Intelligent archival repository monitoring system based on Internet of Things

Abstract

The invention relates to the technical field of intelligent monitoring, and discloses an intelligent archival repository monitoring system based on Internet of Things, comprising: a plurality of edge sensing units deployed in a repository, monitoring parameter drift in real time and triggering a local perturbation event by dynamically learning a personalized stable state baseline of environmental parameters; cooperative judgment is achieved among the units through perturbation wake-up signals, concerned events are reported to a central platform only when preset conditions are met, and full-spectrum perception of slowly changing and sudden risks is formed through baseline learning and a cooperative wake-up mechanism of the edge units in combination with instantaneous event template matching and targeted enhanced monitoring; based on the dynamic health fingerprint analysis of the perturbation resonance characteristic spectrum, the stability degradation trend of the archive microenvironment can be pre-warned in a prospective manner, and the intelligent level and the preventive protection capability of the monitoring system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to an intelligent archive storage monitoring system based on the Internet of Things, belonging to the technical field of Internet of Things environment monitoring. Background Art

[0002] Currently, the environmental monitoring of archive storages mainly relies on a centralized Internet of Things monitoring architecture, that is, a large number of sensors (such as temperature and humidity, gas, vibration, etc.) are deployed to collect environmental data and upload it to the central platform for unified analysis. Although this mode can achieve full coverage, it exposes the inherent contradictions of information redundancy and superficial cognition in actual operation, specifically manifested as follows: 1. Perception overload and ineffective attention: A large number of sensors continuously generate data, and most of them are redundant information under normal conditions, resulting in the ineffective consumption of the computing power of the central platform. Instead, the weak abnormal signals that really need attention are ignored or delayed in response due to data flooding. For example, the continuous small drift of temperature and humidity in a local area is often filtered as noise by the system because it does not exceed the global fixed threshold.

[0003] 2. Mismatch between static thresholds and dynamic risks: The existing system relies on preset rigid thresholds to judge abnormalities and is difficult to adapt to the dynamic risk background caused by factors such as archive materials and seasonal changes. For example, the impact of the same humidity fluctuation on ordinary paper and precious silk archives is significantly different, but the existing technology cannot evaluate them differentially.

[0004] 3. Lack of association and insufficient early prediction: Risk judgment is usually based on independent analysis of a single parameter, lacking the ability to identify the synergistic effect of multiple factors (such as the coupling of temperature and humidity accelerating the deterioration of acidic paper). In addition, sudden instantaneous disturbances (such as impact vibration, gas instantaneous leakage) are often missed by the traditional continuous verification logic due to their short duration.

[0005] The industry has tried to optimize by increasing sensor density or introducing complex algorithms, but such improvements further exacerbate the data transmission and calculation burden, and still cannot solve the core contradictions of the lack of edge intelligence and low-efficiency collaborative perception. For example, some solutions use high-frequency sampling to capture instantaneous events, but due to the limited computing power of edge nodes, the false alarm rate soars and the energy consumption surges. Therefore, how to achieve early and accurate identification of environmental disturbances on low-power and low-cost edge devices, and distinguish local abnormalities from global interferences through a distributed collaborative mechanism, while covering both gradual accumulation and sudden instantaneous risks, has become a key technical problem to be solved in this field. Summary of the Invention

[0006] The present invention provides an intelligent archive storage monitoring system based on the Internet of Things, and its main purpose is to solve the problems of missed detection of weak abnormalities, misjudgment of dynamic risks, and insufficient collaborative identification of multi-source disturbances caused by information redundancy and superficial cognition in the existing centralized monitoring system.

[0007] To achieve the above object, the present invention provides an Internet of Things-based intelligent archive warehouse monitoring system, which includes: A plurality of edge sensing units, which are configured in the archive warehouse, and each edge sensing unit is configured to: Dynamically learn and continuously maintain a personalized steady-state baseline of at least one environmental parameter within its monitoring range, and the personalized steady-state baseline is obtained by processing the historical sampling values of the environmental parameter within a predetermined time window through a weighted moving average or exponential smoothing algorithm; Real-time monitor the current value of the environmental parameter, and when the drift of the current value from the personalized steady-state baseline reaches a predetermined length and deviates from a predetermined drift threshold, determine that a local perturbation event has occurred and enter the local alert state; When entering the local alert state, send a perturbation wake-up signal that does not include specific environmental parameter data but only includes its own identifier and perturbation wake-up status to its predefined neighboring edge sensing units; Receive the perturbation wake-up signals from its neighboring edge sensing units, and when the preset collaborative perturbation determination conditions are met, enter the collaborative attention state and send attention event information to the central processing platform. The preset collaborative perturbation determination conditions include that the edge sensing unit itself is in the local alert state and receives at least one perturbation wake-up signal from its neighboring edge sensing units, or the edge sensing unit receives perturbation wake-up signals from at least two different neighboring edge sensing units; and, a central processing platform, which is configured to: Receive and process the attention event information from the edge sensing units that enter the collaborative attention state to achieve the monitoring of the archive warehouse; According to the attention event information, request relevant edge sensing units to upload more detailed historical environmental parameter data as needed for in-depth risk analysis; Based on the statistics of the attention event information and the in-depth risk analysis results, dynamically adjust the predetermined drift threshold of the edge sensing unit or the parameters in the preset collaborative perturbation determination conditions.

