Inspection method, system and device for fire alarm system

Abstract

The invention discloses an inspection method, system and device for a fire alarm system, relates to the technical field of fire safety, and solves the technical problems of low inspection efficiency and resource waste caused by the fact that a fixed period polling mode is generally adopted in the prior art, a large amount of redundant communication is generated, and system bandwidth and computing resources are occupied. A to-be-queried queue is constructed based on a state response result; generating an inspection result in an instruction sending mode based on the state type in the queue to be queried; based on the real-time load and the load change trend of the to-be-queried queue, an inspection cycle is generated in a self-adaptive mode, lightweight state response results of all fire-fighting components are rapidly obtained through the variable inspection cycle, then the to-be-queried queue is constructed only for the components with the state changed, and refined detailed state query is carried out. And finally, the next inspection period is dynamically and adaptively generated according to the load condition of the queue, so that the generation of redundant data is eradicated from a communication mechanism, and the problems of low inspection efficiency and resource waste are solved.

Description

Technical Field

[0001] This application belongs to the field of fire safety technology, specifically a method, system and device for inspecting a fire alarm system. Background Technology

[0002] As a core component of a building's fire safety system, the proper functioning of the fire alarm system directly impacts timely warnings, effective evacuation, and rapid response in the event of a fire. Regular inspections are crucial for ensuring the long-term stable and reliable operation of this system. They aim to identify potential faults or performance degradation by periodically and systematically checking the working status of detectors, alarm controllers, linkage devices, and related wiring, thereby eliminating safety hazards. Scientific and standardized inspections not only help improve the system's response accuracy and reliability but also play an irreplaceable role in preventing fire accidents and reducing casualties and property damage.

[0003] Existing technologies typically employ a fixed-period polling method, performing a full query regardless of whether the component status changes. This generates a large amount of redundant communication, consuming system bandwidth and computing resources, resulting in low inspection efficiency and resource waste. Therefore, the inspection methods for fire alarm systems still need further improvement. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems existing in the prior art; to this end, this application proposes a method, system and device for inspecting a fire alarm system, which solves the technical problem that the prior art usually adopts a fixed periodic polling method, which generates a large amount of redundant communication, occupies system bandwidth and computing resources, and leads to low inspection efficiency and waste of resources.

[0005] To achieve the above objectives, the first aspect of this application provides a method for inspecting a fire alarm system, comprising: The status response results of several fire protection component IDs are obtained based on the inspection cycle; the status response results include status change flags and no response status. A query queue is constructed based on the processing priority of the fire component ID according to the status response results; The inspection results are generated by sending instructions based on the status type in the queue to be queried. Inspection cycles are adaptively generated based on the real-time load and load change trends of the queue to be queried.

[0006] This application, through the above steps, adopts a variable inspection cycle to quickly obtain the lightweight status response of all fire protection components; then, a query queue is constructed for components whose status has changed, and a refined detailed status query is performed; finally, based on the load of the query queue, the next inspection cycle is dynamically and adaptively adjusted, effectively eliminating the generation of redundant data at the communication level, significantly improving inspection efficiency, and avoiding resource waste.

[0007] Furthermore, the step of constructing a query queue based on the processing priority of fire component IDs according to the status response results includes: Extract the status response results of several fire protection component IDs; the status response results include status change flags and no response status. An initial queue is generated based on the state change flag and the fault type; the initial queue includes an alarm queue and a fault queue. Generate the processing priority corresponding to the fire protection component ID based on the initial queue and status response results; The fire component IDs are sorted in descending order according to processing priority to obtain the queue of queries.

[0008] Furthermore, the step of generating an initial queue based on the state change flag bit according to the fault type includes: Extract the status change flag bit corresponding to the fire protection component ID; the status change flag bit is a two-bit binary value, including a fault change flag bit, an alarm change flag bit, and a status-free flag bit; Set the status type of the fire protection component ID, whose status change flag is the fault change flag, to fault status; Set the status type of the fire protection component ID, whose status change flag is the alarm change flag, to alarm status; Set the status type of the fire protection component ID, where the status change flag is set to "no status", to "no status". A fault queue is formed by combining several fire-fighting component IDs with a status type of fault. A queue is formed by combining several fire-fighting component IDs whose status type is alarm. The initial queue is determined based on the fault queue and the alarm queue.

