Circuit board equipment remote management system and method based on internet of things

Abstract

The present application relates to the technical field of Internet of Things device operation and maintenance and remote management, in particular to a circuit board device remote management system and method based on Internet of Things, which collects circuit board device system interaction information to establish dependency relationship, generates dependency topology and dependency parameter set; combines operation environment, historical behavior and early warning to generate and online update dependency influence index and dynamic threshold; predicts health trend quantity based on historical and real-time data; generates control decision accordingly and executes weight reduction, blockage, freezing or migration; writes relevant information into dependency relationship knowledge graph, triggers update reconstruction optimization according to structural deviation and semantic inconsistency, significantly reduces cross-device chain failure caused by misjudgment by online updating dependency influence index and dynamic threshold through dynamic modeling of dependency relationship between circuit board devices. Combined with health trend prediction and control decision linkage, risk propagation is intervened in advance, and device fault tolerance is improved by dynamically adjusting fault tolerance threshold and fault tolerance level.

Description

Technical Field

[0001] This invention relates to the field of IoT device operation and maintenance and remote management technology, and more specifically, to a remote management system and method for circuit board equipment based on IoT. Background Technology

[0002] In the remote management of circuit board equipment in the Internet of Things (IoT), devices typically form a multi-layered dependency network through data acquisition triggering relationships, control command links, task orchestration, and resource collaboration. For example, circuit board equipment A might trigger data acquisition from circuit board equipment B, or the status of circuit board equipment A might serve as a prerequisite for the execution strategy of circuit board equipment B. In current engineering practices, remote management platforms often employ single-device status monitoring and alarm mechanisms, combined with statically configured dependencies or manually maintained topologies to aid decision-making. Furthermore, to reduce implementation complexity, thresholds and fault tolerance levels are often set using fixed rules or experience, and calibrated through periodic inspections and log backtracking. This approach can provide basic operational assurance in environments with relatively stable dependencies and minimal changes in interaction patterns, and also facilitates location and handling by maintenance personnel.

[0003] However, in scenarios involving edge computing collaboration, long cross-device task links, and significant fluctuations in operating conditions, the strength of dependencies between devices, communication quality, and resource reserves will continuously drift with changes in load and environment, resulting in dynamic and hidden dependencies. On the one hand, an anomaly of a certain device may be "masked" by local indicators. For example, although circuit board device A generates an error, it may not trigger a local alarm or may be misjudged as being normal overall. On the other hand, the platform may still continue to issue configurations or trigger link actions based on the "normal" judgment of circuit board device A, causing circuit board device B, which depends on circuit board device A, to run continuously under erroneous conditions, thereby inducing policy deviations or data pollution in more devices, and ultimately evolving from a single point of failure into a group-level error propagation, forming a cross-device state chain failure. In response, the industry generally needs a more refined approach to remote management: not only should the true dependency topology and dependency parameters be depicted from system interaction information, but the impact of dependencies should also be updated online and dynamically assessed for thresholds. Furthermore, risk propagation should be regulated in conjunction with the prediction results of health trends. At the same time, dependency relationships and interaction evidence should be precipitated into an updatable knowledge representation to support the continuous verification and dynamic reconstruction of dependency relationships. This would improve the ability to identify and handle cascading risks without increasing the amount of manual intervention. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a remote management system and method for circuit board equipment based on the Internet of Things.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The IoT-based remote management system for circuit board equipment includes: The dependency modeling module is used to collect system interaction information of circuit board equipment, and perform dependency relationship modeling based on the system interaction information to generate the dependency topology of circuit board equipment and form a dependency parameter set. The intensity update module is used to generate and update the dependency impact index online based on the dependency parameter set, equipment operating environment parameters, historical operating behavior patterns and abnormal warning information, and output the dependency impact index and dynamic threshold. The health prediction module is used to build a health prediction model based on historical and real-time operating data and output the health trend of the circuit board equipment. The dependency control module is used to generate control decisions and execute control actions based on dependency impact indicators, dynamic thresholds, and health trend data. The fault tolerance adjustment module is used to dynamically adjust the fault tolerance threshold and fault tolerance level in real time or based on a preset period, and output the corresponding fault tolerance strategy. The fault-tolerant strategy driver module is used to drive the circuit board equipment to degrade or protect based on the fault-tolerant strategy in order to deal with abnormal or overload conditions of the equipment. The knowledge graph reconstruction module is used to write dependency topology, dependency parameter set, control actions and system interaction information into the dependency knowledge graph. When the triggering conditions are met, the module will trigger the update of the dependency knowledge graph based on the graph structure offset and semantic inconsistency, so as to complete the dynamic reconstruction and optimization of dependencies.

[0006] Furthermore, the system interaction information includes: circuit board device state vector, data interaction records, control command links, and abnormal event time series.

[0007] Furthermore, the circuit board equipment state vector includes the circuit board equipment's real-time operating parameters, historical operating modes, resource utilization, and communication quality indicators. Data interaction records include the order, frequency, and delay of data transmission between circuit board devices; The control command link includes the command type, command trigger time, and response delay of control signal transmission between circuit board devices.

[0008] Furthermore, the dependency modeling module further includes: The operation coordination and consistency calculation module is used to output the operation coordination and consistency index between circuit board devices based on the circuit board device state vector, data interaction record, control command link and historical operation mode in the system interaction information. The cross-circuit board device abnormal coupling calculation module is used to output the cross-circuit board device abnormal coupling index based on the time sequence of abnormal events and the control command link in the system interaction information. The information dissemination influence calculation module is used to analyze the control command chain and the time series of abnormal events, and output the information dissemination influence index between circuit board devices.

[0009] Furthermore, the dependency modeling module is also used for: By collecting circuit board device state vectors, data interaction records, and control command links from system interaction information, an interaction network diagram between circuit board devices is constructed and the interaction behavior is analyzed. The dependencies between circuit board devices are determined based on the operational coordination consistency index, the cross-circuit board device abnormal coupling index, and the information propagation influence index. The dependency topology is then generated by combining resource utilization, historical operating mode, and communication latency. Calculate the dependency parameter set based on the dependency topology.

[0010] Furthermore, the control measures include at least one of dependency demotion, dependency blocking, configuration freezing, and task migration to suppress the propagation of errors across circuit board devices.

[0011] Furthermore, the fault tolerance adjustment module is used for: The fault tolerance threshold and fault tolerance level are calculated based on the real-time operating status, load, resource reserves, communication quality, dependency impact indicators, and health trend of the circuit board equipment. The fault tolerance threshold and level are dynamically adjusted based on a preset periodic evaluation, and the fault tolerance strategy is updated accordingly.

[0012] Furthermore, the triggering conditions for the knowledge graph reconstruction module to update the dependency relationship knowledge graph include: The graph structure offset is calculated based on system interaction information and dependency relationship knowledge graph. When the graph structure offset exceeds a first preset threshold, the dependency relationship knowledge graph is updated; and / or The semantic inconsistency is calculated based on the system interaction information and the dependency knowledge graph. When the semantic inconsistency is greater than the second preset threshold, the dependency knowledge graph is updated. Specifically, when the dependency knowledge graph is updated, the dependency topology and dependency parameter set are reconstructed, and the updated dependency topology and dependency parameter set are written into the dependency knowledge graph.

