Intelligent management method and device for fire-fighting system based on intelligent property

Abstract

The application discloses a fire-fighting system intelligent management method and device based on intelligent property, and relates to the technical field of data management.The fire-fighting system intelligent management method and device based on intelligent property comprises the following steps: S1, collecting multi-source fire-fighting data sets through multi-protocol dynamic triggering and temporarily storing in edge normalization; S2, constructing an event full-cycle window and a ternary association index of events, time sequence data and original messages; S3, performing double-track storage of semantic analysis data and original messages, and constructing a multi-dimensional index association; S4, synchronizing through edge cloud increment and reconstructing a multi-dimensional time sequence chain to visually replay an alarm process; and S5, performing transient analysis, protocol upgrading and responsibility evidence taking based on a cloud data collaborative link and an original message data set.The application effectively improves the complete collection of fire-fighting data and the reproducibility of an event process, and solves the problem of missing transient processes and message-level evidence caused by only polling a single-point alarm and discarding original messages in the prior art.

Description

Technical Field

[0001] This invention relates to the field of data management technology, specifically to a method and device for intelligent management of fire protection systems based on smart property management. Background Technology

[0002] With the popularization of smart property management and the rapid development of intelligent buildings, traditional fire protection systems face many challenges. Most existing fire protection systems rely on manual inspections and periodic checks, failing to acquire comprehensive data in real time. This results in long response times during emergencies and an inability to accurately determine the status and operational status of fire protection equipment. Furthermore, existing fire protection data is largely based on simple alarm signals, lacking complete event recording and analysis, leading to significant information gaps in post-incident investigations and emergency response. Current technologies cannot effectively solve the problems of real-time data acquisition, early warning, tracking, and subsequent data analysis in fire protection systems.

[0003] For example, invention patent CN118037231A discloses a SaaS-based smart park management system and method, including: inputting, querying, statistically analyzing, and managing basic park information, customer information, contract information, and fee information; monitoring, maintaining, and alarming the park's water, electricity, gas, fire protection, security, and parking facilities; publishing, scheduling, evaluating, and settling payments for the park's property management, cleaning, catering, and courier services; and facilitating communication, exchange, interaction, and collaboration among the park's customers to build a park ecosystem. This can improve park management efficiency, enhance park service quality, and increase park security.

[0004] For example, the invention patent with publication number CN109559031A discloses an intelligent fire protection management system, including: a handheld reader, a backend server, and equipment tags; wherein, the handheld reader includes: a tag reading and writing module, an equipment input module, and an equipment inspection module, and is communicatively connected to the backend server; the tag reading and writing module is used to read and write information on the equipment tags according to the operation instructions input by the operator; the equipment input module is used to receive the equipment information corresponding to the equipment tags entered by the operator; the equipment inspection module is used to display the corresponding equipment inspection information based on the read tag information, and to receive the equipment inspection records and signatures entered by the operator.

[0005] In existing technologies, edge gateways poll the status of fire alarm controllers and PLCs at fixed intervals, inserting only one alarm event record in the cloud and appending records when the status changes. Process data for the instantaneous pump startup is neither collected nor stored. Existing multi-protocol gateways discard the original binary data directly after parsing messages from fire alarm hosts, PLCs, and serial sensor ports. If the protocol is upgraded, fields are added, or the encoding meaning is changed, historical data cannot be re-parsed due to the lack of the original messages, and the semantics of the new fields are permanently lost.

[0006] Therefore, in order to address the above issues, there is an urgent need for intelligent management methods and devices for fire protection systems based on smart property management. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides an intelligent management method and device for fire protection systems based on smart property management, which solves the problem that existing technologies only poll and record single-point alarms and discard original messages, resulting in the lack of transient process and message-level evidence.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a method and device for intelligent management of fire protection systems based on smart property management, comprising: S1, collecting multi-source fire protection datasets through a multi-protocol compatibility and dynamic triggering collaborative acquisition mechanism, and classifying, temporarily storing, and normalizing the multi-source fire protection datasets through a temporary buffer on the edge side; S2, constructing an event full-cycle window based on the multi-source fire protection datasets, performing integrity verification, and constructing a ternary association index of events, time-series data, and original messages; S3, performing dual-track storage of semantically parsed data and original messages based on parsed structured fields, window datasets, and collected original message datasets, and constructing a multi-dimensional index association; S4, constructing an edge and cloud data collaborative link based on an incremental synchronization mechanism, constructing a multi-dimensional time-series chain for full-cycle dynamic reconstruction of alarm events, and visually replaying the process; S5, performing transient process analysis based on the cloud data collaborative link and window datasets, and performing protocol upgrade adaptation and responsibility verification based on the original message datasets and multi-dimensional index associations.

[0011] Furthermore, through a multi-protocol compatibility and dynamic triggering collaborative acquisition mechanism, multi-source fire protection datasets are collected. The specific process of classifying, temporarily storing, and normalizing the multi-source fire protection datasets through a temporary buffer on the edge side is as follows: Multi-source fire protection datasets are collected collaboratively through multi-protocol compatibility and dynamic triggering. These datasets include: equipment status datasets, raw message datasets, and key parameter datasets. A dynamic acquisition mechanism combining regular periodic acquisition and event-triggered acquisition is adopted. The collected multi-source fire protection datasets are classified and temporarily stored through a temporary buffer on the edge side, and a historical labeled sample dataset is constructed as a training data source. The multi-source fire protection datasets are linearly transformed to a unified interval through minimum-maximum normalization mapping. For equipment status data, numerical mapping or one-hot encoding is used to unify the representation dimension. All normalization operations are executed synchronously on the edge side and during data temporary storage, and are uniformly guaranteed through protocol compatibility, time synchronization, and breakpoint resumption.

[0012] Furthermore, the specific process of constructing a full-cycle event window based on a multi-source fire dataset for integrity verification is as follows: inputting device status change signals, transient event trigger commands, and multi-source fire datasets; constructing a full-cycle time window and caching it in a time-series database; constructing a linkage mechanism between the event window and data acquisition; binding the transient event trigger command with the window start and end time through timestamp alignment; adopting interrupt-triggered acquisition logic; immediately switching the acquisition frequency of the corresponding device after the window starts; directionally capturing core parameters such as pressure fluctuations, temperature rises, smoke concentration changes, and starting current; introducing a circular buffer and integrity verification rules; and checking the temporal continuity and amplitude rationality of the cached data.

[0013] Furthermore, the specific process of constructing a ternary association index for events, time-series data, and original messages is as follows: Input the original message dataset and window dataset; construct a ternary association index for events, time-series database, and original messages; construct a dual-write channel and index update mechanism for original messages; simultaneously complete semantic field writing and original message archiving during the protocol parsing stage; introduce protocol adaptation and process restoration algorithms based on original messages; construct a historical event and message association dataset; take the time weight coefficient multiplied by the exponential function term as the time association term; take the time difference between the event occurrence time and the message time as the exponential function term, which is exponentially decayed by the time decay coefficient; multiply the attribute weight coefficient by the attribute matching indicator function as the attribute matching term; sum the time association term and the attribute matching term to obtain the association degree value; output the event archive containing the ternary association index and a re-parseable normalized original message library.

[0014] Furthermore, based on the parsed structured fields, window datasets, and collected raw message datasets, the specific process of dual-track storage of semantic parsing data and raw messages, and the construction of multi-dimensional index associations, is as follows: A hierarchical structured storage mechanism for semantic data is constructed, storing the parsed fields of device status and window datasets according to business scenarios. A full-scale archiving rule for raw messages is established. During the protocol parsing phase, an independent channel for raw messages is opened in parallel. Complete raw messages are archived to edge-side object storage in the order of device ID, timestamp, and protocol version. An association index table is constructed with the event window ID in the window dataset as the core, recording window metadata, semantic data storage path, raw message storage location, and protocol version information. Path normalization constraints are applied to the dual-track storage data. The storage paths of both semantic data and raw messages contain unique identifier fields. An index encoding function is introduced to uniformly map multi-dimensional identifiers, concatenating the multi-dimensional identifier parameters into a string in sequence. Then, a hash operation is performed on this concatenated string to output a globally unique index value. The output includes an edge-side local semantic database, a cloud-based standardized event table, a standardized raw message library, a raw message index, and a multi-dimensional association index table.

