Intelligentized fire-fighting online data acquisition system

Abstract

The application discloses an intelligent fire-fighting online data collection system, and relates to the technical field of intelligent fire fighting, comprising a fire-fighting online data analysis module, which is used for analyzing the pretreated data set to obtain first characteristic coefficients and second characteristic coefficients; an availability judgment module, which is used for inputting the first characteristic coefficients and the second characteristic coefficients into a data verification model, outputting availability evaluation values, and triggering a hierarchical response mechanism according to the availability evaluation values.The application calculates the first characteristic coefficients and the second characteristic coefficients through the fire-fighting online data analysis module, outputs the availability evaluation values by using the data verification model, and triggers the hierarchical response mechanism, and the preset threshold supports multi-level configuration of daily and emergency modes, realizes intelligent evaluation and targeted processing of data quality, improves the adaptability and robustness of the system, and ensures that data abnormal conditions can be effectively coped with in different scenes.

Description

Technical Field

[0001] This invention relates to the field of intelligent fire protection technology, and in particular to an intelligent online fire protection data acquisition system. Background Technology

[0002] With the acceleration of urbanization and the increasing complexity of industrial facilities, real-time monitoring and early warning of fire safety face unprecedented challenges. In modern fire scenarios, the dynamic changes of multi-dimensional data such as temperature, smoke concentration, and pressure are highly uncertain, and traditional fire protection systems often rely on single sensors or localized data collection, making it difficult to comprehensively perceive the risk situation in complex environments. Furthermore, the demand for real-time processing, transmission, and secure storage of massive amounts of data has surged, while traditional systems lag significantly in data quality assessment, anomaly response mechanisms, and security protection, failing to meet the urgent needs of smart fire protection for efficiency, reliability, and intelligence.

[0003] Currently, fire data acquisition technology is gradually evolving towards the Internet of Things (IoT) and cloud computing. Some systems have achieved basic data collection by deploying multiple types of sensors and utilize simple threshold alarm mechanisms for risk warning. Existing technologies include some advanced systems that attempt to introduce data cleaning algorithms and encrypted transmission protocols to improve the standardization and security of data processing. However, these systems still fall short in feature extraction and dynamic evaluation, mostly relying on static models or fixed thresholds, which cannot adapt to the time-varying characteristics and sudden anomalies of data in fire scenarios. Furthermore, the lack of tiered strategies in data verification and response mechanisms often leads to high false alarm rates or delayed responses, making it difficult to achieve accurate decision-making in emergency situations.

[0004] The main problems with existing technologies are as follows: First, data preprocessing methods are rigid, and outlier filtering and missing value repair rely on fixed rules, making it difficult to cope with data fluctuations in complex environments; second, feature analysis lacks dynamic weight adjustment, failing to effectively capture recent data trends and global statistical characteristics, resulting in insufficient robustness of the evaluation model; third, data verification and response mechanisms are simplistic, failing to develop differentiated strategies based on different anomaly levels, easily leading to resource waste or missed risk assessments; finally, data transmission and storage security measures are weak, and static encryption and centralized storage architectures face risks of key leakage and data tampering.

[0005] Therefore, it is imperative to invent an intelligent online fire protection data acquisition system to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent online fire protection data acquisition system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent online fire protection data acquisition system, comprising the following modules:

[0008] The fire-fighting online data acquisition module is used to collect multi-dimensional raw data in fire-fighting scenarios to form a raw dataset.

[0009] The fire online data processing module is used to sequentially perform outlier filtering, missing value repair, and normalization mapping on the original dataset, and output a preprocessed dataset.

[0010] The fire protection online data analysis module is used to analyze the preprocessed dataset to obtain the first feature coefficient and the second feature coefficient;

[0011] The first characteristic coefficient is specifically:

[0012] ,

[0013] Where α is the time decay factor, t i Let s be the normalized temperature data at time i. i p represents the normalized smoke concentration data at time i. i Let t be the normalized pressure data at time i. k For the normalized temperature data at time k, s k p represents the normalized smoke concentration data at time k. k This represents the normalized pressure data at time k.