[0013] Furthermore, the IoT-based remote management method for circuit board equipment includes the following steps: Step 1: Collect system interaction information from the circuit board equipment; Step 2: Model dependency relationships based on system interaction information, generate dependency topology and form dependency parameter set; Step 3: Generate and update dependency impact metrics and dynamic thresholds based on the dependency parameter set, device environment parameters, historical behavior, and anomaly warnings; Step 4: Build a health prediction model based on historical and real-time data, and output health trends; Step 5: Generate and execute regulatory decisions based on dependency impact indicators, dynamic thresholds, and health trends; Step 6: Dynamically adjust the fault tolerance threshold and fault tolerance level, and update the fault tolerance strategy; Step 7: Write the dependency topology, dependency parameter set, control actions and interaction information into the knowledge graph, and trigger updates based on the graph structure offset and semantic inconsistency to complete the dynamic reconstruction and optimization of the dependency relationship.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention models dependencies by collecting system interaction information from circuit board devices, forming a dependency topology and dependency parameter set. During operation, it continuously outputs dependency impact indicators and their dynamic thresholds, enabling the triggering relationships, data transmission relationships, and control associations between devices to be verified and corrected based on real-time interaction evidence, rather than relying solely on static configuration or manual maintenance. This significantly reduces the possibility of circuit board device A exhibiting abnormalities but being mistakenly judged as being in normal condition, and prevents the platform from continuing to issue configurations based on erroneous premises, causing circuit board device B or even more devices to enter an erroneous operating state, thus fundamentally suppressing the formation of cross-device state cascading failures. By generating and updating dependency impact indicators online based on dependency parameter sets, device operating environment parameters, historical operating behavior patterns, and abnormal early warning information, and outputting health trend quantities based on historical and real-time operating data, risk identification not only focuses on instantaneous alarms but also reflects the potential intensity of dependency propagation and the evolution trend of device health. Furthermore, dependency control generates control decisions and executes control actions based on dependency impact indicators, dynamic thresholds, and health trend quantities. Before errors spread into group-level failures, key dependency edges can be downgraded, isolated, configuration frozen, or task migrated, thereby reducing the propagation of errors along dependency paths and reducing the cumulative effects of data pollution and policy deviation. This is especially suitable for long-link linkage tasks in edge computing collaboration scenarios. By dynamically adjusting the fault tolerance threshold and level in real time or based on a preset period to output fault tolerance strategies, and driving degraded or protective operation by the fault tolerance strategies, the intensity of handling can be matched with the risk situation. This enables risk convergence and avoids fault spillover during overload, communication quality degradation, or local anomaly phases. At the same time, the dependency topology, dependency parameter set, control actions, and system interaction information are written into the dependency knowledge graph. When the graph structure offset or semantic inconsistency reaches the trigger condition, an update is triggered, and the dependency topology and dependency parameter set are reconstructed and written. This allows the dependency relationship to be continuously optimized as the actual interaction behavior changes. This not only improves the ability to trace and explain the causes of cascading failures, but also reduces the hidden risks and probability of incorrect decisions caused by dependency drift during long-term operation. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 A schematic diagram illustrating dependency modeling and dependency topology generation; Figure 3 A schematic diagram illustrating the online updates of influencing indicators and dynamic thresholds, and the prediction of health trends. Figure 4 A schematic diagram illustrating the control execution, fault tolerance strategy output, and dependency relationship knowledge graph update and reconstruction. Detailed Implementation

[0016] Example 1: Refer to Figures 1 to 4 The IoT-based remote management system for circuit board equipment includes: Dependency Modeling Module: This module transforms the circuit board equipment group from "independent operation" into an "interpretable dependency network," providing a unified structured foundation for subsequent impact assessment, health prediction and linkage, control decisions, and fault tolerance strategies. By collecting system interaction information from the circuit board equipment, this module identifies the direction and strength of associations between the equipment in data interaction, control commands, and abnormal linkages. It organizes these associations into a dependency topology, shifting remote management from passive processing of individual devices to proactive management of dependency chains. Input and Output: Input is system interaction information; output is the dependency topology and dependency parameter set. The dependency topology describes the dependency connections and propagation paths between circuit board equipment, while the dependency parameter set carries parameterized information related to dependencies, supporting the generation and online updating of dependency impact indicators by the subsequent strength update module. The dependency modeling module transforms system interaction information into a computable and iterable dependency structure, enabling subsequent modules to perform calculations, decisions, executions, and updates around the same dependency topology, ensuring consistent remote management logic and traceable closed-loop processes. In one specific implementation, remote management of circuit board equipment achieves optimized equipment monitoring and management by collecting and analyzing system interaction information. The specific implementation steps are as follows: By collecting the state vectors of the circuit board equipment, a comprehensive understanding of the equipment's operational status can be achieved. The state vectors include real-time operating parameters, historical operating modes, resource utilization, and communication quality indicators. Specifically, real-time operating parameters may include key physical quantities such as voltage, current, and temperature; historical operating modes reflect the equipment's working status and performance over a past period; resource utilization, such as memory, CPU usage, and storage occupancy, provides crucial information for subsequent resource allocation; and communication quality indicators, including data transmission stability and latency, help determine if the equipment is affected by network issues, ensuring efficient and stable operation. By collecting data interaction records between circuit board devices, the collaboration between these devices can be further analyzed. These records include information such as the order, frequency, and latency of data transmission between devices. Collecting this information helps monitor data flow and communication timeliness between devices, identify data transmission bottlenecks or latency issues, and thus adjust the data interaction methods between devices to optimize system response speed and efficiency. For example, if a device frequently sends large amounts of data to other devices, it may cause communication delays. In this case, measures such as reducing the data transmission frequency or adjusting the transmission time can be taken. The control command chain is collected and analyzed to understand the transmission of control signals between devices on the circuit board. The control command chain records the type of command, trigger time, and response delay between devices. By tracking and analyzing these control signals, the timeliness and accuracy of command execution between devices can be understood. For example, if a device's control signal response delay is too long, the overall system's control effect may be poor. In this case, it is necessary to adjust the control strategy or optimize the command transmission process to ensure the timely response of each device when executing control commands, and avoid device failures or performance degradation due to command transmission delays.

[0017] To ensure the comparability of state vectors, data interaction records, control command links, and abnormal event time series of different circuit board devices at the same time scale, this implementation method performs unified time alignment and windowed segmentation on the system interaction information. Let the unified sampling granularity be... , with length as The sliding window aligns various sequences, with the window index as... The corresponding time interval is For any device Continuous observations (including voltage, current, temperature, resource utilization, communication quality, etc.) constitute a state vector sequence. .

[0018] Missing segments are handled using an "interpolation mark" method: a missing mask is used. When observations are missing The missing dimensions are then compensated for using forward preservation or linear interpolation to obtain the interpolated state sequence. Meanwhile, the mask is retained for subsequent calculations to avoid the spread of bias caused by missing values.

[0019] For all scalar characteristics involved in the calculation Normalization is performed using the mean of the baseline window set. with standard deviation Standardization leads to ; in To prevent extremely small positive numbers with a denominator of zero, the above-mentioned time alignment, windowing, missing data interpolation, and standardization processes ensure that the state changes, interactive behaviors, and command responses of different circuit board devices within the same window are comparable, thus providing a consistent input basis for subsequent dependency modeling and index calculation.

