Fire extinguishing system activation method based on multiple signal interlocking

Abstract

The application discloses a fire extinguishing system starting method based on multi-signal interlocking, and belongs to the technical field of fire extinguishing control, and comprises the following steps: synchronously collecting original signals of three types of circuits of fire detection, power protection and environmental auxiliary monitoring to generate an effective fire signal set; completing bidirectional binding of multi-source signals and device identities by relying on a protection object identity association table, and constructing multi-source heterogeneous signal association evidence clusters; calculating a basic confidence by a hierarchical weight decision model, identifying signal mutual exclusion conflicts, introducing a compensation correction coefficient to obtain a modified confidence, dividing four confidence levels, and issuing an actuator according to the grade matching corresponding linkage locking control words; continuously collecting signals to update the confidence level during the fire extinguishing process, outputting a rekindling additional injection or system reset instruction as required, synchronously archiving full-amount data to form an irreversible traceability chain; and the application fuses flame, electrical and environmental multi-dimensional signal interlocking verification, solves the single detection false alarm and missed alarm problem, and realizes hierarchical control by taking into account safety and response efficiency.

Description

Technical Field

[0001] This invention belongs to the field of fire extinguishing control technology, specifically a method for activating a fire extinguishing system based on multi-signal interlocking. Background Technology

[0002] Currently, fires in key protected areas such as electrical rooms and power equipment are mostly caused by electrical faults. Traditional fire extinguishing control systems rely on single temperature or flame detectors to determine the fire situation. The signal dimension is limited and is easily affected by environmental interference such as dust, airflow, and temperature and humidity fluctuations, which can lead to false alarms. At the same time, there is a lack of monitoring for fault precursors such as electrical arcs and residual currents, which poses a risk of automatic fire extinguishing systems malfunctioning.

[0003] Existing technologies do not perform hardware address verification, time synchronization, or device identity association for multi-source detection signals. This can easily lead to data corruption and signal mismatch to non-target protection devices due to signals from different loops. Furthermore, they do not distinguish between signal conflict and mutual exclusion scenarios, lack conflict resolution mechanisms in confidence assessment, and result in insufficient reliability in fire situation determination.

[0004] Conventional fire suppression systems only have simple two-level start-stop logic, which cannot be used to classify and handle fires according to the actual situation. Either the fire suppression system will malfunction and damage the protected equipment, or the manual confirmation window will be too long and delay the handling of emergency fires. After the fire suppression is started, there is no dynamic monitoring of reignition and additional spray control logic. After the flame is extinguished, there is no standardized reset procedure. Furthermore, the monitoring signals and control commands do not have a complete archive and traceability link, making it difficult to trace the source of faults and malfunctions.

[0005] To address the industry pain points mentioned above, such as low reliability of single detection, lack of multi-signal fusion judgment, crude handling strategies, and lack of full-process traceability, this invention proposes a fire extinguishing system activation method based on multi-signal interlocking. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a fire extinguishing system activation method based on multi-signal interlocking. It simultaneously collects raw signals from three types of circuits: fire detection, power protection, and environmental auxiliary monitoring, generating a valid fire signal set. It utilizes a protected object identity association table to achieve bidirectional binding between multi-source signals and device identities, constructing a cluster of evidence for multi-source heterogeneous signals. A hierarchical weighted adjudication model calculates the basic confidence level, identifies signal mutual exclusion conflicts, introduces compensation correction coefficients to obtain a corrected confidence level, and classifies the system into four confidence levels. Based on the level, corresponding linkage interlocking control words are matched to issue execution mechanisms. During the fire extinguishing process, signals are continuously collected to update the confidence level, and additional spraying or system reset commands are output as needed. Simultaneously, all data is archived to form an irreversible traceability chain. This invention integrates multi-dimensional signal interlocking verification of flame, electrical, and environmental signals, solving the problem of false alarms and missed alarms from single detectors, and providing hierarchical control that balances safety and response efficiency.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a fire extinguishing system activation method based on multi-signal interlocking, comprising: The system collects the raw monitoring signals output from the fire detection circuit, power protection circuit, and environmental auxiliary monitoring circuit, performs physical isolation verification of the signal sources, and combines them to generate an effective fire signal set. Retrieve the pre-constructed protection object identity association table, and perform bidirectional binding between each signal element in the effective fire signal set and the identity code of the same protected device to generate a multi-source heterogeneous signal association evidence cluster; The preset hierarchical weighted adjudication model is invoked to initially assign confidence values ​​to the multi-source heterogeneous signal association evidence cluster, obtain a basic confidence score, simultaneously detect the logical mutual exclusion state between each signal element within the multi-source heterogeneous signal association evidence cluster, perform compensation correction on the basic confidence score according to the conflict resolution rules, and output a four-level discretized confidence level. Based on the four levels of discretized confidence levels, four execution strategies are matched to generate a linkage interlocking control word; The linkage interlock control word is sent to the fire extinguishing actuator to initiate the fire extinguishing action. After activation, the original monitoring signal is re-acquired and the four-level discretized confidence level is updated. Based on the updated four-level discretized confidence level, the additional spray for reignition suppression or the system reset command after the flame is extinguished is triggered. At the same time, all original signal logs, control command flow records and on-site image data are archived synchronously to build an irreversible operation traceability chain.

[0008] In a preferred embodiment of the present invention, the process of generating the effective fire signal set includes: Acquire the first raw monitoring signal output from the fire detection circuit; the first raw monitoring signal includes at least the temperature change rate value and the flame spectrum characteristic value; Acquire the second raw monitoring signal output by the power protection circuit; the second raw monitoring signal includes at least the residual current value, voltage distortion rate, and arc pulse count. Acquire the third raw monitoring signal output by the environmental auxiliary monitoring loop; the third raw monitoring signal includes at least the gas concentration value, the gas pressure change, and the humidity compensation value; Based on the unique hardware address of the signal acquisition card corresponding to each loop, the first original monitoring signal, the second original monitoring signal, and the third original monitoring signal are compared bit by bit with the pre-stored list of valid addresses, and valid signal channels with matching addresses are retained. Using the system's reference clock as a synchronization reference, the original monitoring signals output from the retained valid signal channels are timestamped and aligned, and valid fire signal sets are generated by combining them according to the signal source category.

[0009] As a preferred embodiment of the present invention, the process of generating the multi-source heterogeneous signal correlation evidence cluster includes: Based on the physical source identifier carried by each signal element in the effective fire signal set, query the protected object identity association table to obtain the protected equipment identity code and spatial coordinate information corresponding to each signal element; Using the protected device identification code as the primary key for querying, all corresponding signal acquisition channel numbers are obtained from the protected object identity association table to form an identity anchor mapping table. The physical source identifier of each signal element in the valid fire signal is matched with the signal acquisition channel number in the identity anchor mapping table. The matched signal elements are stored in the corresponding flame sign subset, electrical anomaly subset and environmental disturbance subset in the identity anchor mapping table according to their monitoring type. When the sum of the number of signal elements in the flame sign subset and the electrical anomaly subset is not less than the preset joint trigger threshold, the signal elements in all subsets are fused with the protected device identification code as the anchor point to generate a multi-source heterogeneous signal association evidence cluster.

[0010] In a preferred embodiment of the present invention, the step of invoking a preset hierarchical weighted adjudication model to initially assign confidence values ​​to the multi-source heterogeneous signal association evidence clusters and obtain a basic confidence score includes: A list of signal elements is extracted from the multi-source heterogeneous signal correlation evidence cluster, and each signal element is classified into flame sign signals, electrical anomaly signals, and environmental disturbance signals according to the monitoring type. For each signal element in each type of signal, the corresponding quality weight coefficient is calculated by calling the preset signal quality evaluation function based on the signal quality indicator code and the estimated signal-to-noise ratio carried by the signal element. For each signal element in each type of signal, the corresponding feature significance score is calculated by calling the preset feature intensity mapping function based on the signal element value. The product of the quality weight coefficient and the feature significance score of each signal element in the same category is accumulated and then divided by the total number of signal elements in that category to obtain the confidence components of flame signs, electrical anomalies and environmental disturbances. Each confidence component is multiplied by the corresponding pre-set weight coefficient of the hierarchical weighted adjudication model, then weighted summed and normalized to obtain the basic confidence score.

[0011] In a preferred embodiment of the present invention, the preset feature intensity mapping function sets corresponding feature thresholds and mapping rules for different types of signal elements. For signal elements of the temperature change rate type, the corresponding feature significance score is obtained by linear interpolation based on the relative position of its value between the preset low and high temperature change rate thresholds. For signal elements of the flame spectrum feature value type, the correlation coefficient between its value and the preset reference spectrum template is calculated, and the correlation coefficient is converted into the corresponding feature significance score. For signal elements of the residual current value, voltage distortion rate, and arc pulse count type, the corresponding feature significance score is obtained after amplitude limiting based on the ratio of its value to its respective preset reference threshold.

[0012] As a preferred embodiment of the present invention, the output process of the four-level discretized confidence levels includes: Extract the acquisition time stamp sequence from the multi-source heterogeneous signal association evidence cluster. Determine whether there is a signal sudden intrusion mutual exclusion state based on whether the difference between adjacent time stamps is less than a preset sudden intrusion time threshold. Determine whether there is a photoelectric signal homogeneity mutual exclusion state based on whether the flame spectrum feature value in the flame sign signal is in the infrared band and whether the arc pulse count in the electrical anomaly signal is not zero. Determine whether there is an airflow interference mutual exclusion state based on whether the air pressure change in the environmental disturbance signal is greater than a preset air pressure drastic change threshold. Summarize the identified mutual exclusion states and count the total number of mutual exclusion states; The corresponding compensation correction coefficient is obtained by querying the preset conflict resolution rule table based on the total number of mutual exclusion states; the compensation correction coefficient is negatively correlated with the total number of mutual exclusion states. Multiply the base confidence score by the compensation correction coefficient to obtain the corrected confidence score; Based on the preset numerical range into which the corrected confidence score falls, four levels are mapped and output: observation level, early warning level, action level, and emergency level.

[0013] As a preferred embodiment of the present invention, the step of generating a linkage interlocking control word by matching four execution strategies according to the four-level discretized confidence levels includes: The preset strategy mapping table is invoked to obtain the corresponding control bit combination rules based on the current four-level discretization confidence level; the control bit includes at least the status indication channel activation bit, the pre-suppression device start bit, the fire extinguishing main valve lockout release bit, the manual confirmation time window control bit, and the spray command trigger bit. When the confidence level is observation level, a first linkage interlocking control word is generated that only sets the activation bit of the status indicator channel to the valid state. When the confidence level is the warning level, a second linkage interlocking control word is generated to set the status indicator channel activation bit and the pre-suppression device start bit to the effective state, and to set the manual confirmation time window control bit to the open timed confirmation state. When the confidence level is action level, a third linkage interlocking control word is generated that sets the status indicator channel activation bit, the pre-suppression device start bit, the fire extinguishing main valve lockout bit, and the spray command trigger bit to the effective state. When the confidence level is emergency, a fourth linkage interlocking control word is generated, which sets the status indicator channel activation position, the pre-suppression device start position, the fire extinguishing main valve lockout release position, and the spray command trigger position to the effective state, and sets the manual confirmation time window control position to the bypass state.

