigger time, parameter value at the time of triggering, duration of the first-level intervention, response curve of parameters after intervention, and final judgment conclusion; this data can serve as valuable samples for subsequent optimization of probability models and adjustment of thresholds, further improving the system's recognition accuracy.

[0047] As an optional embodiment, the method may further include: step 305: dynamically adjusting the duration of the injection of the gas-liquid two-phase medium according to the amplitude of the time derivative value.

[0048] In this step, when initiating the first-level intervention, the duration of the jetting of the gas-liquid two-phase medium is calculated based on the amplitude of the time derivative. The larger the amplitude, the more severe the temperature rise, the more cooling energy is required, and the longer the jetting duration is accordingly. The specific implementation can be achieved using a lookup table method or a function mapping method: Lookup table method: A table corresponding to the amplitude range and the jetting duration is established in advance; Function mapping method: A linear or nonlinear functional relationship is established.

[0049] If the amplitude is extremely high, a second spray can be performed immediately after the first spray to ensure a thorough cooling effect.

[0050] As an optional embodiment, the method may further include: simultaneously initiating the first-level intervention action, transmitting a probe beam to the target area via an ultrasonic sensor array and receiving the reflected echo; based on the reflected echo, reconstructing in real time a three-dimensional point cloud model of the target area and the surrounding preset distance space, the three-dimensional point cloud model including the surface morphology of the monitored object, the spatial distribution of the sprayed medium, and the coverage boundary of the expanded flame-retardant material; comparing the three-dimensional point cloud model with a pre-stored normal state three-dimensional reference model to calculate the actual coverage and effectiveness of the current intervention action; if the actual coverage and effectiveness do not reach the preset intervention effect target value, adjusting the spray direction, spray angle, or spray pressure of the second-level or third-level intervention action in real time according to the deviation, until the actual coverage and effectiveness reach the intervention effect target value.

[0051] In this step, while initiating the first-level intervention action, the ultrasonic sensor array is triggered to emit a detection beam towards the target area and receive the reflected echo. Based on the reflected echo, a three-dimensional point cloud model of the target area and the surrounding preset distance space is reconstructed in real time. This model includes at least the surface morphology of the monitored object, the spatial distribution of the sprayed medium, and the coverage boundary of the expanded flame-retardant material. The real-time reconstructed three-dimensional point cloud model is compared with a pre-stored normal state three-dimensional reference model to calculate the actual coverage range and effectiveness of the current intervention action. The effectiveness is a weighted comprehensive value of multiple evaluation indicators such as coverage range, coverage uniformity, effective depth of action, char layer coverage area, and char layer thickness, used to quantitatively characterize the actual effect of the fire intervention action. If the actual coverage range and effectiveness do not reach the preset intervention effect target value, the spray direction, spray angle, or spray pressure of the subsequent second-level or third-level intervention action is adjusted in real time according to the deviation. The adjustment process continues until the actual coverage and effectiveness reach the target value of the intervention effect, or the current level of intervention ends. Through the above-mentioned real-time imaging and dynamic adjustment mechanism, it is ensured that the intervention actions at all levels can accurately hit the target area and achieve the expected blocking effect, avoiding intervention failure caused by spray deviation, media drift or coverage dead angles.

[0052] Step 306: Dynamically adjust the injection pressure and duration of the high-pressure inert gas according to the amplitude of the acoustic vibration signal intensity.

[0053] In this step, when initiating the second-level intervention, the injection pressure and duration of the high-pressure inert gas are calculated based on the amplitude at the trigger moment. A larger amplitude of the acoustic vibration signal intensity indicates a stronger discharge or mechanical vibration, requiring a greater degree of insulation reinforcement. Therefore, a higher injection pressure and a longer injection time are needed to ensure a sufficiently concentrated and thick inert gas barrier is formed around the fault point. This can be achieved using a lookup table method or a function mapping method. The relationship between the amplitude of the vibration signal intensity and the injection pressure is established: the base pressure is used when the amplitude of the vibration signal intensity is within 1-2 times the threshold; 1.5 times is used when it is within 2-3 times, and the injection duration is correspondingly extended.