[0020] In one specific implementation, to obtain quantifiable and updatable dependencies between circuit board devices, the system interaction information is processed according to the following steps to output an operational coordination consistency index, a cross-circuit board device abnormal coupling index, and an information propagation influence index: The system interaction information timeline is uniformly collected and aligned. The circuit board equipment state vector, data interaction record, control command link and abnormal event time series are windowed according to the same sampling granularity. Missing segments are interpolated and marked. The historical operation mode is used as window label for subsequent comparative analysis, so that the state changes, interaction behaviors and command responses of different circuit board equipment in the same window are comparable. The operational coordination consistency index is calculated based on the state vector of the circuit board equipment, data interaction records, control command links and historical operating modes. Within each window, the synchronization degree of state vector changes, the consistency degree of data interaction record sequence and the matching degree of trigger and response delay of control command links are measured simultaneously. Cooperative behaviors consistent with historical operating modes are given higher confidence weights, thereby distinguishing between accidental simultaneous changes and stable cooperative relationships. For example, when the resource utilization of circuit board equipment A increases and the data interaction frequency of circuit board equipment B increases and occurs simultaneously for a long time, and the trigger time of control commands and the response delay remain stable, the operational coordination consistency index will remain high. The cross-circuit board device anomaly coupling index is calculated based on the time series of abnormal events and the control command chain. The sequence and time interval of abnormal events are identified within the window. Combined with the command type, command trigger time and response delay in the control command chain, it is determined whether the anomaly presents a repeatable linkage pattern. At the same time, the accompanying relationship of anomalies under the same command type is grouped and statistically analyzed to improve the ability to identify anomaly coupling caused by command linkage. For example, when circuit board device A has an abnormal event after triggering a specific control command, and circuit board device B has an abnormal event within a fixed time difference immediately afterward, and the event is repeated in multiple windows, the cross-circuit board device anomaly coupling index increases significantly. The information propagation influence index is calculated by analyzing the control command chain and the time series of abnormal events. The command triggering time and response delay in the control command chain are regarded as propagation time characteristics, and the diffusion range of the abnormal event time series is regarded as propagation result characteristics. The ability of a certain circuit board device to control and spread abnormalities to other circuit board devices is comprehensively evaluated. Higher influence scores are assigned to situations where "abnormalities spread faster after command triggering, more devices are covered, and response delays are shorter". For example, after circuit board device A issues a command, multiple circuit board devices experience abnormal events in a shorter time and the data interaction record delay increases simultaneously. In this case, the information propagation influence index of circuit board device A is judged to be higher, thus providing an interpretable quantitative basis for determining subsequent dependencies.

[0021] In each window Within, for any two devices and Operational Coordination Consistency Index It consists of state synchronization components, interaction sequence consistency components, and instruction delay matching components, and is normalized to an interval. .

[0022] (a) State synchronization components: Let the normalized state vector sequences within the window be as follows: and ,in and These respectively represent in the window Internal and The time-mean vector. Define the state synchronization degree as... ;in For cosine similarity, this formula maps synchronization degree to Interval.

[0023] (ii) Consistent interaction sequence components: Assume that within the window The data interaction event timestamp sequence is ,as well as The corresponding event timestamp sequence is The two sequences are merged into an event sequence according to their order of occurrence, and the order correlation coefficient (Kendall) is used as the criterion. Or Spearman To measure the stability of an interaction sequence, the consistency of the interaction sequence is defined as follows: ;in The rank correlation coefficient represents the order of interactions within a window.

[0024] (iii) Command delay matching component: Assume that the device in the control command link For equipment The instruction trigger-response delay sample set is as follows The delay matching degree is defined as ; in The time delay variance within the window. The baseline time delay variance is used for scale normalization. To prevent extremely small positive numbers with a denominator of zero.

[0025] The above three components are combined to obtain the operational coordination consistency index: ; in Default is acceptable. The reason is that the three components (state synchronization, interaction sequence consistency, and instruction latency matching) characterize the collaborative relationship from three complementary dimensions: state change, interaction behavior, and control link response. Furthermore, before baseline statistical calibration or supervised learning, there is no a priori provable preference among these three components for the importance of the collaborative relationship. Using equal-weighted fusion avoids the over-amplification of any dimension due to manual weighting, thereby reducing the risk of misjudgment and ensuring better robustness and transferability of the index under different operating conditions and equipment types. Subsequently, the weights are calibrated based on the baseline period to minimize the false alarm / false negative rate, gradually transitioning the weights from "unbiased default" to "data-driven optimal." This default is consistent with the parallel design in the specification of "simultaneously measuring synchronization, sequence consistency, and latency matching within each window." This indicates that the result will be truncated to... Within the range.

[0026] In the window Inside, there are equipment The set of abnormal events is ,equipment The set of abnormal events is Given an anomaly accompanied by a time difference threshold. Define the "abnormal pair" count as: ; Let $N_u(k) = |E_u(k)|$ be the number of exceptions occurring within the window. Then, the exception coupling strength can be expressed as a conditional proportion: ; Further consider the "repeatable linkage mode", and set near The coupling strength sequence within each window is The stable coupling degree is defined as: ; in The minimum coupling threshold (which can be taken as the high quantile of the baseline distribution) is used. The final cross-circuit board device abnormal coupling index is defined as: ; in , This indicates that the result will be truncated to... The range. The default allocation example is to take... , The reason is: It reflects the immediate coupling strength of "sequential occurrence of anomalies" within the window, is more sensitive to the propagation of sudden faults, and is suitable for rapid early warning and real-time control triggering; while This indicates the stability of linkages, representing whether they "recur" within the last L windows. It leans more towards identifying long-term structural associations and has a relatively delayed response. Therefore, in scenarios where the goal is to quickly detect risks, prioritizing this degree of stability is crucial. To reduce underreporting; while retaining This is used to suppress false positives caused by occasional noise (enhancing coupling judgment only when multiple windows repeat). If used for offline topology reconstruction, it can be retrieved in reverse. , Enhance stability.

[0027] In the window Inside, for the equipment The system assesses the system's external control and dissemination capabilities, defines the characteristics of dissemination time and dissemination results, and calculates the information dissemination influence index accordingly. .

[0028] Propagation time characteristics: Assume equipment The average response delay from the issued command to each receiving device is The mean baseline delay is The propagation time characteristic is defined as follows: ; in To prevent extremely small positive numbers with a denominator of zero; The smaller the value, the faster the spread. The larger.

[0029] Characteristics of propagation results: set in the window An abnormality occurred within the equipment. The instruction triggering has a causal proximity relationship (satisfying) The equipment set is as follows The total number of devices is The coverage rate of the propagation result is defined as follows: ; The final information dissemination influence index is defined as follows: ; in This indicates that the result will be truncated to... Interval. A common allocation example is taking... , The reason lies in the characteristics of propagation time. This reflects the device's immediate responsiveness to control commands, directly impacting the speed of information transmission. It's a key indicator, and transmission speed is often more critical in urgent situations, thus warranting higher weighting. The characteristics of the transmission outcome... This reflects how many devices a device's u-command can cover after propagation, involving the breadth of influence. While equally important, for some high-efficiency devices, coverage may reach a high level in a short period, thus receiving a relatively lower weight. This allocation balances the contributions of rapid response and breadth expansion, suitable for scenarios that prioritize rapid response and aim to increase system efficiency through breadth of influence. For long-term analysis, the weights can be adjusted according to specific needs.