[0014] The time decay factor adopts a dynamic adjustment strategy, specifically:

[0015] ,

[0016] Where α0 is the initial decay factor, λ is the adjustment coefficient, and t i Let t be the normalized temperature data at time i. i-1 The normalized temperature data is at time i-1, and n is the total number of sampling points, i.e., the number of temperature data points within the time window.

[0017] The second characteristic coefficient is specifically:

[0018] ,

[0019] Where, σ t σ represents the standard deviation of the normalized temperature data. s σ represents the standard deviation of the normalized smoke concentration data. p The standard deviation of the normalized pressure data, μ t The normalized average temperature data, μ sThe average value of the normalized smoke concentration data, μ p δ represents the normalized average pressure data, δ is the adjustment coefficient, and max is the maximum value function;

[0020] The availability judgment module is used to input the first feature coefficient and the second feature coefficient into the data verification model, output the availability assessment value, and trigger the hierarchical response mechanism based on the availability assessment value.

[0021] The graded response mechanism includes:

[0022] When the availability assessment value M exceeds the preset threshold M1, the first-level response is activated: a data verification pass command is generated, a high-speed data upload channel is started, and the data is backed up to the cloud server in real time.

[0023] When the availability assessment value M is within the preset threshold M2 range, the secondary response is activated: triggering a re-inspection of local acquisition points, performing data fusion verification through the Kalman filter algorithm, and if the data is still abnormal after the re-inspection, marking the relevant sensors and reducing their acquisition frequency;

[0024] When the availability assessment value M is lower than the preset threshold M3, a level 3 response is activated: the full node self-inspection protocol is started, the data source is traced based on blockchain technology, suspicious data sources are isolated, and an equipment maintenance alarm is sent to the fire monitoring center, while data collection in the area is suspended.

[0025] The fire protection online data storage module adopts a distributed storage architecture that supports timestamp indexing, and stores the original dataset, preprocessed dataset, feature coefficients and availability evaluation values ​​in categories.

[0026] The fire protection online data communication module integrates the AES-256 encrypted transmission protocol and dynamic key negotiation mechanism to encrypt and upload the assessment value and associated raw data to the cloud server.

[0027] Preferably, the original dataset includes temperature sequence data, smoke concentration sequence data, and pressure sequence data.

[0028] Preferably, the data verification model is as follows:

[0029] ,

[0030] Where γ is the bias coefficient, C1 is the first feature coefficient, C2 is the second feature coefficient, β1 and β2 are dimension weight coefficients, β1+β2=1 and β1, β2∈[0,1].

[0031] Preferably, the preset thresholds M1, M2, and M3 support multi-level configuration, specifically:

[0032] When the system is in daily data collection mode, the preset thresholds are M1=m1, M2=m2, and M3=m3.

[0033] When the system detects that the temperature or smoke concentration exceeds the preset warning value during the data collection process, it automatically switches to emergency mode and sets preset thresholds M1=n1, M2=n2, and M3=n3, where n1>m1, n2>m2, and n3>m3.

[0034] The technical effects and advantages of this invention are as follows:

[0035] This invention calculates the first and second characteristic coefficients through the fire online data analysis module, outputs the availability assessment value using the data verification model and triggers the hierarchical response mechanism, and the preset threshold supports multi-level configuration for daily and emergency modes, realizing intelligent assessment and targeted processing of data quality, improving the system's adaptability and robustness, and ensuring that it can effectively deal with data anomalies in different scenarios.

[0036] This invention achieves dynamic weighted analysis and model optimization of data features by dynamically adjusting the time decay factor and the adaptive update mechanism of the data verification model parameters. The former adjusts the weight of recent data according to the temperature data change trend, which is more in line with the real-time requirements in fire-fighting scenarios; the latter continuously optimizes the model parameters based on historical data and manual annotation, so that the system can adapt to changes in sensor characteristics and environmental interference, and maintain the accuracy of data quality assessment in the long term.

[0037] This invention ensures the confidentiality and integrity of data during transmission by employing the AES-256 encrypted transmission protocol and a dynamic key negotiation mechanism; the distributed storage architecture combined with timestamp indexing enables efficient classification management and rapid retrieval of historical data; blockchain technology provides an immutable chain record of data sources, accurately tracing suspicious data, and, in conjunction with a full-node self-inspection protocol, forms a closed loop of end-to-end security protection and trusted verification from data collection, transmission to storage, significantly improving the system's anti-attack capability and fault diagnosis efficiency. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the system architecture of the present invention.