[0054] In addition, the intensity of the vibration signal can be further refined based on the characteristic frequency of the vibration signal intensity. If the vibration signal intensity is concentrated in the ultrasonic frequency band, it indicates that partial discharge is the main cause, and the spray direction can be strengthened towards the weak point of insulation. If the vibration signal intensity is concentrated in the low frequency band, it indicates that mechanical loosening is the main cause, and the overall spray intensity of the annular nozzle can be strengthened.

[0055] Step 307: Project the encapsulated expandable flame retardant material onto the surface of the target area. After contacting the high-temperature surface, the expandable flame retardant material forms a fireproof isolation layer.

[0056] As an optional embodiment, the encapsulated expandable flame-retardant material is projected onto the surface of the target area. Specifically, this may include: uniformly doping the expandable flame-retardant material with magnetic particles; while projecting the expandable flame-retardant material, adjusting the field strength and direction of a magnetic field generator deployed around the nozzle according to the target area location and obstacle distribution identified by the three-dimensional point cloud model; under the action of the magnetic field force, the expandable flame-retardant material moves along a preset obstacle avoidance path to adhere to the surface of the target area.

[0057] In this step, before encapsulation, the expandable flame retardant material is uniformly doped with magnetic particles with high magnetic permeability through physical mixing or chemical coating. The doped flame retardant material exhibits controllable magnetic response characteristics under the action of an external magnetic field, and the magnetization intensity is proportional to the doping concentration.

[0058] As an alternative, flame-retardant materials can be electrostatically sprayed after encapsulation to charge their surface and thus respond to an electric field. However, this embodiment prefers a magnetic doping scheme because it is not affected by humidity and has strong magnetic field penetration.

[0059] At least one set of magnetic field generating devices is deployed around the nozzle or near the target area. The magnetic field generating devices are preferably electromagnetic coil arrays. By adjusting the magnitude and direction of the current in each coil, a controllable magnetic field with arbitrary direction and gradient in space can be generated. The number and layout of the magnetic field generating devices are pre-optimized according to the geometric characteristics of the target area to ensure that an effective magnetic control space can be formed in the target area and possible injection paths.

[0060] During the system installation and commissioning phase, the magnetic field generator is calibrated using a three-dimensional point cloud model of the ultrasonic wave, establishing the correspondence between the magnetic field intensity distribution map and spatial coordinates. When initiating the third-level intervention action, the model already includes the surface topography of the monitored object, the location of the target area, and the distribution of obstacles within a preset distance in the surrounding space. A path planning algorithm is used to calculate the optimal obstacle avoidance path from the nozzle outlet to the surface of the target area. The optimal obstacle avoidance path must at least satisfy the following constraints: the path length is minimized to reduce flight time and energy consumption. The path is kept at a distance greater than a preset safety threshold to avoid collisions; the angle between the path's end direction and the surface normal of the target area is less than a preset angle to ensure effective material adhesion.

[0061] In this step, during the third-level intervention, the expandable flame-retardant material is released via projection. The encapsulation material, such as a thin-walled plastic shell or aluminum foil bag, ruptures instantly under the action of the electrically controlled release mechanism. The flame-retardant material is projected onto the surface of the fault point by inertia or a micro-propulsion device. The projection angle and distance can be pre-calibrated according to the location of the target area to ensure that the material accurately reaches the expected position. For complex-shaped joint areas, multiple projection points can be set or steerable nozzles can be used. When the expandable flame-retardant material comes into contact with a high-temperature surface, for expandable graphite: the interlayer inserted acid radicals or other compounds rapidly decompose and vaporize, generating gas pressure that pushes the graphite layers apart, expanding in volume by tens to hundreds of times to form a worm-like expanded char layer. For expandable flame-retardant coatings: the charring agent, foaming agent, and acid source in the material undergo a synergistic reaction at high temperatures to form a porous expanded char layer. The formed expanded char layer has good heat insulation, insulation, and oxygen barrier properties, and can tightly cover the surface of the fault point, forming a permanent fireproof isolation layer that remains stable even under water erosion or mechanical vibration, achieving long-term effective protection.