[0030] In one specific implementation, to construct and optimize the dependencies between circuit board devices, the interaction network diagram between the devices is first analyzed and constructed by collecting the state vectors, data interaction records, and control command links of the circuit board devices. The system interaction information, including the state vectors, data interaction records, and control command links of the circuit board devices, is collected and organized. The state vectors of the circuit board devices include indicators such as real-time operating parameters, historical operating modes, resource utilization, and communication quality. The data interaction records include the data transmission order, frequency, and delay between devices, while the control command links record the transmission type, trigger time, and response delay of control signals between the circuit board devices. This information serves as key data points, mapped to nodes and edges in the interaction network diagram to form a comprehensive interaction view for further analysis. Based on the aforementioned interactive network diagram, the dependencies between circuit board devices are assessed by combining the operational coordination consistency index, the cross-circuit board device anomaly coupling index, and the information propagation influence index. The operational coordination consistency index measures the stability of collaboration between devices during joint operation; the cross-circuit board device anomaly coupling index identifies and quantifies abnormal linkages between devices; and the information propagation influence index reflects a device's control over other devices and its ability to propagate anomalies. By integrating these three indices, it is possible to identify which circuit board devices have strong dependencies and determine their potential linkage risks in terms of fault propagation and performance impact. Based on the aforementioned dependency index, resource utilization, historical operating patterns, and communication latency are taken into account to generate a dependency topology between circuit board devices. This dependency topology not only presents the basic dependency structure between devices but also incorporates factors such as resource consumption, historical behavior patterns, and communication latency to provide a deeper understanding of the collaboration and interaction between devices. For example, if device A and device B have high operational coordination consistency, and device A has low resource utilization and short communication latency, their dependency relationship will appear tighter, and the connecting edges in the topology will reflect this. The dependency parameter set is calculated based on the dependency topology. This set includes specific parameters for each dependency, such as resource allocation ratios, data transmission latency, and tolerable fault impact ranges. These parameters serve as the basis for subsequent operational management decisions, fault warnings, and fault tolerance strategy formulation, ensuring precise control over the dependencies between circuit board devices in actual operation. For example, if the topology reveals that device A has a significant dependency on device B, and device B has a high load, the system can adjust the task scheduling or load allocation for device A based on the dependency parameter set, thereby preventing overload or failure. By implementing these steps, a scientific and quantitative solution for dependency management and optimization between circuit board devices can be provided.

[0031] In the window Within, for any device with evidence of interaction... Based on the operational coordination consistency index Abnormal coupling index of cross-circuit board devices and the information dissemination influence index Define the dependency edge weight (dependency strength) as ;in , This indicates that the result will be truncated to... Interval. When At that time, directed dependency edges are established in the dependency topology. And write the following parameters into the dependency parameter set: (1) Propagation delay parameter: Let the window contain The average command trigger-response latency of the link is ,but ; (2) Resource coupling strength: denoted as For equipment and The degree of coupling of shared resource occupancy within the window can be estimated by the correlation coefficient or mutual information of the shared resource occupancy sequence.

[0032] (3) Tolerance range of influence: for right The tolerable range of impact can be determined from the historical window. Abnormality caused The proportion of "alarms" is estimated. The default allocation example is to take... , , The reason is: It reflects the coordination and consistency under normal operation (state synchronization, interaction order, and time delay matching). It usually has more samples and lower noise, making it suitable as the "primary evidence" of the dependent edge, hence it has the highest weight. Capturing the sequential and recurring linkages of anomalies can reveal the path of risk propagation, but anomalies are sparse and easily affected by occasional disturbances, so this method is secondary. Describing the propagation capability and coverage of device u is more like an "amplifier / spillover potential". The existence of opposite edges is auxiliary information, so it has the lowest weight.

[0033] The Intensity Update Module reflects the "real-time effectiveness" and "actual impact" of dependencies under different operating conditions. The impact intensity of the same dependency topology may change under different equipment operating environment parameters, different historical operating behavior patterns, and different anomaly warning information. The Intensity Update Module generates dependency impact indicators based on the dependency parameter set, combined with equipment operating environment parameters, historical operating behavior patterns, and anomaly warning information, and updates them online during operation. This allows the remote management system to dynamically identify "which dependencies are more likely to amplify risks, propagate anomalies, or cause overload." Inputs and Outputs: Inputs include the dependency parameter set, equipment operating environment parameters, historical operating behavior patterns, and anomaly warning information; outputs are dependency impact indicators and dynamic thresholds. Dependency impact indicators characterize the degree of influence of dependencies, while dynamic thresholds provide reference boundaries that match the operating state for anomaly judgment, control triggering, and fault tolerance adjustment at different operating stages. The Intensity Update Module transforms dependencies from static graphs into dynamic quantities that change with time, environment, behavior, and warning status, providing the dependency control module and fault tolerance adjustment module with input data closer to the actual situation. In one specific implementation, to generate and update dependency impact indicators online based on dependency parameter sets, device operating environment parameters, historical operating behavior patterns, and anomaly warning information, and to output dependency impact indicators and dynamic thresholds, the following steps can be performed: The dependency parameter set is structured and organized, and the association direction, association strength, propagation delay, resource coupling characteristics, etc. of each dependency relationship are bound to the circuit board equipment identifier. At the same time, the equipment operating environment parameters, historical operating behavior patterns and abnormal warning information are collected. The above information is aligned and normalized with a preset time window so that data from different sources can participate in the calculation at the same time scale and have comparability. Within each time window, the basic computational workload of dependency impact indicators is constructed. The dependency relationship features represented by the dependency parameter set are used as structural items, the temperature fluctuation, power supply stability, network jitter, etc. represented by equipment operating environment parameters are used as environmental items, the stable segment, fluctuating segment, and high load segment represented by historical operating behavior patterns are used as behavioral items, and the warning level, warning frequency, and warning duration represented by abnormal warning information are used as risk items. The initial value of the dependency impact indicator for the current window is obtained by fusing the structural items, environmental items, behavioral items, and risk items, so that the dependency impact indicator can simultaneously reflect the propagation potential of the dependency structure and the amplification or inhibition effect of external operating conditions on the propagation potential. The dependency impact index is updated online. When a new window arrives, the statistics of structural items, environmental items, behavioral items, and risk items are updated incrementally, and the dependency impact index is rolled over. This allows the dependency impact index to continuously evolve with changes in the operating environment and early warning situation. For example, when the dependency relationship from circuit board equipment A to circuit board equipment B shows a high correlation strength and a shortened propagation delay in the dependency parameter set, and at the same time the equipment operating environment parameters show a decline in communication quality and an increase in the frequency of abnormal early warning information, the dependency impact index will be updated online to a higher level to indicate that the dependency relationship is more likely to amplify the risk of abnormal propagation at the current stage. Dynamic thresholds are generated based on online updated dependency impact indicators. Within each time window, the normal fluctuation range is adaptively characterized by combining historical operating behavior patterns. The dynamic thresholds are set as judgment boundaries that match the current window's operating status, enabling the dynamic thresholds to automatically adjust with different operating stages. For example, when the historical operating behavior pattern is in a high-load segment and abnormal warning information is short-term and dense but quickly dissipates, the dynamic thresholds are adjusted upwards to avoid misjudging short-term fluctuations as continuous risks. Conversely, when the historical operating behavior pattern is in a stable segment but abnormal warning information continues to accumulate, the dynamic thresholds are adjusted downwards to improve sensitivity to early abnormal propagation. This achieves synchronous output of dependency impact indicators and dynamic thresholds, providing a more accurate basis for subsequent control decisions based on the actual situation.

[0034] In the window Inside, for each dependent edge Define dependency impact metrics It is obtained by integrating structural, environmental, behavioral, and risk items.

[0035] (a) Structural terms: characterized by dependency strength, propagation delay, and resource coupling strength. ;in These are the normalized propagation delay and coupling strength, respectively. The allocation example is to take... , , The reason is that in the structural items... It is the comprehensive main evidence of "edge existence and strength" obtained by the fusion of three indices. It has the most concentrated information and the most stable updates, and should be used as the dominant term. (Normalized latency) reflects transmission speed / link efficiency more, but latency is greatly affected by network jitter and sampling error, and fast does not necessarily mean strong dependence. Therefore, its weight is moderate to low. (Resource coupling) directly reflects the structural binding relationship of shared resources, and can explain the root cause of "rising and falling together / mutual restraint". Its stability is usually higher than that of latency, so it can be given a slightly higher weight than latency.