[0014] As a preferred embodiment of the present invention, the interlocking control word is sent to the fire extinguishing actuator to initiate the fire extinguishing action, including: The state of each control bit in the linkage interlock control word is analyzed. When the injection command trigger bit is invalid, the linkage interlock control word is converted into a state update command and sent to the state indication channel driver and the front suppression device driver respectively. When the spray command trigger bit is in an active state, the linkage interlock control word is converted into a spray execution command and a status update command; the spray execution command includes the fire extinguishing main valve opening duration and spray nozzle angle adjustment parameters; The spray execution command is sent to the fire extinguishing main valve driver and the spray nozzle driver through a high-priority data queue, and the status update command is sent to the status indication channel driver and the pre-suppression device driver through a low-priority data queue, so that each driver responds to the corresponding command and executes the action.

[0015] As a preferred embodiment of the present invention, the step of triggering the additional injection for reignition suppression or the system reset command after flameout based on the updated four-level discretized confidence level includes: When the updated four-level discretized confidence level is observation level or warning level, it is determined that the flame has been extinguished and a system reset command is generated; the system reset command includes a reset position for instructing the main fire extinguishing valve to close, the pre-suppression device to be cut off, and the status indication channel to be reset; When the updated four-level discretized confidence level is action level or emergency level, it is determined that there is a reignition state, and a reignition suppression additional injection command is generated. The system reset command or the reignition suppression additional injection command is sent to the corresponding actuator, and after the reignition suppression additional injection command is executed, it is executed repeatedly until the flame extinguishing determination condition is met and a system reset command is generated.

[0016] In a preferred embodiment of the present invention, the reignition suppression additional injection command includes an additional injection duration and an additional injection flow rate adjustment coefficient, which are determined as follows: The additional injection duration in the reignition suppression additional injection command is determined based on the updated four-level discretized confidence level. When the updated confidence level is the action level, it is set as the first additional injection duration; when the updated confidence level is the emergency level, it is set as the second additional injection duration. The second additional injection duration is greater than the first additional injection duration. The additional injection flow rate adjustment coefficient is determined based on the concentrated temperature change rate value of the re-collected effective fire signal. The ratio of the temperature change rate value to the preset high threshold is calculated as the basic flow rate adjustment coefficient. The basic flow rate adjustment coefficient is compared with the preset minimum flow rate coefficient, and the larger of the two is taken as the additional injection flow rate adjustment coefficient to control the opening degree of the fire extinguishing main valve.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention proposes a fire extinguishing system activation method based on multi-signal interlocking. This invention employs the coordinated acquisition of three signals: fire detection, power protection, and environmental auxiliary monitoring. It achieves standardized fusion of multi-source signals through hardware address verification, timestamp alignment, and device identity anchor binding. A hierarchical weighted adjudication model is constructed to quantify fire confidence. Simultaneously, it identifies various mutually exclusive conflicts such as sudden signal intrusion and airflow interference, and introduces compensation correction coefficients to calibrate the scoring. This significantly eliminates false alarms and missed alarms caused by dust, airflow, temperature, and humidity disturbances from a single sensor, accurately distinguishing between real fires and environmental interference. Based on a four-level discrete confidence level matching and graded handling strategy, the observation level provides only status indication, the early warning level activates pre-emptive suppression and reserves a manual confirmation window, the action level automatically sprays, and the emergency level bypasses manual confirmation for rapid fire extinguishing. This approach balances equipment protection and fire response efficiency, avoiding indiscriminate spraying that wastes extinguishing agents and causes secondary corrosion damage to electrical equipment.

[0018] 2. This invention proposes a fire extinguishing system activation method based on multi-signal interlocking. After fire extinguishing is activated, this invention continuously and cyclically collects on-site monitoring signals to dynamically update the confidence level. It can automatically identify the risk of reignition and adaptively adjust the additional spray duration and spray flow according to the fire level to continuously suppress reignition hazards. After the fire is extinguished, it automatically outputs a system reset command to achieve a closed-loop process. The system synchronously archives all original monitoring signals, control commands, and on-site image data, establishing an irreversible operation traceability chain to facilitate post-event fault and malfunction tracing and troubleshooting. The entire solution achieves multi-dimensional signal interlocking judgment of electrical fire precursors, open flames, and environmental disturbances. The fire judgment logic is rigorous, and the hierarchical control is flexible. It also has the ability to dynamically suppress reignition and trace the entire process data, significantly improving the intelligence, reliability, and controllability of fire extinguishing systems in scenarios such as electrical equipment and enclosed computer rooms. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the fire extinguishing system activation method based on multi-signal interlocking according to the present invention; Figure 2 This is a flowchart illustrating the principle of the fire extinguishing system activation method based on multi-signal interlocking according to the present invention. Detailed Implementation

[0020] Please see Figures 1-2 This invention provides an embodiment of a fire extinguishing system activation method based on multi-signal interlocking. This embodiment selects a large data center high-voltage power distribution room as the application scenario. The power distribution room contains 12 high-voltage power distribution cabinets, two UPS uninterruptible power supply units, cable trays, and a battery protection compartment. All electrical equipment is the protected equipment described in this invention. Each independent electrical device is assigned a unique device identification code and exclusive spatial coordinates. The power distribution room is simultaneously equipped with three types of signal acquisition hardware: fire detection circuits, power protection circuits, and environmental auxiliary monitoring circuits. It is also equipped with an edge signal processing control cabinet, a hierarchical weighted decision-making unit, a linkage interlocking control output module, a perfluorohexanone gas fire extinguishing actuator, a local monitoring terminal, and a cloud-based irreversible log storage server. This fully replicates the logic of this invention. The method includes steps S1-S5, comprising the following steps: S1: Collect the raw monitoring signals output from the fire detection circuit, power protection circuit and environmental auxiliary monitoring circuit respectively, perform physical isolation verification of the signal source, and combine them to generate an effective fire signal set; The process of generating the effective fire signal set includes: S1.1: Acquire the first raw monitoring signal output by the fire detection circuit; the first raw monitoring signal includes at least the temperature change rate value and the flame spectrum characteristic value; Furthermore, in the data center high-voltage power distribution room scenario of this embodiment, the fire detection circuit is a dedicated signal circuit with independent wiring. Each detection device in the circuit is equipped with an independent acquisition card hardware address, and there is no design of sharing acquisition channels across circuits, which fully meets the underlying hardware requirements for physical isolation of signal sources. Inside the power distribution room, three types of independent detection devices are deployed for each high-voltage power distribution cabinet, UPS unit, and battery compartment, respectively, and output three types of basic fire monitoring data. The three types of data are uniformly integrated into the first original monitoring signal, which fully includes the three core data dimensions of temperature change rate value, flame spectrum characteristic value and documented data. Each type of value is continuously collected in real time by the corresponding detection hardware, and the collection frequency is uniformly set to 20 times per second to ensure that the signal value is continuous without interruption.

[0021] Furthermore, regarding the hardware for acquiring temperature change rate values, distributed fiber optic temperature sensors are deployed inside the cabinet of each protected electrical equipment and on both the upper and lower layers of the cable interlayer. The fiber optic sensors are arranged along the cable routing and the busbar of the distribution cabinet. The fiber optic demodulator reads the real-time temperature values ​​at each point on the fiber optic cable in real time. The embedded processing unit reads the temperature sampling values ​​from two consecutive measurements in a fixed one-second time window, automatically calculates the temperature rise rate per unit time, and generates a standardized temperature change rate value. The value distinguishes between two states: positive temperature rise and negative temperature drop. During the stage of local overheating and fire, the device will continuously output a large positive temperature rise value. During normal heat dissipation fluctuations, the device will only output a small positive or negative change value. The fiber optic demodulator is also bound to an independent acquisition card hardware address. The signal transmission line is physically separated from the other two types of detection hardware, so there is no line crosstalk that could interfere with signal accuracy.

[0022] Furthermore, regarding the hardware for acquiring flame spectrum characteristic values, infrared and ultraviolet composite flame detectors are installed at the four corners of the key fire-resistant partitions in the power distribution room and inside the enclosed space of the battery compartment. The detectors have built-in spectroscopic detection modules that can capture specific bands of infrared spectrum and ultraviolet pulse signals generated during flame combustion. The acquired spectral waveforms are compared in real time with the standard combustion flame spectrum templates pre-stored locally on the device, and standardized flame spectrum characteristic values ​​are output. The magnitude of the characteristic value directly corresponds to the degree of matching between the on-site flame spectrum and the standard fire spectrum. In scenarios without open flame, only dust, light, or direct sunlight, the flame spectrum characteristic value will remain in an extremely low reference range. Only when a real combustion flame appears will the characteristic value rise rapidly. The detector output signal also follows an independent fire detection circuit wiring, which is completely physically isolated from the power protection circuit and environmental monitoring circuit, and there is no problem of line coupling interference.

[0023] Furthermore, the temperature change rate and flame spectrum characteristic values ​​collected by the three types of detection hardware are transmitted in real time to the corresponding signal acquisition card inside the edge signal processing control cabinet through a dedicated shielded cable for the fire detection circuit. The acquisition card binds the current hardware address, device location identifier, and original sampling timestamp to each set of sampling data, packages it to generate a complete first original monitoring signal, and sends it synchronously to the edge processing unit to wait for address verification. There is no problem of numerical distortion caused by mutual interference between the three types of signals.

[0024] S1.2: Obtain the second original monitoring signal output by the power protection circuit; the second original monitoring signal includes at least the residual current value, voltage distortion rate and arc pulse count; in this embodiment, it is also set to sample 20 times per second to ensure that the power abnormality signal and the fire signal are aligned in time dimension. Furthermore, the residual current value is collected in real time by the residual current transformer connected to the outgoing line of the distribution cabinet. The transformer is connected in series in the main power supply circuit of the equipment, continuously collecting the current difference between the phase line and the neutral line, and outputting a standardized residual current value through a dedicated conversion module. When the equipment insulation ages, the cable is damaged, or the internal components are short-circuited and leak current, the residual current value will quickly exceed the normal operating reference range. The transformer is equipped with a data acquisition card bound to a unique hardware address, and the leakage current signal is transmitted independently and will not share a channel with the voltage and arc detection signal lines. The transformer has a built-in self-check function for disconnection. Once the data acquisition line is disconnected or the power supply to the transformer is interrupted, a hardware fault mark will be added to the end of the residual current value signal.

[0025] Furthermore, the voltage distortion rate is calculated and output in real time by the bus power quality monitoring module. The module continuously collects the three-phase voltage waveform of the equipment power supply bus, compares it with the standard sinusoidal voltage waveform to calculate the degree of waveform distortion, and generates a standardized voltage distortion rate value. When there is a short circuit, arc discharge, or breakdown of high-power components inside the equipment, the power supply voltage waveform will be severely distorted, and the distortion rate value will increase significantly. The signal output line of the power quality monitoring module is laid out separately and does not share the transmission channel with other power monitoring hardware, ensuring that the accuracy of the distortion rate value acquisition is not affected by the load of other hardware.