[0008] Preferably, each edge sensing unit is further configured to: store at least one predefined feature sequence template corresponding to a specific type of instantaneous environmental disturbance event; while continuously monitoring the current value of the environmental parameter and performing personalized steady-state baseline learning and comparison, concurrently and in real-time match the current sampling data of the environmental parameter or the short-term feature sequence formed thereby with at least one feature sequence template; when the current sampling data or the short-term feature sequence formed thereby successfully matches any one of the feature sequence templates, it is determined that an instantaneous critical disturbance event corresponding to the template has occurred locally, and regardless of whether the edge sensing unit enters the local alert state or the collaborative attention state, directly send an instantaneous alarm message to the central processing platform to indicate the type, occurrence time, and location of the instantaneous critical disturbance event.

[0009] Preferably, the central processing platform is further configured to: after receiving the instantaneous alarm message, send a targeted enhanced monitoring instruction to the edge sensing unit where the instantaneous critical disturbance event occurs and its predefined neighboring edge sensing units, and the instruction is configured to cause the instructed edge sensing unit to temporarily increase its monitoring sensitivity or adjust the parameters for determining the local perturbation event in step (b) within a predetermined short time period.

[0010] Preferably, the power supply module of the edge sensing unit is configured to support long-life battery power supply or energy harvesting technology power supply.

[0011] Preferably, the communication module of the edge sensing unit is configured to communicate with neighboring edge sensing units and / or the central processing platform through low-power Bluetooth (BLE) or LoRa wireless technology.

[0012] Preferably, the power consumption of the local alert state of each edge sensing unit is higher than that of the silent monitoring state, and the power consumption of the collaborative attention state is higher than that of the local alert state.

[0013] Preferably, the central processing platform is further configured to: for one or more specific monitoring areas in a preset archive storage, or for one or more specific edge sensing units, long-term record and statistically analyze the resonance mode time series feature spectrum data composed of the frequency of entering the collaborative attention state within a predetermined time period, the combination pattern and quantity characteristics of the perturbation wake-up signals sent by neighboring edge sensing units required to trigger the collaborative attention state, the response lag time of entering the collaborative attention state, or the duration of the collaborative attention state.

[0014] Preferably, the central processing platform is further configured to: based on the evolution of the resonance mode time series feature spectrum data of a specific monitoring area or a specific edge sensing unit over time, construct and continuously update a dynamic health fingerprint for the area or unit that can characterize the comprehensive response characteristics and inherent stability of the corresponding archive microenvironment to external disturbances, where the feature vector of the dynamic health fingerprint It can be represented in the following form: , wherein, is the average number of triggered neighbors, is the average resonance brewing period, is the average duration of perturbation, is the recent occurrence frequency; is a preset weight coefficient.

[0015] Preferably, the central processing platform is further configured to analyze the long-term evolution trend of the dynamic health fingerprint, and when it detects that the dynamic health fingerprint presents a predetermined negative evolution pattern indicating that the potential risk of the file microenvironment is accumulating or its internal stability is continuously decreasing, generate a prospective trend-based risk warning information different from that based on the abnormality of the immediate environmental parameters.

[0016] Preferably, after generating the prospective trend-based risk warning information, the central processing platform is further configured to dynamically adjust the file inspection priority, optimize the allocation of preventive maintenance resources, or guide the scientific partition storage of files with different sensitivity levels.

[0017] Compared with the problems of the background technology, the beneficial effects of the present invention are: 1. Through the continuous learning of the personalized baseline of the monitoring parameters by the edge sensing unit, the system can capture the weak and continuous environmental drift that is difficult to detect by traditional fixed thresholds. When multiple neighboring units trigger the cooperative interaction of the perturbation wake-up signal based on the local baseline drift, the system can autonomously identify the potentially risky areas with spatial correlation. This spatio-temporal double verification mechanism not only avoids the misjudgment of isolated perturbations, but also can accurately locate the progressive environmental anomalies with low communication overhead through the logical superposition of the extremely simple signals between units, providing a critical time window for risk intervention.

[0018] 2. While continuously monitoring the baseline drift, the edge unit parallelly matches the predefined instantaneous perturbation feature templates to construct an independent alarm channel for sudden and highly dynamic risks (such as shock vibration, gas instantaneous leakage). When the central platform receives the instantaneous alarm, it immediately triggers the targeted enhanced monitoring mechanism to dynamically adjust the detection sensitivity and judgment parameters of the relevant area. This nested design of fast and slow monitoring logics enables the system to not only capture the trend characteristics of slow-changing risks, but also immediately respond to and secondarily verify sudden perturbations, forming a closed loop of multi-dimensional risk perception.

[0019] 3. Based on the collaborative attention event feature data recorded over a long period (such as the number of triggering neighbors, response lag time), the system constructs dynamic health fingerprints for different regions to reflect their environmental stability. By analyzing the temporal evolution of the perturbation resonance pattern, this fingerprint reveals the tolerance decay of the archival carrier to external perturbations or the trend of internal structural deterioration. Compared with directly monitoring environmental parameters, this reverse inference mechanism based on the system's behavior pattern can detect earlier the accumulation of microenvironment vulnerability caused by hidden factors such as material aging and improper storage density, providing a forward-looking guidance for preventive maintenance.