[0015] Furthermore, the specific process of building a data collaboration link between the edge and the cloud based on the incremental synchronization mechanism is as follows: based on the window dataset, semantic data and the original message index, incremental synchronization and cross-end penetration query are performed between the edge and the cloud, a synchronization priority and status control mechanism is built, a cross-end query interface is designed based on a multi-dimensional association index table, an access control and encrypted transmission mechanism is introduced, and incremental updates from the cloud are output to the time-series database, window metadata and cross-end linkage query interface.

[0016] Furthermore, the specific process of constructing a multi-dimensional time-series chain for dynamic reconstruction of the entire alarm event lifecycle and visual replay is as follows: Based on window data, semantic data, and window metadata synchronized in the cloud, the alarm data reported asynchronously from multiple sources is uniformly aggregated and modeled, constructing a multi-dimensional time-series chain covering before, during, and after the alarm. Under the constraints of a unified time base and spatial topology, the time-series alignment and association logic of dynamic time regularization for multiple devices and regions is executed, elevating single-point curves to regional and system-level linked process views. After completing the reconstruction of the multi-dimensional time-series chain, the reconstruction results are transformed into a visually rendering-friendly dynamic process description and configuration file, driving the front end or monitoring screen to generate an interactive process playback interface, outputting a visual dynamic process curve and event replay view, and synchronously storing the multi-dimensional time-series chain organized by event window ID, device ID, and spatial location into the cloud time-series database to form a cloud-based multi-dimensional time-series chain.

[0017] Furthermore, the specific process of transient process analysis based on cloud-based data collaboration links and window datasets is as follows: Input the cloud-based multidimensional time series chain, window dataset, and equipment operation logs; perform feature analysis and optimization on the fire transient process; obtain the weight vector by minimizing the prediction fit training on the historical labeled sample dataset; multiply the weight vector with the key feature vector of the transient process to obtain a linear term; multiply the feature vector by the weight matrix on the left and then by the feature vector on the right to obtain half of the quadratic term; add the bias term to obtain the independent variable of the model; send the independent variable into the compression function for activation; and output a comprehensive transient risk score. It is necessary to collect operating data covering the entire working condition and life cycle of the target equipment, and select an appropriate training method according to the data scale and distribution characteristics; construct a process anomaly detection and root cause analysis mechanism; trace the cause of anomalies to carry out linkage strategy optimization simulation; and output a transient process feature analysis report and linkage strategy iteration scheme.

[0018] Furthermore, the specific process of protocol upgrade adaptation and responsibility evidence collection based on the association of the original message dataset and multi-dimensional index is as follows: Input the standardized original message library, multi-dimensional association index table and cross-end linkage query results, pre-screen and classify the acquired data according to scenarios, and enter the protocol upgrade adaptation process and the full-link responsibility evidence collection process respectively. Complete the semantic supplementation of historical data and the construction of event evidence chain, build a responsibility evidence chain generation mechanism, and perform hierarchical association and cross-verification of data at the protocol layer, time sequence layer and physical process layer by comparing protocol differences and analyzing mutual relationships. Use hash algorithm and blockchain technology to introduce evidence chain anti-tampering mechanism and evidence storage strategy, and uniformly include key original message fragments, index entries and reconstructed time sequence summaries into the verification scope. Automatically generate various forms of output results according to different usage scenarios, and obtain the historical data supplementation results after protocol upgrade, field comparison table, parsing difference list and responsibility evidence chain report.

[0019] Furthermore, a second aspect of the present invention provides an intelligent management device for fire protection systems based on smart property management, applied to an intelligent management method for fire protection systems based on smart property management, comprising: a multi-source fire protection data full-volume acquisition module, used to acquire multi-source fire protection datasets through a multi-protocol compatible and dynamically triggered collaborative acquisition mechanism, and to classify, temporarily store, and normalize the multi-source fire protection datasets through an edge-side temporary buffer; an event window caching and timing management module, used to construct an event full-cycle window based on the multi-source fire protection dataset, perform integrity verification, and construct a ternary association index of events, timing data, and original messages; and a dual-track data storage and indexing module. The association module is used for dual-track storage of semantically parsed data and original messages based on parsed structured fields, window datasets, and collected raw message datasets, and to build multi-dimensional index associations; the edge-cloud data collaboration and time-series reconstruction module is used to build edge-cloud data collaboration links based on incremental synchronization mechanisms, construct multi-dimensional time-series chains for full-cycle dynamic reconstruction of alarm events, and visualize process replay; the intelligent analysis and full-link traceability application module is used for transient process analysis based on cloud data collaboration links and window datasets, and for protocol upgrade adaptation and responsibility verification based on raw message datasets and multi-dimensional index associations.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This invention achieves comprehensive collection of multi-source fire protection data through multi-protocol compatibility and dynamic triggering mechanisms, avoiding the omission of key transient processes in traditional collection methods and improving the integrity and accuracy of data collection. Through the dual-track storage mechanism of event window collection and original message, the problem that alarm process and transient process data cannot be reproduced in the prior art is solved, ensuring the traceability and replayability of the entire alarm event process.

[0023] (2) This invention achieves effective data association and query optimization by constructing a ternary association index of events, time-series data and original messages. Furthermore, through edge cloud incremental synchronization and dynamic reconstruction of multi-dimensional time-series chains, it further enhances the availability and flexibility of data in protocol upgrade adaptation, responsibility evidence collection and alarm process visualization replay, and improves response speed and data processing capabilities.

[0024] (3) This invention, through protocol upgrade adaptation and accountability evidence collection mechanisms, ensures backward compatibility of equipment protocols and data integrity, enabling the provision of an immutable original chain of evidence during protocol upgrades and event tracing, thus meeting audit and compliance requirements. Combined with data storage and anti-tampering mechanisms, it achieves hash verification and secure storage of key data, ensuring the security and reliability of multi-source fire protection data during collection, transmission, and storage.

[0025] (4) This invention provides accurate early warning and optimization suggestions for equipment maintenance through intelligent analysis and transient process optimization, timely detection of potential problems in real-time monitoring, and improvement of equipment operating efficiency and maintenance cycle. By realizing complete recording and real-time analysis of the process, it solves the limitations of the existing technology in the collection and analysis of instantaneous events, and significantly improves the emergency response and management capabilities of the fire protection system.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 This is a flowchart of the intelligent management method for fire protection systems based on smart property management, as described in this invention.

[0028] Figure 2 This is a structural diagram of the intelligent management device for fire protection system based on smart property management according to the present invention;

[0029] Figure 3 This is a diagram illustrating the multi-stage intelligent sensing transient event analysis of the smoke sensor of the present invention.

[0030] Figure 4 This is a graph illustrating the multi-dimensional temporal chain reconstruction and dynamic process of this invention.

[0031] Figure 5 This is a flowchart illustrating the processing of data storage and protocol adaptation based on the present invention. Detailed Implementation

[0032] 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.

[0033] Please see Figures 1-5This invention provides a technical solution: a method and device for intelligent management of fire protection systems based on smart property management, comprising: S1, collecting multi-source fire protection datasets through a multi-protocol compatibility and dynamic triggering collaborative acquisition mechanism, and classifying, temporarily storing, and normalizing the multi-source fire protection datasets through an edge-side temporary buffer; S2, constructing an event full-cycle window based on the multi-source fire protection datasets, performing integrity verification, and constructing a ternary association index of events, time-series data, and original messages; S3, performing dual-track storage of semantically parsed data and original messages based on parsed structured fields, window datasets, and collected original message datasets, and constructing a multi-dimensional index association; S4, constructing an edge and cloud data collaborative link based on an incremental synchronization mechanism, constructing a multi-dimensional time-series chain for full-cycle dynamic reconstruction of alarm events, and visually replaying the process; S5, performing transient process analysis based on the cloud data collaborative link and window datasets, and performing protocol upgrade adaptation and responsibility verification based on the original message datasets and multi-dimensional index associations.