[0036] (ii) Environment Item: Let the environment parameter vector associated with this dependency edge be . (e.g., temperature fluctuations, power supply stability, network jitter, etc.), its normalized vector is ,but ;in This is an environmental weight vector, which can be set hierarchically according to equipment category.

[0037] (iii) Behavioral Items: Mapping historical operational behavior patterns to discrete states (Stable phase / volatile phase / high load phase, etc.), let its risk amplification factor be... ,but ; (iv) Risk items: The abnormal warning level is set as follows: The frequency of warnings is The warning duration is After normalization, it becomes ,definition ; The allocation example is to take... , , The reason is: (Warning level) is usually most directly related to the severity of the event, can quickly distinguish between "fatal / high-risk / moderate" risks, and is relatively less affected by the sampling frequency, making it suitable as the primary risk item; (Frequency of warnings) reflects the activity and recurrence of risks. It is very effective in identifying "continuous disturbances / repeated alarms", but it may be amplified by jitter or false alarms, so its weight is secondary. (Duration) reflects the "tail" of the impact and the ability to recover. It can supplement situations where the frequency is insufficient but the impact is long. However, it fluctuates greatly under window truncation and alarm merging strategies, so its weight is slightly lower.

[0038] Initial values ​​for the dependency impact index obtained from the fusion: ;in The allocation example is to take... , , , The reasons are as follows: structural items directly characterize the "objective dependency pattern" (strength / delay / coupling) of edges, are relatively stable and highly interpretable, and are suitable as core evidence; environmental items are often exogenous disturbances that can significantly affect dependencies but are phased, and have moderate weights; behavioral items are mostly amplification coefficients of pattern labels, with coarser information and are easily affected by state division errors, and are usually given smaller weights; risk items reflect the current alarm situation and can show "which edge should be paid more attention to at the moment," but to avoid short-term alarms overwhelming structural facts, they can be slightly lower than structural items and jointly dominate with structural items.

[0039] Online updates (rolling corrections): Statistical changes are updated incrementally and corrected on a rolling basis using Exponential Moving Average (EWMA). ;in This is the smoothing coefficient; the default value is [value]. This is to suppress frequent jitter caused by short-term noise.

[0040] Health Prediction Module: This module is used to form trend predictions of the future operating status of circuit board equipment, enabling the system to intervene early and manage proactively. Based on historical and real-time operating data, this module establishes a health prediction model, expanding the changes in equipment operating status from "current values" to "trend quantities." This allows the remote management system to identify the direction and extent of equipment health degradation before anomalies evolve into faults or overloads. Inputs and Outputs: Inputs are historical and real-time operating data; output is the health trend quantity of the circuit board equipment. The health trend quantity characterizes the changing trend of equipment health status over time, supporting the dependency control module in considering both "dependency impact" and "health trend" when making control decisions. The health prediction module provides the system with a predictive perspective, avoiding delayed processing based solely on current anomalies or current loads, making control actions and fault-tolerant strategies more balanced in terms of risk mitigation and continuous operation.

[0041] In one specific implementation, to establish a health prediction model based on historical and real-time operating data and output the health trend of the circuit board equipment, the following steps can be performed: Historical and real-time operating data are homogenized and quality controlled. Monitoring quantities with different sampling frequencies are aligned with a unified time granularity. Missing segments are interpolated and marked, and abnormal peaks are retained. At the same time, key observations reflecting the health status of circuit board equipment are constructed into a continuously traceable health feature sequence. The health feature sequence includes at least temperature, voltage, current, clock stability, resource utilization, communication quality indicators, and load-related fluctuation characteristics. Based on this, multi-scale time windows are generated to simultaneously cover short-term jitter and long-term drift, so that subsequent modeling can identify the difference between short-term fluctuations caused by transient disturbances and long-term trends caused by device aging or heat dissipation degradation. A health prediction model is established using health feature sequences. First, the normal evolution trajectory and fluctuation boundaries are learned from historical operating data. Then, real-time operating data is injected in a rolling window for continuous correction, enabling the health prediction model to adaptively update its understanding of the current state during operation. The prediction target is defined as the health trend quantity facing the future window. The health trend quantity is used to characterize the direction, rate of change, and stability of the health status. For example, when the average temperature of the circuit board equipment rises week by week under the same load conditions and the rate of temperature drop slows down, while the latency distribution in the communication quality index gradually shifts to the right and more frequent retransmissions appear, the health prediction model will learn the correlation between this combination of features and health degradation. Thus, when subsequent real-time operating data arrives, it will give a continuously decreasing health trend quantity and reflect its downward slope. The output of the health prediction model is standardized into a health trend quantity expression that can be used for remote management decision-making. Within each rolling window, the trend direction, trend strength, and trend stability are simultaneously output, and consistency is verified with real-time operational data to suppress the interference of occasional noise on trend judgment. This ensures that the health trend quantity reflects both short-term risk escalation and long-term degradation accumulation. For example, when a circuit board device experiences a momentary temperature increase due to external interference but quickly recovers while resource utilization and communication quality indicators remain stable, the trend direction of the health trend quantity remains stable, and the trend strength is limited to a low level. Conversely, when the circuit board device exhibits increased voltage fluctuations, persistently near-saturated resource utilization, and continuously deteriorating communication quality indicators across multiple consecutive windows, the health trend quantity will show a clear downward direction and provide a higher trend strength, indicating a gradually evolving health risk for the circuit board device.

[0042] The Dependency Control Module transforms "assessment results" into "executable remote management and control behaviors." When dependency impact indicators show that certain dependencies have a strong tendency for risk propagation or amplification, and dynamic thresholds and health trend indicators suggest the need for intervention, the Dependency Control Module generates control decisions and executes control actions. This intervenes in system operation at the dependency level, suppressing the risks of abnormal propagation, overload spread, or cascading failures. Inputs and Outputs: Inputs are dependency impact indicators, dynamic thresholds, and health trend indicators; outputs are control decisions and executed control actions. These control actions, as key execution results in the system management closed loop, are further written into the dependency knowledge graph, becoming an important basis for the dynamic reconstruction and optimization of the knowledge graph reconstruction module. The Dependency Control Module aggregates the information output from the Intensity Update Module and the Health Prediction Module into decision-making criteria, enabling the system to control the propagation chain from the perspective of dependencies and the direction of risk evolution from the perspective of health trends, forming a coordinated control system oriented towards "structure-state-trend."