[0062] Step 40: After performing the intervention, obtain the feedback parameters and determine whether to repeat the intervention based on the feedback parameters.

[0063] In this step, the complexity of cable faults means that any single intervention action is uncertain. It may be that the intervention intensity is insufficient and fails to completely block the fault chain, or the fault itself may be stubborn or recurrent. Therefore, the intervention should not be regarded as a one-time action. Instead, the intervention effect should be quantitatively evaluated by monitoring the status parameters after the intervention in real time, and a decision should be made on whether supplementary intervention is needed based on the evaluation results.

[0064] Furthermore, an ultrasonic sensor array is introduced as a key feedback monitoring method. Ultrasonic waves can penetrate smoke, steam, and some obstacles to achieve non-contact imaging of the target area and the distribution of the sprayed medium, thereby quantitatively assessing the actual coverage and effective depth of the intervention action, making up for the shortcomings of single temperature and gas sensors that can only provide point measurements.

[0065] As an optional embodiment, the target area includes multiple adjacent monitoring points, each monitoring point corresponding to an independent blocking device. The blocking devices are connected via a wireless network to form a multi-agent system. The method further includes: broadcasting the monitored time derivative value, acoustic vibration signal intensity, characteristic gas concentration, and remaining medium storage to adjacent devices via a wireless network through the blocking devices; and using distributed reinforcement learning based on the received neighbor information to make collaborative decisions to determine the activation sequence, injection direction, and injection intensity of the device.

[0066] In this step, each device receives broadcast information from neighboring devices to construct a global state based on local observations. Each device cannot know all the information about the entire utility tunnel, but it can gradually expand its sensing range through multi-hop propagation. Each device uses a distributed reinforcement learning algorithm for collaborative decision-making, employing either a value decomposition network or independent Q-learning. Each device maintains a local action-value function Q. i (oi, ai), Q is selected based on the local observation oi. i The AI ​​for maximizing action includes the action space encompassing activation timing, spray direction, spray intensity, and intervention level. The reward function integrates fire suppression effectiveness, coverage efficiency, resource consumption, and collaborative behavior to guide the system in learning the optimal collaborative strategy.

[0067] As an optional embodiment, after the intervention action is performed, feedback parameters are obtained, and it is determined whether to repeat the intervention action based on the feedback parameters. Specifically, it may include: Step 401: Monitoring the local temperature, gas concentration and insulation status after the intervention action through an array of temperature sensors, gas sensors and ultrasonic sensors.

[0068] In this step, the existing thin-film thermocouples or infrared temperature sensors continue to operate, collecting the absolute temperature value and temperature change rate of the target area after intervention. Attention is paid to whether the temperature has dropped below the safe threshold, whether the temperature shows a rebound trend, and whether the temperature distribution is uniform. If an infrared thermal imaging sensor is used, a two-dimensional temperature distribution map of the target area can also be obtained, allowing for a direct assessment of the cooling effect. The existing gas sensors continue to operate, collecting the concentration values ​​of characteristic gases, including methane, ethylene, and carbon monoxide. Attention is paid to whether the gas concentration has dropped below the detection limit, whether the rate of concentration decrease meets expectations, and whether new characteristic gases appear. This may indicate that the intervention has triggered a new pyrolysis reaction. The continued presence or rebound of gas concentration may mean that the fault point is still generating heat or the material is still decomposing.

[0069] Furthermore, by deploying an array of ultrasonic sensors around the target area, a specific frequency ultrasonic detection beam is actively emitted into the target area, and the reflected echoes are received. Utilizing the propagation characteristics of ultrasound in gaseous media and the differences in reflection across different media, the media distribution of the target area and its surrounding space can be inverted. Specifically, when ultrasound propagates in air, it generates reflected echoes when encountering media of different densities. By analyzing the intensity, arrival time, and Doppler shift of the echoes, it can be determined whether a high-pressure inert gas has successfully formed a continuous gas barrier around the target, manifested as a highly reflective annular region; whether expandable flame-retardant material has successfully covered the surface of the fault point, manifested as the surface reflection characteristics changing from metallic or insulating material characteristics to carbon layer characteristics; and whether an electric arc discharge has occurred, as an electric arc generates a strong broadband acoustic emission signal, which can serve as indirect evidence.