[0034] Specifically, through a multi-protocol compatibility and dynamic triggering collaborative acquisition mechanism, multi-source fire protection datasets are collected. The specific process of classifying, temporarily storing, and normalizing the multi-source fire protection datasets through a temporary buffer on the edge side is as follows: Multi-source fire protection datasets are collected collaboratively through multi-protocol compatibility and dynamic triggering. These datasets include: equipment status datasets, raw message datasets, and key parameter datasets. The equipment status dataset includes: location number, status bit, and time, and is obtained through full multi-protocol access, compatible with alarm hosts, PLCs, and sensor fire protection equipment. The raw message dataset includes: binary format messages and text format messages, which are uncompressed messages captured synchronously during the acquisition process. Data collection is prioritized to prevent information loss. Key parameter datasets, including pressure, current, and temperature, are acquired through a targeted acquisition mechanism. A dynamic acquisition mechanism combining regular periodic acquisition and event-triggered acquisition is employed. Regular periodic acquisition involves polling according to a configured cycle to balance resource consumption and status monitoring requirements. Event-triggered acquisition triggers acquisition immediately upon detecting equipment status changes or transient events, capturing transient process data of key parameters. The acquired multi-source fire protection datasets are temporarily stored and categorized using an edge-side temporary buffer, preserving complete original messages and parsed fields without any compression. This provides comprehensive raw data support for subsequent data processing, storage, and analysis. A historical labeled sample dataset is constructed as a training data source. This dataset includes historical events and their corresponding transient feature vectors and risk levels labeled with rules. The samples must cover various operating scenarios, including different equipment models and operating conditions (e.g., full-load start-up, no-load start-up, low-temperature start-up, and high-humidity environment operation).

[0035] The annotation process employs an automated initial annotation method, initially matching risk levels based on transient feature-related standards. Domain experts then verify and correct the results, focusing on the core dimensions of transient features and the actual impact range of the event. Measured data is collected from operational data of actually deployed equipment in different application scenarios, while simulation data is supplemented by simulating extreme conditions to address scarce samples. A balance is achieved by synthesizing and expanding minority class samples based on similar features, selecting core representative data from the majority class samples, and combining feature engineering optimization to improve the recognizability of differentiated features. A stratified sampling method is used to ensure consistent distribution of samples across equipment models, operating conditions, and risk levels in the training, validation, and testing phases, avoiding data bias that could affect the accuracy of model evaluation. Simultaneously, duplicate samples are excluded, and data from different parts is ensured to be non-overlapping, guaranteeing the reliability of the model's generalization ability assessment.

[0036] For multi-source fire protection datasets, a linear transformation to a unified interval is performed using minimum-maximum normalization mapping. For equipment status data, a unified representation dimension is achieved using numerical mapping or one-hot encoding. All normalization operations are executed synchronously at the edge and during data storage, preserving the bidirectional mapping relationship between the original and normalized values. Unified protection is ensured through protocol compatibility, time synchronization, and breakpoint resumption. Protocol compatibility includes support for RS-485, Modbus, OPCUA industrial protocols, and the T26875 fire protection equipment standard protocol, enabling full access for multiple devices. Time synchronization ensures the consistency of the time axis between conventional and high-frequency data acquisition, guaranteeing data time sequence consistency. Breakpoint resumption relies on edge caching nodes to locally cache data during communication interruptions, and retransmits it in timestamp order after the link is restored, ensuring data continuity.

[0037] In this implementation plan, through multi-protocol compatibility, dynamic triggering and collaboration, and dynamic acquisition mechanisms, the full capture of multi-source fire protection data and accurate recording of transient processes are achieved. The complete original data is temporarily stored and preserved at the edge, and the data dimensions are unified by normalization and encoding processing. Relying on multi-protocol compatibility, time synchronization, and breakpoint resume mechanisms, comprehensive, consistent, and continuous high-quality data support is provided for data processing, storage, and analysis.

[0038] Specifically, the process of constructing a full-cycle event window based on a multi-source fire protection dataset for integrity verification is as follows: Inputting equipment status transition signals, transient event trigger commands, and the multi-source fire protection dataset addresses the issue of missing process data caused by the sampling cycle. Equipment status transition signals refer to the transition from normal to alarm and from alarm to recovery. Transient event trigger commands include sudden increases in pump starting current and sudden changes in pipeline pressure. Based on event characteristics and equipment operating patterns, corresponding pre-alarm and post-alarm duration parameters are matched and dynamically adjusted according to equipment type and actual application scenario. A full-cycle time window is constructed and cached in a time-series database. A linkage mechanism between the event window and data acquisition is established, binding the transient event trigger command with the window start and end times through timestamp alignment to achieve targeted acquisition of key parameters within the window. An interrupt-triggered acquisition logic is adopted, immediately switching the acquisition frequency of the corresponding equipment after the window starts to target and capture key parameters such as pressure fluctuations, temperature rises, smoke concentration changes, and starting current, avoiding the omission of transient processes by the sampling cycle. A circular buffer and integrity verification rules are introduced to check the temporal continuity and amplitude rationality of the cached data. The circular buffer adopts a non-volatile storage design, supports data retention during power outages, and its capacity is configurable to adapt to concurrent event processing needs; timing verification identifies missing data through millisecond-level timestamp difference analysis and triggers a local data acquisition mechanism; amplitude verification constructs a dynamic threshold system based on the rated parameter range of the equipment, marks out-of-range data with anomaly tags, and retains the original values; it outputs standardized window datasets and window metadata; the window metadata includes: window start and end times, trigger event type, and a list of associated devices, so that each alarm event has reproducible physical process data support.

[0039] like Figure 3The multi-stage intelligent sensing transient event analysis diagram of the smoke sensor shown is used to completely record the dynamic process of smoke concentration from normal to alarm and then to recovery. Using time as the horizontal axis and smoke concentration as the vertical axis, it clearly presents real-time monitoring data and threshold triggering logic. The blue curve represents real-time smoke concentration changes, and the green, orange, and red dashed lines correspond to the normal threshold, warning threshold, and alarm threshold, respectively. The entire process is divided into stages by background color and key node markings: The initial stage is the normal monitoring stage (low sampling rate), where the smoke concentration is stable near the normal threshold; when the concentration approaches the warning threshold, it enters the warning monitoring stage (medium sampling rate), triggering a warning when the concentration crosses the orange dashed line (marked with an orange dot); subsequently, the concentration continues to rise and crosses the red dashed line, triggering a formal alarm at the red dot, and simultaneously entering the alarm processing stage (high-frequency sampling); thereafter, the smoke concentration rapidly climbs, reaching a peak of 14.82 at the purple dot; after the concentration begins to decrease, it sequentially enters the rapid recovery stage and the stable recovery stage (sampling rate gradually decreasing) until it falls back below the normal threshold near the green dot, marked as recovered to normal. It intuitively demonstrates the core working logic, accurately capturing the instantaneous peak of smoke concentration while fully preserving the dynamic trajectory of changes from normal state, early warning, alarm to recovery and stabilization, clearly reflecting the effectiveness of the multi-stage intelligent sensing strategy.

[0040] In this implementation plan, an event lifecycle window covering the entire physical process before and after an alarm is constructed based on device status change signals and transient event triggering commands, which significantly improves event processing capabilities. It fully tracks the dynamic changes of the alarm process, and adopts a circular buffer and integrity verification mechanism to ensure data continuity and consistency. It provides accurate event tracing and historical data re-parsing capabilities, which greatly improves reliability, data processing efficiency and emergency response capabilities.

[0041] Specifically, the process of constructing a ternary association index for events, time-series data, and raw messages is as follows: Input the raw message dataset and window dataset; construct a ternary association index for events, time-series data, and raw messages, using the window ID as the core association key; when storing window data in the cloud time-series database, simultaneously record the storage path of the raw messages for the corresponding time period, forming a chained index of event records, window data, and raw messages; archive raw messages in layers and blocks according to device ID and time dimension, with each message accompanied by a millisecond-level timestamp and parsing status marker; construct a dual-write channel and index update mechanism for raw messages, simultaneously completing semantic field writing and raw message archiving during the protocol parsing stage, and the parsing module extracts points... When the structured fields of number and status bits are written to the business table, the complete binary and text messages are written to the standardized raw message library through an independent I / O channel, and the message index is embedded in the structured record. The index information is updated synchronously with the data. A strict consistency guarantee strategy is used to build a solid foundation for data reliability: the writing process adopts a logical sequence of raw message priority and structured confirmation lag. The raw message is archived and the index is temporarily stored first. The writing of structured fields is triggered only after successful confirmation to avoid data loss. For failure rollback, a transactional verification mechanism is introduced. Dual write operations either succeed or roll back completely. Instantaneous failures trigger a limited number of retries. Logs are recorded and alarms are set after failure to ensure the spatiotemporal correspondence between structured data and raw messages.