[0009] Furthermore, the process of generating the processing priority corresponding to the fire component ID based on the initial queue and status response results includes: Extract the fire component IDs with a no-response status from the status response results, and set the processing priority corresponding to the fire component IDs to the maximum value; Extract the status types corresponding to several fire-fighting component IDs from the initial queue; the status types include fault types and alarm types. The status anomaly level corresponding to the fire protection component ID is extracted from the status anomaly level table based on the fault type and alarm type; the status anomaly level table is set by experts according to the degree of harm caused by the fire alarm system based on the fault type and alarm type. Obtain the component ID wiring diagram of the fire alarm system; Extract the fire-fighting component IDs and their corresponding status types from the initial queue; The component ID wiring diagram and the fire protection component IDs and their corresponding status types are integrated into impact analysis data; The impact analysis data is input into the impact assessment model to obtain the impact range level corresponding to several fire protection component IDs; the impact assessment model is constructed through an artificial intelligence model and is used to evaluate the degree of diffusion impact of the state type of fire protection component IDs and the comprehensive result of the diffusion impact range between fire protection component IDs; The historical number of times the fire protection component ID is acquired within a time window; the number of times the fire protection component ID has been in alarm or fault state. Through formula Calculate the status frequency score of fire protection component ID Where i represents the ID of the fire-fighting component in the initial queue, and min() represents the minimum value operation. This represents the number of historical states corresponding to the i-th fire protection component ID. This is represented as the threshold for the number of states corresponding to the i-th fire protection component ID; The processing priority corresponding to the fire component ID is obtained by weighting and summing the status anomaly level, impact range level, and status frequency scores corresponding to the fire component ID using weighting coefficients.

[0010] Furthermore, the impact assessment model is constructed using an artificial intelligence model, including: Obtain historical impact analysis data and its corresponding fire protection component IDs and historical impact range levels; Several historical impact analysis data and their corresponding fire protection component IDs and historical impact range levels are divided into training data, validation data, and test data; the training data, validation data, and test data are preprocessed to obtain the training set, validation set, and test set; Choose an artificial intelligence model as the base model; The base model is trained on the training set, and the learning rate and hyperparameters are adjusted on the validation set to obtain the pre-trained model. By validating the pre-trained model on the test set, we finally obtained an impact assessment model that takes impact analysis data as input and outputs the impact range level corresponding to several fire protection component IDs.

[0011] Furthermore, the weighting coefficients are obtained in the following ways: Obtain the current state feature vector of the fire alarm system; the state feature vector includes a system state distribution feature vector and a system environment feature vector; the system state distribution feature vector refers to the statistical characteristics of all fire component IDs in the current initial queue, including the proportion of severely abnormal components, the proportion of components with high impact range, the proportion of high state frequency, and the total number of fire components; the system environment feature vector includes working mode labels and working time feature labels; The state feature vector is input into the weight decision model to obtain the weight coefficients corresponding to the state anomaly level, the impact range level, and the state frequency score; the weight decision model is constructed through reinforcement learning and is used to generate the optimal weight coefficients for calculating the processing priority in the current fire alarm system. The weighted decision model is constructed through reinforcement learning, including: Acquire a number of training data; the training data includes the state feature vector S(t) at time t, the weight coefficient set QX(t), the processing effect reward value R(t), and the new state feature vector S(t+1) after state processing; t represents the time number; the weight coefficient set refers to the set of weight coefficients corresponding to the state anomaly level, influence range level, and state frequency score at time t; the new state feature vector is data with the same structure as the state feature vector but at different times; By using reinforcement learning algorithms and aiming to maximize the reward value of long-term processing effect, the agent model is trained to obtain a weighted decision model. The reward function R of the weighted decision model is a weighted sum of risk reward, efficiency reward, and resource reward. Wherein, the reward function R satisfies: ; This is represented as a risk reward, and as the change in the anomaly level of the initial queue before and after status processing. This is represented as an efficiency reward, and as the change in the number of fire-fighting components in the initial queue before and after state processing. This is represented as a resource reward; , and This is represented as the reward weight.

[0012] This application rapidly constructs alarm and fault queues as initial processing queues by parsing state change flags, and assigns the highest priority to unresponsive components to ensure immediate response. Based on this, a multi-factor weighted priority calculation model is introduced, comprehensively considering the state anomaly level, the impact range level assessed by an artificial intelligence model, and the state change frequency score to dynamically generate the processing priority of each component. A weighted decision model constructed through reinforcement learning can adaptively adjust the weight coefficients of various factors, thereby achieving an optimal balance between risk control, processing efficiency, and resource consumption. This not only ensures priority response to alarm events and critical faults, significantly improving the safety and real-time performance of the fire alarm system, but also effectively reduces redundant processing and greatly improves inspection efficiency through intelligent priority scheduling. Simultaneously, it enables the fire alarm system to possess continuous self-learning capabilities, flexibly adapting to different operating environments and conditions, and achieving dynamic optimization and long-term synergy of performance and resource utilization.