[0043] In one specific implementation, to generate control decisions and execute control actions based on dependency impact indicators, dynamic thresholds, and health trend quantities to suppress error propagation across circuit board devices, the following steps can be followed: The system uniformly aggregates the dependency impact indicators, dynamic thresholds, and health trend quantities corresponding to each circuit board device and aligns them with a preset time window. Then, it performs risk classification judgment on the dependency relationship between each circuit board device. The risk classification judgment must simultaneously satisfy the judgment of the excess magnitude of the dependency impact indicator relative to the dynamic threshold and the judgment of the deterioration direction of the health trend quantity, so as to distinguish between short-term fluctuations and continuous risks. For example, when the dependency impact indicator from circuit board device A to circuit board device B is higher than the dynamic threshold for multiple consecutive windows and the health trend quantity is in a downward direction, the dependency relationship is judged as high risk and enters the control candidate set. Based on the risk classification results, control decisions are generated. Consistent action selection rules are set for different risk levels and action parameters are output. For low risk, dependency reduction is prioritized to reduce the intensity of impact propagation. For medium risk, configuration freezing is superimposed on dependency reduction to avoid configuration drift amplifying the risk. For high risk, dependency blocking is prioritized to cut off the error propagation link. When the health trend of the target circuit board device continues to deteriorate, task migration is triggered to unload the risk source. At the same time, each action is accompanied by trigger conditions, duration and rollback conditions to ensure controllability. For example, when the health trend of circuit board device B decreases and the resource utilization is close to saturation, the control decision can choose to perform dependency blocking from circuit board device A to circuit board device B and migrate the high-frequency tasks carried by circuit board device B to circuit board device C with a more stable health trend. Based on the control decisions, control actions are executed and their effects are verified. During execution, dependency downgrading or dependency blocking is first issued to immediately suppress error propagation. Then, configuration freezing is issued to stabilize key parameters. Finally, when the migration window and resource availability conditions are met, task migration is executed to restore overall stability. After execution, the dependency impact indicators are continuously monitored to see if they fall back to near the dynamic threshold and whether the health trend is stabilizing. If the verification passes, the action is maintained until the end of the duration and gradually released according to the rollback conditions. If the verification fails, the action intensity is increased or the blocking scope is expanded. For example, when the control behavior issued by circuit board device A causes multiple circuit board devices to issue abnormal warnings and show accelerated information propagation, dependency blocking can be executed on the relevant dependencies of circuit board device A to immediately cut off the propagation. Then, configuration freezing can be executed on key configurations to prevent misconfiguration from continuing to spread. At the same time, key business tasks are migrated from the most affected circuit board device B to circuit board device C, thereby suppressing error propagation while maintaining the continuous operation of key functions.

[0044] Dynamic threshold generation (hierarchical baseline + window adaptation): for each dependent edge In the window Generate dynamic threshold It is used for anomaly detection and control of trigger boundaries. The threshold adopts a "layered baseline + window adaptive" approach to adaptively characterize the normal fluctuation range and adjust the threshold according to the operating stage (such as stable / fluctuating / high load).

[0045] (a) Baseline distribution: Selected during the stable operation phase A set of historical windows forms a baseline. (or use the most recent) Scroll baseline of each window ), to obtain the dependency impact index sample set ; (ii) Threshold calculation method one: Quantile threshold: based on the rolling baseline sample set Quantile generation threshold: ;in for Quantile operator. To reflect stage-specific adaptation, we can let Values ​​are assigned in segments based on behavior patterns: Values ​​are increased during high-load segments. To suppress jitter false alarms, the adjustment is lowered during the stable phase. To improve sensitivity and capture persistent offsets.

[0046] (III) Threshold Calculation Method Two: Mean + Standard Deviation; Thresholds are generated based on the mean and standard deviation of the rolling baseline: ;in They are respectively The rolling mean and rolling standard deviation; multiples Segment settings based on behavior pattern: Higher load segments take larger values. To suppress false alarms caused by short-term noise, a smaller stable range is selected. To enhance the ability to detect slow drift and persistent anomalies.

[0047] Fault Tolerance Adjustment Module: This module dynamically adjusts the system's tolerance strategy for anomalies and overloads in response to fluctuations in equipment operating status, load changes, communication quality variations, or changes in dependencies. By dynamically adjusting fault tolerance thresholds and levels in real-time or based on preset periods, the module enables the system to have different fault tolerance configurations at different operational stages. This ensures safety while maintaining availability as much as possible, avoiding overly conservative or overly aggressive fixed threshold strategies. Input and Output: This module dynamically calculates and adjusts based on operating status and management objectives, outputting corresponding fault tolerance strategies. The fault tolerance strategy clarifies the appropriate degradation or protection method to be adopted when anomalies or overloads occur under the current threshold and level configuration. The fault tolerance adjustment module complements the dependency control module: dependency control focuses on "intervention at the dependency level," while fault tolerance adjustment focuses on "adaptive adaptation of system capacity and response methods," jointly supporting the fault tolerance strategy-driven module in implementing specific operational control.

[0048] In one specific implementation, in order to achieve real-time or dynamic adjustment of the fault tolerance threshold and fault tolerance level of the circuit board device based on a preset period, and to output an executable fault tolerance strategy, the following steps can be followed: The system uniformly collects real-time operating status, load, resource margin, communication quality, dependency impact indicators, and health trend quantities of circuit board equipment, and aligns and normalizes them using a consistent time window. Then, it constructs a combined expression of fault tolerance assessment quantities, taking the fluctuation amplitude and stability of key physical quantities in the real-time operating status as the status item, the rate of change of load and resource margin as the bearing item, the latency, packet loss, and jitter in communication quality as the link item, the dependency impact indicators as the propagation item, and the direction and slope of the health trend quantity as the lifetime item. By fusing the status item, bearing item, link item, propagation item, and lifetime item, the fault tolerance threshold and fault tolerance level are obtained. The fault tolerance threshold is used to give the tolerable deviation boundary, and the fault tolerance level is used to give the redundancy and isolation strength to be adopted. Fault tolerance strategies are generated and implemented online based on fault tolerance thresholds and fault tolerance levels. A fault tolerance strategy includes at least triggering conditions, action sets, and fallback conditions. Triggering conditions are defined by the fault tolerance threshold and link items. The action set is mapped from the fault tolerance level and may include rate limiting, load reduction, retry count adjustment, redundancy switching priority adjustment, and isolation range expansion. Fallback conditions are defined by the stabilization of health trend and the decline of dependency impact indicators, enabling the strategy to tighten rapidly when risks increase and gradually loosen when risks ease. For example, when the load on circuit board device A increases and resource margin decreases, while communication quality jitter increases, leading to an increase in dependency impact indicators, and the health trend shows a continuous decline, the fault tolerance level is raised, triggering a stricter fault tolerance threshold. This causes the fault tolerance strategy to lower the upper limit of retry counts to avoid storm-like retransmissions, while simultaneously increasing redundancy switching priority and expanding the isolation range to prevent the cascading spread of errors between multiple circuit board devices. The fault tolerance threshold and level are evaluated and dynamically adjusted according to a preset cycle, and the fault tolerance strategy is updated accordingly. The evaluation performs statistical review of the status, load, link, propagation and lifetime items of the recent window in each cycle to identify long-term drift and phased characteristic changes. The fault tolerance threshold is adaptively converged or relaxed, and the fault tolerance level is raised and lowered in stages to avoid frequent jitter caused by short-term noise. For example, if the health trend is stable and the dependent impact index is consistently below the judgment boundary during the low-load period at night, the fault tolerance threshold will be appropriately relaxed and the fault tolerance level will be lowered to reduce redundancy overhead. During the high-concurrency period during the day, if the link and propagation items fluctuate more, the fault tolerance level can be raised and the fault tolerance threshold tightened even if there is no obvious failure. This makes the output fault tolerance strategy more in line with the risk profile of different operating stages and keeps it stable and controllable.

[0049] For equipment The health-related feature sequences (temperature, voltage, current, clock stability, resource utilization, communication quality, etc.) are constructed into a unified feature vector. Then, input the prediction model in the window. (Set in the window...) At any given moment, the model outputs the future. Step-by-step health score prediction sequence ; Among them, health score It can be calculated by weighting key features or directly output from the model's latent variables. Let the predicted mean be... ; The health trend measure is defined as consisting of the following three parts: (a) Trend direction volume: ;in These represent downward, stable, and upward trends, respectively.

[0050] (ii) Trend strength (regression slope magnitude): Let The strength of the trend is characterized by the slope magnitude of a univariate linear regression of the predicted sequence against a time index:

[0051] in This is a numerically stable term used to avoid the denominator being zero.