[0042] A protocol adaptation and process reconstruction algorithm based on original messages is introduced to support historical data re-parsing and event detail tracing, constructing a historical event and message association dataset. This dataset includes: event timestamp, event type, associated device ID, confirmed associated original message timestamps, and message attribute identifiers. The time association term is obtained by multiplying a time weight coefficient by an exponential function term, and the exponential function term is obtained by applying an exponential function to the time difference between the event occurrence time and the message time, with the time decay coefficient applied to it. The attribute weight coefficient is multiplied by an attribute matching indicator function to obtain the attribute matching term. The association degree is obtained by summing the time association term and the attribute matching term. The specific formula for calculating the association degree is as follows:

[0043]

[0044] In the formula, This represents the correlation value between event k and message i, used to measure the overall matching degree between the event and message in terms of time and attributes; This represents the time-related weighting coefficient, obtained through statistical analysis of historical event and message association datasets. Its value ranges from 0.3 to 0.7 and is used to adjust the contribution weight of time in the correlation calculation. The time decay term is an exponential decay function built based on the time difference between the event and the message, used to quantify the proximity of the two in the time dimension; This indicates the time when event k occurred, which is the point in time when the event was recorded through event monitoring. This indicates the time when message i was generated, which can be directly obtained from the message's timestamp field, reflecting the specific time when the message was generated; This represents the time decay coefficient, which is determined by analyzing the time difference distribution between events and associated messages in a historical event and message association dataset. The value ranges from 100 to 1000 and is used to control the decay rate of the time difference on the association degree. This represents the attribute matching weight coefficient, which is also calibrated through correlation analysis of historical events and message association datasets. The value range is 0.3-0.7, and it is used to adjust the contribution weight of attribute identifier matching in the correlation calculation. This indicates an attribute matching indicator function, which is used when the characteristic attribute identifier of event k is... The characteristic attribute identifier of message i When they are completely identical, the function value is 1; otherwise, it is 0. It is used to quantify whether the two match in key attribute dimensions. The characteristic attribute identifier of event k is a key attribute used to distinguish the category or source of the event (such as device ID, event type code), which is extracted from the event's metadata information; The characteristic attribute identifier of message i is the key identification information carried by the message (such as the sending device ID and message type code), which is obtained by parsing the fields of the message and is used to match and verify the characteristic attributes of the event.

[0045] When the protocol is upgraded or new fields are added, the original message for the target time period is located by indexing, and the new parser is called to perform batch re-parsing to supplement the missing new fields; in the responsibility evidence collection scenario, the original message is retrieved according to the event timestamp and device ID, the original instruction sequence of host linkage command and PLC fault code is restored, and the integrity of the structured record is verified; the output includes an event archive containing a ternary association index and a standardized original message library that can be re-parsed.

[0046] This implementation scheme effectively solves the problems of data loss and untraceability of information in traditional technologies by constructing a ternary association index of events, time-series data, and original messages. By introducing dual-write channels and an index update mechanism, the consistency and integrity of original messages and structured data are ensured, and a transaction rollback mechanism guarantees data reliability. By calculating the correlation degree through time and attribute matching, protocol adaptation and process reconstruction are achieved, providing reliable data support for accountability and protocol upgrades.

[0047] Specifically, the process of implementing dual-track storage of semantic parsed data and raw messages based on parsed structured fields, window datasets, and collected raw message datasets, and constructing a multi-dimensional index association, is as follows: A hierarchical structured storage mechanism for semantic data is constructed, storing the parsed fields of device status and window datasets according to business scenarios. A relational database on the edge side stores local linkage control-related data, with tables partitioned by device ID and date to improve query efficiency. A standardized event table is synchronized to the cloud to support global statistics, optimizing aggregation analysis speed by region and time partitioning. A full archiving rule for raw messages is established, opening an independent channel for raw messages in parallel during the protocol parsing phase. Complete raw messages are archived to edge-side object storage in the order of device ID, timestamp, and protocol version, employing a local dual-copy and scheduled cloud backup strategy. Local copies are stored on different physical media, and cloud synchronization is incrementally based on timestamps to avoid message loss due to single-point failures. An association index table is constructed with the event window ID in the window dataset as the core, recording window elements... The system includes data and semantic data storage paths, original message storage locations, and protocol version information. The index table is updated in real-time as data is generated, ensuring immediate association with new data. A multi-condition query interface is designed to support retrieving associated data by window ID, device ID, and timestamp, enabling chained calls to events, window data, and original messages. Path normalization constraints are applied to dual-track data storage; both semantic data and original message storage paths include unique identifier fields. An index encoding function is introduced to uniformly map multi-dimensional identifiers, concatenating multi-dimensional identifier parameters into a string in sequence, and then performing a hash operation on this concatenated string to output a globally unique index value. To address the risk of SHA-256 hash collisions, a dual verification and collision marking mechanism is adopted: after generating the globally unique index value, uniqueness is verified by bidirectional comparison between the original concatenated string and the index value. If a collision is detected, a random collision marker is automatically appended to the index value, and the hash operation log is updated synchronously to trace the source of the collision. The specific calculation formula for the globally unique index value is as follows:

[0048] ,

[0049] In the formula, It represents a globally unique index value, which is generated by hashing a multidimensional identifier. It has a fixed length and is used as a unique identifier for data records in edge and cloud storage. It ensures that the path prefix on different storage media corresponds one-to-one with the index key, enabling efficient location and related querying of cross-end data. The index encoding function receives four types of multi-dimensional identifier parameters: device ID, event window ID, timestamp, and protocol version number. It outputs a unique index key through string concatenation and hash operation, realizing a standardized mapping from multi-dimensional information to a single index. This represents a string concatenation operation for multidimensional identifiers. It combines information from each dimension into a continuous string in the order of "Device ID, Event Window ID, Normalized Time Identifier, Protocol Version Number" to serve as the input source for a hash function, ensuring the uniqueness of the input. The device ID is a unique identifier that distinguishes different physical devices or logical units. It is assigned when the device leaves the factory and is used to locate the device source dimension of the data. The event window ID is a unique identifier for a window that divides consecutively occurring events according to time or logical rules. It is used to aggregate discrete events by window, which facilitates the grouping and analysis of batch data. Representing a millisecond-level timestamp, it records the precise time when an event or data was generated, provided by the device clock, and reflects the temporal dimension characteristics of the data; Represents a time-normalized integer identifier, obtained by using the original timestamp. Divide by time normalization precision It also rounds down the time, discretizing the continuous millisecond-level time into fixed-granularity time units, reducing the precision redundancy in the time dimension, and balancing index precision and storage efficiency. This indicates the time normalization precision, which is obtained based on the time granularity requirements of the business scenario and is used to control the granularity of time discretization. This indicates the protocol version number, identifying the protocol version followed by data transmission or processing. It is used to ensure compatibility with data format differences under different protocol versions and to avoid index conflicts caused by protocol iterations.

[0050] Ensure consistent data location across levels and devices; the index table field format is uniformly structured key-value pairs, compatible with the retrieval syntax of both the edge and cloud, guaranteeing rapid access to semantic data for local linkage control, supporting historical data re-parsing based on original messages during protocol upgrades, and process backtracking during accountability investigations. Outputs include a local semantic database on the edge side, a standardized event table on the cloud, a standardized original message library, an original message index, and a multi-dimensional relational index table.

[0051] This implementation scheme achieves efficient data storage and precise data location by constructing a multi-dimensional indexing and association mechanism. Through precise calculations using timestamp and attribute matching, it supports protocol upgrades and historical data re-parsing, providing a reliable basis for accountability and data traceability. It offers comprehensive support for equipment management, fault detection, and process optimization, ensuring data accuracy and security.

[0052] Specifically, the process of building a collaborative data link between the edge and the cloud based on the incremental synchronization mechanism is as follows: Based on window datasets, semantic data, and original message indexes, incremental synchronization and cross-end penetration queries are performed between the edge and the cloud to solve the problems of data silos, full transmission bandwidth consumption, and low efficiency in tracing original data; the synchronized content includes edge-side window data, semantic structured fields, and original message index metadata, adopting a timestamp incremental strategy, and only transmitting data added after the last synchronization breakpoint when the network recovers to avoid duplicate transmission; a synchronization priority and state control mechanism is constructed, with the following state definitions: the initial state after data generation is unsynchronized, and after triggering the synchronization process, it changes to synchronized, at which point the data is locked and modification is prohibited; if the transmission is completed and the verification passes, it is updated to synchronized; if the transmission is interrupted or the verification fails, it is marked as synchronization failure, and the reason for the failure is recorded; a retry is initiated for data in the synchronization failure state, and if the retry is successful, it changes to synchronized; if the number of retries exceeds the threshold, the synchronization failure is maintained and an alarm is triggered.