[0039] Figure 2 This is a timing diagram for dynamic key negotiation in this invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] This invention provides, for example Figure 1 The intelligent online fire protection data acquisition system shown includes the following modules:

[0042] The fire-fighting online data acquisition module is used to collect multi-dimensional raw data in fire-fighting scenarios to form a raw dataset.

[0043] Furthermore, in the above technical solution, the original dataset includes temperature sequence data, smoke concentration sequence data, and pressure sequence data.

[0044] It should be noted that the fire-fighting online data acquisition module includes two or more temperature sensors used to collect real-time temperature data of the monitored area, forming a temperature sequence data T={t1, t2, ..., t...} n Two or more smoke concentration sensors are used to collect real-time smoke concentration data of the monitored area, forming a smoke concentration sequence data S={s1, s2, ..., s...} n Two or more pressure sensors are used to collect real-time pressure data from fire-fighting pipelines, forming a pressure sequence data P={p1, p2, ..., p...} n The original dataset is D={T, S, P}, where n is the number of sampling points within a preset time window;

[0045] It should be noted that the temperature sensor is an infrared temperature sensor; the smoke concentration sensor is a photoelectric sensor; and the pressure sensor is selected according to different scenarios, such as explosion-proof pressure sensors used in flammable and explosive places like chemical plants and oil depots, high-temperature resistant pressure sensors used in high-temperature places like steel plants and boiler rooms, and waterproof pressure sensors used in humid environments like underground fire pump rooms.

[0046] The fire online data processing module is used to sequentially perform outlier filtering, missing value repair, and normalization mapping on the original dataset, and output a preprocessed dataset.

[0047] It's important to understand that the outlier filtering process involves first constructing a dynamic threshold model. This model determines the outlier range based on the characteristics of the data within the current time window. The upper and lower thresholds are obtained by calculating the mean μ and standard deviation σ of the data within the current time window, using the formula μ ± 3σ. The window width m and confidence coefficient k can be dynamically adjusted according to the sensor type and scenario. For example, in industrial plant scenarios, due to the complex environment and large data fluctuations, the window width m and confidence coefficient k can be appropriately increased to reduce false positives. In practice, data is monitored in real time. Once data at a certain moment exceeds the calculated upper and lower threshold range, it is identified as an outlier and marked for removal, preventing abnormal data from interfering with subsequent analysis.

[0048] The missing value repair employs a spatiotemporal joint interpolation algorithm. For missing points in the time dimension, a time-decay-based weighted prediction model is constructed using data from before and after the missing point. Data closer to the missing point in time has a larger weight, and the weight decreases exponentially with the time interval. For example, if the time interval between the data before the missing point and the missing point is t1, and the time interval between the data after the missing point and the missing point is t2, then the weight of the data before the missing point is α^t1, and the weight of the data after the missing point is α^t2, where α is the time decay factor, with a value between (0, 1). For missing points in the spatial dimension, a distance-weighted fusion model is constructed by combining data from adjacent sensors. The inverse squared distance weighting method is used, where if the distance between an adjacent sensor and the missing point is d, then the weight of the sensor data is 1 / d. 2 The data from surrounding nodes is merged to calculate missing values. The results from the time and spatial dimensions are combined to obtain and fill in the missing values.

[0049] The normalization mapping process uses a dynamic interval scaling algorithm to normalize the data to the [0, 1] interval, as shown in the formula. , where x ω min x is the minimum value of the data within the current time window. ω max The maximum value of the data within the current time window is represented and changes dynamically as the data is updated. ε is a very small positive constant used to prevent the denominator from being zero. This method ensures that temperature, smoke concentration, and pressure data of different dimensions have a unified quantitative standard, eliminating the impact of dimensional differences on subsequent analysis and ultimately outputting a preprocessed dataset suitable for further analysis.