[0070] Furthermore, by processing the echo signals received by multiple ultrasonic sensors using beamforming and imaging algorithms, a three-dimensional point cloud model of the target area and its surrounding space can be reconstructed in real time, intuitively displaying key information such as the spatial distribution of the jetting medium and the coverage boundary of the expanded carbon layer.

[0071] Step 402: If the feedback parameters determine that the fault chain has not been completely blocked, trigger the repeated execution of at least one intervention action.

[0072] In this step, if any condition is not met, the fault chain is determined to be not completely blocked, including: the temperature drops but is still above the safety threshold or rebounds; the gas concentration is still at a high level or drops slowly; ultrasonic imaging shows that the expanded carbon layer is not completely covered and there are exposed areas; there are still intermittent weak acoustic emission signals, indicating that there may be residual discharge.

[0073] Step 403: Based on the deviation between the feedback parameters and the preset target state, trigger the repeated execution of at least one intervention action.

[0074] As an optional embodiment, before generating a trigger signal when the temperature change rate, acoustic vibration signal characteristic value, or gas concentration value reaches a preset critical value, the method may further include: acquiring a historical monitoring data sequence of the target area; inputting the historical monitoring data sequence into a time-series prediction model, and outputting a predicted trajectory for the temperature change rate, acoustic vibration signal characteristic value, and gas concentration value within a future preset time window; performing similarity matching between the predicted trajectory for the temperature change rate, acoustic vibration signal characteristic value, and gas concentration value and multiple fire paths in a corresponding preset fire evolution path library, selecting the fire path with the highest similarity, identifying the evolution mode type of the current fault, and predicting the remaining available intervention time window; and selecting a corresponding intervention strategy from multiple preset intervention strategies based on the evolution mode type and the remaining available intervention time window, wherein the intervention strategy includes the initiation timing relationship, intensity, and duration of the intervention action.

[0075] In this step, based on the type label of the matched fire path, the evolution mode type of the current fault is identified. If the matched path is an overheating fault, it is determined that the current fault is dominated by overheating. Remaining available intervention time window prediction: On the matched fire path, the stage corresponding to the current predicted trajectory is located. Specifically, the predicted trajectory is optimally aligned with the fire path. Using the optimal path of DTW (Dynamic Time Warping), the time corresponding to the end of the predicted trajectory on the fire path is found. The time length from the corresponding time on the fire path to the critical point is the remaining available intervention time window. If the end of the predicted trajectory has already crossed the critical point, the intervention time window is 0 or negative, indicating that it is about to enter or has already entered a state of loss of control, requiring immediate emergency intervention. The physical meaning of the remaining available intervention time window is how much time is left for intervention from the current moment until the fault enters an irreversible state of loss of control, according to the current evolution trend. The longer the time window, the more suitable a gentle intervention method can be; the shorter the time window, the more decisive and forceful intervention is required.

[0076] Table 1 Intervention Strategy Library

[0077] Based on the identified evolutionary pattern type, a subset of applicable strategies is selected. Based on the predicted remaining available intervention time window, a strategy matching the time window is selected from the subset: if several pre-time windows are extremely short, the highest intensity synchronous startup strategy is directly selected; if several pre-time windows are long and have high confidence, a mild strategy can be selected to save resources; if the confidence is low, a conservative strategy can be selected, such as default sequential startup, to avoid erroneous actions.

[0078] Furthermore, the selected strategy will be stored in a register as a preset intervention strategy for use in subsequent step 30. When step 20 generates a trigger signal, step 30 will directly call the preset strategy to perform the intervention without having to calculate it from scratch. If the type of trigger signal generated by step 20 conflicts with the preset strategy, such as a probability trigger signal but the preset strategy is an emergency mode, then the signal from step 20 will be used to override it.