[0013] Furthermore, the step of generating inspection results based on the status type in the queue to be queried using an instruction sending method includes: Extract the queue to be queried; Extract the fire component IDs and their corresponding status types from the queue to be queried in sequence; When the status type is no response, obtain the most recent successful communication time point and the communication signal strength of the most recent several times for the fire component ID; use the status type, successful communication time point and the communication signal strength of the several times as the inspection result of the fire component ID, and remove the fire component ID and its corresponding status type from the query queue; Otherwise, the sending instruction is determined based on the fire protection component ID and its corresponding status type; the sending instruction includes a parameter acquisition instruction and a status query instruction; The inspection results are retrieved from the fire protection component ID based on the sent command.

[0014] Furthermore, the adaptive generation of the inspection cycle based on the real-time load and load change trend of the queue to be queried includes: Get the queue to be queried; Extract the queue length Q corresponding to the queue to be queried; Calculate the inspection cycle using a formula. The formula satisfies: ;in, This is represented as the baseline inspection cycle; Represented as a queue load function, it satisfies: ; This represents the current queue length. This represents the preset maximum queue capacity, and b is a constant, where b>0; Represented as a queue load trend function, it satisfies: ; Represented as the forgetting factor, ∈(0,1); This indicates the previous inspection cycle. value, This represents the queue length from the previous inspection cycle; and These are represented as queue load weight and queue trend weight, respectively. and ∈(0,1); Represented as a random fluctuation factor, it introduces small-amplitude random variations to prevent cycle-locking effects, satisfying: rand() represents a random function. This represents the maximum value of the random fluctuation factor.

[0015] This application introduces a multi-factor dynamic calculation formula to adaptively generate inspection cycles. It utilizes a hyperbolic tangent function to non-linearly map queue length to a load factor, ensuring smooth response. An exponentially weighted moving average is used to calculate the load trend factor, enabling the fire alarm system to perceive changes in busy / idle conditions. Furthermore, a random fluctuation factor breaks the synchronization lock effect, allowing the fire alarm system to intelligently accelerate or decelerate inspections based on real-time load. This automatically improves real-time response when queues are congested and reduces resource consumption when the system is idle. Simultaneously, the memory effect and random fluctuation achieve an optimal balance between response speed and system stability, completely resolving the resource waste and response delay problems of fixed-cycle inspection mechanisms.

[0016] A second aspect of the present invention provides an inspection system for a fire alarm system, comprising: a data acquisition module and a data analysis module; the data acquisition module and the data analysis module are connected together; The data acquisition module acquires the status response results of several fire-fighting component IDs through data acquisition equipment. The data analysis module includes a queue construction unit, a result generation unit, and a periodic feedback unit; The queue construction unit: constructs a queue to be queried based on the processing priority of the fire component ID according to the status response result; The result generation unit generates inspection results based on the status type in the queue to be queried by sending instructions. The periodic feedback unit adaptively generates the inspection cycle based on the real-time load and load change trend of the queue to be queried.

[0017] Another aspect of the present invention provides an inspection device for a fire alarm system, comprising: a processor and a storage medium; the storage medium includes instructions, and the processor is configured to execute the instructions to implement the method described in the first aspect and any possible implementation thereof; the inspection device for the fire alarm system may be an electronic device or a chip in an electronic device.

[0018] Compared with the prior art, the beneficial effects of this application are: 1. This application constructs a query queue based on the processing priority of fire component IDs according to the status response results; generates inspection results by sending instructions based on the status types in the query queue; and adaptively generates an inspection cycle based on the real-time load and load change trend of the query queue. With a variable inspection cycle, it quickly obtains lightweight status response results of all fire components, and then constructs a query queue only for components whose status has changed and performs refined and detailed status queries. Finally, it dynamically and adaptively generates the next inspection cycle according to the load of the queue. This eliminates the generation of redundant data from the communication mechanism and solves the problems of low inspection efficiency and resource waste.

[0019] 2. This application rapidly constructs alarm and fault queues as initial queues by parsing state change flags, and sets the highest priority for unresponsive components to ensure immediate processing. Based on this, a multi-factor weighted priority calculation model is introduced, comprehensively considering the state anomaly level, the impact range level assessed by an artificial intelligence model, and the state frequency score to dynamically generate the processing priority for each component. A weighted decision model constructed through reinforcement learning adaptively adjusts the weight coefficients to optimize the balance between risk, efficiency, and resource consumption. This not only ensures priority response to alarms and critical faults, improving the safety and real-time performance of the fire alarm system, but also reduces redundant processing and improves inspection efficiency through intelligent priority scheduling. Simultaneously, it enables the fire alarm system to have self-learning capabilities, adapting to different operating states and achieving an optimal balance between resource consumption and performance. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of an inspection method for a fire alarm system according to this application; Figure 2 This is a schematic diagram of the inspection system principle of a fire alarm system according to this application. Detailed Implementation

[0022] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0023] Please see Figure 1 The first aspect of this application provides a method for inspecting a fire alarm system, including: The status response results of several fire protection component IDs are obtained based on the inspection cycle; the status response results include status change flags and no response status. A query queue is constructed based on the processing priority of the fire component ID according to the status response results; The inspection results are generated by sending instructions based on the status type in the queue to be queried. Inspection cycles are adaptively generated based on the real-time load and load change trends of the queue to be queried.