[0050] The fire protection online data analysis module is used to analyze the preprocessed dataset to obtain the first feature coefficient and the second feature coefficient;

[0051] Furthermore, in the above technical solution, the first characteristic coefficient is specifically:

[0052] ,

[0053] Where α is the time decay factor, reflecting the weight of recent data on the feature, and t i Let s be the normalized temperature data at time i. i p represents the normalized smoke concentration data at time i. i Let t be the normalized pressure data at time i. k For the normalized temperature data at time k, s k p represents the normalized smoke concentration data at time k. k This represents the normalized pressure data at time k.

[0054] Furthermore, in the above technical solution, the time decay factor adopts a dynamic adjustment strategy, specifically as follows:

[0055] ,

[0056] Where α0 is the initial decay factor, λ is the adjustment coefficient, and t i Let t be the normalized temperature data at time i. i-1 The normalized temperature data is at time i-1, and n is the total number of sampling points, i.e., the number of temperature data points within the time window.

[0057] It should be noted that the initial value of the initial attenuation factor α0 is set to 0.9, and the value of the adjustment coefficient λ is set to 0.1. The adjustment coefficient λ is adjusted by monitoring the cumulative difference in temperature changes; if the difference significantly exceeds the expected range, then λ is adjusted accordingly. new =λ old The method of increasing λ by +Δλ enhances the sensitivity to sudden changes, where Δλ is a constant, Δλ = 0.05; the specific difference is... The expected range is [μ-k×σ, μ+k×σ], where μ is the average value of the normalized temperature change, σ is the standard deviation of the normalized temperature change, and k is the confidence coefficient, ranging from 1 to 3.

[0058] Furthermore, in the above technical solution, the second characteristic coefficient is specifically:

[0059] ,

[0060] Where, σ t σ represents the standard deviation of the normalized temperature data. s σ represents the standard deviation of the normalized smoke concentration data. p The standard deviation of the normalized pressure data, μ t The normalized average temperature data, μs The average value of the normalized smoke concentration data, μ p δ represents the normalized average pressure data, δ is the adjustment coefficient, and max is the maximum value function.

[0061] It should be noted that the adjustment coefficient δ can be set based on industry experience.

[0062] The availability judgment module is used to input the first feature coefficient and the second feature coefficient into the data verification model, output the availability assessment value, and trigger the hierarchical response mechanism based on the availability assessment value.

[0063] Furthermore, in the above technical solution, the data verification model specifically refers to:

[0064] ,

[0065] Where γ is the bias coefficient, C1 is the first feature coefficient, C2 is the second feature coefficient, β1 and β2 are dimension weight coefficients, β1+β2=1 and β1, β2∈[0,1].

[0066] Furthermore, in the above technical solution, the hierarchical response mechanism includes:

[0067] When the availability assessment value M exceeds the preset threshold M1, the first-level response is activated: a data verification pass command is generated, a high-speed data upload channel is started, and the data is backed up to the cloud server in real time.

[0068] When the availability assessment value M is within the preset threshold range M2, the secondary response is activated: triggering a re-inspection of local acquisition points, performing data fusion verification through the Kalman filter algorithm, and if the data is still abnormal after the re-inspection, marking the relevant sensors and reducing their acquisition frequency;

[0069] When the availability assessment value M is lower than the preset threshold M3, a level 3 response is activated: the full node self-inspection protocol is started, the data source is traced based on blockchain technology, suspicious data sources are isolated, and an equipment maintenance alarm is sent to the fire monitoring center, while data collection in the area is suspended.