[0079] For the matched fire path, not only can preset strategies be selected, but the strategy parameters can also be further optimized based on the specific evolution curves along the path. For example, the required cooling energy can be accurately calculated based on the curvature of the predicted trajectory near the critical point; the injection pressure curve of the inert gas can be optimized based on the predicted peak discharge intensity; and the required amount of flame-retardant material can be estimated based on the predicted total amount of pyrolysis gas. This dynamic optimization makes the intervention strategy no longer a fixed template, but a customized solution for the current fault.

[0080] In other words, this step runs continuously in the background, updating the prediction results and pre-selected strategies at a low frequency. Once the predicted intervention time window is lower than a certain warning threshold, a warning message can be sent to the monitoring platform in advance. Steps 10 and 20 maintain real-time monitoring, and once the parameters reach the critical value, a trigger signal is generated immediately. After receiving the trigger signal, step 30 first checks whether there is an intervention strategy pre-selected in step 00. If so, the strategy is executed directly. If not, for example, if it is the first run or the prediction confidence is too low, the default response logic based on the signal type is executed, such as steps 301-304. The feedback evaluation results of step 40 can be used as new samples for subsequent optimization of the prediction model and path library.

[0081] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0082] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0083] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0085] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0086] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0087] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0088] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0089] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0090] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0091] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for proactive fire prevention in underground utility tunnels, characterized in that, The method includes: Acquire composite physical field characteristic parameters of the target area, including temperature change rate, acoustic vibration signal characteristic value and gas concentration value; A trigger signal is generated when the temperature change rate, acoustic vibration signal characteristic value, or gas concentration value reaches a preset critical value; The intervention action is responded to based on the trigger signal; After performing the intervention, feedback parameters are obtained, and it is determined whether to repeat the intervention based on the feedback parameters.

2. A fire prevention method for underground utility tunnels according to claim 1, characterized in that, When the temperature change rate, the characteristic value of the acoustic vibration signal, or the gas concentration value reaches a preset critical value, a trigger signal is generated, specifically including: The temperature change rate includes the time derivative of the temperature, the acoustic vibration signal characteristic value includes the acoustic vibration signal intensity of a preset frequency band, and the gas concentration value is the characteristic gas concentration generated by the pyrolysis of the cable insulation material. When the time derivative, acoustic vibration signal intensity, and characteristic gas concentration all exceed the critical values, it is determined that the fire has entered a critical state, and an emergency trigger signal is generated.

3. A fire prevention method for underground utility tunnels according to claim 2, characterized in that, The method further includes: When one or both of the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration are detected to exceed a preset threshold, the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration are input into a pre-trained probability model. Obtain the probability value representing the occurrence of a fire, output by the probability model; When the probability value exceeds a preset action probability threshold, a probability trigger signal is generated.

4. A fire prevention method for underground utility tunnels according to claim 3, characterized in that, The intervention action based on the trigger signal specifically includes: According to the emergency trigger signal, the first-level intervention action, the second-level intervention action and the third-level intervention action are initiated in sequence. The intervention actions include spraying ultra-high-speed gas-liquid two-phase medium, spraying high-pressure inert gas and releasing expandable flame-retardant material. The first-level intervention action is initiated according to the probability trigger signal, and the time differential value, acoustic vibration signal intensity and characteristic gas concentration are continuously monitored after the first-level intervention action. If, within the preset confirmation time window, at least one of the time derivative value, acoustic vibration signal intensity, and characteristic gas concentration continues to rise or exceeds the corresponding auxiliary threshold, the second-level intervention action and the third-level intervention action will be initiated. If, within the confirmation time window, the time derivative, acoustic vibration signal intensity, and characteristic gas concentration all decrease or return to normal ranges, the intervention action is terminated and recorded as an alarm event.