[0053] Data is prioritized based on importance, with core process data taking precedence over regular data, and critical equipment data taking precedence over ordinary sensor data. Each data entry is marked as unsynchronized, in synchronization, or synchronized. In case of synchronization failure, the breakpoint is recorded and a resume is triggered. Edge caching ensures no data loss. A cross-platform query interface is designed based on a multi-dimensional relational index table. The cross-platform penetration query interface uses HTTPS protocol for communication, with data exchange format in JSON. The interface protocol defines request parameters including query conditions and user tokens, and response parameters including semantic data storage URL, original message storage path, data generation timestamp, and metadata version number. The consistency model adopts eventual consistency, and metadata version number verification ensures that the query results reflect the latest synchronization state between the edge and the cloud. It supports multi-condition queries based on event window ID, device ID, timestamp, and spatial location. The cloud metadata database is used to match the original message index to generate the edge-side access path. Access control and encrypted transmission mechanisms are introduced.

[0054] The encrypted transmission uses the TLS 1.3 protocol, with keys uniformly generated by a cloud-based key management server. Symmetric session keys are distributed to edge devices via asymmetric encryption, and keys are automatically rotated daily. Access control is based on the RBAC model, with user roles categorized as administrator (full permissions), operations and maintenance personnel (device-level query permissions), and auditors (read-only permissions). Permission verification is implemented using JWT tokens, which contain role identifiers and validity periods. Querying requires verification of user permissions. Original messages are transmitted through an encrypted channel to ensure data security. The system outputs incremental updates to the time-series database, window metadata, and cross-device linked query interfaces. This forms a traversal link, allowing administrators to directly trace original messages in the cloud without operating edge devices. The cross-device linked query interface, after receiving the event window ID, device ID, timestamp, and spatial location query conditions, returns query results containing the semantic data storage location associated with the target event, the original message storage location, and its access path. These results are recorded as cross-device query results for subsequent protocol upgrades, adaptations, and accountability verification.

[0055] As shown in Table 1, at 18:24 on November 5, 2025, an event window EW-20251105-00-008 was generated on floors 5-15 of Zone A. The corresponding data type is critical equipment data, the synchronization status is not synchronized, the semantic structured fields show that the equipment type is fire controller, the data priority is 2, the permission level is high, and the original message index is RAW-202511051824-00-008. This indicates that the event involves core fire control equipment, the data priority is high (priority 2), and it is only authorized for access by personnel with high-level permissions. Currently, it has not been uploaded to the cloud due to network or synchronization queue reasons. It is necessary to pay close attention to the subsequent synchronization progress to avoid the loss of critical equipment operation data. At the same time, the original instructions and status messages of the fire controller during this period can be traced through the original message index to provide the original basis for equipment operation status analysis. At 20:12 on November 5, 2025, an event window EW-20251105-02-006 was generated on the 3rd floor of Building C. The data type is ordinary sensor data, the original message index is RAW-202511052012-02-006, the synchronization status is synchronized, and the semantic structured fields indicate that the device type is fire controller, the data priority is 4, and the permission level is ordinary. This reflects the routine monitoring data of ordinary sensors in this area. The data priority is low (priority 4), suitable for ordinary administrators to view, and it has been successfully synchronized to the cloud. It can be included in the global fire status statistics. The original message index provides a traceability basis for subsequent sensor data accuracy verification and protocol parsing adaptation. At 17:01 on November 6, 2025, an event window EW-20251106-01-001 was created on the 18th floor of Building B (4th floor). The data type is regular data, the synchronization status is synchronized, and the semantic structured fields show the device type as fire controller, data priority as 3, the original message index as RAW-202511061701-01-001, and the permission level as ordinary. This indicates that the event is a routine data record of the fire controller's daily operation. The data priority (priority 3) is between critical data and ordinary sensor data, which can support daily status monitoring. The data has been synchronized to the cloud to ensure its integrity and reusability. The original message index ensures that if it is necessary to review the details of the device operation during this period (such as changes in status bits or fluctuations in basic parameters), the original message can be accurately located and retrieved for analysis.

[0056] Table 1 Multi-dimensional Association Index Table

[0057]

[0058] This implementation plan establishes a data collaboration link between the edge and the cloud through an incremental synchronization mechanism, ensuring real-time synchronization and efficient storage of fire protection data. To guarantee data security and privacy, encrypted transmission and access control mechanisms are employed. This ensures data integrity and consistency, enabling administrators to directly trace original messages in the cloud without operating edge devices, thus providing strong support for protocol upgrades and adaptations, as well as accountability verification.

[0059] Specifically, the process of constructing a multi-dimensional time-series chain for dynamic reconstruction of the entire alarm event cycle and visual replay of the process is as follows: Based on window data, semantic data, and window metadata synchronized in the cloud, alarm data reported asynchronously from multiple sources is uniformly aggregated and modeled, constructing a multi-dimensional time-series chain covering the pre-alarm, alarm, and post-alarm stages. This achieves refined reconstruction and unified expression of the dynamic process of alarm events, solving the problem of traditional processes being unreproducible. A unified time reference is adopted using a time synchronization mechanism, synchronizing the clocks of each device to the cloud reference time via NTPv4 or PTP protocols. A tree structure is used to record the parent-child relationship between devices and areas, ensuring that data is associated according to spatial hierarchy. Using event window ID, device ID, and spatial location as multi-dimensional association keys, scattered window data, state change records, and linkage execution results are concatenated with millisecond-level timestamps to form a complete time-series sequence. Time axis normalization and resampling processing are performed on data sources from different sampling frequencies. Linear interpolation is used to preserve abrupt change characteristics, and cubic spline interpolation is used to smooth trends. At the same time, a hysteresis compensation correction strategy is used to eliminate transmission delays between devices. Pressure, current, and temperature are integrated into the time-series chain. Key quantities such as temperature and smoke concentration are aligned to a unified time step, and the original sampling point markers are retained for subsequent detailed analysis and evidence collection. Under the constraints of a unified time base and spatial topology, dynamic time warping of multiple devices and multiple regions is performed to align and associate time sequences, elevating single-point curves to regional and system-level linked process views. Data from different devices are aligned along the time axis, and data from detection points in the same region are associated through spatial location. Key nodes for alarm triggering or pump start-up are marked using window metadata. For complex scenarios involving cross-windows, multiple alarms, and long linkage chains, strategies such as front-and-back window splicing, overlapping time period deduplication, and master-slave event marking are adopted to avoid splitting or double-counting the same physical process, ensuring the consistency of the time sequence chain in logic and time. After completing the reconstruction of the multi-dimensional time sequence chain, the reconstruction results are transformed into a visually appealing dynamic process description and configuration file, driving the front end or monitoring screen to generate an interactive process playback interface, outputting visual dynamic process curves and event replay views, and synchronously storing the multi-dimensional time sequence chain organized by event window ID, device ID, and spatial location into a cloud time sequence database to form a cloud-based multi-dimensional time sequence chain.

[0060] The visualization description includes the type of physical quantity of the curve, the range of the coordinate axis, the position of the threshold line, the background color segment of the stage, and the information of the key event annotation points. It supports one-click playback of the fire development process by event window ID. Through time axis dragging, speed playback, curve overlay and device filtering interaction, it intuitively shows the whole process of smoke or temperature slowly rising from near the normal threshold, exceeding the warning or alarm threshold, triggering the linkage equipment and until it returns to stability. It realizes the full-cycle dynamic reconstruction of alarm events from the data layer to the visualization layer, and provides visual evidence support for false alarm identification, strategy optimization and responsibility determination.