[0070] It's important to understand that the specific process of the secondary response is as follows: First, a local sampling point re-inspection is triggered. This involves identifying the local area where a sampling point with an abnormal availability assessment value is located, such as a sensor group within the same fire compartment. Then, sensor data related to the abnormal point within this local area is extracted, such as time-series data from temperature, smoke concentration, and pressure sensors, forming a local dataset. Next, a Kalman filter algorithm is used to perform data fusion verification. First, state-space modeling is performed, constructing state equations that treat sensor data as state variables, and establishing a state transition model X. t =A×Xt-1 +W t Where A is the state transition matrix, usually set as the identity matrix, and the data sequence is assumed to have first-order Markov property, W t The process noise is represented by a Gaussian distribution; the observation equation Z is then constructed. t =H×X t +V t Where H is the observation matrix, set as the identity matrix to directly observe the state variables, V t The covariance matrix of the observed noise is known. Then, a recursive filtering process is initiated, first estimating the state from the previous time step. Covariance P t-1|t-1 Predict the current state Covariance P t|t-1 =A×P t-1|t-1 ×A T +Q, where Q is the process noise covariance; combined with the current measured value Z t Calculate the Kalman gain K t =P t|t-1 ×H T ×(H×P t|t-1 ×H T +R) -1 Where R is the observation noise covariance, and the updated state estimate is... Covariance P t|t =(IK t ×H)×P t|t-1 Where I is the identity matrix, and the residuals between the measured and predicted values ​​are calculated simultaneously. If the residual exceeds a preset confidence interval, such as 3σ, the data point is considered abnormal. After recursive filtering, data fusion verification is performed. Kalman filtering is applied to the time series data of all sensors in the local area to generate a fused estimated value sequence. The original measured values ​​are then compared with the fused estimated values ​​to identify abnormal data points with continuous deviations, such as residuals exceeding the limit for three consecutive time points. If the data is still abnormal after re-inspection, the relevant sensors are marked and their acquisition frequency is reduced. If more than 30% of the data points of a sensor are still judged as abnormal after re-inspection, or if there are five consecutive abnormal time points, it is marked as suspicious, and the abnormality type is recorded, such as temperature sensor drift or smoke sensor false alarm. At the same time, the sensor is highlighted on the system monitoring interface, and an abnormality log is generated, including information such as abnormal time, data deviation amplitude, and correlation characteristic coefficient. Then, the acquisition frequency of the abnormal sensor is set to the default value, such as once per second, and reduced to a low-power mode, such as once every ten seconds, to reduce invalid data acquisition and reduce system resource consumption. At the same time, the sensor self-test program is triggered. If the self-test fails, the low-frequency acquisition is maintained until manual maintenance. This execution process features localized processing, dynamic fault tolerance, and adaptive adjustment. It only re-inspects sensors in abnormal areas, avoiding full system scanning and improving efficiency. Kalman filtering effectively filters short-term noise, distinguishes between random interference and persistent faults, and reduces false alarms. It can also dynamically adjust the acquisition frequency according to the degree of anomaly, optimizing system energy consumption while ensuring data availability. It achieves accurate location, intelligent verification, and hierarchical processing of data anomalies, ensuring stable system operation through local optimization when data quality is moderately reliable, while providing clear fault clues for subsequent maintenance.

[0071] The specific process of the three-level response is as follows: First, a full-node self-check protocol is initiated to perform a comprehensive self-check on all sensor nodes, data acquisition modules, and communication equipment. This covers hardware status, such as power supply stability and signal transmission quality; software operating parameters, such as acquisition frequency, algorithm module integrity, and data integrity, such as consistency verification between raw and preprocessed data; and identifying potential fault points by traversing the node self-check logs. Then, the data source is traced using blockchain technology. Each sensor node synchronously generates a hash value containing a timestamp, sensor ID, data value, and device status when acquiring data and embeds it into a blockchain block, forming an immutable chain record. When an anomaly is detected, the system extracts the historical data chain of the nodes in that area through a smart contract, verifies data integrity and compliance of the acquisition process, compares the differences between multi-sensor data, and locates suspicious data sources based on hash value matching. After completing the tracing and identifying the suspicious data source, the system immediately implements isolation measures, on the one hand through intelligent routing calculations... The system adds the IP address or device ID of the suspicious data source to the isolation list, cutting off its connection to the data acquisition network. Simultaneously, it highlights the location of the suspicious data source on the system monitoring interface and triggers an alarm on the device indicator light, generating a log containing the isolation time and abnormal characteristics for easy maintenance. Next, it sends an equipment maintenance alarm to the fire monitoring center, pushing detailed information including the location of the abnormal area, the type of suspicious data source, the type of abnormality, the blockchain tracing results, and the system status via a dedicated encrypted link. At the same time, it notifies relevant personnel via SMS to ensure the monitoring center receives real-time details of the abnormality. Finally, while isolating the suspicious data source, the system suspends data acquisition in that area, stops all nodes, and sends a status report to the cloud server, marking the area as suspended for maintenance to avoid cloud misjudgment. The system remains operational until maintenance personnel repair the fault and pass a self-test before reactivation. This forms a complete closed loop from fault detection, tracing and location, isolation alarm to suspension for maintenance, ensuring the system's stability and reliability under extreme abnormal conditions.