[0026] Furthermore, the arc pulse count is statistically output in real time by arc light sensors installed inside the distribution cabinet and in the cable interlayer. The sensors capture the strong light pulse signal generated by the arc of the circuit fault. Each time an arc pulse is captured, the count is automatically accumulated. A rapid increase in the arc pulse count per unit time indicates that a continuous fault arc has occurred inside the equipment, which is the most critical early sign of an electrical fire. The output signal of the arc sensor is also connected to an independent acquisition card with a unique and identifiable hardware address. The shielded cable for signal transmission is laid independently to prevent crosstalk and distortion of the arc pulse signal with other circuit signals.

[0027] Furthermore, the residual current value, voltage distortion rate, and arc pulse count output by the three types of power monitoring hardware are packaged into a second original monitoring signal. The signal data packet includes the dedicated acquisition card hardware address of the power protection circuit, the corresponding protected equipment identifier, and the original sampling timestamp, and is continuously transmitted to the dedicated power signal acquisition channel inside the edge signal processing control cabinet.

[0028] S1.3: Obtain the third raw monitoring signal output by the environmental auxiliary monitoring loop; the third raw monitoring signal includes at least the gas concentration value, the gas pressure change, and the humidity compensation value; in this embodiment, the sampling frequency in the data center power distribution room scenario is also uniformly 20 times per second; Furthermore, the gas concentration value is collected by combustible organic volatile substances sensors installed inside the enclosed space of the battery compartment and distribution cabinet. Combustible and harmful gases are released when the insulation material inside the electrical equipment decomposes at high temperature or when the battery thermally runs away. The sensors capture the concentration of the corresponding gas molecules in the air in real time and output standardized gas concentration values. The values ​​increase synchronously with the amount of harmful gas volatilization. The sensors are equipped with independent acquisition card hardware addresses, and the signal transmission lines are laid out separately to avoid numerical drift of the gas concentration signal due to harmonic interference from the power circuit.

[0029] Furthermore, the pressure change is collected by high-precision pressure transmitters installed inside each enclosed fire-resistant zone of the power distribution room. When equipment catches fire, extinguishing agents are released prematurely in sealed compartments, or airflow impacts occur due to ventilation system failure, the pressure inside the sealed space will rise or fall rapidly. The transmitter continuously collects the real-time pressure value of the space, calculates the pressure difference before and after with a one-second time window, and generates a standardized pressure change value. The pressure transmitter signal channel is independent and does not share the acquisition channel with other environmental monitoring hardware, ensuring stable and noise-free acquisition of pressure change.

[0030] Furthermore, the humidity compensation value is collected by a distributed integrated temperature and humidity probe. Excessive humidity in the power distribution room can lead to decreased insulation of electrical equipment and abnormally increased leakage current, while excessively low humidity can increase the risk of static electricity fire. The probe collects the relative humidity value of the environment in real time. The system pre-stores the standard safe humidity reference range for the power distribution room and generates a standardized humidity compensation value based on the difference between the real-time humidity and the standard reference range. This value is used to correct the power and fire protection signal values ​​and eliminate the signal value deviation caused by environmental humidity. The temperature and humidity probe is bound to a dedicated acquisition card hardware address, and the shielded cable for signal transmission is laid independently and is not affected by the other two types of circuit signals.

[0031] Furthermore, the gas concentration values, air pressure changes, and humidity compensation values ​​output by the three types of environmental monitoring hardware are uniformly packaged into a third original monitoring signal. The data packet includes the dedicated acquisition card hardware address of the environmental auxiliary monitoring loop, the corresponding protective space identifier, and the original sampling timestamp. It is continuously sent to the dedicated environmental signal acquisition channel of the edge signal processing control cabinet, waiting for the hardware address validity verification of S1.4.

[0032] S1.4: Based on the unique hardware address of the signal acquisition card corresponding to each loop, the first original monitoring signal, the second original monitoring signal, and the third original monitoring signal are compared bit by bit with the pre-stored list of valid addresses, and valid signal channels with matching addresses are retained. Furthermore, in this embodiment, the embedded processing unit inside the edge signal processing control cabinet has a built-in local read-only memory chip. The chip pre-deposits a list of legal addresses for all acquisition cards in the three types of circuits (fire protection, power, and environment) of the power distribution room. The list of legal addresses is stored in partitions according to circuit type, with separate partitions for fire detection circuits, power protection circuits, and environmental auxiliary monitoring circuits. Each partition records the unique hardware address that is factory-fixed for all acquisition cards in the corresponding circuit. The address encoding format is uniformly hexadecimal fixed-length encoding, eliminating the problem of inconsistent address lengths or chaotic formats. At the same time, the list of legal addresses supports online updates from local terminals. When new or replaced monitoring and acquisition hardware is added to the power distribution room, maintenance personnel can enter the hardware address of the new acquisition card through the local monitoring terminal, which will simultaneously write to the read-only memory chip to update the list of legal addresses, adapting to on-site equipment expansion and renovation scenarios.

[0033] Furthermore, after receiving the first raw monitoring signal data packet, the edge processing unit first extracts the hardware address code of the fire detection loop acquisition card attached to the end of the data packet. It reads the code character by character from the first character to the last character and compares it character by character with each legal address code in the legal address partition of the fire detection loop in the storage chip. If any character does not match during the comparison, the acquisition card from which the first raw monitoring signal originates is determined to be illegal hardware, the signal channel is determined to be invalid, and the raw signal is discarded. If every character of the entire hardware address code completely matches an address in the list of legal addresses, the signal channel is determined to be a valid channel. The first raw monitoring signal containing the temperature change rate and flame spectrum characteristics is completely retained, the signal source is marked as a valid fire channel, and it is stored in the local temporary signal buffer to await timestamp alignment.

[0034] Furthermore, similarly, when the edge processing unit receives the output second raw monitoring signal data packet, it extracts the hardware address code of the power protection circuit acquisition card bound to the data packet, compares each character with the internal address of the power partition in the list of valid addresses, and retains the second raw monitoring signal containing residual current, voltage distortion rate, and arc pulse count if the address codes match completely, marking it as a valid power channel signal and storing it in a temporary buffer; if any character of the address does not match, the entire raw power signal is discarded directly, filtering out invalid noise signals generated by line faults, damaged acquisition cards, and unauthorized external acquisition devices.

[0035] Furthermore, when receiving the third raw monitoring signal data packet, the hardware address code of the environmental auxiliary monitoring loop acquisition card is extracted and compared bit by bit with the environmental partition address in the list of legal addresses. If the codes match completely, the third raw monitoring signal containing gas concentration, air pressure change, and humidity compensation value is retained and marked as an effective environmental channel signal and stored in a temporary buffer. If the addresses do not match, the raw environmental signal is discarded directly to prevent external illegal acquisition devices from accessing the system and sending false environmental monitoring data, which could lead to misjudgment.

[0036] Furthermore, the bit-by-bit address comparison operation is completed in parallel by the embedded hardware logic unit. The comparison of the three types of loop signal addresses is carried out simultaneously, and the time for comparing multiple channels of a single batch is controlled at the millisecond level, without causing signal processing delay. At the same time, the embedded unit will synchronously record each discarded illegal address signal log, recording the illegal hardware address code, signal reception time, and the loop type to which the signal belongs. The log is synchronized to the local storage unit in real time. The operation and maintenance personnel can retrieve the illegal signal records through the monitoring terminal to quickly locate on-site problems such as line damage, acquisition card failure, and illegal external access. After all the original signal address verifications are completed, only the three types of original monitoring signals corresponding to the valid signal channels with all successfully matched addresses are retained in the temporary buffer area, completing the signal physical isolation verification.

[0037] S1.5: Using the system reference clock source as a synchronization reference, perform timestamp alignment on the original monitoring signals output from the retained valid signal channels, and generate a set of valid fire signals according to the signal source category.

[0038] S2: Retrieve the pre-constructed protection object identity association table, and perform bidirectional binding between each signal element in the effective fire signal set and the identity code of the same protected device to generate a multi-source heterogeneous signal association evidence cluster; The process of generating the multi-source heterogeneous signal correlation evidence cluster includes: S2.1: Based on the physical source identifier carried by each signal element in the effective fire signal set, query the protected object identity association table to obtain the protected equipment identity code and spatial coordinate information corresponding to each signal element; It should be noted that in this embodiment, each valid fire signal set output contains an independent original monitoring signal data packet with a unique physical source identifier. The physical source identifier is generated by combining the hardware address of the acquisition card and the spatial sub-number of the acquisition point, which can uniquely correspond to a certain monitoring point inside the power distribution room and accurately point to one or more protected electrical devices within the monitoring coverage area of ​​that point.

[0039] S2.2: Using the protected device identification code as the primary key, obtain all corresponding signal acquisition channel numbers from the protected object identification association table to form an identification anchor mapping table; Iterate through all signal elements generated in S2.1 and temporary device association records, extract the protected device identification code from each record, set the device identification code as the retrieval primary key, retrieve the protected object identification association table again, accurately match the complete entry of the corresponding device in the form with the primary key, and read all the signal acquisition channel numbers bound to the device that were pre-entered in the entry. The channel numbers are completely distinguished into three categories: fire detection channel number, power protection channel number, and environmental auxiliary monitoring channel number. There are no cases of chaotic classification without numbering.

[0040] Furthermore, for each successfully matched protected device, a separate identity anchor mapping table is created. The header of the identity anchor mapping table is fixed as the unique identity code of the current device, i.e., the anchor identifier. The table is divided into three columns to store all fire detection acquisition channel numbers, all power protection acquisition channel numbers, and all environmental auxiliary monitoring acquisition channel numbers bound to the device. At the same time, two auxiliary fields, device spatial coordinates and device fire risk level, are added.

[0041] Furthermore, if the same device is simultaneously associated and matched by signal elements identified by multiple different physical sources, only one identity anchor mapping table corresponding to this device will be generated. Multiple identity anchor mapping tables for the same device will not be created repeatedly. The identity anchor mapping table fully summarizes all the data acquisition channel numbers bound to the device, covering all fire protection, power, and environmental monitoring points. For scenarios where multiple different devices are successfully matched, the system automatically generates an independent identity anchor mapping table with the same number of matched devices. Each identity anchor mapping table corresponds to only a single device identity anchor, and the data in each identity anchor mapping table is independent and does not interfere with each other.

[0042] S2.3: Perform an equal-value match between the physical source identifier of each signal element in the effective fire signal set and the signal acquisition channel number in the identity anchor mapping table. The successfully matched signal elements are stored in the flame sign subset, electrical anomaly subset and environmental disturbance subset corresponding to the identity anchor mapping table according to their monitoring type. Furthermore, in S2.3, each independent identity anchor point mapping table is retrieved sequentially, and all fire protection, power, and environmental acquisition channel numbers recorded in the identity anchor point mapping table are traversed. The channel number is matched with the physical source identifier carried by each signal element in the current valid fire signal set. The physical source identifier contains the complete acquisition channel number field. If the characters are completely consistent, the match is considered successful, indicating that the current signal element is valid fire data acquired by the dedicated monitoring channel of this device. According to the original signal loop type to which the signal element belongs, the corresponding storage subset is divided.