[0020] 4. Through edge intelligence processing such as local baseline calculation and collaborative judgment of perturbation signals, the system triggers detailed data inspection and communication link activation in specific regions under the state of collaborative attention or instantaneous alarm. This event-triggered dynamic power consumption strategy, combined with low-power power supply and communication technologies, enables the system to support ultra-large-scale node deployment and continuous operation for more than several years while maintaining high monitoring accuracy, significantly superior to the centralized architecture that relies on continuous data transmission.

[0021] 5. The central platform remotely optimizes the baseline tolerance and collaborative judgment threshold of the edge unit according to historical event data and manual verification results. This feedback mechanism not only avoids the interference of seasonal environmental changes on fixed parameters but also enables differential configuration for the archival characteristics (such as material sensitivity) of different regions. Thus, the system forms an autonomous evolution ability of monitoring - analysis - optimization, maintaining stable monitoring efficiency in the complex and changeable warehouse environment and avoiding the sensitivity drift problem caused by parameter solidification in traditional systems. Brief Description of the Drawings

[0022] Figure 1 It is the timing diagram of the interaction process for the present invention to identify and trigger alarms and enhance monitoring based on the instantaneous perturbation feature template; Figure 2 It is the schematic diagram of the process for generating trend warnings based on feature extraction and dynamic health fingerprint analysis of the present invention; Figure 3 It is the schematic diagram of the perturbation event triggering and collaborative attention process of the edge perception node of the present invention.

[0023] The implementation, functional features, and advantages of the present invention will be further described in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Embodiments

[0024] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0025] The embodiment of the present application provides an intelligent archival warehouse monitoring system based on the Internet of Things, and the system includes: A plurality of edge perception units, which are configured in the archive storage, and each edge perception unit is configured to: Dynamically learn and continuously maintain a personalized steady-state baseline of at least one environmental parameter within its monitoring range, and the personalized steady-state baseline is obtained by processing the historical sampling values of the environmental parameter within a predetermined time window through a weighted moving average or exponential smoothing algorithm; Real-time monitor the current value of the environmental parameter, and when the drift between the current value and the personalized steady-state baseline reaches a predetermined length and deviates from a predetermined drift threshold, determine that a local perturbation event has occurred and enter the local alert state; When entering the local alert state, send a perturbation wake-up signal that does not contain specific environmental parameter data but only contains its own identifier and the perturbation wake-up state to its predefined neighboring edge perception units; Receive a perturbation wake-up signal from its neighboring edge perception units, and when the preset collaborative perturbation determination conditions are met, enter the collaborative attention state and send attention event information to the central processing platform. The preset collaborative perturbation determination conditions include that the edge perception unit itself is in the local alert state and receives at least one perturbation wake-up signal from its neighboring edge perception units, or the edge perception unit receives perturbation wake-up signals from at least two different neighboring edge perception units; and, a central processing platform, which is configured to: Receive and process the attention event information from the edge perception units that enter the collaborative attention state to achieve the monitoring of the archive storage; According to the attention event information, request the relevant edge perception units to upload more detailed historical data of environmental parameters as needed for in-depth risk analysis; Based on the statistics of the attention event information and the in-depth risk analysis results, dynamically adjust the predetermined drift threshold of the edge perception units or the parameters in the preset collaborative perturbation determination conditions.

[0026] Preferably, each edge perception unit is further configured to: store at least one predefined feature sequence template corresponding to a specific type of instantaneous environmental perturbation event; while continuously monitoring the current value of the environmental parameter and performing personalized steady-state baseline learning and comparison, concurrently and in real-time match the current sampling data of the environmental parameter or the short-term feature sequence formed thereby with at least one feature sequence template; when the current sampling data or the short-term feature sequence formed thereby successfully matches any one of the feature sequence templates, it is determined that a corresponding instantaneous key perturbation event has occurred locally, and regardless of whether the edge perception unit enters the local alert state or the collaborative attention state, directly send instantaneous alarm information indicating the type, occurrence time, and location of the instantaneous key perturbation event to the central processing platform.

[0027] Preferably, the central processing platform is further configured to: after receiving the instantaneous warning information, send targeted enhanced monitoring instructions to the edge sensing unit where the instantaneous critical disturbance event occurs and its predefined neighboring edge sensing units, and the instructions are configured to enable the instructed edge sensing units to temporarily increase their monitoring sensitivity or adjust the parameters for determining local perturbation events in step (b) within a predetermined short time period.

[0028] Preferably, the power supply module of the edge sensing unit is configured to support long-life battery power supply or power supply by energy harvesting technology.

[0029] Preferably, the communication module of the edge sensing unit is configured to communicate with neighboring edge sensing units and / or the central processing platform through low-power Bluetooth (BLE) or LoRa wireless technology.

[0030] Preferably, the local alert state power consumption of each edge sensing unit is higher than the silent monitoring state power consumption, and the collaborative attention state power consumption is higher than the local alert state power consumption.

[0031] Preferably, the central processing platform is further configured to: for one or more specific monitoring areas in a preset archive storage, or for one or more specific edge sensing units, long-term record and statistically analyze the resonance mode time series feature spectrum data composed of the frequency of entering the collaborative attention state within a predetermined time period, the combined mode and quantity characteristics of the perturbation wake-up signals sent by neighboring edge sensing units required to trigger the collaborative attention state, the response lag time of entering the collaborative attention state, or the duration of the collaborative attention state.