[0024] In this embodiment, a query queue is constructed based on the processing priority of the fire component ID according to the status response result, including: Extract the status response results of several fire protection component IDs; the status response results include status change flags and no response status. An initial queue is generated based on the state change flag and the fault type; the initial queue includes an alarm queue and a fault queue. Generate the processing priority corresponding to the fire protection component ID based on the initial queue and status response results; The fire component IDs are sorted in descending order according to processing priority to obtain the queue of queries.

[0025] In this embodiment, under normal system conditions without any abnormalities, the inspection only requires millisecond-level flag bit screening, resulting in extremely simple communication data volume and an adaptively extended inspection cycle, minimizing system load and resource consumption. Even when there are state changes, the system only performs precise queries on a few components, avoiding the time and resource waste caused by traversing all components in the traditional method, thereby achieving a qualitative leap in inspection efficiency.

[0026] In this embodiment, the initial queue is generated based on the state change flag bit according to the fault type, including: Extract the status change flag bit corresponding to the fire protection component ID; the status change flag bit is a two-bit binary value, including a fault change flag bit, an alarm change flag bit, and a status-free flag bit; in this embodiment, the flag bits corresponding to the fault change flag bit, the alarm change flag bit, and the status-free flag bit are set to 01, 10, and 00, respectively. Set the status type of the fire protection component ID, whose status change flag is the fault change flag, to fault status; Set the status type of the fire protection component ID, whose status change flag is the alarm change flag, to alarm status; Set the status type of the fire protection component ID, where the status change flag is set to "no status", to "no status". A fault queue is formed by combining several fire-fighting component IDs with a status type of fault. A queue is formed by combining several fire-fighting component IDs whose status type is alarm. The initial queue is determined based on the fault queue and the alarm queue.

[0027] In this embodiment, the processing priority for generating the fire component ID based on the initial queue and status response results includes: Extract the fire component IDs with a no-response status from the status response results, and set the processing priority corresponding to the fire component ID to the maximum value; in this embodiment, the maximum range of processing priority is 1, so the processing priority corresponding to the fire component ID with a no-response status is set to 1; Extract the status types corresponding to several fire protection component IDs in the initial queue; the status types include fault types and alarm types. The status anomaly level corresponding to the fire alarm component ID is extracted from the status anomaly level table based on the fault type and alarm type. The status anomaly level table is set by experts according to the degree of harm caused by the fire alarm system based on the fault type and alarm type. In this embodiment, the fault types include power failure and sensor failure, etc.; the alarm types include fire alarm, early warning, and monitoring alarm, etc.; the status anomaly level in this embodiment is the result of quantifying the degree of harm caused by the fault type and alarm type to the fire alarm system. In this embodiment, the status anomaly levels corresponding to fire alarm, early warning, and monitoring alarm are set to 1, 0.8, and 0.6, respectively; and the status anomaly levels corresponding to power failure and sensor failure are set to 0.7 and 0.5, respectively. Obtain the component ID wiring diagram of the fire alarm system; Extract the fire-fighting component IDs and their corresponding status types from the initial queue; The component ID wiring diagram and the fire protection component IDs and their corresponding status types are integrated into impact analysis data; The impact analysis data is input into the impact assessment model to obtain the impact range level corresponding to several fire protection component IDs; the impact assessment model is constructed through an artificial intelligence model to evaluate the degree of diffusion impact of the state type of fire protection component IDs and the comprehensive result of the diffusion impact range between fire protection component IDs; The historical number of times the fire component ID is obtained within the time window is specified. In this embodiment, the size of the time window is set based on experience, and in this embodiment, the time window is set to 30 days. The number of states refers to the total number of times the fire component ID has an alarm state and a fault state. Through formula Calculate the status frequency score of fire protection component ID Where i represents the ID of the fire-fighting component in the initial queue, and min() represents the minimum value operation. This represents the number of historical states corresponding to the i-th fire protection component ID. This represents the state frequency threshold corresponding to the i-th fire component ID. In this embodiment, considering the significant differences between fire component IDs, different state frequency thresholds are set for different fire component IDs to calculate the state frequency score. The state frequency threshold is set by experts based on experience. The processing priority corresponding to the fire component ID is obtained by weighting and summing the status anomaly level, impact range level, and status frequency scores corresponding to the fire component ID using weighting coefficients.