[0043] Furthermore, successfully matched fire detection circuit signal elements, including temperature change rate and flame spectrum characteristic values, are uniformly stored in the flame sign subset built into the current identity anchor point mapping table. Within the subset, all successfully matched fire-related signal elements are stored in order of sampling time, and all values, timestamps, and physical source identifiers of each signal are completely preserved. Successfully matched power protection circuit signal elements, such as residual current, voltage distortion rate, and arc pulse count, are uniformly stored in the electrical anomaly subset built into the mapping table, and all power monitoring values ​​and signal metadata are also completely preserved. Successfully matched environmental auxiliary monitoring circuit signal elements, such as gas concentration, air pressure change, and humidity compensation value, are uniformly stored in the environmental disturbance subset built into the mapping table, and all data fields related to environmental monitoring are completely preserved.

[0044] Furthermore, if the physical source identifier of any signal element cannot match any acquisition channel number in the current mapping table, it means that the signal does not belong to the data acquired by the dedicated monitoring channel of this device, and will not be stored in any subset of this mapping table. This signal element will be skipped directly, and the system will continue to match the next signal. After the three subsets of a single identity anchor point mapping table are stored, the system automatically counts the total number of signal elements in the flame sign subset and the total number of signal elements in the electrical anomaly subset, adds the two values ​​together, and generates the total number of joint trigger signals, which serves as the basis for threshold judgment calculation.

[0045] S2.4: When the sum of the number of signal elements in the flame sign subset and the electrical anomaly subset is not less than the preset joint trigger threshold, the signal elements in all subsets are fused with the protected device identification code as the anchor point to generate a multi-source heterogeneous signal association evidence cluster.

[0046] Furthermore, in this embodiment, for electrical fire scenarios in high-voltage power distribution rooms of data centers, the preset joint trigger threshold is fixed at 2. That is, when the total number of flame symptom subset signals plus the number of electrical anomaly subset signals in the identity anchor point mapping table of a single device is greater than or equal to two, it is determined that the precondition for fusion generation of evidence clusters is met. This threshold value is set according to the ignition pattern of electrical fires. A single fire signal or a single power signal is easily falsely triggered by normal heat dissipation fluctuations of equipment, instantaneous power grid harmonics, and environmental dust interference. At least one fire symptom signal and one power anomaly signal must appear simultaneously to preliminarily determine that there are real electrical fire precursors, thereby reducing the probability of system malfunction. The threshold value can be adjusted by the local monitoring terminal according to the risk level of the protection scenario. The threshold can be lowered to 1 for high-risk areas such as battery compartments, and raised to 3 for ordinary low-voltage power distribution cabinets, to adapt to the differentiated safety requirements of different protection scenarios.

[0047] Furthermore, the summation of the number of signal elements in the flame symptom subset and electrical anomaly subset within each identity anchor point mapping table is read and compared with a preset joint trigger threshold. If the summation value does not reach the threshold standard, signal fusion is not performed; only the signal data of the three subsets in the current mapping table is saved and sent to S3 for low-confidence adjudication. Only an observation-level status prompt is triggered, and a complete multi-source heterogeneous signal association evidence cluster is not generated. If the summation value is greater than or equal to the preset joint trigger threshold, signal fusion is initiated. Using the protected equipment identity code corresponding to the current identity anchor point mapping table as the globally unique anchor point identifier, all fire signal elements in the flame symptom subset, all power signal elements in the electrical anomaly subset, and all environmental signal elements in the environmental disturbance subset are uniformly encapsulated and fused to generate an independent and complete multi-source heterogeneous signal association evidence cluster.

[0048] Furthermore, the top layer of each evidence cluster data package is fixedly labeled with the device identity anchor point. The internal structure consists of three major blocks that orderly store all signal elements of three subsets. Each signal element fully retains the original monitoring value, physical source identifier, sampling alignment timestamp, and basic signal quality markers. It also includes auxiliary information such as the current device spatial coordinates and the device's fire risk level. This fully records all synchronous multi-source fire monitoring data corresponding to the same protected device, enabling deep association and binding of three types of heterogeneous signals—fire protection, power, and environment—based on the same device identity. This eliminates the technical defects of scattered multi-loop signal sources and the inability to accurately correspond to the fire-stricken device. Each device that meets the joint trigger threshold generates an independent and exclusive evidence cluster. When multiple devices trigger a fire simultaneously, multiple independent evidence clusters that do not interfere with each other are generated. Each evidence cluster is separately sent to the S3 hierarchical weighted adjudication model to perform confidence calculation.

[0049] S3: Call the preset hierarchical weighted adjudication model to initially assign confidence values ​​to the multi-source heterogeneous signal association evidence cluster, obtain the basic confidence score, simultaneously detect the logical mutual exclusion state between each signal element within the multi-source heterogeneous signal association evidence cluster, perform compensation correction on the basic confidence score according to the conflict resolution rules, and output a four-level discretized confidence level. The process of invoking a pre-set hierarchical weighted adjudication model to initially assign confidence values ​​to the multi-source heterogeneous signal association evidence clusters and obtain a basic confidence score includes: S3.1: Extract a list of signal elements from the multi-source heterogeneous signal association evidence cluster, and classify each signal element into flame sign signals, electrical anomaly signals, and environmental disturbance signals according to the monitoring type; Furthermore, upon receiving the evidence cluster of multi-source heterogeneous signals, the system first reads all signal elements stored in the three major blocks within the evidence cluster. These three blocks include the flame sign subset block, the electrical anomaly subset block, and the environmental disturbance subset block. All signal elements are extracted and organized into an ordered list of signal elements. Each entry in the list is accompanied by a label indicating the original circuit type of the signal. Based on the circuit label, the three types of signals are accurately classified.

[0050] Furthermore, all signal elements extracted from the flame sign subset block of the multi-source heterogeneous signal correlation evidence cluster are uniformly classified as flame sign signals, including all temperature change rate signals and flame spectrum feature signals, which directly reflect the intuitive combustion characteristics such as open flame, high temperature, and smoke on site. They are assigned the highest level weight in the hierarchical weight model and are the core basis for determining the actual fire situation. All signal elements extracted from the electrical anomaly subset block are uniformly classified as electrical anomaly signals, including three types of power monitoring signals: residual current, voltage distortion rate, and arc pulse count. They represent pre-fire faults such as internal insulation damage, short circuit, and arc discharge in equipment and are assigned a medium level weight. All signal elements extracted from the environmental disturbance subset block are uniformly classified as environmental disturbance signals, including three types of environmental monitoring signals: gas concentration, air pressure change, and humidity compensation value. They are only used as auxiliary evidence data for the fire situation and are easily affected by external interference such as ventilation, temperature, humidity, and dust, and are assigned the lowest level weight.

[0051] Furthermore, after classification, separate lists of flame sign signals, electrical anomaly signals, and environmental disturbance signals are generated. These three lists are independent of each other and have no duplicate signal elements. Each list completely retains all original values, signal quality indicator codes, signal-to-noise ratio estimates, sampling timestamps, and other metadata for all signal elements in the corresponding category.

[0052] S3.2: For each signal element in each type of signal, calculate the corresponding quality weight coefficient by calling the preset signal quality evaluation function based on the signal quality indicator code and signal-to-noise ratio estimate carried by the signal element. Furthermore, in this embodiment, during the acquisition and packaging stage, the acquisition card synchronously calculates the signal transmission signal-to-noise ratio (SNR) value for each signal element and generates a standardized signal quality indicator code. The standardized signal quality indicator code segments and marks the signal transmission status, including four basic markers: no line interference, slight noise, signal attenuation, and momentary disconnection. The SNR estimate quantifies the ratio between effective signal data and transmission noise.

[0053] Furthermore, the pre-set signal quality evaluation function is embedded in the underlying computational logic of the hierarchical weighted adjudication model. The function's computational logic relies entirely on the signal quality indicator code segmentation benchmark and the signal-to-noise ratio (SNR) value range division, requiring no external input parameters. It uses the same set of quality evaluation logic for the three types of signals, only outputting standardized quality weight coefficients that are adapted to the various signal ranges. The coefficient values ​​are uniformly defined between zero and one. The better the signal quality and the higher the SNR, the closer the quality weight coefficient value is to 1. When the signal is attenuated or has noise interference, the coefficient value decreases synchronously, weakening the impact of distorted signals on the overall confidence level.

[0054] Furthermore, the process iterates through each signal element in the list of flame sign signals, sequentially reading the signal quality indicator code and signal-to-noise ratio estimate of the current signal, substituting them into the signal quality evaluation function to complete the calculation, and obtaining the unique quality weight coefficient for this flame sign signal. This is recorded and stored one by one. The process iterates through all signal elements in the list of electrical anomaly signals, repeating the same operation to calculate and store the quality weight coefficient for each power signal. The process iterates through all signal elements in the list of environmental disturbance signals, similarly calling the quality evaluation function to calculate the environmental signal quality weight coefficient for each signal. After the single signal quality weight calculation is completed for all three types of signals, the feature saliency score for each signal is calculated.

[0055] S3.3: For each signal element in each type of signal, calculate the corresponding feature significance score by calling the preset feature intensity mapping function based on the signal element value; The preset feature intensity mapping function sets corresponding feature thresholds and mapping rules for different types of signal elements; S3.3.1: For signal elements of the temperature change rate type, the corresponding feature significance score is obtained by linear interpolation based on the relative position of its value between the preset low and high thresholds of temperature change rate. Furthermore, the steps in S3.3.1 include: (1) Read the list of flame signs, extract individual signal elements in order of their storage order, read the signal type marker and real-time temperature change rate value built into the individual signal element, and generate a single temperature change original data cache; retrieve two sets of fixed threshold parameters pre-stored in the signal feature mapping parameter partition of the hierarchical weight adjudication model, read the fixed value of the low threshold of temperature change rate of 0.3℃ / s and the fixed value of the high threshold of temperature change rate of 2.5℃ / s, compare the two sets of threshold parameters with the real-time temperature change rate value in the single temperature change original data cache, and generate interval judgment identifier data; (2) When the real-time temperature change rate of the interval determination marker data is lower than the low threshold of temperature change rate of 0.3℃ / s, the standardized extremely low feature significance score is directly output. The score is fixed at 0.00 and the score storage precision retains two decimal places, generating a single temporary data of temperature change score. When the real-time temperature change rate of the interval determination marker data is greater than or equal to the low threshold of temperature change rate of 0.3℃ / s and less than or equal to the high threshold of temperature change rate of 2.5℃ / s, the processing unit starts the built-in linear interval matching operation of the preset feature intensity mapping function, using the low threshold of temperature change rate of 0.3℃ / s as the lower limit of the interval and the high threshold of temperature change rate as the upper limit of the interval. Using 2.5℃ / s as the upper limit of the interval, the relative proportion of the real-time temperature change rate within the complete interval formed by the two thresholds is calculated. The processing unit performs standardized numerical conversion based on the calculated relative proportion and outputs the corresponding feature significance score. The score range is strictly limited to 0.00 to 1.00, generating a single temporary temperature change score data. When the interval judgment indicator data marks the real-time temperature change rate as higher than the high threshold of 2.5℃ / s, the processing unit directly outputs the standardized highest level feature significance score, with a fixed value of 1.00, generating a single temporary temperature change score data. (3) Store the generated single temperature change score temporary data into the temporary score buffer according to the order of the current traversed signal elements. Then, traverse all temperature change rate signal elements in the flame sign signal list until all temperature change rate signal elements in the list have completed all calculations. Read all single temperature change score temporary data in the temporary score buffer and classify all temperature change corresponding scores into the generated flame sign signal score list according to the order of signal traversal. This completes the unified collection and storage of scores.