[0061] like Figure 4 The multi-dimensional time-series reconstruction and dynamic process curves shown are used to present the dynamic changes and state transitions of multi-sensor data throughout the entire operation process. They display the real-time monitoring curves (different types of curves are distinguished) of four types of sensors: smoke, temperature, pressure, and current, along with corresponding warning and alarm thresholds (marked with dashed lines). Using time as the horizontal axis and sensor values ​​as the vertical axis, the correlation and change patterns of each parameter over time are intuitively presented. The operating stages are clearly divided by background color: the green area represents the initial normal operation stage, where the sensor curves fluctuate slightly around normal values, reflecting the initial stable state; the light yellow area represents the warning monitoring stage, where the smoke and temperature curves begin to rise slowly, while the pressure and current parameters remain stable, indicating potential abnormal signals; the red area represents the coordinated response stage, where the current and pressure curves rise synchronously and remain high, while smoke and temperature continue to rise, corresponding to the proactive response process of pump startup and valve action; subsequently, the background returns to green, and the various sensor curves gradually fall back to the stable range, indicating that the equipment has completed abnormal response and entered the recovery stage. This diagram clearly reflects the dynamic response logic of the entire process from normal operation, abnormal warning, coordinated handling to restoration of stability. Through the time-series correlation and stage division of multi-sensor data, it intuitively shows the changing characteristics of each parameter under different operating conditions and the effect of coordinated handling.

[0062] This implementation plan effectively solves the problems of unreproducible alarm events and data inconsistencies by constructing multi-dimensional time-series chains and reconstructing dynamic processes. By transforming the reconstructed time-series chains into visualized dynamic process curves, it provides an intuitive display of the entire fire development process, supporting interactive playback and data analysis. It offers complete dynamic reconstruction of alarm events and provides strong visual evidence support for false alarm identification, strategy optimization, and accountability.

[0063] Specifically, the process of transient process analysis based on cloud-based data collaboration links and window datasets is as follows: Input the cloud-based multidimensional time series chain, window dataset, and equipment operation logs to perform feature analysis and optimization on the fire transient process, solving the problems of difficult equipment fault diagnosis and lagging linkage strategies caused by the inability to quantify process details; the equipment operation logs include: equipment start-up and shutdown records, operation mode switching records, fault code and alarm code reporting records, and manual intervention operation records; the transient process includes: pump start-up current peak change, pipeline water hammer pressure fluctuation, smoke concentration rise rate, and temperature change gradient. By extracting the peak value, slope, and duration characteristic parameters of the time series curves, a correlation model between process characteristics and equipment status is established.

[0064] The weight vector is obtained by training a minimized prediction fit on a historical labeled sample dataset. This weight vector is then multiplied by the key feature vector of the transient process to obtain a linear term. The feature vector is then multiplied by the weight matrix on the left and then by the feature vector on the right to obtain half of a quadratic term. Adding the bias term yields the model's independent variables. These independent variables are then fed into a compression function for activation, outputting a comprehensive transient risk score. The specific formula for calculating the comprehensive transient risk score is as follows:

[0065]

[0066] In the formula, The comprehensive transient risk score of device d under event e can be directly used as the core basis for risk level classification. The S-shaped compression function, which uses the sigmoid function, maps the calculation results of linear combination and quadratic interaction terms to a unified [0,1] interval. This avoids excessively large score ranges caused by differences in the dimensions of different features and numerical fluctuations. At the same time, it gives the score a probabilistic meaning, which is convenient for formulating a unified risk classification standard. This represents the key feature vector of the transient process, with a dimension of 5×1, composed of five transient features. All features have undergone interval normalization, specifically including: the normalized ratio of the pump starting current peak value, calculated by dividing the actual peak current of device d during the starting phase in event e by the rated current of the device (taken from the device's manufacturer's manual), specifically characterizing the risk of motor damage caused by current overload during the starting phase; the normalized ratio of the effective value of water hammer pressure fluctuation, obtained by calculating the root mean square value of pipeline pressure fluctuation in event e and dividing it by the rated operating pressure of the device, used to quantify the risk of pipeline leakage and loosening of interfaces caused by water hammer effect; and the time normalized slope of the smoke concentration rise rate, calculated by taking the smoke concentration within the event window from the pre-set value in the fire protection code. The time interval between the alarm threshold rising to the alarm threshold is used as the denominator, and the concentration difference is used as the numerator to calculate the original slope. This slope is then normalized by dividing by the safe ramp-up rate threshold, accurately characterizing the speed risk of fire spread or smoke leakage. The temperature change gradient safety normalization ratio is calculated by dividing the maximum temperature change per unit time in event e by the safe temperature rise gradient allowed by the equipment or environment. This is used to characterize the risk of equipment insulation aging and fire expansion caused by a sudden temperature rise. The abnormal duration standard normalization ratio is calculated by dividing the duration for which equipment parameters in event e exceed the safe threshold by the standard fire response time and normalizing. This measures the cumulative risk of abnormal states not being dealt with in a timely manner. Together, these five factors comprehensively cover the core risk dimensions of transient processes. It is a 5×1 one-dimensional weight vector, with each element corresponding to the linear contribution weight of the five major features. It is obtained by minimizing the prediction fitting training on the historical labeled sample dataset. The samples need to cover the operation data of different equipment models and different operating conditions (such as full load start, no load start, low temperature start, and high humidity environment operation), and the labeling categories need to include four categories: normal operation, minor abnormality, moderate risk, and serious failure. This is a 5×5 symmetric quadratic weight matrix. The diagonal elements are used to amplify the nonlinear contribution of individual features, while the off-diagonal elements specifically characterize the interactive amplification effect between different features. The matrix needs to be compared with the weight vector. Collaborative training based on the same batch of historical labeled sample datasets ensures accurate capture of feature collaboration risks in actual operation; The bias term is determined by calibrating the distribution characteristics of historical labeled sample datasets and risk assessment requirements. Its value range is [-2, 2], which is used to flexibly adjust the scoring baseline and avoid the majority of samples being concentrated in the low score segment, resulting in insufficient risk discrimination.

[0067] Operational data covering the entire operating conditions and lifecycle of the target equipment needs to be collected. This data includes both measured and simulated data. Measured data is obtained from multi-source fire protection data and maintenance records collected in real time by sensors throughout the entire process, from factory commissioning, daily operation, abnormal alarms, maintenance, to decommissioning. Simulation data is generated through a digital twin model of the equipment, supplementing the scarce extreme operating condition data from actual measurements, such as ultra-high pressure start-up and extreme temperature and humidity. Each set of operational data must include complete data on the five major transient characteristics and corresponding risk level labels. An appropriate training method should be selected based on the data scale and distribution characteristics. Supervised learning as the foundation, self-supervised learning for blind spot filling, and thresholding are employed. The learning calibration hybrid framework employs the least squares method to fit the mapping relationship between features and risk levels when the sample size is sufficient and the distribution is balanced. When the sample size is insufficient or there are long-tailed outliers, the latent features are extracted by self-supervised learning, and the robustness of parameters is improved by Bayesian estimation. At the same time, a dynamic threshold learning mechanism is introduced to calibrate the risk level judgment boundary and adapt to different scenarios. After training, a portion of the records not involved in the training should be reserved as validation data and multi-dimensional verification should be performed to ensure the accuracy of risk level judgment before it can be put into actual use. It also supports quarterly iteration updates based on newly added operational data to achieve adaptive matching for different equipment models and different application scenarios.

[0068] Risk classification standards according to The values ​​are clearly divided into four levels; when the transient comprehensive risk score is within the range of the initial value to the first threshold, it is considered a safe level, indicating that the transient process is stable and without abnormalities, the equipment is operating well, and no additional intervention is required; when When the temperature falls between the first and second thresholds, it is considered a sub-healthy level, indicating an abnormality that has not yet affected the core functions of the equipment. During the next routine maintenance, the components corresponding to the abnormal characteristics should be carefully inspected. For example, if the current is too high, the motor windings should be checked; if the pressure fluctuations are too large, the pipelines and valves should be checked. When the value falls within the range of the second to third thresholds, it indicates a risk level, signifying that the abnormal characteristics have shown a trend of synergistic amplification, accelerating equipment aging or reducing the reliability of the linkage response, requiring immediate special testing and parameter calibration; when When the value falls within the range of the third threshold to the maximum score, it indicates a severe risk level, signifying that there are significant safety hazards in the transient process, which may trigger equipment failure or secondary safety accidents at any time. The system should be shut down immediately to investigate the root cause of the failure, and the linkage strategy should be optimized simultaneously to prevent the risk from spreading.