[0032] Preferably, the central processing platform is further configured to: based on the evolution of the resonance mode time series feature spectrum data of a specific monitoring area or a specific edge sensing unit over time, construct and continuously update a dynamic health fingerprint for the area or unit that can characterize the comprehensive response characteristics and internal stability of the corresponding archive microenvironment to external perturbations, where the eigenvector of the dynamic health fingerprint can be represented in the following form: , where, is the average number of triggered neighbors, is the average resonance incubation period, is the average perturbation duration, is the recent occurrence frequency; is the preset weight coefficient.

[0033] Preferably, the central processing platform is further configured to analyze the long-term evolution trend of the dynamic health fingerprint, and generate a prospective trend-based risk warning information different from that based on the abnormality of the immediate environmental parameters when it detects that the dynamic health fingerprint presents a predetermined negative evolution pattern indicating that the potential risk of the file microenvironment is accumulating or its internal stability is continuously decreasing.

[0034] Preferably, after generating the prospective trend-based risk warning information, the central processing platform is further configured to dynamically adjust the file inspection priority, optimize the allocation of preventive maintenance resources, or guide the scientific partition storage of files with different sensitivity levels.

[0035] Meanwhile, to further improve the response accuracy and judgment consistency of the system in actual application scenarios, the central processing platform introduces a multi-period difference and normalization offset analysis mechanism by combining historical vector data when processing the evolution trend of the dynamic health fingerprint. Specifically, if any two core indicators (such as the average perturbation duration, collaborative brewing period) in three consecutive statistical periods deviate negatively by more than 10% from the historical mean, it is regarded as a significant evolution trend, and the system will immediately generate a prospective warning prompt to assist manual intervention judgment; and in the enhanced monitoring stage, the targeted enhanced monitoring instructions not only include the increase of the sampling frequency (for example, from once every 60 seconds to once every 20 seconds), but also synchronously adjust the judgment parameters, such as temporarily tightening the perturbation event trigger threshold to ±5% and shortening the baseline update window to 60 minutes, and enabling the high-sensitivity template matching logic. The specific setting of these parameters is dynamically optimized based on the historical response data of the same type of events to ensure the pertinence and effectiveness of the instruction execution, which all belong to the extended implementation methods known to those of ordinary skill in the art.

[0036] Embodiment 1: For the intelligent archive storage monitoring system based on the Internet of Things provided in this embodiment, the edge sensing unit can adopt a modular embedded design, and internally integrate a multi-parameter composite sensor array with the capabilities of temperature and humidity, gas components (including carbon dioxide and volatile organic compounds), vibration, and light detection. The data acquisition module wakes up the sensor array regularly at preset time intervals through a highly stable sampling control logic. In the silent monitoring state, the system defaults to obtaining environmental parameters at a sampling frequency of once per minute, and this frequency can be dynamically adjusted according to the alert state or enhanced monitoring requirements. In the default state, the system maintains ultra-low power operation, and only the low-power microprocessor maintains the timing wake-up mechanism. At the same time, to achieve early perception of subtle environmental changes, the edge sensing unit embeds a lightweight time series modeling mechanism locally, which is used to generate and continuously update personalized stable state baselines for each independent environmental parameter. This baseline is not dependent on the unified distribution from the central platform, but is a dynamic mean curve independently constructed by the edge unit based on its own collected historical data window. For each environmental parameter, weighted moving average or exponential smoothing methods are respectively used for processing, and the smoothing coefficient is set, for example, between 85% and 95%. The specific value is automatically determined by the local algorithm according to the fluctuation characteristics of the environmental parameter itself during the system initialization stage. The system adopts a non-uniform weighting strategy during the baseline maintenance process, assigning higher weights to recent data to improve the response ability to short-term changes and avoid the interference of extreme values on the overall trend judgment.

[0037] When the system is running, the edge unit continuously compares the deviation degree between the currently sensed value and its baseline. If a certain parameter shows the same-direction deviation in three consecutive sampling periods, and the maximum deviation value exceeds the preset tolerance range of the baseline (this range is set to plus or minus 7%, determined according to the historical fluctuation range of the typical archive storage environment), the preliminary judgment logic of the perturbation event is triggered, and the unit then enters the local alert state. In this state, the edge unit does not need to activate the complete data upload process, but only sends a micro-perturbation wake-up signal with a very simple structure to its predefined three neighboring edge units. This signal includes the unique identifier of this unit, the event trigger timestamp, and the current status flag bit, and does not include specific environmental parameter data, so as to complete the preliminary wake-up action of local event collaborative recognition with extremely low power consumption. After receiving this type of micro-perturbation wake-up signal, the neighboring edge unit makes a judgment according to its local current status and the preset collaborative judgment rules. If the current unit itself is also in the local alert state and receives at least one wake-up signal from a neighboring unit in this state, it enters the collaborative attention state and determines the existence of a possible regional risk; another triggering mechanism is: even if this unit is not yet in the alert state, but if it continuously receives wake-up signals from two different neighboring units within ten minutes, it will also directly enter the collaborative attention state and send a concern event report including the event time, location, and relevant edge unit information to the central processing platform.