[0056] S3.3.2: For signal elements of flame spectrum feature value type, calculate their correlation coefficient with the preset reference spectrum template, and convert the correlation coefficient into the corresponding feature significance score; Furthermore, the steps in S3.3.2 include: (1) Read the list of flame sign signals, extract individual signal elements in sequence according to the storage order of the signal elements in the list, read the signal type mark and real-time spectral sampling sequence built into the individual signal element, and generate a single spectrum raw data cache; retrieve the standard flame spectrum reference template for electrical combustion in the power distribution room built into the standard template storage partition of the hierarchical weight adjudication model, and match the 256 standard spectral sampling points built into the template with the 256 real-time spectral sampling points in the single spectrum raw data cache one by one; start the correlation coefficient calculation logic built into the preset feature intensity mapping function, complete the waveform comparison between the real-time spectral sampling sequence and the standard flame spectrum reference template for each sampling point, and calculate the correlation coefficients corresponding to the two sets of spectral sequences; (2) Retrieve the correlation coefficient conversion mapping rule and fixed boundary parameter 0.2 built into the signal feature mapping parameter partition of the hierarchical weight adjudication model, substitute it into the calculated correlation coefficient, and complete the feature significance score conversion; when the correlation coefficient value is greater than or equal to 0.2, perform linear conversion according to the corresponding proportion of the correlation coefficient in the interval 0.2 to 1, and output the corresponding feature significance score, with the score range strictly limited to 0.00 to 1.00; when the correlation coefficient value is less than 0.2, the processing unit outputs a standardized feature significance score close to 0, with the minimum output score fixed at 0.00; after completion, a single spectrum score temporary data is generated; (3) Store the generated single spectrum score temporary data into the temporary score buffer according to the order of the current traversed signal elements; traverse all flame spectrum feature value signal elements in the flame sign signal list in turn until all flame spectrum feature value signal elements in the list have completed all operations; read all single spectrum score temporary data in the temporary score buffer, and classify all spectrum corresponding scores into the updated flame sign signal score list according to the order of signal traversal, and complete the unified collection and storage of scores.

[0057] S3.3.3: For signal elements of residual current value, voltage distortion rate and arc pulse count type, the corresponding feature significance score is obtained after amplitude limiting processing based on the ratio of their values ​​to their respective preset reference thresholds.

[0058] Furthermore, the steps in S3.3.3 include: (1) Read the list of electrical abnormality signals, extract a single independent signal element in the order of natural storage of signal elements in the list of electrical abnormality signals, and simultaneously read the signal type mark field and the corresponding real-time sampling value attached to the current single signal element. After reading, generate a single electrical raw data cache, which serves as the output data generated in the current stage. (2) Based on the signal type marker recorded in the current electrical raw data cache, retrieve the corresponding exclusive reference threshold parameter in the storage partition of the hierarchical weight adjudication model signal feature mapping parameter; if the current signal type marker corresponds to the residual current value signal element, retrieve the pre-calibrated residual current reference threshold of 50mA; if the current signal type marker corresponds to the voltage distortion rate signal element, retrieve the pre-calibrated voltage distortion rate reference threshold of 5%; if the current signal type marker corresponds to the arc pulse count signal element, retrieve the pre-calibrated arc pulse count reference threshold of 5 times; after retrieving the corresponding exclusive reference threshold, start the numerical ratio calculation logic inside the preset feature intensity mapping function, use the real-time sampled value carried by the electrical raw data cache as the numerator, and use the retrieved exclusive reference threshold as the denominator to complete the ratio conversion; (3) Retrieve two sets of pre-set amplitude limiting boundary parameters within the model, namely the lower limit of amplitude limiting with a value of zero and the upper limit of amplitude limiting with a value of one. Perform boundary comparison judgment on the dimensionless single ratio data. When the dimensionless single ratio data is less than the lower limit of amplitude limiting with a value of zero, directly correct the current ratio data to the fixed value of the lower limit of amplitude limiting with a value of zero. When the dimensionless single ratio data is greater than the upper limit of amplitude limiting with a value of one, directly correct the current ratio data to the fixed value of the upper limit of amplitude limiting with a value of one. When the dimensionless single ratio data is between the two sets of values ​​of the lower limit of amplitude limiting and the upper limit of amplitude limiting, do not perform the numerical correction operation, directly retain the original converted dimensionless single ratio data. After completing all boundary constraint correction operations, obtain the standardized ratio data after amplitude limiting processing. This standardized ratio data is directly used as the feature significance score corresponding to the current single signal element, and simultaneously generate temporary data of the single electrical score. (4) Store the generated single electrical score temporary data in the local temporary score cache area according to the order of the current traversed signal elements, and complete the full set of reading and conversion processing of the current single electrical signal element; sequentially and completely traverse all the remaining current value signal elements, voltage distortion rate signal elements, and arc pulse count signal elements stored in the electrical abnormality signal list until all three types of electrical signal elements in the list have completed the full set of calculation processing; (5) After all signal element traversal operations are completed, read all the temporary electrical score data stored in the local temporary score cache area, collect and organize them in the order of signal element traversal, and generate a complete and standardized list of electrical anomaly signal scores.

[0059] S3.4: The product of the quality weight coefficient and the feature significance score of each signal element in the same category is accumulated and then divided by the total number of signal elements in that category to obtain the confidence components of flame signs, electrical anomalies and environmental disturbances. Furthermore, component calculations are performed independently for the three types of signals. First, flame sign signals are processed: each signal in the flame sign signal list is traversed, and the quality weight coefficient of this signal is numerically superimposed with the corresponding feature significance score. The superposition results of all signals are summarized and accumulated to obtain the total weighted score of flame signs. The total number of current flame sign signals is counted, and the total weighted score is divided by the total number of signals. The average is then used to obtain the standardized flame sign confidence component. This component comprehensively considers the signal transmission quality and fire characteristic strength of all fire protection signals of this equipment to uniformly quantify the overall fire risk level in the fire protection dimension.

[0060] Furthermore, all power anomaly signals are traversed, and the results of the summation of the quality weight and feature significance of each individual signal are accumulated to obtain the total weighted electrical score. This score is then divided by the total number of power signals to output the average electrical anomaly confidence component, thus uniformly quantifying the overall fire precursor risk of equipment power failure. Finally, environmental disturbance signals are processed: all environmental monitoring signals are traversed, and the results of the summation of the quality weight and significance score of each individual signal are accumulated to obtain the total weighted environmental score. This score is then divided by the total number of environmental signals to output the average environmental disturbance confidence component, thus quantifying the risk level supported by environmental auxiliary dimensions.

[0061] Furthermore, the values ​​of the three confidence components are all standardized to the range of zero to one. The higher the component value, the stronger the fire risk in the corresponding dimension. The three components are stored independently and do not interfere with each other.

[0062] S3.5: Multiply each confidence component by the corresponding pre-set hierarchical weight coefficient in the hierarchical weighted adjudication model, then perform weighted summation and normalization to obtain the basic confidence score.

[0063] Furthermore, the three sets of fixed-level weight coefficients pre-stored in the internal hierarchical weight parameter storage partition of the hierarchical weight adjudication model are retrieved in sequence: the weight coefficient of flame sign level (0.65), the weight coefficient of electrical anomaly level (0.25), and the weight coefficient of environmental disturbance level (0.10). The weighted summation operation logic built into the hierarchical weight adjudication model is activated, and independent numerical multiplication operations are performed for each set of confidence components in the component cache data. Specifically, the confidence component of flame sign is multiplied by the value 0.65, the confidence component of electrical anomaly is multiplied by the value 0.25, and the confidence component of environmental disturbance is multiplied by the value 0.10. After all three sets of independent multiplication operations are completed, the three sets of independent product values ​​are all accumulated. After the accumulation calculation is completed, a weighted original total score is generated.

[0064] Furthermore, the normalization processing logic built into the hierarchical weighted adjudication model is activated. The model pre-stores a fixed normalization conversion benchmark, which is obtained by adding three sets of hierarchical weight coefficients with fixed values. The sum of the three sets of hierarchical weight coefficients (0.65, 0.25, and 0.10) is fixed at 1. The weighted original total score is converted to the sum of the weight coefficients. The conversion process uses the sum of the weight coefficients as the unified conversion denominator and the weighted original total score as the conversion numerator to complete the standardization and compression process. After the conversion, a standardized value in the interval [0,1] is obtained. This standardized value is the basic confidence score corresponding to the current multi-source signal. A batch of basic score temporary storage data is generated simultaneously. After all three sets of confidence components in a batch are multiplied, accumulated, and normalized, the generated basic score temporary storage data is completely stored in the local data cache area.

[0065] The output process for the four-level discretized confidence levels includes: S3.6: Extract the acquisition time tag sequence from the multi-source heterogeneous signal correlation evidence cluster; Furthermore, in S3.6, the global reference timestamps after unified alignment of all signal elements within the current evidence cluster are retrieved, and a complete acquisition time tag sequence is generated according to the sampling order. Each time tag in the sequence corresponds to the synchronous sampling time of signals from different loops of the same device. The tag precision is retained to the millisecond level, and the sequence completely records the acquisition time sequence of all signals within a certain time window.

[0066] S3.7: Determine whether there is a signal burst intrusion mutual exclusion state based on whether the difference between adjacent time tags is less than the preset burst intrusion time threshold; Furthermore, the steps in S3.7 include: (1) Retrieve two sets of fixed judgment parameters pre-stored in the storage partition of the mutual exclusion judgment parameter of the hierarchical weight adjudication model, namely a sudden intrusion time threshold of 20ms and a continuous judgment count standard of 3 sets; start the adjacent time difference calculation logic built into the hierarchical weight adjudication model, select two sets of time tags adjacent to each other in the sequence according to the order of the tags inside the collection time tag sequence, read the millisecond time values ​​recorded by each of the two sets of time tags, complete the difference calculation of the two sets of values, and generate a single set of adjacent time difference temporary data after the single calculation is completed; continuously loop the difference calculation operation until all adjacent tag combinations inside the collection time tag sequence have completed the difference calculation, and summarize to generate a complete set of adjacent difference cache data; (2) Retrieve the complete set of adjacent difference cache data, and perform numerical comparison operations group by group according to the order of the difference corresponding to the labels. Compare the temporary data of each group of adjacent time difference with the fixed value of 20ms sudden intrusion time threshold, and continuously count the cumulative number of difference combinations that are continuously less than the threshold standard. During the entire process of group comparison, if a group of adjacent time difference is greater than or equal to 20ms, the continuous cumulative group count is cleared to zero and the counting starts again. If the adjacent time difference is continuously less than 20ms, the continuous cumulative group count is continuously added upward. The real-time updated continuous cumulative group count is recorded synchronously. When the real-time cumulative group count reaches the pre-calibrated 3-group continuous judgment count standard, it is directly determined that there is a batch of instantaneous interference signal in the multi-source heterogeneous signal association evidence cluster corresponding to the current collection time label sequence, and a standardized mark representing the existence of signal sudden intrusion mutual exclusion state is generated. If after completely traversing all adjacent difference combinations, the real-time cumulative group count can never reach the 3-group continuous judgment count standard, a standardized mark representing the absence of signal sudden intrusion mutual exclusion state is generated. (3) Store the generated standardized mutual exclusion state tags completely into the temporary summary cache area of ​​mutual exclusion states.