[0072] Furthermore, in the above technical solution, the preset thresholds M1, M2, and M3 support multi-level configuration, specifically as follows:

[0073] When the system is in daily data collection mode, the preset thresholds are M1=m1, M2=m2, and M3=m3.

[0074] When the system detects that the temperature or smoke concentration exceeds the preset warning value during the data collection process, it automatically switches to emergency mode and sets preset thresholds M1=n1, M2=n2, and M3=n3, where n1>m1, n2>m2, and n3>m3.

[0075] It is important to know that in the normal data collection mode, m1=0.85, m2∈[0.6, 0.85], m3=0.6; when the system detects that the temperature exceeds 60℃ or the smoke concentration exceeds 500ppm during the data collection process, it automatically switches to emergency mode, n1=0.9, n2∈[0.75, 0.9], n3=0.75;

[0076] The fire protection online data storage module adopts a distributed storage architecture that supports timestamp indexing, and stores the original dataset, preprocessed dataset, feature coefficients and availability evaluation values ​​in categories.

[0077] The fire protection online data communication module integrates the AES-256 encrypted transmission protocol and dynamic key negotiation mechanism to encrypt and upload the assessment value and associated raw data to the cloud server.

[0078] It should be noted that the system within the cloud server extracts historical data from the system database at preset intervals, such as daily or weekly. This includes historically calculated first feature coefficients C1 and second feature coefficients C2, and combines this data with manually labeled usable data tags, such as y=1 indicating usable and y=0 indicating unusable. This constructs a training dataset containing the first feature coefficients C1 and C2 and their corresponding labels. Where N is the total number of samples, C1 （i） C2 is the first characteristic coefficient of the i-th data point. （i） Let y be the second characteristic coefficient of the i-th data point. （i） The label corresponding to the i-th data point provides the data foundation for model parameter optimization. Based on the training dataset, the gradient descent algorithm is used to adaptively update the bias coefficient γ, dimensional weight coefficients β1 and β2 in the data validation model. First, the cross-entropy loss function is defined to measure the difference between the predicted value and the true label. Then, the gradient of the loss function with respect to each parameter is calculated through backpropagation, specifically: , , Then, the parameters are iteratively adjusted along the reverse direction of the gradient using a preset learning rate η, specifically as follows: , , ,in, N is the total number of samples in the training dataset, and the process continues until the loss function converges or the maximum number of iterations is reached. After parameter optimization, the system will update β1. final β2 final and γ finalThe data is stored on a cloud server and synchronized to the availability assessment module through mechanisms such as timed polling and message triggering. After receiving new parameters, the module automatically loads and replaces the original parameters and restarts the assessment logic, so that subsequent availability assessments of the collected data are based on the optimized model, thereby continuously improving the accuracy and robustness of data quality assessment and effectively adapting to dynamic factors such as changes in sensor characteristics and environmental interference in fire scenarios.

[0079] It should be noted that the dynamic key negotiation mechanism is as follows: Figure 2 As shown, a unique session key is dynamically generated between the communicating parties to achieve high-strength encryption and security protection for data transmission, avoiding the risk of static key leakage. Its core principle is: using the Diffie-Hellman key exchange algorithm and its variants, through four steps—parameter initialization, secret value generation, public value exchange, and session key calculation—the communicating parties independently generate the same temporary session key without directly transmitting the key. This key is only valid within a single session and is destroyed immediately after the session ends, achieving a one-time pad secure transmission mode. The specific implementation process is as follows: Parameter initialization: The device and the cloud pre-agree on public mathematical parameters, such as a large prime number p and a generator g, determined through secure channel configuration or system initialization negotiation; Secret and public value generation: The device generates a random private key a and calculates the public key A=g. a The modulo p is sent to the cloud, where a random private key b is generated, and the public key B = g is calculated. b mod p and send it back to the device; session key generation, the device uses B and a to calculate the key K=B. a Mod p ensures that the keys of both parties are consistent; data is transmitted in encrypted form. The device uses K to encrypt the original data, such as temperature sequences and availability assessment values, using the AES-256 algorithm. The cloud receives the data and decrypts it with the same key, ensuring the confidentiality and integrity of the data; key lifecycle management ensures that the session key is destroyed immediately after the end of a single session or after the preset time interval is reached, and a new key is renegotiated for the next communication, eliminating the risk of key reuse.