[0038] The system has a built-in independent instantaneous disturbance recognition mechanism to cope with high-dynamic risk events such as vibration shocks, instantaneous gas leaks, or sudden changes in light. The edge perception unit pre-stores feature sequence templates for various typical instantaneous disturbance events, including features such as frequency mutations in vibration signals and abnormal upward slopes in gas concentrations. These templates can be refined based on historical data, for example. During actual operation, the edge unit performs real-time sliding processing on the current sensed data to form a short-time sequence and compares it with the preset templates. Usually, a sliding window with a length of four to six sampling periods is used for feature matching. When the similarity between the current sequence and any template exceeds the preset matching threshold (generally set at 80%), it is determined that a corresponding type of instantaneous critical disturbance event has occurred, and an alarm message is immediately reported to the central platform. The content includes the event type, trigger time, and geographical location code. After receiving such instantaneous alarm information, the central platform immediately issues targeted enhanced monitoring instructions to the event-occurring unit and its adjacent units. The instructions require the target unit to increase the sampling frequency (up to once every 20 seconds) and appropriately relax the micro-disturbance event judgment threshold within the next five to ten minutes to capture possible subsequent disturbances or aftershock interferences. The system also constructs a dynamic health assessment mechanism for the archival storage environment through the archiving and analysis of long-term monitoring data. The central platform automatically statistically analyzes the resonance feature data of collaborative attention events in each monitoring area or unit every week, including parameters such as the average number of adjacent units triggered by each event, the collaborative brewing period (i.e., the time from the first alert state to the formation of collaborative attention), the average duration of the disturbance, and the event occurrence frequency per unit time. After normalization processing, the above indicators are organized into a structured vector, which serves as an important reference information reflecting the stability of the micro-environment in this area or unit. For example, when a certain area has ten collaborative attention events within a month, the average number of adjacent units triggered is 2.6, the collaborative brewing period is seven minutes, the average duration of the disturbance is five minutes, and the event frequency per unit time is 2.5 times per week. The current environmental stability state vector of this area is: [2.6, 7, 5, 2.5]. Through trend modeling and change analysis of such vectors, if a significant negative evolution trend is identified, a trend warning is issued to prompt the operation and maintenance personnel to conduct unscheduled inspections, adjust the archival density, or optimize the operation strategies of local air conditioners and dehumidification equipment in a timely manner.

[0039] In terms of key parameter setting, in this embodiment, by analyzing the historical monitoring data of four typical document archive storages, the key judgment thresholds for system perturbation events are obtained. The initial judgment threshold for temperature and humidity perturbation is recommended to be set within plus or minus seven percent of the baseline. The effective identification template for vibration perturbation mainly focuses on the instantaneous peak change characteristics in the frequency range of ten to twenty hertz. For gas perturbation, the abnormal criterion is that the concentration rise rate exceeds eight units per minute. The above parameter settings are double-verified by the laboratory simulation platform and the feedback data of on-site deployment, and have sufficient engineering basis and technical feasibility. In terms of system energy consumption, the typical power consumption of the edge sensing unit in the silent state is ten to thirty microwatts; the power consumption rises to within two hundred microwatts in the local alert state; and the peak power consumption does not exceed five hundred microwatts in the enhanced monitoring state. All units are powered by energy harvesting components such as photovoltaic cells or thermoelectric generators, which are all extended implementation methods known to those of ordinary skill in the art.

[0040] Embodiment 2: In the actual application scenario of the microenvironment monitoring system for archive storages, in this embodiment, by constructing the environment of a typical document archive storage, common microenvironment fluctuation situations and typical perturbation events are simulated to evaluate the response ability, judgment accuracy rate, and false alarm control level of the system under actual deployment conditions.

[0041] In this embodiment, a document archive storage on the second floor of an archive is selected as the test site. The internal size of the storage is 28 meters × 12 meters, with a floor height of 3.2 meters. It is equipped with a centralized ventilation system and a zoning air-conditioning control device. The storage stores paper originals, manuscripts, and some precious document materials for a long time. The environmental control indicators require that the temperature be maintained in the range of 18°C to 25°C, and the relative humidity be maintained between 40% and 60%. A total of 12 edge sensing units are deployed in the storage, evenly arranged at a spacing of 3 meters. Among them, 4 units are set as intervention control nodes to simulate human perturbation situations. Each edge unit integrates a multi-parameter composite sensor array, including temperature and humidity, vibration, and total volatile organic compounds (TVOC) detection modules. The data acquisition frequency is default set to once every 60 seconds, and it has local processing capabilities and can implement functions such as baseline learning, drift judgment, adjacent communication, and template matching. The parameter settings are shown in Table 1: Table 1: Parameter Setting Table

[0042] Within the first 24 hours of the start of the experiment, a constant environmental condition (temperature 22.5°C, relative humidity 48%) was maintained in the storage room to test the baseline modeling ability of the system in a stable environment. The experimental phenomenon was that all edge sensing units completed the preliminary baseline modeling of their respective environmental parameters within 6 hours after the system initialization. The modeling curve showed a monotonically convergent characteristic. After that, the deviation of the collected value from the baseline was stably controlled within the range of ±1.2%. The experimental conclusion was that the dynamic modeling algorithm based on exponential smoothing had good convergence and anti-interference capabilities and was applicable to the dynamic baseline construction requirements in the actual engineering environment.

[0043] Starting from the 36th hour, manual temperature and humidity perturbation treatment was carried out on the local environment of node #04, increasing the temperature by 1.8°C and the relative humidity by 5%. The perturbation duration was 90 minutes. The system reaction was that node #04 detected that the deviation value from the baseline exceeded 7% in three consecutive sampling periods and entered the local alert state. Immediately, it sent a perturbation wake-up signal to the adjacent nodes #03 and #05. Node #03 entered the collaborative attention state at the 12th minute. Node #05 did not trigger a response due to the stable environment. The central platform received the event report information at the 15th minute. The analysis conclusion was that the local baseline deviation trigger mechanism effectively identified the temperature and humidity slow change event, and the response delay was controlled within the expected range. The collaborative sensing mechanism had the ability to identify regional anomalies.