[0067] S3.8: Determine whether there is a photoelectric signal homogeneity mutual exclusion state based on whether the flame spectrum feature value in the flame sign signal is in the infrared band and whether the arc pulse count in the electrical anomaly signal is not zero; It should be noted that a real combustion flame will simultaneously generate infrared spectral features and a continuous electric arc pulse. The two are synchronized in timing and have a unified source, so there is no logical conflict. However, high-power infrared heating equipment, infrared thermometers, and temporary welding maintenance inside the power distribution room will output high-value infrared flame spectrum signals separately, while there is no arc pulse count. The two types of signal features are logically disconnected, which is a typical interference conflict scenario and is defined as a photoelectric signal homogeneity mutual exclusion state. All flame spectrum feature values ​​within the evidence cluster are extracted to determine whether the values ​​fall within the infrared feature range. At the same time, the total number of arc pulse counts in the electrical anomaly subset is read. If the infrared spectral features are significant but the arc pulse count is zero, it is determined that there is a photoelectric signal homogeneity mutual exclusion state, and a corresponding mutual exclusion mark is generated and stored. If the infrared spectral signal and the arc pulse appear synchronously, there is no such mutual exclusion state mark.

[0068] S3.9: Determine whether there is a mutually exclusive airflow interference state based on whether the air pressure change in the environmental disturbance signal is greater than the preset air pressure change threshold. In this embodiment, the air pressure change threshold is set to 50Pa. Furthermore, when the real-time air pressure change is greater than 50 Pa and all conditions exceeding the threshold are met, a standardized marker representing the existence of airflow interference and mutual exclusion states is generated; if the comparison result does not meet the conditions exceeding the threshold, or if the conditions are not met, a standardized marker representing the absence of airflow interference and mutual exclusion states is generated, and the newly generated airflow interference and mutual exclusion state markers are completely stored in the mutual exclusion state temporary summary cache area.

[0069] S3.10: Summarize the identified mutual exclusion states and count the total number of mutual exclusion states; S3.11: Query the preset conflict resolution rule table based on the total number of mutual exclusion states to obtain the corresponding compensation correction coefficient; the compensation correction coefficient is negatively correlated with the total number of mutual exclusion states; Furthermore, the conflict resolution rule table is pre-configured manually. The table is divided into multiple ranges corresponding to the total number of mutually exclusive states. Each range is individually bound to a fixed compensation correction coefficient. The parameter configuration rule is set so that the larger the total number of mutually exclusive states, the smaller the corresponding compensation correction coefficient, maintaining a negative correlation. Specifically, the conflict resolution rule table has four pre-set fixed value ranges and corresponding compensation correction coefficients. All parameters are manually entered and stored, with the following settings: In the first tier, the total number of mutually exclusive states is 0, and the corresponding compensation correction coefficient is set to 1.0; The second level has a total number of mutually exclusive states of 1 to 2, with a corresponding compensation correction coefficient of 0.8. The third level has a total number of mutually exclusive states of 3 to 5, with a corresponding compensation correction coefficient of 0.5. In the fourth tier, the total number of mutually exclusive states is greater than 5, and the corresponding compensation correction coefficient is set to 0.2.

[0070] S3.12: Multiply the base confidence score by the compensation correction coefficient to obtain the corrected confidence score; S3.13: Based on the preset numerical range into which the corrected confidence score falls, map and output four levels: observation level, early warning level, action level, and emergency level.

[0071] Furthermore, when the adjusted confidence score falls within the first segment's range of 0 to 0.3, the matched level is designated as "observation level"; when the adjusted confidence score falls within the second segment's range of greater than 0.3 and less than or equal to 0.6, the matched level is designated as "warning level"; when the adjusted confidence score falls within the third segment's range of greater than 0.6 and less than or equal to 0.85, the matched level is designated as "action level"; and when the adjusted confidence score falls within the fourth segment's range of greater than 0.85 and up to 1, the matched level is designated as "emergency level".

[0072] S4: Based on the four levels of discretized confidence levels, match the four execution strategies to generate a linkage interlocking control word; The step of matching four execution strategies based on the four-level discretized confidence levels to generate a linkage interlocking control word includes: S4.1: Call the preset strategy mapping table and obtain the corresponding control bit combination rules according to the current four-level discretization confidence level; the control bit includes at least the status indication channel activation bit, the pre-suppression device start bit, the fire extinguishing main valve lockout release bit, the manual confirmation time window control bit, and the spray command trigger bit; Furthermore, the pre-set strategy mapping table is divided into two core fields: a level column and a control bit combination rule column. The level column sequentially records four discrete confidence levels: observation level, early warning level, action level, and emergency level. The control bit combination rule column fully records the four status definitions of five types of control bits corresponding to each level: activation, deactivation, timed opening, and bypass. The five types of control bits respectively manage different supporting hardware equipment in the power distribution room: the status indicator channel activation position manages the on-site audible and visual warning lights and the pop-up prompts on the local monitoring terminal; the pre-suppression device activation position manages the power distribution room ventilation fire damper and the equipment power supply disconnection pre-module; the fire extinguishing main valve lockout release position manages the mechanical lockout mechanism of the gas fire extinguishing main valve; the manual confirmation time window control position manages the reserved countdown window for maintenance personnel to manually confirm the fire extinguishing action on-site; and the spray command trigger position manages the fire extinguishing main valve opening and nozzle spraying action.

[0073] S4.2: When the confidence level is observation level, generate a first linkage interlocking control word that only sets the activation bit of the status indicator channel to the valid state; Furthermore, the confidence score was adjusted to fall within the lowest risk range, indicating that there were only slight environmental fluctuations and transient noise interference, with no real fire risk. This corresponds to the observation-level confidence level, and the observation-level exclusive control bit rules in the matching strategy mapping table are applied: only the status indicator channel activation bit is switched to the effective state, while the remaining pre-suppression device activation bit, fire extinguishing main valve lockout bit, manual confirmation time window control bit, and spray command trigger bit are all kept in an ineffective closed state.

[0074] Furthermore, a standardized first linkage interlocking control word code is generated based on the combination of control bit states. In the first linkage interlocking control word code, only the bit corresponding to the status indication channel is marked as a valid trigger identifier, while the bits corresponding to the other four types of control bits are all marked as invalid. The control word code only carries local status prompt instructions and does not have any fire extinguishing or equipment interlocking action instructions. After being issued, it only triggers the low-frequency intermittent flashing of the on-site warning light and the pop-up of a low-risk prompt window on the local monitoring terminal. It does not start any suppression or spraying equipment to avoid triggering equipment interlocking actions due to slight interference. The first linkage interlocking control word code is cached.

[0075] S4.3: When the confidence level is the warning level, generate a second linkage interlocking control word that sets the status indicator channel activation bit and the pre-suppression device start bit to the effective state, and sets the manual confirmation time window control bit to the open timed confirmation state. Furthermore, the confidence score was revised to a medium-low risk range, indicating the presence of minor fire precursors and a low probability of electrical ignition. This corresponds to the warning level confidence level, and the matching mapping table warning level control bit rules are implemented: the status indicator channel activation bit and the pre-suppression device activation bit are synchronously switched to effective; the fire extinguishing main valve lockout bit and the spray command trigger bit remain ineffective and closed; the manual confirmation time window control bit is switched to open timed confirmation state, and the system presets a fixed countdown time for manual confirmation. During the countdown period, maintenance personnel can go to the site to verify and manually terminate the fire extinguishing action.

[0076] Furthermore, based on this rule, a standardized second linkage interlocking control word code is generated. The status indication and pre-suppression control bits within the code are marked as valid, while the spray and main valve interlocking bits are invalid. The manual confirmation bit is marked with a timed opening indicator. After the control word is issued, multiple pre-warning actions are executed simultaneously: the on-site audible and visual warning lights flash continuously at high frequency, the monitoring terminal pushes the warning point and risk information in a pop-up window, automatically closes the ventilation fire damper of the corresponding zone in the power distribution room, cuts off the non-fire auxiliary power supply to the protective equipment, and at the same time starts a local countdown pop-up window, reserving a window period for manual intervention. The fire extinguishing main valve will not be unlocked or the spray will not be started until the countdown ends.

[0077] S4.4: When the confidence level is action level, generate a third linkage interlocking control word that sets the status indicator channel activation bit, the pre-suppression device start bit, the fire extinguishing main valve lockout bit, and the spray command trigger bit to the effective state. Furthermore, the confidence score was revised to the medium-high risk range, indicating that clear electrical fire characteristics have appeared and a real fire is highly likely. The corresponding action-level confidence level was matched with the action-level control bit rules in the mapping table: the status indicator channel activation bit, the pre-suppression device activation bit, the fire extinguishing main valve lockout release bit, and the spray command trigger bit were all switched to be effective; the manual confirmation time window control bit remained open for timed confirmation, and the short-term manual emergency stop permission was retained.

[0078] Furthermore, a standardized third-linkage interlocking control word code is generated, with all control bits related to the four types of actions marked as valid, and the manual confirmation bit maintaining a timed open indicator. After the control word is issued, the entire set of pre-interlocking + automatic spraying process is fully executed: continuous high-intensity audible and visual alarm, shutdown of all zone fire dampers, disconnection of main power supply to equipment, unlocking of the mechanical interlocking mechanism of the main fire extinguishing valve, and simultaneous opening of nozzles to spray extinguishing agent. At the same time, a short manual confirmation window is retained, during which maintenance personnel can manually terminate the spraying action if they verify on-site that there is no fire.

[0079] S4.5: When the confidence level is emergency, generate a fourth linkage interlocking control word that sets the status indicator channel activation position, the pre-suppression device start position, the fire extinguishing main valve lockout release position, and the spray command trigger position to the effective state, and sets the manual confirmation time window control position to the bypass state.

[0080] Furthermore, the confidence score was corrected to fall within the highest risk range, indicating that the multi-dimensional signals fully confirmed the intense fire. There was no buffer time for manual on-site verification, corresponding to the emergency level confidence level. The matching mapping table for emergency level control position rules showed that all four types of control positions—status indication, pre-suppression, main valve lockout, and spray trigger—were effective. The manual confirmation time window control position was forcibly switched to bypass status, directly skipping the manual confirmation countdown process. There was no manual abort buffer window, and the system immediately executed the complete fire extinguishing spray action.