[0028] The impact assessment model in this embodiment is constructed using an artificial intelligence model, including: Obtain historical impact analysis data and its corresponding fire protection component IDs and historical impact range levels; Several historical impact analysis data and their corresponding fire protection component IDs and historical impact range levels are divided into training data, validation data, and test data; the training data, validation data, and test data are preprocessed to obtain training sets, validation sets, and test sets; the ratio between the training set, test set, and validation set is 7:2:1; Choose an artificial intelligence model as the base model; artificial intelligence models include DeepSeek models, etc. The base model is trained on the training set, and the learning rate and hyperparameters are adjusted on the validation set to obtain the pre-trained model. By validating the pre-trained model on the test set, we finally obtained an impact assessment model that takes impact analysis data as input and outputs the impact range level corresponding to several fire protection component IDs.

[0029] The weighting coefficients in this embodiment are obtained in the following ways: Obtain the current state feature vector of the fire alarm system; the state feature vector includes the system state distribution feature vector and the system environment feature vector; the system state distribution feature vector refers to the statistical characteristics of all fire component IDs in the current initial queue, including the proportion of severely abnormal components, the proportion of high-impact-range components, the proportion of high-state-frequency components, and the total number of fire components; in this embodiment, the proportion of severely abnormal components refers to the proportion of fire components with a state abnormality level greater than or equal to 0.7; the proportion of high-impact-range components refers to the proportion of fire components with an impact range level greater than or equal to 0.6; the proportion of high-state-frequency components refers to the proportion of fire components with a state frequency score greater than or equal to 0.7; the system environment feature vector includes the working mode label and the working time feature label; in this embodiment, the working mode label includes daytime mode and nighttime mode; the working time feature label includes weekdays, weekends, and holidays; The state feature vector is input into the weight decision model to obtain the weight coefficients corresponding to the state anomaly level, the impact range level, and the state frequency score; the weight decision model is constructed through reinforcement learning to generate the optimal weight coefficients for calculating the processing priority in the current fire alarm system; The weighted decision model is constructed using reinforcement learning, including: Acquire some training data; the training data includes the state feature vector S(t) at time t, the weight coefficient set QX(t), the processing effect reward value R(t), and the new state feature vector S(t+1) after state processing; t represents the time number; the weight coefficient set refers to the set of weight coefficients corresponding to the state anomaly level, influence range level, and state frequency score at time t; the new state feature vector is data with the same structure as the state feature vector but at different times; in this embodiment, state processing refers to the operation of fault and alarm processing on the initial queue in the current fire protection system; A weighted decision model is obtained by training the agent model using a reinforcement learning algorithm with the objective of maximizing the long-term processing effect reward value. In this embodiment, the agent model is a Q-network model. The objective of maximizing the long-term processing effect reward value in this embodiment can be expressed as: ; The reward function R of the weighted decision model is a weighted sum of risk reward, efficiency reward, and resource reward; whereby the reward function R satisfies: ; The risk reward is represented as the change in the anomaly level of the initial queue before and after state processing. In this embodiment, the risk reward can be represented as... n and m both represent the numbers corresponding to the fire component IDs in the initial queue. n is the number corresponding to the fire component IDs in the initial queue before status processing, m is the number corresponding to the fire component IDs in the initial queue after status processing, and ZYD represents the status abnormality level corresponding to the fire component ID. This is represented as an efficiency reward, specifically the change in the number of fire-fighting components in the initial queue before and after state processing. In this embodiment, the efficiency reward can be represented as... N and M represent the total number of fire-fighting components in the initial queue before and after state processing, respectively. Represented as resource reward, in this embodiment, the resource reward can be represented as... γ represents the penalty coefficient, γ>0, and the specific value is set according to experience. In this embodiment, γ is set to 0.2; MS represents the number of commands sent when the initial queue performs state processing, PMS represents the average command length of the sent commands, and DTL represents the theoretical maximum communication volume. , and Represented as reward weight, this embodiment considers resource rewards as a way to penalize excessive resource consumption, therefore <0; In this embodiment, the relationship between reward weights can be expressed as: The specific values ​​are set based on experience; in this embodiment, they will be... , and They were set to 0.7, 0.25, and -0.05 respectively.