[0069] A process anomaly detection and root cause analysis mechanism is constructed. By comparing actual transient curves with equipment rated parameter curves, anomalies such as current exceeding thresholds and excessive pressure fluctuations are identified. Combined with trigger event types in window metadata (e.g., manual start, alarm-linked start), the causes of anomalies are traced to conduct linkage strategy optimization simulations. The complete link is reconstructed based on the time sequence chain, and the response delay of each link (e.g., the time difference between alarm and sprinkler start) is quantified. Response speed is improved by adjusting linkage logic parameters (e.g., shortening valve opening delay). Transient process characteristic analysis reports, equipment optimization suggestions, and linkage strategy iteration schemes are output. This provides data-driven equipment management and fire linkage optimization, reducing fault misjudgments and strategy inefficiencies caused by process ambiguity.

[0070] This implementation plan effectively solves the problems of difficult equipment fault diagnosis and lagging linkage strategies in transient process analysis by constructing a process based on incremental synchronization mechanism and multi-dimensional time-series chain. It also achieves intelligent monitoring of equipment status through risk level classification. The linkage strategy has been optimized, improving response speed. By dynamically adjusting linkage logic and parameters, the operating efficiency of equipment and the linkage effectiveness of the fire protection system have been significantly improved, ensuring efficient coordination of fault identification, strategy optimization, and safety protection.

[0071] Specifically, the process of protocol upgrade adaptation and liability verification based on the association between the original message dataset and the multi-dimensional index is as follows: Figure 5As shown, the input includes a standardized raw message library, a multi-dimensional related index table, and the results of cross-platform linked queries. The acquired data undergoes pre-screening and scenario classification, then proceeds to the protocol upgrade and adaptation process and the full-link responsibility evidence collection process. This completes the semantic supplementation of historical data and the construction of the event evidence chain, resolving the issues of historical data not being re-parseable and the lack of direct evidence for event responsibility determination. In the protocol upgrade and adaptation phase: when the equipment manufacturer updates the protocol (e.g., adding fault codes or adjusting status bit definitions), the original message for the target time period is located through the index table, and the new parser is called to re-parse it, supplementing the missing semantics of new fields. The parsing results are stored in parallel with the original structured data, without affecting the existing business table structure. The responsibility evidence chain is then constructed. The mechanism employs protocol-based differential comparison and cross-verification of data at the protocol, time-series, and physical process layers, performing hierarchical association and cross-validation. In false alarm tracing and accident analysis scenarios, it associates original messages with event window IDs to reconstruct details and forms a complete chain of evidence by combining the physical process curves of the window data. Utilizing hash algorithms and blockchain technology, it introduces an evidence chain anti-tampering mechanism and storage strategy, uniformly including key original message fragments, index entries, and reconstructed time-series summaries within the verification scope. The anti-tampering mechanism uses hash algorithms to calculate hash values ​​for data at the protocol, time-series, and physical process layers and signs the data using digital signature technology, ensuring that the data is not tampered with during storage, transmission, and access. The storage strategy, combined with blockchain technology, stores data in a distributed ledger, generating immutable historical data records, and generating new blocks with each update, ensuring that timestamps cannot be forged. The hierarchical correlation and cross-validation of data are performed using the following rules: Protocol layer validation rules use protocol differentiation comparison to check whether changes to protocol fields comply with specifications; Time sequence layer validation rules use time sequence alignment algorithms to verify timestamp consistency, ensuring the accuracy of data order and timestamps; Physical process layer validation rules use threshold detection and abnormal fluctuation detection to check the validity of sensor data, ensuring that the data conforms to physical laws; Cross-validation rules use cross-relationship analysis to verify the logical consistency of data between different devices; Hash value verification is performed on the original message and associated index to generate corresponding fingerprint values ​​and write them to read-only storage or third-party trusted evidence storage media in chronological order, ensuring that the evidence data has not been tampered with throughout the entire process of collection, transmission, storage, and retrieval, meeting the compliance requirements for auditing and liability determination.

[0072] Based on different usage scenarios, various output formats are automatically generated, including supplementary historical data results after protocol upgrades, field comparison tables, parsing difference lists, and responsibility evidence chain reports. The responsibility evidence chain report includes: original message fragments, time-series curves, command reconstruction records, and key hash verification information. This provides backward compatibility and forward scalability data support for fire protection protocol evolution, and provides an immutable original evidence foundation for responsibility determination, event auditing, and third-party assessments. It achieves an upgrade from single-point recording to a full-link, verifiable, and reproducible data management and evidence collection system.

[0073] This implementation plan addresses the issues of historical data being unresolvable and the lack of direct evidence for event liability determination by associating the original message dataset with a multi-dimensional index. It ensures data integrity and consistency, guarantees data tamper-proofing and evidence preservation strategies, and provides robust data support for liability identification, event auditing, and protocol evolution, achieving a full-chain, verifiable, and reproducible data management and evidence collection system upgrade.

[0074] Specifically, the second aspect of this invention provides an intelligent management device for fire protection systems based on smart property management, applied to an intelligent management method for fire protection systems based on smart property management. It includes: a multi-source fire protection data full-volume acquisition module, used to acquire multi-source fire protection datasets through a multi-protocol compatible and dynamically triggered collaborative acquisition mechanism. Using edge computing technology, the data is temporarily classified and stored in a temporary buffer at the edge side, and the data is normalized to ensure the consistency and availability of data from different sources. Through a dynamic triggering mechanism, the module automatically adjusts the acquisition frequency when equipment status changes or transient events occur, accurately capturing key transient data and avoiding information loss. An event window caching and time-series management module is used to construct a full-cycle event window based on the multi-source fire protection dataset, performing integrity verification on the data within each event window to ensure no data loss or corruption. A precise indexing system is established through a ternary association index of events, time-series data, and original messages, supporting efficient querying and analysis of original data, time-series data, and event data. The dual-track data storage and indexing module is used for dual-track storage of semantically parsed data and raw messages based on parsed structured fields, window datasets, and collected raw message datasets. It establishes multi-dimensional indexing associations to ensure rapid data location and retrieval. This dual-track storage and indexing mechanism effectively solves the problem of being unable to parse historical data during protocol upgrades and ensures accurate reconstruction of each event and its associated data when accountability is required. The edge-cloud data collaboration and time-series reconstruction module is used to build an edge-cloud data collaboration link based on an incremental synchronization mechanism. It constructs a multi-dimensional time-series chain for dynamic reconstruction of the entire alarm event lifecycle. Through dynamic reconstruction of the event process using the multi-dimensional time-series chain, it achieves full-cycle visual replay of alarm events. Whether it's alarm triggering, process evolution, or recovery handling, all data can be clearly displayed through the time-series chain. The intelligent analysis and full-link traceability application module is used for transient process analysis based on the cloud data collaboration link and window datasets. It can also adapt and re-parse historical data after protocol upgrades. Protocol upgrades and adaptations, as well as responsibility verification, are performed based on the association of the original message dataset with multi-dimensional indexes. Through precise event timestamps and data indexes, the specific process of the failure can be reconstructed, ensuring a complete chain of evidence when tracing the incident.

[0075] This implementation plan ensures efficient collection, storage, and processing of fire protection data through multi-protocol compatibility, dynamic triggering acquisition, multi-dimensional time-series chain construction, and incremental synchronization mechanisms. The device includes modules for multi-source data acquisition, event window caching and time-series management, dual-track data storage and indexing, edge-cloud data collaboration and time-series reconstruction, and intelligent analysis. It enables dynamic reconstruction of the entire event process, adaptation of historical data after protocol upgrades, and accountability tracing, thereby improving the real-time performance, traceability, and fault diagnosis capabilities of the fire protection system.