[0080] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent online fire protection data acquisition system, characterized in that, Includes the following modules: The fire-fighting online data acquisition module is used to collect multi-dimensional raw data in fire-fighting scenarios to form a raw dataset. The fire online data processing module is used to sequentially perform outlier filtering, missing value repair, and normalization mapping on the original dataset, and output a preprocessed dataset. The fire protection online data analysis module is used to analyze the preprocessed dataset to obtain the first feature coefficient and the second feature coefficient; The first characteristic coefficient is specifically: ， Where α is the time decay factor, t i Let s be the normalized temperature data at time i. i p represents the normalized smoke concentration data at time i. i Let t be the normalized pressure data at time i. k For the normalized temperature data at time k, s k p represents the normalized smoke concentration data at time k. k The normalized pressure data at time k; The time decay factor adopts a dynamic adjustment strategy, specifically: ， Where α0 is the initial decay factor, λ is the adjustment coefficient, and t i Let t be the normalized temperature data at time i. i-1 The normalized temperature data is at time i-1, and n is the total number of sampling points, i.e., the number of temperature data points within the time window; The second characteristic coefficient is specifically: ， Where, σ t σ represents the standard deviation of the normalized temperature data. s σ represents the standard deviation of the normalized smoke concentration data. p The standard deviation of the normalized pressure data, μ t The normalized average temperature data, μ s The average value of the normalized smoke concentration data, μ p δ represents the normalized average pressure data, δ is the adjustment coefficient, and max is the maximum value function; The availability judgment module is used to input the first feature coefficient and the second feature coefficient into the data verification model, output the availability assessment value, and trigger the hierarchical response mechanism based on the availability assessment value. The graded response mechanism includes: When the availability assessment value M exceeds the preset threshold M1, the first-level response is activated: a data verification pass command is generated, a high-speed data upload channel is started, and the data is backed up to the cloud server in real time. When the availability assessment value M is within the preset threshold M2 range, the secondary response is activated: triggering a re-inspection of local acquisition points, performing data fusion verification through the Kalman filter algorithm, and if the data is still abnormal after the re-inspection, marking the relevant sensors and reducing their acquisition frequency; When the availability assessment value M is lower than the preset threshold M3, a level 3 response is activated: the full node self-inspection protocol is started, the data source is traced based on blockchain technology, suspicious data sources are isolated, and an equipment maintenance alarm is sent to the fire monitoring center, while data collection in the area is suspended. The fire protection online data storage module adopts a distributed storage architecture that supports timestamp indexing, and stores the original dataset, preprocessed dataset, feature coefficients and availability evaluation values ​​in categories. The fire protection online data communication module integrates the AES-256 encrypted transmission protocol and dynamic key negotiation mechanism to encrypt and upload the assessment value and associated raw data to the cloud server.

2. The intelligent online fire protection data acquisition system according to claim 1, characterized in that, The original dataset includes temperature sequence data, smoke concentration sequence data, and pressure sequence data.

3. The intelligent online fire protection data acquisition system according to claim 1, characterized in that, The data verification model is specifically as follows: ， Where γ is the bias coefficient, C1 is the first feature coefficient, C2 is the second feature coefficient, β1 and β2 are dimension weight coefficients, β1+β2=1 and β1, β2∈[0,1].

4. The intelligent online fire protection data acquisition system according to claim 1, characterized in that, The preset thresholds M1, M2, and M3 support multi-level configuration, specifically: When the system is in daily data collection mode, the preset thresholds are M1=m1, M2=m2, and M3=m3. When the system detects that the temperature or smoke concentration exceeds the preset warning value during the data collection process, it automatically switches to emergency mode and sets preset thresholds M1=n1, M2=n2, and M3=n3, where n1>m1, n2>m2, and n3>m3.

Images