[0044] At the 48th hour, nodes #07, #08, and #09 were synchronously perturbed (temperature increased by 1.6°C, TVOC concentration increased by 80 ppm), and the perturbation duration was set to 25 minutes. The system reaction was that the three nodes successively entered the local alert state within the 4th to 6th minutes. After being mutually perturbed and awakened, they all entered the collaborative attention state. The central platform received the event report at the 10th minute and initiated an enhanced monitoring instruction. In the subsequent 10 minutes, the sampling frequency of the relevant nodes was increased to once every 20 seconds, and the rising trend of TVOC and the micro-vibration signal were successfully captured. The analysis conclusion was that the system had the ability to identify regional resonance risks under the condition of multi-point synchronous perturbation, the event response mechanism was efficient, and the instruction issuance was in good synchronization with the on-site status.

[0045] At the 60th hour, an artificial ground vibration event was simulated, with an input perturbation frequency of 12 Hz, a peak acceleration of 0.6 g, and a duration of 6 seconds. The system reaction was that node #10 immediately identified the vibration event and generated an instantaneous alarm, and the matching template similarity reached 0.88, exceeding the preset threshold. The central platform completed the instruction issuance within 5 seconds, and no subsequent anomalies were detected by the system, and the node status automatically recovered. The analysis conclusion was that the preset template mechanism accurately identified instantaneous events of the vibration type, the system response time was fast, and the false alarm control was good.

[0046] Table 2: Summary Table of Test Data

[0047] The test data clearly verifies the system's baseline modeling ability for dynamic changes in environmental parameters, the triggering and recognition effect of local anomalies, and the collaborative response logic under multi-point disturbances in the actual deployment scenario. The collaborative perception mechanism constructed by the system has the ability to accurately identify regional micro-disturbances, and cooperates with the event-driven enhanced monitoring process to achieve closed-loop processing. In addition, the system realizes a rapid response to high-dynamic disturbance events through template matching, effectively compensating for the delay problem of the static judgment mechanism in case of emergencies. The overall false alarm rate of the system is controlled within 2%, and the instantaneous response delay is significantly lower than that of traditional centralized monitoring systems, with excellent real-time performance and engineering feasibility.

[0048] Example 3: This example combines Figures 1 to 3 to illustrate the implementation of an intelligent archives storage monitoring system based on the Internet of Things. As Figure 1 shown, when the edge sensing unit A continuously monitors the current value of environmental parameters and parallelly and real-time matches the short-term feature sequence formed by the current sampling data with at least one feature sequence template, under the condition of successfully matching the instantaneous disturbance feature template, it can determine that an instantaneous key disturbance event corresponding to the template occurs locally and immediately send the instantaneous alarm information (type, occurrence time, and location) to the central processing platform. After receiving the instantaneous alarm information, the central processing platform will separately send targeted enhanced monitoring instructions to the adjacent edge sensing unit B and adjacent edge sensing unit C to trigger them to temporarily improve the monitoring sensitivity. At the same time, the central processing platform will also send a targeted enhanced monitoring instruction to the edge sensing unit A, and the edge sensing unit A will therefore temporarily improve the monitoring sensitivity and adjust the parameters for determining local micro-disturbance events to enhance the subsequent disturbance recognition ability. If no feature sequence template is matched, the edge sensing unit A will continue baseline learning and comparison.

[0049] As Figure 2 shown, through the feature extraction module, key parameters including the number of triggered neighbors, resonance brewing period, duration, and occurrence frequency are extracted from the collaborative attention status long-term recorded by the edge sensing unit. These parameters are used as multi-dimensional inputs and sent to the dynamic health fingerprint calculation module to construct a dynamic health fingerprint that can reflect the response characteristics and internal stability evolution of a specific monitoring area or edge sensing unit to external disturbances in the micro-disturbance resonance event. Subsequently, the calculation result of this dynamic health fingerprint is input into the trend analysis module, and by identifying the change trend of the fingerprint data in the time dimension, the early identification of the potential risk evolution path is realized. If the trend analysis result indicates that the system stability has decreased significantly or potential risks are accumulating, it enters the warning signal generation link, and the system outputs the corresponding trend risk warning signal accordingly.

[0050] As Figure 3As shown in the figure, nodes #01, #02, #05, and #07 are all in the silent monitoring state, and no abnormalities are detected; node #03 enters the local alert state based on the determination logic of parameter deviation ≥ 7% and sends a perturbation wake-up signal to neighboring nodes; node #06 also enters the local alert state after independently determining an abnormality. Subsequently, due to meeting the collaborative determination conditions, node #04 enters the collaborative attention state and reports the attention event to the central processing platform. After receiving the attention event, the central processing platform sends a targeted enhanced monitoring instruction to relevant nodes, thereby guiding the local area to improve the monitoring sensitivity. In the figure, the solid arrow marks the propagation path of the perturbation wake-up signal, the dashed arrow indicates the direction of the enhanced monitoring instruction, and the solid arrow shows the data stream of the reported attention event.