[0052] (iii) Trend-stabilizing quantities (measured by residual fluctuations): Let the residual sequence of "actual score and predicted score" within the recent window be: ;in Let be the set of indices of the most recent windows; let the variance of the baseline residuals be . The trend stabilizing quantity is defined as follows: ;in for Internal residual variance Truncate the result to The closer the stable quantity is to This indicates that the smaller the deviation between the prediction and the actual result, the more stable the trend; the closer the prediction is to the actual result. This indicates that the residual fluctuates significantly and the trend is unstable.

[0053] Fault-Tolerance Strategy Driven Module: This module implements fault-tolerance strategies into the operational mode of the circuit board equipment, ensuring the strategies are executable and produce tangible results. Based on the fault-tolerance strategy, this module drives the circuit board equipment to perform degraded or protective operations to cope with abnormal or overload conditions, thereby maintaining system stability and critical functional continuity in uncertain operating environments and reducing the impact of anomaly propagation on the overall system. Input and Output: The input is the fault-tolerance strategy; the output reflects the control results of the circuit board equipment's degraded or protective operations. The execution results, along with system interaction information, constitute one of the factual bases for subsequent dependency knowledge graph updates and reconstructions. The Fault-Tolerance Strategy Driven Module receives the output from the fault-tolerance adjustment module, ensuring that the "strategy" not only remains at the rule level but is transformed into observable changes in the equipment's operating state, providing traceable execution feedback for the knowledge graph reconstruction module.

[0054] In one specific implementation, in order to drive the circuit board equipment to degrade or protect it based on a fault-tolerant strategy to cope with abnormal or overload conditions of the equipment, the following steps can be taken: Based on the preset fault tolerance strategy, the trigger conditions and standards for degradation or protection are determined for different abnormal or overload states. The trigger conditions include excessive device load, degraded communication quality, and excessive resource utilization. When the trigger conditions are met, each device will enter a degradation or protection state according to the fault tolerance strategy. For example, when the device load exceeds a set threshold and the health trend value decreases, the system will automatically trigger a degradation operation. Based on the set of actions defined by the fault-tolerant strategy, the system drives the device to perform corresponding degradation or protection actions. These actions may include task migration, service degradation, load balancing, and functional isolation to reduce the pressure on the device. For example, when circuit board device A is overloaded and its health trend continues to decline, the system can migrate some tasks to circuit board device B, which has a lighter load, or reduce some non-core functions of device A to avoid complete failure. After implementing degradation or protection measures, continuously monitor the equipment's operating status and health trends, and adjust fault-tolerant strategies based on real-time feedback to ensure the equipment operates as stably as possible. If the equipment's health trends continue to deteriorate, further protection measures can be strengthened, such as completely eliminating the source of the problem through redundancy switching or fault isolation to ensure the overall stability of the equipment group. For example, when degradation measures are ineffective, a more stringent protection strategy can be implemented for circuit board device A to ensure that the fault does not spread to other devices.

[0055] The knowledge graph reconstruction module is used to precipitate the dependency structure, parameter states, control actions, and interaction facts during system operation into a continuously evolving dependency knowledge graph. It updates the knowledge graph as needed, enabling dynamic reconstruction and optimization of dependencies. This module writes dependency topology, dependency parameter sets, control actions, and system interaction information into the dependency knowledge graph, making dependencies not just current calculation results, but possessing historical traceability and iterative optimization capabilities. When the system operating environment changes, device behavior patterns change, or interaction relationships shift, the knowledge graph may deviate from the latest state. Therefore, when triggering conditions are met, this module triggers an update of the dependency knowledge graph based on graph structure offset and semantic inconsistency. This update mechanism ensures that the knowledge graph remains consistent with the actual operating relationships and supports dynamic reconstruction and optimization of dependencies. Input and Output: Inputs include dependency topology, dependency parameter sets, control actions, and system interaction information, combined with graph structure offset and semantic inconsistency to determine whether to trigger an update. The output is the updated dependency knowledge graph and the supported dynamic reconstruction and optimization results of dependencies. The knowledge graph reconstruction module runs through the closed loop: it absorbs both the front-end modeling and update results, as well as the control and fault-tolerant execution results, enabling the system to have the ability to continuously evolve in the long run by "updating with facts - optimizing with updates - making decisions again with optimizations".

[0056] In one specific implementation, to write dependency topology, dependency parameter set, control actions, and system interaction information into the dependency knowledge graph, and to trigger an update of the dependency knowledge graph based on the graph structure offset and semantic inconsistency when triggering conditions are met, so as to complete the dynamic reconstruction and optimization of dependencies, the following steps can be followed: The knowledge writing and alignment process involves uniformly identifying and aligning the circuit board device nodes, dependency direction edges and their weights in the dependency topology, the resource allocation ratio, data transmission delay, and tolerable fault impact range in the dependency parameter set, as well as event elements in the control actions such as dependency weight reduction, dependency blocking, configuration freezing, and task migration, with the device state vector, data interaction records, and control command links in the system interaction information. This information is then recorded with a traceable version number, enabling the continuous characterization of the relationship changes of the same circuit board device in different time windows. For example, the dependency edge from circuit board device A to circuit board device B, along with its propagation delay, the number of historical control actions, and the triggering reasons, are written together to form a dependency relationship knowledge expression that integrates structural and semantic information. Subsequently, trigger condition evaluation is performed. Within each evaluation window, graph structure offset and semantic inconsistency are calculated. Graph structure offset can be quantified based on factors such as the rate of change of the dependency edge set, the drift magnitude of the in-degree and out-degree of key nodes, and the statistical distance of edge weight distribution. It is used to characterize the degree of structural migration of the dependency topology relative to the existing graph. Semantic inconsistency can be obtained based on the difference between the actual data flow and causal relationship of the control link reflected in the system interaction information and the matching difference with the semantics of existing relationships in the dependency relationship knowledge graph. It is used to characterize whether the meaning of the relationship has deviated. The first and second preset thresholds can be set based on existing... In engineering practice, statistical process control and reliability management standards are established, which prioritize the historical window of the stable operation phase as the baseline distribution, set alarm limits according to quantile boundaries or the mean plus the standard deviation, and perform tiered calibration in combination with equipment category, load level and communication quality level to avoid false updates under high fluctuation conditions, while ensuring sensitivity to small but continuous offsets under stable conditions. For example, the graph structure offset threshold can be set to the high quantile boundary of the baseline distribution to capture significant structural migration, and the semantic inconsistency threshold can be set to a stricter boundary of the baseline distribution to prioritize the capture of causal semantic drift. Finally, when the graph structure offset exceeds the first preset threshold and / or the semantic inconsistency exceeds the second preset threshold, an update is triggered and a reconstruction write-back is executed. After triggering, the dependency topology is reconstructed first, and the direction, weight, and critical path of the dependency edge are re-evaluated based on the latest system interaction information. Then, the dependency parameter set is reconstructed, and parameters such as propagation delay, resource coupling strength, and tolerance range are updated. Subsequently, the updated dependency topology and dependency parameter set are written into the dependency relationship knowledge graph as a new version, and the control actions and triggering reasons associated with this update are archived synchronously for backtracking and self-verification. For example, if it is recently observed that circuit board device A originally mainly outputs data to circuit board device B, but the actual interaction record shows that the data path gradually shifts to circuit board device C and the triggering order of the control command link changes, resulting in a significant increase in the graph structure offset, and the dependency relationship knowledge graph still retains the strong dependency semantics from A to B, which is inconsistent with reality and increases the semantic inconsistency, then after triggering the update, the dependency edge from A to B will be downweighted or redirected, and the new dependency edge from A to C, along with its propagation delay and risk tolerance range, will be written, thereby realizing the dynamic reconstruction and continuous optimization of the dependency relationship.