5. A fire prevention method for underground utility tunnels according to claim 4, characterized in that, The method further includes: The duration of the injection of the gas-liquid two-phase medium is dynamically adjusted based on the amplitude of the time derivative value. The injection pressure and duration of the high-pressure inert gas are dynamically adjusted based on the amplitude of the acoustic vibration signal intensity. The encapsulated expandable flame retardant material is projected onto the surface of the target area, and the expandable flame retardant material forms a fireproof isolation layer after contacting the high-temperature surface.

6. A fire prevention method for underground utility tunnels according to claim 5, characterized in that, The method further includes: Simultaneously with initiating the first-level intervention action, a probe beam is emitted towards the target area via an ultrasonic sensor array, and the reflected echo is received. Based on the reflected echo, a three-dimensional point cloud model of the target area and the surrounding preset distance space is reconstructed in real time. The three-dimensional point cloud model includes the surface morphology of the monitored object, the spatial distribution of the sprayed medium, and the coverage boundary of the expanded flame retardant material. The three-dimensional point cloud model is compared with a pre-stored three-dimensional baseline model of normal state to calculate the actual coverage and effectiveness of the current intervention action; If the actual coverage and effectiveness do not reach the preset fire intervention effect target value, the spray direction, spray angle or spray pressure of the second-level intervention action or the third-level intervention action shall be adjusted in real time according to the deviation until the actual coverage and effectiveness reach the fire intervention effect target value.

7. A fire prevention method for underground utility tunnels according to claim 6, characterized in that, Projecting the encapsulated expandable flame-retardant material onto the surface of the target area specifically includes: The expandable flame-retardant material is uniformly doped with magnetic particles; While projecting the expandable flame-retardant material, the field strength and direction of the magnetic field generating device deployed around the nozzle are adjusted according to the target area location and obstacle distribution identified by the three-dimensional point cloud model. Under the influence of a magnetic field, the expandable flame-retardant material moves along a preset obstacle avoidance path to adhere to the surface of the target area.

8. A fire prevention method for underground utility tunnels according to claim 4, characterized in that, The target area includes multiple adjacent monitoring points, each corresponding to an independent blocking device. These blocking devices are connected via a wireless network to form a multi-agent system. The method further includes: The time derivative value, acoustic vibration signal intensity, characteristic gas concentration, and remaining medium storage quantity monitored by the blocking device are broadcast to adjacent devices via the wireless network. Based on the received neighbor information, distributed reinforcement learning is used to make collaborative decisions to determine the activation sequence, spray direction and spray intensity of the device.

9. A fire prevention method for underground utility tunnels according to claim 5, characterized in that, After performing the intervention, obtaining feedback parameters and determining whether to repeat the intervention based on the feedback parameters specifically includes: The local temperature, gas concentration, and insulation status after the intervention are monitored using an array of temperature sensors, gas sensors, and ultrasonic sensors. If the feedback parameters determine that the fault chain has not been completely blocked, at least one intervention action is triggered to be repeated. Based on the deviation between the feedback parameters and the preset target state, at least one intervention action is triggered to be repeated.

10. A method for fire prevention in advance for underground utility tunnels according to claim 3, characterized in that, Before generating a trigger signal when the temperature change rate, acoustic vibration signal characteristic value, or gas concentration value reaches a preset critical value, the method further includes: Obtain the historical monitoring data sequence of the target area; The historical monitoring data sequence is input into the time series prediction model, and the predicted trajectory of temperature change rate, acoustic vibration signal characteristic value, and gas concentration value within the future preset time window is output. The temperature change rate prediction trajectory, acoustic vibration signal feature value prediction trajectory, and gas concentration value prediction trajectory are matched with multiple fire paths in the corresponding preset fire evolution path library for similarity. The fire path with the highest similarity is selected to identify the current fire evolution mode type and predict the remaining available intervention time window; Based on the evolution mode type and the remaining available intervention time window, a corresponding intervention strategy is selected from a plurality of preset intervention strategies. The intervention strategy includes the timing relationship of the intervention action, the intensity of the effect, and the duration of the intervention.

Images