[0076] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0077] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for intelligent management of a fire-fighting system based on a smart property, characterized in that, Comprise the following steps: S1, through the multi-protocol compatible and dynamic trigger cooperative collection mechanism, collect multi-source fire data set, through the edge side temporary buffer to multi-source fire data set for classification and normalization processing; S2, based on the multi-source fire data set, construct the event full cycle window, carry out the integrity check, construct the event, time sequence data and original message three related index; S3, based on the analyzed structured field, window data set and collected original message data set, carry out semantic analysis data and original message double track storage, and construct multi-dimensional index association; S4, based on the incremental synchronization mechanism, construct the edge and cloud data cooperative link, construct the multi-dimensional time sequence chain to carry out the alarm event full cycle dynamic reconstruction, process visualization replay; S5, based on the cloud data cooperative link and window data set, carry out transient process analysis, based on the original message data set and multi-dimensional index association, carry out protocol upgrade adaptation and responsibility evidence collection; The specific process of the double track storage of the analyzed structured field, window data set and collected original message data set, semantic analysis data and original message, and the construction of multi-dimensional index association is: Construct semantic data hierarchical structured storage mechanism, store the analyzed field of device state and window data set according to business scenario classification, establish original message full quantity archiving rule, open original message independent channel in parallel in protocol analysis link, archive complete original message to edge side object storage according to device ID, time stamp, protocol version order, construct associated index table with event window ID in window data set as core, record window metadata, semantic data storage path, original message storage location and protocol version information; Apply path normalization constraint to double track storage data, the storage path of semantic data and original message contains unique identification field, introduce index coding function to unify mapping of multi-dimensional identification, splice multi-dimensional identification parameters into a string in order, then execute hash operation on the spliced string, output global unique index value; Output edge side local semantic database, cloud standardization event table, normalized original message library, original message index and multi-dimensional associated index table. 2.The intelligent management method of the intelligent property-based fire-fighting system according to claim 1, characterized in that: The specific process of collecting multi-source fire data set through multi-protocol compatible and dynamic trigger cooperative collection mechanism, and classifying and temporarily storing multi-source fire data set through edge side temporary buffer is: Collect multi-source fire data set through multi-protocol compatible and dynamic trigger cooperative collection, multi-source fire data set includes: device state data set, original message data set, key parameter data set; Dynamic collection mechanism of regular cycle collection and event triggered collection is adopted, the collected multi-source fire data set is classified and temporarily stored through edge side temporary buffer, and historical labeled sample data set is constructed as training data source; Through minimum maximum normalization mapping linear conversion to uniform interval for multi-source fire data set; For device state data, numerical mapping or one-hot encoding is adopted to unify the representation dimension, all normalization operations are executed synchronously with data temporary storage on the edge side, and unified guarantee is carried out through protocol compatibility, time synchronization and breakpoint continuation. 3.The intelligent management method of the intelligent property-based fire-fighting system according to claim 1, characterized in that: The specific process of constructing an event full-cycle window based on the multi-source fire data set and performing integrity checking is as follows: The input device state jump signal, transient event trigger instruction and multi-source fire data set are used to construct a full-cycle time window and cache to a time sequence database, a linkage mechanism of event window construction and collection is constructed, the transient event trigger instruction is bound to the start and end time of the window through timestamp alignment, an interrupt trigger type collection logic is adopted, the collection frequency of the corresponding device is switched immediately after the window is started, directional capture of pressure fluctuation, temperature rise, smoke concentration change and starting current core parameters is performed, a ring buffer and integrity checking rules are introduced, and the cached data is checked for time sequence continuity and amplitude rationality. 4.The intelligent management method of the intelligent property-based fire-fighting system according to claim 1, characterized in that: The specific process of constructing a ternary association index of events, time sequence data and original messages is as follows: The original message data set and window data set are input, a ternary association index of events, time sequence database and original messages is constructed, a double-writing channel of original messages and an index updating mechanism are constructed, semantic field writing and original message archiving are simultaneously completed in the protocol analysis link, a protocol adaptation and process restoration algorithm based on original messages is introduced, and a historical event and message association data set is constructed; A time weight coefficient is multiplied by an exponential function item as a time association item, and a time difference between event occurrence time and message time is an exponential function of exponential decay as an exponential function item; an attribute weight coefficient is multiplied by an attribute matching indication function as an attribute matching item, and a sum of the time association item and the attribute matching item is an association degree value; an event archive containing the ternary association index and a normalized original message library capable of being re-analyzed are output. 5.The intelligent management method of the intelligent property-based fire-fighting system according to claim 1, characterized in that: The specific process of constructing an edge and cloud data collaborative link based on an incremental synchronization mechanism is as follows: Based on the window data set, semantic data and original message index, incremental synchronization and cross-end penetration query between the edge and the cloud are performed, a synchronization priority and state control mechanism is constructed, a cross-end query interface is designed based on a multi-dimensional association index table, a permission control and encrypted transmission mechanism is introduced, and cloud incremental updates are output to the time sequence database, window metadata and cross-end linkage query interface. 6.The intelligent management method of the intelligent property-based fire-fighting system according to claim 1, characterized in that: The specific process of constructing a multi-dimensional time sequence chain for alarm event full-cycle dynamic reconstruction and process visualization replay is as follows: Based on the window data, semantic data and window metadata synchronized by the cloud, alarm data reported by multiple sources asynchronously is uniformly aggregated and modelized organized in space topology, a multi-dimensional time sequence chain covering before, during and after the alarm is constructed, under the constraints of a unified time reference and space topology, time sequence alignment and association logic of dynamic time warping of multiple devices and multiple regions are executed, and a single-point curve is promoted to a regional and system-level linkage process view; After the multi-dimensional time sequence chain reconstruction is completed, the reconstruction result is converted into a visual rendering-friendly dynamic process description and configuration file, a front end or a monitoring large screen is driven to generate an interactive process playback interface, a visual dynamic process curve and an event replay view are output, and the multi-dimensional time sequence chain organized according to the event window ID, device ID and spatial position is synchronously stored into the cloud time sequence database, forming a cloud multi-dimensional time sequence chain. 7.The intelligent management method of the intelligent property-based fire-fighting system according to claim 1, characterized in that: The specific process of transient process analysis based on the cloud data collaborative link and the window data set is as follows: The input cloud multi-dimensional time sequence chain, window data set and equipment operation log are used for feature analysis and optimization of the fire transient process. The weight vector is obtained by minimizing the prediction fitting training of the historical labeled sample data set. The weight vector is multiplied by the key feature vector of the transient process to obtain the linear term. The feature vector is left multiplied by the weight matrix and then right multiplied by the feature vector to obtain half of the quadratic term. Then, the bias term is added to obtain the independent variable of the model. The independent variable is sent to the compression function for activation, and the output comprehensive transient risk score value is output. The operation data covering the whole working condition and whole life cycle of the target equipment need to be collected. According to the data size and distribution characteristics, a suitable training method is selected. An abnormal process detection and root cause analysis mechanism is constructed. The abnormal reason is traced back to carry out linkage strategy optimization simulation. The transient process feature analysis report and linkage strategy iteration scheme are output. 8.The intelligent management method of the intelligent property-based fire-fighting system according to claim 1, characterized in that: The specific process of the protocol upgrade adaptation and responsibility evidence based on the original message data set and multi-dimensional index association is as follows: The standardized original message library, multi-dimensional association index table and cross-end linkage query result are input. The obtained data are pre-screened and classified into scene. They are respectively input into the protocol upgrade adaptation process and the whole link responsibility evidence process. The historical data semantic supplement and event evidence chain construction are completed. The responsibility evidence chain generation mechanism is constructed. The data of the protocol layer, time sequence layer and physical process layer are analyzed by protocol differentiation comparison and cross correlation analysis for step-by-step association and cross verification. The hash algorithm and block chain technology are used to introduce the evidence chain tamper-proof mechanism and storage strategy. The key original message segment, index item and reconstructed time sequence summary are uniformly included in the verification range. According to different use scenarios, various forms of output results are automatically generated. The historical data supplement results after protocol upgrade, field comparison table, analysis difference list and responsibility evidence chain report are obtained.

9. The intelligent management device for the fire-fighting system based on the intelligent property, which applies the intelligent management method for the fire-fighting system based on the intelligent property as claimed in any one of claims 1-8, characterized in that, It includes: A multi-source fire data full collection module is used to collect multi-source fire data sets through multi-protocol compatibility and dynamic trigger cooperative collection mechanism. The multi-source fire data sets are classified, temporarily stored and normalized by the edge side temporary buffer. An event window buffer and time sequence management module is used to construct event full cycle window based on multi-source fire data set, perform integrity verification, and construct three-element association index of event, time sequence data and original message. A dual-track data storage and index association module is used to perform semantic analysis data and original message dual-track storage based on the analyzed structured field, window data set and collected original message data set, and construct multi-dimensional index association. An edge cloud data cooperation and time sequence reconstruction module is used to construct edge and cloud data cooperation link based on incremental synchronization mechanism, construct multi-dimensional time sequence chain for alarm event full cycle dynamic reconstruction, and process visual replay. An intelligent analysis and whole link tracing application module is used for transient process analysis based on cloud data cooperation link and window data set, and protocol upgrade adaptation and responsibility evidence based on original message data set and multi-dimensional index association.

Images