[0081] Furthermore, a standardized fourth-linkage interlocking control word code is generated, with all action-related control bits being valid, and the manual confirmation bit marked with a bypass skip flag. After the control word is issued, there is no delay or buffer, and all interlocking and spraying actions are executed synchronously to minimize the fire response time. The fourth-linkage interlocking control word code is then cached and sent to S5.

[0082] S5: The linkage interlock control word is sent to the fire extinguishing actuator to start the fire extinguishing action. After the start, the original monitoring signal is re-acquired and the four-level discrete confidence level is updated. Based on the updated four-level discrete confidence level, the additional spray for reignition suppression or the system reset command after the flame is extinguished is triggered. At the same time, all original signal logs, control command flow records and on-site image data are archived synchronously to build an irreversible operation traceability chain.

[0083] Sending the interlocking control word to the fire extinguishing actuator to initiate the fire extinguishing action includes: S5.1: Analyze the state of each control bit in the linkage interlocking control word. When the injection command trigger bit is invalid, convert the linkage interlocking control word into a state update command and send it to the state indication channel driver and the pre-suppression device driver respectively. Furthermore, the standardized linkage interlocking control word code is received and output, and the bit identifiers corresponding to the five types of control bits inside the code are parsed segment by segment. The status identifier of the spray command trigger bit is read first. If the identifier is an invalid closed state, it means that the current level is observation level or warning level. There is no need to start the fire extinguishing agent spraying process. Only the status prompt and front-end equipment interlocking operation are performed.

[0084] Furthermore, the effective control bit information within the complete control word is extracted and converted into a standardized status update instruction data packet. This standardized status update instruction data packet contains two independent sub-instructions. The first sub-instruction corresponds to the status indication channel driver and includes the warning light flashing frequency, monitoring pop-up prompt content, and local audible and visual alarm volume parameters. The second sub-instruction corresponds to the pre-suppression device driver and includes fire damper closing instructions, equipment auxiliary power supply disconnection instructions, and ventilation fan shutdown instructions. Both types of sub-instructions are uniformly grouped into a low-priority data transmission queue and sequentially sent to the corresponding driver hardware according to the queue order. After receiving the instruction, the driver synchronously executes the status prompt and pre-fire interlock actions, without triggering any action on the main fire extinguishing valve or spray nozzles throughout the process. After the instruction is sent, the status update instruction flow log is synchronously recorded and stored in the local cache for archiving.

[0085] S5.2: When the spray command trigger bit is in an active state, the linkage interlock control word is converted into a spray execution command and a status update command; the spray execution command includes the fire extinguishing main valve opening duration and spray nozzle angle adjustment parameters; Furthermore, when parsing the control word bit identifier, if the spray command trigger bit flag is valid, it indicates that the current level is action level or emergency level, requiring simultaneous execution of two sets of actions: pre-interlock and fire extinguishing agent spraying. The module splits the control word to generate two types of independent instruction data packets: the first type is the status update instruction, which is similar to S5.1, controlling pre-equipment devices such as audible and visual prompts, fire dampers, and power cut-off; the second type is the dedicated spray execution instruction, which carries complete dynamic adjustment parameters in the data packet, including the continuous opening duration of the fire extinguishing main valve, the up, down, left, and right adjustment angle of the spray nozzles in each zone of the power distribution room, and the baseline value of the extinguishing agent distribution flow rate in each zone. The parameters are dynamically matched and generated based on the current equipment identity anchor point and the corrected confidence level. The main valve opening duration is longer and the nozzle coverage angle is larger for high-risk emergency levels, adapting to the full coverage fire extinguishing needs of intense combustion scenarios.

[0086] S5.3: The spray execution command is sent to the fire extinguishing main valve driver and the spray nozzle driver through the high-priority data queue, and the status update command is sent to the status indication channel driver and the pre-suppression device driver through the low-priority data queue, so that each driver responds to the corresponding command execution action.

[0087] Furthermore, the steps in S5.3 include: (1) Retrieve the hierarchical transmission partition of the internal instruction of the linkage interlocking control output module, and complete the partition storage operation of the two types of instructions according to the queue storage admission rules. First, write the injection execution instruction completely into the internal storage area of ​​the high priority data queue, and then write the status update instruction completely into the internal storage area of ​​the low priority data queue. After the two types of instructions have completed the storage writing of the corresponding queues, a complete queue storage cache is generated. The complete queue storage cache is used as the output content of the current link and is directly sent to the instruction issuance and scheduling logic built into the module. (2) Start the internal instruction dispatching logic of the linkage interlocking control output module. The dispatching logic arranges the dispatching timing according to the offline calibrated priority sorting parameters. The high priority data queue corresponding to the first priority first occupies the allocated 70% of the transmission bandwidth. The independent transmission link is opened to continuously send the spray execution instructions stored in the queue to the first group of hardware drivers. The transmission objects are fixed as the fire extinguishing main valve driver and the spray nozzle driver. The two types of drivers continuously receive the complete spray execution instructions transmitted by the link and read all the adjustment parameters in the data packet. Based on the parameter content, they complete the corresponding hardware actions. (3) During the synchronization period when the high-priority data queue starts transmission, the scheduling logic allocates 30% of the transmission bandwidth to the low-priority data queue corresponding to the second-level priority, and opens another completely independent transmission link to send the status update instructions stored in the queue to the second group of hardware drivers. The transmission targets are fixed as the status indication channel driver and the pre-suppression device driver. The two types of drivers continuously receive the complete status update instructions transmitted by the link and read all the early warning control parameters inside the data packet, and complete the corresponding hardware matching actions according to the parameter content. (4) The two types of independent transmission links operate synchronously throughout the process. The bandwidth allocation parameters are fixed according to the offline locking standard. The transmission link corresponding to the high priority queue will not occupy the bandwidth resources allocated to the low priority queue, and the transmission link corresponding to the low priority queue will not occupy the transmission bandwidth of the high priority queue. The data transmission of the two links will not interfere with each other. Various types of drivers only receive the exclusive instructions issued by the corresponding queue and will not read other instruction data packets that are not within their own recognition range. (5) When all injection execution instructions in the high priority data queue have completed the complete link transmission, and all status update instructions in the low priority data queue have completed the complete link transmission, all drivers send back hardware action execution receipts to the linkage interlocking control output module. All receipt records are uniformly collected and stored in the instruction issuance completion cache.

[0088] The system reset command triggered by the additional injection for reignition suppression or flame extinguishing based on the updated four-level discretized confidence level includes: S5.4: When the updated four-level discretization confidence level is observation level or warning level, it is determined that the flame has been extinguished and a system reset command is generated; the system reset command includes a reset position for indicating that the main fire extinguishing valve is closed, the pre-suppression device is cut off, and the status indication channel is reset; Furthermore, after the extinguishing agent spraying is initiated, the system continuously cycles through S1 to S3, re-acquiring the original monitoring signals of the three types of circuits in the power distribution room in real time, and synchronously updating the effective fire signal set, multi-source signal evidence cluster, corrected confidence score, and four-level discrete confidence level, continuously monitoring the changes in the fire situation on site after spraying; if the updated confidence level falls back to the observation level or warning level, it means that the on-site fire is completely extinguished, and the smoke, high temperature, and power anomalies have all returned to the safe baseline range. The system determines that the fire extinguishing is complete and automatically generates a standardized system reset command. The standardized system reset command carries multiple sets of reset control bits, corresponding to the main extinguishing valve closing reset position, the pre-suppression device power restoration and ventilation opening reset position, the audible and visual alarm device shutdown reset position, and the local monitoring pop-up window clearing reset position; the standardized system reset command is uniformly issued to all driver hardware, and the driver sequentially executes the main valve closing, fire damper opening, equipment auxiliary power restoration, audible and visual alarm shutdown, and monitoring prompt interface reset of the entire set of reset actions, and the equipment returns to normal duty monitoring status, while recording a complete log of this reset operation.

[0089] S5.5: When the updated four-level discretized confidence level is action level or emergency level, it is determined that there is a reignition state, and a reignition suppression additional injection command is generated; The additional injection command for reignition suppression includes the additional injection duration and the additional injection flow rate adjustment coefficient, which are determined as follows: S5.5.1: The additional injection duration in the reignition suppression additional injection command is determined according to the updated four-level discretized confidence level. When the updated confidence level is the action level, it is set as the first additional injection duration; when the updated confidence level is the emergency level, it is set as the second additional injection duration. The second additional injection duration is greater than the first additional injection duration. Furthermore, the system pre-sets two fixed additional spray duration benchmarks. For the action level with a medium risk of reignition, the first level with a shorter additional spray duration is matched; for the emergency level with an extremely high risk of reignition, the second level with a longer additional spray duration is matched. The second level has a duration longer than the first level throughout. In the case of severe reignition, the fire extinguishing agent spray suppression cycle is extended to avoid secondary spread of the fire. The duration parameter is directly written into the additional spray command data packet and sent to the main valve driver. The driver controls the main valve to continuously open the spray according to the duration parameter. After the set duration is reached, the main valve is automatically temporarily closed, and the signal is re-acquired to update the confidence level and determine the reignition status again.

[0090] S5.5.2: The additional injection flow rate adjustment coefficient is determined based on the temperature change rate value in the effective fire signal set for re-extraction, and the ratio of the temperature change rate value to the preset high threshold is calculated as the basic coefficient for flow rate adjustment. Furthermore, the steps in S5.5.2 include: (1) Read the real-time temperature change rate value stored in the flame sign subset inside the re-collected effective fire signal set, and integrate it to generate a real-time fire value cache after reading. (2) Retrieve the two sets of offline calibrated high threshold parameters within the partition of the hierarchical weighted adjudication model signal feature mapping parameter, retrieve the high threshold for ignition with a temperature change rate of 2.5℃ / s, and generate a high threshold reference cache after retrieving the two sets of threshold parameters; (3) Perform a single-dimensional ratio conversion logic on the real-time temperature change rate value, use the real-time temperature change rate value as the conversion numerator and the temperature change rate ignition high threshold of 2.5℃ / s as the conversion denominator to complete the conversion and obtain dimensionless ratio data. This value is the basic coefficient of flow regulation. Store the basic coefficient of flow regulation completely in the temporary cache of the basic coefficient of flow regulation.

[0091] S5.5.3: Compare the basic flow rate adjustment coefficient with the preset minimum flow rate coefficient, and take the larger value of the two as the additional spray flow rate adjustment coefficient to control the opening degree of the fire extinguishing main valve.

[0092] Furthermore, the steps in S5.5.3 include: (1) Retrieve the minimum flow coefficient completed offline calibration in the dedicated storage partition of the fire extinguishing spray parameters of the hierarchical weight adjudication model, read the standard parameter with a fixed value of 0.2, generate the minimum flow coefficient reference cache after reading, and integrate the minimum flow coefficient reference cache with the input flow adjustment basic coefficient temporary cache into a dual coefficient comparison cache set; (2) Based on the dual-value comparison and filtering logic, two sets of dimensionless values ​​are read simultaneously: the basic flow rate adjustment coefficient and the minimum flow rate coefficient of 0.2, and the complete comparison operation of the two sets of values ​​is completed. If the value of the basic flow rate adjustment coefficient is greater than 0.2, the dual-value comparison and filtering logic selects the basic flow rate adjustment coefficient as the output value. If the value of the basic flow rate adjustment coefficient is less than or equal to 0.2, the dual-value comparison and filtering logic selects the minimum flow rate coefficient of 0.2 as the output value. The single value obtained after comparison and filtering is the additional injection flow rate adjustment coefficient.