[0030] In this embodiment, generating inspection results based on the status type in the queue to be queried via instruction sending includes: Extract the queue to be queried; Extract the fire component IDs and their corresponding status types from the queue to be queried in sequence; When the status type is no response, obtain the most recent successful communication time point and the communication signal strength of the most recent several times for the fire component ID; use the status type, successful communication time point and the communication signal strength of the several times as the inspection result of the fire component ID, and remove the fire component ID and its corresponding status type from the query queue; Otherwise, the sending instruction is determined based on the fire protection component ID and its corresponding status type; the sending instruction includes a parameter acquisition instruction and a status query instruction; in this embodiment, when the status type corresponding to the fire protection component ID is fire alarm, the sending instruction includes a parameter acquisition instruction, specifically a "get detailed fire alarm status" instruction; when the status type corresponding to the fire protection component ID is sensor fault, the sending instruction includes a parameter acquisition instruction and a status query instruction, specifically a "get detailed fault information" instruction, and a status query instruction specifically a "query sensor sub-status" instruction; The inspection results are obtained from the fire protection component ID based on the sent command. In this embodiment, the inspection results are composed of the parameter results returned by the controller through sending the command and the fire protection component ID checking its own status according to the sent command. When the status type corresponding to the fire protection component ID is fire alarm, the inspection results returned by the fire protection component ID include alarm level, sensor value, its own status and trigger timestamp. When the status type corresponding to the fire protection component ID is warning, the inspection results returned by the fire protection component ID include warning type, sensor value and sensor historical data.

[0031] In this embodiment, the adaptive generation of the inspection cycle based on the real-time load and load change trend of the queue to be queried includes: Get the queue to be queried; Extract the queue length Q corresponding to the queue to be queried; Calculate the inspection cycle using a formula. The formula satisfies: ;in, This is represented as the baseline inspection cycle. The specific value is set based on experience. In this embodiment, it is set to 2000ms. Represented as a queue load function, it satisfies: ; This represents the current queue length. This represents the preset maximum queue capacity, and b represents a constant, where b>0. The specific value is set based on experience; in this embodiment, b is set to 0.5. Represented as a queue load trend function, it satisfies: ; Represented as the forgetting factor, ∈(0,1); The specific value is set based on experience, and in this embodiment it is set to 0.7; This indicates the previous inspection cycle. value, This represents the queue length from the previous inspection cycle; and These are represented as queue load weight and queue trend weight, respectively. and ∈(0, 1); the specific values ​​are set based on experience, and in this embodiment, they are respectively... and Set to 0.7 and 0.3; Represented as a random fluctuation factor, it introduces small-amplitude random variations to prevent cycle-locking effects, satisfying: rand() represents a random function. Represented as the maximum value of the random fluctuation factor, in this embodiment, it will be... Set to 0.1; In this embodiment, the inspection cycle required for the next inspection is calculated based on the data obtained in the current inspection phase, that is, the inspection cycle calculated in the current inspection is used for the next inspection.

[0032] This embodiment proposes a multi-factor dynamic calculation formula for adaptively generating inspection cycles, significantly improving the response intelligence and resource utilization efficiency of fire alarm systems. Specifically, firstly, the hyperbolic tangent function is used to nonlinearly map the queue length into a load factor, ensuring that the system can achieve a smooth and stable response under different load levels. Secondly, an exponentially weighted moving average algorithm is introduced to calculate the load trend factor, enabling the system to proactively perceive the current load's upward or downward trend and thus actively adjust the inspection rhythm. In addition, by introducing a random fluctuation factor, the synchronization lock effect that may occur between multiple devices is effectively broken, avoiding performance bottlenecks caused by periodic resonance. Through the above mechanisms, the system can intelligently adjust the inspection frequency according to the real-time load: automatically accelerating the inspection speed when the queue is backlogged to improve real-time response; and appropriately reducing the inspection intensity when the system is idle to reduce unnecessary resource consumption, achieving the optimal balance between response speed and operational stability. This fundamentally overcomes the resource waste and response delay problems caused by the traditional fixed-cycle inspection mechanism, providing a highly efficient, robust, and adaptive inspection strategy for fire alarm systems.

[0033] Please see Figure 2 A second aspect of this application provides an inspection system for a fire alarm system, comprising: a data acquisition module and a data analysis module; the data acquisition module and the data analysis module are electrically and / or communicatively connected. Data acquisition module: Acquires the status response results of several fire protection component IDs through data acquisition equipment; The data analysis module includes a queue construction unit, a result generation unit, and a periodic feedback unit; Queue construction unit: Construct a queue to be queried based on the processing priority of fire component ID according to the status response results; Result generation unit: Generates inspection results based on the status type in the queue to be queried by sending instructions; Periodic feedback unit: Adaptively generates inspection cycles based on the real-time load and load change trends of the queue to be queried.