[0051] Embodiment 4: In a typical application scenario, such as an intelligent transformation of a special collection archive warehouse with a first-level security level by an archive management agency, the system deploys multiple composite environment perception units with edge intelligent processing capabilities and constructs a full-spectrum intelligent monitoring system with collaborative perception, fingerprint modeling, and dynamic early warning functions through a central processing platform. In the specific implementation process, each edge perception unit is installed at key monitoring points in the archive warehouse, including below the central ventilation outlet, in the area near the outer wall, the passage between the high-density data storage racks, and the skylight area. Since the above positions are respectively faced with various types of environmental disturbances such as wind speed changes, temperature gradient fluctuations, local moisture accumulation, or sudden changes in light intensity during daily operation, the perception unit needs to have good scene adaptation ability and dynamic parameter self-adjustment ability; a set of lightweight local data modeling logics are preset inside each edge perception unit for continuously monitoring multiple environmental parameters such as temperature and humidity, gas concentration, and micro-vibration signals, and constructing a personalized steady-state baseline based on a data window of the most recent ninety minutes. This modeling process uses an exponential smoothing algorithm, combined with a non-uniform weight update strategy, to assign higher weights to the data of the most recent thirty minutes in the time series to enhance the recognition ability of the early form of sudden disturbances. The smoothing factor value is set within the range of 85% to 95%. The specific value is automatically set by the perception unit according to the historical volatility of various sensors during the system initialization phase and can also be adjusted according to the data feedback returned by the central platform during subsequent operation, thereby effectively avoiding modeling errors caused by differences in sensor accuracy or different service life.

[0052] Taking the high-temperature and high-humidity environment as an example, node A is deployed near the air outlet of the air conditioner. Due to the condensation effect, the area where it is located is prone to small periodic fluctuations in humidity. During the modeling process, node A identifies the frequency change characteristics of the humidity parameter, and then automatically adjusts the smoothing factor to a lower value, and raises the judgment threshold of humidity disturbance by 1% on the basis of the original 7%, so as to enhance the tolerance ability to this type of background disturbance, thereby effectively suppressing false alarms caused by the start-stop or switching of the air conditioner in the actual measurement environment. In the default state of the system operation, if a certain edge unit shows an offset in the same direction as its stable-state baseline and exceeding the tolerance range in three consecutive sampling periods (about three minutes), the unit is determined to have a local perturbation event and enters the local alert state. In this state, the node will synchronously send a perturbation wake-up signal to its three predefined neighboring nodes. The above signal adopts a fixed-length low-byte structure, only contains the sender identification, trigger timestamp and current status flag bit, and is broadcast through the low-power Bluetooth communication channel to reduce the communication bandwidth occupancy and ensure the collaborative response efficiency in the scenario of dense node deployment; if one of the neighboring nodes is in the local alert state when receiving the wake-up signal, or a node continuously receives wake-up signals from two different neighboring nodes within ten minutes, then the node enters the collaborative attention state and uploads a data packet including location coding, collaborative relationship information and event summary content to the central processing platform through an encrypted communication channel. The above collaborative judgment mechanism is designed based on the common risk propagation path model of document archives storage rooms, and gives priority to identifying multi-point perturbation patterns with spatial correlation trends.

[0053] In an actual deployment test, due to a sudden failure of the air-conditioning equipment in the area where Node C is located, the local temperature increased. The node detected that the temperature offset value exceeded the tolerance range within four minutes and sent perturbation wake-up signals to adjacent Node B and Node D. Node B had entered the local alert state in advance due to detecting a slight humidity change. Therefore, the two nodes quickly formed a collaborative relationship and uploaded event data to the central processing platform. After receiving the event information, the central platform identified through the event recording module that two collaborative events had occurred in this area within one hour, and immediately issued an enhanced monitoring instruction, requiring all relevant nodes in this area to increase the sampling frequency to once every fifteen seconds and temporarily shorten the baseline update time window to sixty minutes to improve the perception ability of trend perturbation changes. In addition to the above detection mechanism based on baseline drift, the edge sensing unit also presets several instantaneous perturbation feature templates, including the shock pulse width distribution feature, the sudden change pattern of gas concentration slope, etc., and performs parallel feature matching on the monitoring data during operation. The above templates are derived from the features extracted by professional archival institutions through spectral, slope, duration, and multi-dimensional behavior statistical analysis of environmental accident data over the past decade. The system uses a short sequence window of fixed length for matching and defaults to setting the template similarity determination threshold at 82%. The matching algorithm uses a time-weighted coincidence comparison logic to ensure the recognition ability for approximate but non-standard perturbations.

[0054] In the simulated earthquake interference scenario, node E successfully matched the high-frequency pulse interference template within two seconds after the vibration occurred, immediately generated instantaneous alarm information and uploaded it to the central processing platform. The report content included the event type, occurrence time, and node location code. After receiving this information, the platform immediately sent targeted enhanced monitoring instructions to node E and its three neighboring nodes. The instruction content included enabling the high-frequency vibration detection subroutine and expanding the detection bandwidth of the gas sensor. Subsequently, the system recorded a slight upward trend in TVOC concentration within fifteen minutes, confirming the possibility of structural disturbance accompanied by material volatilization. The platform then triggered the manual review process for this area. The central processing platform incorporated the above events into the dynamic health fingerprint database and constructed a feature vector based on key parameters such as the number of responding neighbors, average trigger lag time, disturbance duration period, and event frequency per unit time. This vector was compared with the historical mean for trend comparison. If any two of the parameters showed a negative offset of more than ten percent in three consecutive statistical periods, a trend warning was automatically triggered, prompting the maintenance personnel to perform pre-emptive maintenance on the relevant area. This warning mechanism enables the system to identify the downward trend of regional stability in advance without obvious parameter anomalies, providing a warning basis for archival resource protection. At the same time, the system also adopts a cyber-physical fusion control technology architecture with real-time state perception, edge intelligent processing, and centralized coordinated feedback as the core structure, having the ability to operate dynamically and adaptively. And in specific implementation, for example, it can combine a low-power embedded computing platform with a low-power communication module and supplement it with an energy harvesting power supply method, enabling the system to achieve long-term, continuous, and efficient intelligent operation without additional wiring or frequent maintenance. All these belong to the extended implementation methods known to those of ordinary skill in the art.