[0093] S5.6: Send the system reset command or the reignition suppression additional injection command to the corresponding actuator, and after executing the reignition suppression additional injection command, repeat the execution until the flame extinguishing determination condition is met and a system reset command is generated.

[0094] Furthermore, after issuing the additional injection command, the driver performs a secondary enhanced injection action. After the injection cycle ends, the system runs the signal acquisition and confidence update process from S1 to S3 again to re-determine the reignition status: if the updated level is still at the action or emergency level, the additional injection command is repeatedly generated to continue injection suppression; if the updated level falls back to the observation or warning level and meets the flame extinguishing determination conditions, the additional injection cycle is terminated, and a system reset command is issued to complete the equipment recovery.

[0095] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of the present invention without departing from the spirit and scope of the present invention. All of these variations are within the protection scope of the present invention.

Claims

1. A method for activating a fire extinguishing system based on multi-signal interlocking, characterized in that, include: The system collects the raw monitoring signals output from the fire detection circuit, power protection circuit, and environmental auxiliary monitoring circuit, performs physical isolation verification of the signal sources, and combines them to generate an effective fire signal set. Retrieve the pre-constructed protection object identity association table, and perform bidirectional binding between each signal element in the effective fire signal set and the identity code of the same protected device to generate a multi-source heterogeneous signal association evidence cluster; The preset hierarchical weighted adjudication model is invoked to initially assign confidence values ​​to the multi-source heterogeneous signal association evidence cluster, obtain a basic confidence score, simultaneously detect the logical mutual exclusion state between each signal element within the multi-source heterogeneous signal association evidence cluster, perform compensation correction on the basic confidence score according to the conflict resolution rules, and output a four-level discretized confidence level. Based on the four levels of discretized confidence levels, four execution strategies are matched to generate a linkage interlocking control word; The linkage interlock control word is sent to the fire extinguishing actuator to initiate the fire extinguishing action. After activation, the original monitoring signal is re-acquired and the four-level discretized confidence level is updated. Based on the updated four-level discretized confidence level, the additional spray for reignition suppression or the system reset command after the flame is extinguished is triggered. At the same time, all original signal logs, control command flow records and on-site image data are archived synchronously to construct an irreversible operation traceability chain.

2. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 1, characterized in that, The process of generating the effective fire signal set includes: Acquire the first raw monitoring signal output from the fire detection circuit; the first raw monitoring signal includes at least the temperature change rate value and the flame spectrum characteristic value; Acquire the second raw monitoring signal output by the power protection circuit; the second raw monitoring signal includes at least the residual current value, voltage distortion rate, and arc pulse count. Acquire the third raw monitoring signal output by the environmental auxiliary monitoring loop; the third raw monitoring signal includes at least the gas concentration value, the gas pressure change, and the humidity compensation value; Based on the unique hardware address of the signal acquisition card corresponding to each loop, the first original monitoring signal, the second original monitoring signal, and the third original monitoring signal are compared bit by bit with the pre-stored list of valid addresses, and valid signal channels with matching addresses are retained. Using the system's reference clock as a synchronization reference, the original monitoring signals output from the retained valid signal channels are timestamped and aligned, and valid fire signal sets are generated by combining them according to the signal source category.

3. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 1, characterized in that, The process of generating the multi-source heterogeneous signal correlation evidence cluster includes: Based on the physical source identifier carried by each signal element in the effective fire signal set, query the protected object identity association table to obtain the protected equipment identity code and spatial coordinate information corresponding to each signal element; Using the protected device identification code as the primary key for querying, all corresponding signal acquisition channel numbers are obtained from the protected object identity association table to form an identity anchor mapping table. The physical source identifier of each signal element in the effective fire signal set is matched with the signal acquisition channel number in the identity anchor mapping table. The matched signal elements are stored in the corresponding flame sign subset, electrical anomaly subset and environmental disturbance subset in the identity anchor mapping table according to their monitoring type. When the sum of the number of signal elements in the flame sign subset and the electrical anomaly subset is not less than the preset joint trigger threshold, the signal elements in all subsets are fused with the protected device identification code as the anchor point to generate a multi-source heterogeneous signal association evidence cluster.

4. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 1, characterized in that, The process of invoking a pre-set hierarchical weighted adjudication model to initially assign confidence values ​​to the multi-source heterogeneous signal association evidence clusters and obtain a basic confidence score includes: A list of signal elements is extracted from the multi-source heterogeneous signal correlation evidence cluster, and each signal element is classified into flame sign signals, electrical anomaly signals, and environmental disturbance signals according to the monitoring type. For each signal element in each type of signal, the corresponding quality weight coefficient is calculated by calling the preset signal quality evaluation function based on the signal quality indicator code and the estimated signal-to-noise ratio carried by the signal element. For each signal element in each type of signal, the corresponding feature significance score is calculated by calling the preset feature intensity mapping function based on the signal element value. The product of the quality weight coefficient and the feature significance score of each signal element in the same category is accumulated and then divided by the total number of signal elements in that category to obtain the confidence components of flame signs, electrical anomalies and environmental disturbances. Each confidence component is multiplied by the corresponding pre-set weight coefficient of the hierarchical weighted adjudication model, then weighted summed and normalized to obtain the basic confidence score.

5. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 4, characterized in that, The preset feature intensity mapping function sets corresponding feature thresholds and mapping rules for different types of signal elements; For signal elements of the temperature change rate type, the corresponding feature significance score is obtained by linear interpolation based on the relative position of its value between the preset low and high thresholds of temperature change rate. For signal elements of flame spectral feature value type, calculate their correlation coefficient with the preset reference spectral template, and convert the correlation coefficient into the corresponding feature significance score; For signal elements of residual current value, voltage distortion rate and arc pulse count type, the corresponding feature significance score is obtained after amplitude limiting processing based on the ratio of their values ​​to their respective preset reference thresholds.

6. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 1, characterized in that, The output process for the four-level discretized confidence levels includes: Extract the acquisition time tag sequence from the multi-source heterogeneous signal association evidence cluster, and determine whether there is a signal sudden intrusion mutual exclusion state based on whether the difference between adjacent time tags is less than a preset sudden intrusion time threshold. Based on whether the flame spectrum feature value in the flame sign type signal is in the infrared region and whether the arc pulse count in the electrical anomaly type signal is not zero, it is determined whether there is a photoelectric signal homogeneity mutual exclusion state. Based on whether the air pressure change in the environmental disturbance signal is greater than a preset air pressure drastic change threshold, it is determined whether there is a mutually exclusive airflow interference state. Summarize the identified mutual exclusion states and count the total number of mutual exclusion states; The corresponding compensation correction coefficient is obtained by querying the preset conflict resolution rule table based on the total number of mutual exclusion states; the compensation correction coefficient is negatively correlated with the total number of mutual exclusion states. Multiply the base confidence score by the compensation correction coefficient to obtain the corrected confidence score; Based on the preset numerical range into which the corrected confidence score falls, four levels are mapped and output: observation level, early warning level, action level, and emergency level.

7. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 1, characterized in that, The step of matching four execution strategies based on the four-level discretized confidence levels to generate a linkage interlocking control word includes: The preset strategy mapping table is invoked to obtain the corresponding control bit combination rules based on the current four-level discretization confidence level; the control bit includes at least the status indication channel activation bit, the pre-suppression device start bit, the fire extinguishing main valve lockout release bit, the manual confirmation time window control bit, and the spray command trigger bit. When the confidence level is observation level, a first linkage interlocking control word is generated that only sets the activation bit of the status indicator channel to the valid state. When the confidence level is the warning level, a second linkage interlocking control word is generated to set the status indicator channel activation bit and the pre-suppression device start bit to the effective state, and to set the manual confirmation time window control bit to the open timed confirmation state. When the confidence level is action level, a third linkage interlocking control word is generated that sets the status indicator channel activation bit, the pre-suppression device start bit, the fire extinguishing main valve lockout bit, and the spray command trigger bit to the effective state. When the confidence level is emergency, a fourth linkage interlocking control word is generated, which sets the status indicator channel activation position, the pre-suppression device start position, the fire extinguishing main valve lockout release position, and the spray command trigger position to the effective state, and sets the manual confirmation time window control position to the bypass state.

8. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 1, characterized in that, Sending the interlocking control word to the fire extinguishing actuator to initiate the fire extinguishing action includes: The state of each control bit in the linkage interlock control word is analyzed. When the injection command trigger bit is invalid, the linkage interlock control word is converted into a state update command and sent to the state indication channel driver and the front suppression device driver respectively. When the spray command trigger bit is in an active state, the linkage interlock control word is converted into a spray execution command and a status update command; the spray execution command includes the fire extinguishing main valve opening duration and spray nozzle angle adjustment parameters; The spray execution command is sent to the fire extinguishing main valve driver and the spray nozzle driver through a high-priority data queue, and the status update command is sent to the status indication channel driver and the pre-suppression device driver through a low-priority data queue, so that each driver responds to the corresponding command and executes the action.

9. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 1, characterized in that, The system reset command triggered by the additional injection for reignition suppression or flame extinguishing based on the updated four-level discretized confidence level includes: When the updated four-level discretized confidence level is observation level or warning level, it is determined that the flame has been extinguished and a system reset command is generated; the system reset command includes a reset position for instructing the main fire extinguishing valve to close, the pre-suppression device to be cut off, and the status indication channel to be reset; When the updated four-level discretized confidence level is action level or emergency level, it is determined that there is a reignition state, and a reignition suppression additional injection command is generated. The system reset command or the reignition suppression additional injection command is sent to the corresponding actuator, and after the reignition suppression additional injection command is executed, it is executed repeatedly until the flame extinguishing determination condition is met and a system reset command is generated.

10. The fire extinguishing system activation method based on multi-signal interlocking as described in claim 1, characterized in that, The additional injection command for reignition suppression includes the additional injection duration and the additional injection flow rate adjustment coefficient, which are determined as follows: The additional injection duration in the reignition suppression additional injection command is determined according to the updated four-level discretized confidence level. When the updated confidence level is the action level, it is set as the first additional injection duration, and when the updated confidence level is the emergency level, it is set as the second additional injection duration. The duration of the second additional injection is longer than the duration of the first additional injection; The additional injection flow rate adjustment coefficient is determined based on the concentrated temperature change rate value of the re-collected effective fire signal. The ratio of the temperature change rate value to the preset high threshold is calculated as the basic flow rate adjustment coefficient. The basic flow rate adjustment coefficient is compared with the preset minimum flow rate coefficient, and the larger of the two is taken as the additional injection flow rate adjustment coefficient to control the opening degree of the fire extinguishing main valve.

Images