[0034] Another embodiment of this application provides an inspection device for a fire alarm system, comprising: a processor and a storage medium; the storage medium includes instructions, and the processor is configured to execute the instructions to implement the method described in the first aspect embodiment and any possible implementation thereof; an inspection device for a fire alarm system may be an electronic device or a chip in an electronic device.

[0035] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0036] The working principle of this application is as follows: Status response results of several fire protection component IDs are obtained based on an inspection cycle; a query queue is constructed based on the processing priority of the fire protection component IDs according to the status response results; inspection results are generated by sending instructions based on the status types in the query queue; the inspection cycle is adaptively generated based on the real-time load and load change trend of the query queue. With a variable inspection cycle, lightweight status response results of all fire protection components are quickly obtained. Then, a query queue is constructed only for components whose status has changed, and a refined, detailed status query is performed. Finally, the next inspection cycle is dynamically and adaptively generated based on the load of the queue. This eliminates the generation of redundant data from the communication mechanism, solving the problems of low inspection efficiency and resource waste. It avoids the problems of low inspection efficiency and resource waste caused by the fixed-cycle polling method commonly used in existing technologies, which generates a large amount of redundant communication, occupies system bandwidth and computing resources, and leads to low inspection efficiency.

[0037] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A method of inspecting a fire alarm system, characterized in that The method comprises the following steps: acquiring state response results of a plurality of fire-fighting component IDs based on a patrol cycle; the state response results comprise a state change flag bit and a non-response state; constructing a to-be-queried queue based on the state response results and the processing priority of the fire-fighting component IDs; generating a patrol result based on the state type in the to-be-queried queue and the instruction sending mode; self-adaptively generating a patrol cycle based on the real-time load and the load change trend of the to-be-queried queue.

2. The method of claim 1, wherein, The method of constructing the to-be-queried queue based on the state response results and the processing priority of the fire-fighting component IDs comprises the following steps: extracting the state response results of a plurality of fire-fighting component IDs; dividing and generating an initial queue based on the state change flag bit according to the fault type; the initial queue comprises an alarm queue and a fault queue; generating the processing priority corresponding to the fire-fighting component IDs based on the initial queue and the state response results; descendingly sorting the fire-fighting component IDs according to the processing priority to obtain the to-be-queried queue.

3. The method of claim 1, wherein the method further comprises: The method of dividing and generating the initial queue based on the state change flag bit according to the fault type comprises the following steps: extracting the state change flag bit corresponding to the fire-fighting component IDs; the state change flag bit is a two-bit binary value, comprising a fault change flag bit, an alarm change flag bit and a non-state flag bit; setting the state type of the fire-fighting component ID with the state change flag bit as the fault change flag bit as a fault state; setting the state type of the fire-fighting component ID with the state change flag bit as the alarm change flag bit as an alarm state; setting the state type of the fire-fighting component ID with the state change flag bit as the non-state flag bit as a non-state; grouping a plurality of fire-fighting component IDs with the state type as the fault state to form a fault queue; grouping a plurality of fire-fighting component IDs with the state type as the alarm state to form an alarm queue; determining the initial queue based on the fault queue and the alarm queue.

4. The method of claim 2, wherein the method further comprises: The method of generating the processing priority corresponding to the fire-fighting component IDs based on the initial queue and the state response results comprises the following steps: extracting the fire-fighting component IDs with the state type as the non-response state in the state response results, and setting the processing priority corresponding to the fire-fighting component IDs as a maximum value; extracting the state type corresponding to a plurality of fire-fighting component IDs in the initial queue; the state type comprises a fault type and an alarm type; extracting the state abnormality level corresponding to the fire-fighting component IDs from a state abnormality level table based on the fault type and the alarm type; acquiring a component ID line relationship diagram of the fire alarm system; extracting the fire-fighting component IDs in the initial queue and the state types corresponding thereto; integrating the component ID line relationship diagram and the fire-fighting component IDs and the state types corresponding thereto into influence analysis data; inputting the influence analysis data into an influence evaluation model to obtain the influence range level corresponding to a plurality of fire-fighting component IDs; the influence evaluation model is constructed by an artificial intelligence model; acquiring the historical state frequency of the fire-fighting component IDs within a time window; the state frequency refers to the total number of times of the alarm state and the fault state of the fire-fighting component IDs; Through formula Calculate the status frequency score of fire protection component ID Where i represents the ID of the fire-fighting component in the initial queue, and min() represents the minimum value operation. This represents the number of historical states corresponding to the i-th fire protection component ID. This is represented as the threshold for the number of states corresponding to the i-th fire protection component ID; weighting and summing the state abnormality level, the influence range level and the state frequency score corresponding to the fire-fighting component IDs by a weight coefficient to obtain the processing priority corresponding to the fire-fighting component IDs.