[0055] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent archives storage monitoring system based on the Internet of Things, characterized in that, The system includes: A plurality of edge perception units, which are configured in the archives storage room. Each edge perception unit is configured to: dynamically learn and continuously maintain a personalized steady-state baseline of at least one environmental parameter within its monitoring range; monitor the current value of the environmental parameter in real time, and when the current value drifts from the personalized steady-state baseline for a duration reaching a predetermined length and deviates from a predetermined drift threshold, determine that a local perturbation event has occurred and enter the local alert state; when entering the local alert state, send a perturbation wake-up signal that does not contain specific environmental parameter data but only contains its own identifier and the perturbation wake-up state to its predefined neighboring edge perception units; receive the perturbation wake-up signals from its neighboring edge perception units, and when the preset collaborative perturbation determination conditions are met, enter the collaborative attention state and send attention event information to the central processing platform. In addition, the system further includes a central processing platform, which is configured to: receive and process the attention event information from the edge perception units that enter the collaborative attention state; request the relevant edge perception units to upload more detailed historical environmental parameter data according to the attention event information; dynamically adjust the predetermined drift threshold of the edge perception units or the parameters in the preset collaborative perturbation determination conditions based on the statistics of the attention event information and the results of in-depth risk analysis.

2. The intelligent file storage monitoring system based on the Internet of Things according to claim 1, characterized in that, Each edge perception unit is further configured to: store at least one predefined feature sequence template corresponding to a specific type of instantaneous environmental perturbation event; while continuously monitoring the current value of the environmental parameter and performing personalized steady-state baseline learning and comparison, match the current sampling data of the environmental parameter or the short-term feature sequence formed thereby with at least one feature sequence template in parallel and in real time; when the current sampling data or the short-term feature sequence formed thereby successfully matches any feature sequence template, determine that a local instantaneous critical perturbation event corresponding to the template has occurred, and directly send instantaneous alarm information indicating the type, occurrence time, and location of the instantaneous critical perturbation event to the central processing platform, regardless of whether the edge perception unit enters the local alert state or the collaborative attention state.

3. The intelligent archive storage monitoring system based on the Internet of Things according to claim 2, characterized in that, The central processing platform is further configured to: after receiving the instantaneous alarm information, send targeted enhanced monitoring instructions to the edge perception unit where the instantaneous critical perturbation event occurs and its predefined neighboring edge perception units. The instructions are configured to enable the instructed edge perception units to temporarily increase their monitoring sensitivity within a predetermined short time period.

4. An intelligent archive warehouse monitoring system based on the Internet of Things according to claim 1, characterized in that The communication module of the edge perception unit is configured to communicate with neighboring edge perception units and / or the central processing platform through low-power Bluetooth or LoRa wireless technology.

5. The intelligent file storage monitoring system based on the Internet of Things according to claim 1, characterized in that, The central processing platform is further configured to: for one or more specific monitoring areas in a preset archival storage, or for one or more specific edge sensing units, long-term record and statistically analyze the resonance pattern time-series feature data composed of the frequency of entering the collaborative attention state within a predetermined time period, the combined pattern and quantity characteristics of the perturbation wake-up signals sent by the neighboring edge sensing units required when triggering the collaborative attention state, the response lag time of entering the collaborative attention state, or the duration of the collaborative attention state.

6. The intelligent archive repository monitoring system based on the Internet of Things according to claim 5, wherein, The central processing platform is further configured to: based on the evolution of the resonance mode time series feature spectrum data of a specific monitoring area or a specific edge sensing unit over time, construct and continuously update a dynamic health fingerprint for the area or unit, which can characterize the comprehensive response characteristics and inherent stability of the corresponding archival microenvironment to external disturbances, where the feature vector of the dynamic health fingerprint can be represented in the following form: ， Among them, is the average number of triggered neighbors, is the average resonance incubation period, is the average disturbance duration, is the recent occurrence frequency; is the preset weight coefficient.

7. The intelligent archive storage monitoring system based on the Internet of Things according to claim 6, characterized in that, The central processing platform is further configured to: analyze the long-term evolution trend of the dynamic health fingerprint, and generate trend-based risk warning information different from that based on immediate environmental parameter anomalies when detecting that the dynamic health fingerprint presents a predetermined negative evolution pattern indicating that the potential risks of the archival microenvironment are accumulating or its internal stability is continuously decreasing.

8. An intelligent archive warehouse monitoring system based on the Internet of Things according to claim 7, characterized in that, After generating the trend-based risk warning information, the central processing platform is further configured to dynamically adjust the archival inspection priority, optimize the allocation of preventive maintenance resources, or guide the scientific partitioned storage of archives with different sensitivity levels.

Images