5. The method of claim 4, wherein the method further comprises: The influence evaluation model is constructed by an artificial intelligence model, comprising the following steps: Obtaining a plurality of historical influence analysis data and a plurality of corresponding fire-fighting component IDs and historical influence range levels; Divide the plurality of historical influence analysis data and the plurality of corresponding fire-fighting component IDs and historical influence range levels into training data, validation data and test data; and perform data preprocessing on the training data, validation data and test data to obtain a training set, a validation set and a test set; Selecting an artificial intelligence model as a base model; Training the base model through the training set, and adjusting the learning rate and hyperparameters on the validation set to obtain a pre-trained model; Through the validation of the pre-trained model on the test set, an influence evaluation model is finally obtained, which inputs the influence analysis data and outputs the influence range levels corresponding to the plurality of fire-fighting component IDs.

6. The method of claim 4, wherein the method further comprises: The weight coefficient is obtained by the following method, comprising: Obtaining a state feature vector of the current fire alarm system; the state feature vector comprises a system state distribution feature vector and a system environment feature vector; the system state distribution feature vector refers to the statistical characteristics of all fire-fighting component IDs in the current initial queue, including the proportion of severely abnormal components, the proportion of high-influence-range components, the proportion of high-state-frequency components and the total number of fire-fighting components; the system environment feature vector comprises a working mode label and a working time feature label; Inputting the state feature vector into the weight decision model to obtain the weight coefficients corresponding to the state abnormality level, the influence range level and the state frequency score; The weight decision model is constructed by reinforcement learning, comprising: Obtaining a plurality of training data; the training data comprises a state feature vector S(t) at time t, a weight coefficient group QX(t), a processing effect reward value R(t) and a new state feature vector S(t+1) after state processing; t represents the time number; Training the agent model by reinforcement learning algorithm and taking maximizing the long-term processing effect reward value as the target to obtain the weight decision model; The reward function R of the weight decision model is the weighted sum of risk reward, efficiency reward and resource reward.

7. The method of claim 1, wherein the method further comprises: The generation of the inspection result in the instruction sending mode based on the state type in the to-be-queried queue, comprising: Extracting the to-be-queried queue; Extracting the fire-fighting component ID and its corresponding state type in the to-be-queried queue in turn; When the state type is the non-response state, obtaining the last successful communication time point and the communication signal strength of the last several times of the fire-fighting component ID; taking the state type, the successful communication time point and the communication signal strength of the several times as the inspection result of the fire-fighting component ID, and removing the fire-fighting component ID and its corresponding state type from the to-be-queried queue; Otherwise, determining the sending instruction according to the fire-fighting component ID and its corresponding state type; the sending instruction comprises an acquisition parameter instruction and a query state instruction; Obtaining the inspection result returned from the fire-fighting component ID based on the sending instruction.

8. The method of claim 1, wherein, The self-adaptive generation of the inspection cycle based on the real-time load and load change trend of the to-be-queried queue, comprising: Obtaining the to-be-queried queue; Extracting the queue length Q corresponding to the to-be-queried queue; calculating the inspection cycle by a formula which satisfies: ; wherein, is expressed as a reference patrol period; is expressed as a queue load function, satisfying: ; denotes the current queue length, denotes a preset maximum queue capacity, b denotes a constant, b > 0; denotes a queue load trend function, satisfying: ; is expressed as a forgetting factor, ∈(0, 1); is expressed as the queue length of the last polling period; value, is expressed as the queue length of the last polling period; and are respectively expressed as the queue load weight and the queue trend weight, and ∈(0, 1); is expressed as a random fluctuation factor, a small amplitude random change is introduced to prevent the period locking effect, and satisfies: ; rand() represents a random function, represents the maximum value of the random fluctuation factor.

9. A patrol system for a fire alarm system, characterized in that Comprising: A data acquisition module and a data analysis module; the data acquisition module and the data analysis module are connected; The data collection module: obtains the state response results of a plurality of fire-fighting component IDs through a data collection device; The data analysis module comprises a queue construction unit, a result generation unit and a periodic feedback unit; The queue construction unit: constructs a to-be-queried queue based on the processing priorities of the fire-fighting component IDs according to the state response results; The result generation unit: generates an inspection result in an instruction sending mode based on the state types in the to-be-queried queue; The periodic feedback unit: adaptively generates an inspection period based on the real-time load and load change trend of the to-be-queried queue.

10. A patrol device for a fire alarm system, comprising: A processor and a storage medium; the storage medium comprises instructions, and the processor is configured to execute the instructions to implement the inspection method of the fire alarm system according to any one of claims 1-8.

Images