[0057] When the actual dependencies reflected in the system interaction information deviate from the existing associations in the knowledge graph, the graph structure offset is calculated. Semantic inconsistency And it is compared with the threshold to trigger an update.

[0058] (a) Graph Structure Offset: Let the set of dependency edges of the current version of the knowledge graph be... The candidate edge set obtained by re-estimation based on the latest system interaction information is: Define the rate of change of the edge set (Jaccard distance): Let the in-degree / out-degree distribution vectors of the two graphs be respectively... Defining degree drift: ;in This is the normalization factor (which can take the maximum possible value or the baseline mean). Let the edge weight distributions be as follows: Measured by statistical distance: ;in Either Wasserstein distance or KL distance can be used (either can be chosen). The final graph structure offset is: ;in The allocation example is to take... , , The reason is: (Rate of change of edge set / Jaccard distance) directly reflects "whether there are new / disappearing dependent edges", which is the most sensitive to whether the knowledge graph structure needs to be reconstructed and has the most intuitive semantics, so it is used as the dominant factor; (In-degree and out-degree drift) can characterize the migration of the center of gravity from "point to surface" depending on the topology, such as a node changing from the edge to the hub. It is an important supplement to structural changes, and weight is secondary. (Edge weight distribution distance) reflects the overall fluctuations or changes in the distribution pattern of intensity more. It may fluctuate in stages due to external factors such as load and sampling window. Therefore, the weight is slightly lower to avoid misjudging "intensity jitter" as "structural reconstruction".

[0059] (ii) Semantic inconsistency: The "causal candidate edge set" extracted from system interaction information. This includes: control command trigger-response causal edges, exception sequence causal edges, and data flow edges. The corresponding set of semantic relationship edges in the knowledge graph is as follows: Define the weighted mismatch rate: ;in The evidence weight for an edge can be obtained by combining factors such as "number of observations, time delay stability, and proportion of repeated windows".

[0060] Example 2: A remote management method for circuit board equipment based on the Internet of Things, comprising the following steps: Step 1: Collect system interaction information from the circuit board equipment; Step 2: Model dependency relationships based on system interaction information, generate dependency topology and form dependency parameter set; Step 3: Generate and update dependency impact metrics and dynamic thresholds based on the dependency parameter set, device environment parameters, historical behavior, and anomaly warnings; Step 4: Build a health prediction model based on historical and real-time data, and output health trends; Step 5: Generate and execute regulatory decisions based on dependency impact indicators, dynamic thresholds, and health trends; Step 6: Dynamically adjust the fault tolerance threshold and fault tolerance level, and update the fault tolerance strategy; Step 7: Write the dependency topology, dependency parameter set, control actions and interaction information into the knowledge graph, and trigger updates based on the graph structure offset and semantic inconsistency to complete the dynamic reconstruction and optimization of the dependency relationship.

[0061] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A remote management system for circuit board equipment based on the Internet of Things, characterized in that, Include: The dependency modeling module is used to collect system interaction information of circuit board equipment, and perform dependency relationship modeling based on the system interaction information to generate the dependency topology of circuit board equipment and form a dependency parameter set. System interaction information includes: circuit board device status vector, data interaction records, control command links, and abnormal event time series; The circuit board equipment state vector includes the circuit board equipment's real-time operating parameters, historical operating modes, resource utilization, and communication quality indicators; Data interaction records include the order, frequency, and delay of data transmission between circuit board devices; The control command link includes the command type, command triggering time, and response delay of control signal transmission between circuit board devices; The dependency modeling module further includes: The operation coordination and consistency calculation module is used to output the operation coordination and consistency index between circuit board devices based on the circuit board device state vector, data interaction record, control command link and historical operation mode in the system interaction information. The cross-circuit board device abnormal coupling calculation module is used to output the cross-circuit board device abnormal coupling index based on the time sequence of abnormal events and the control command link in the system interaction information. The information dissemination influence calculation module is used to analyze the control command link and the time series of abnormal events, and output the information dissemination influence index between circuit board devices; The dependency modeling module is also used for: By collecting circuit board device state vectors, data interaction records, and control command links from system interaction information, an interaction network diagram between circuit board devices is constructed and the interaction behavior is analyzed. The dependencies between circuit board devices are determined based on the operational coordination consistency index, the cross-circuit board device abnormal coupling index, and the information propagation influence index. The dependency topology is then generated by combining resource utilization, historical operating modes, and communication latency. The dependency modeling module obtains the dependency edge weight by fusing the operational coordination consistency index, the cross-circuit board device abnormal coupling index and the information propagation influence index. When the dependency edge weight is greater than the edge establishment threshold, a directed dependency edge is established, and the propagation delay parameter, resource coupling strength and tolerable influence range are written into the dependency parameter set. Calculate the dependency parameter set based on the dependency topology; The intensity update module is used to generate and update the dependency impact index online based on the dependency parameter set, equipment operating environment parameters, historical operating behavior patterns and abnormal warning information, and output the dependency impact index and dynamic threshold. The health prediction module is used to build a health prediction model based on historical and real-time operating data and output the health trend of the circuit board equipment. The dependency control module is used to generate control decisions and execute control actions based on dependency impact indicators, dynamic thresholds, and health trend data. The fault tolerance adjustment module is used to dynamically adjust the fault tolerance threshold and fault tolerance level in real time or based on a preset period, and output the corresponding fault tolerance strategy. The fault-tolerant strategy driver module is used to drive the circuit board equipment to degrade or protect based on the fault-tolerant strategy in order to deal with abnormal or overload conditions of the equipment. The knowledge graph reconstruction module is used to write dependency topology, dependency parameter set, control actions and system interaction information into the dependency knowledge graph. When the triggering conditions are met, the module will trigger the update of the dependency knowledge graph based on the graph structure offset and semantic inconsistency, so as to complete the dynamic reconstruction and optimization of dependencies.

2. The IoT-based remote management system for circuit board equipment according to claim 1, characterized in that, The control measures include at least one of dependency demotion, dependency blocking, configuration freezing, and task migration to suppress the propagation of errors across circuit board devices.

3. The IoT-based remote management system for circuit board equipment according to claim 1, characterized in that, The fault tolerance adjustment module is used for: The fault tolerance threshold and fault tolerance level are calculated based on the real-time operating status, load, resource reserves, communication quality, dependency impact indicators, and health trend of the circuit board equipment. The fault tolerance threshold and level are dynamically adjusted based on a preset periodic evaluation, and the fault tolerance strategy is updated accordingly.

4. The IoT-based remote management system for circuit board equipment according to claim 1, characterized in that, The knowledge graph reconstruction module triggers dependency relationships, and the triggering conditions for knowledge graph updates include: The graph structure offset is calculated based on system interaction information and dependency relationship knowledge graph. When the graph structure offset exceeds a first preset threshold, the dependency relationship knowledge graph is updated; and / or The semantic inconsistency is calculated based on the system interaction information and the dependency knowledge graph. When the semantic inconsistency is greater than the second preset threshold, the dependency knowledge graph is updated. Specifically, when the dependency knowledge graph is updated, the dependency topology and dependency parameter set are reconstructed, and the updated dependency topology and dependency parameter set are written into the dependency knowledge graph.

5. A method for remote management of circuit board equipment based on the Internet of Things (IoT), applied to the remote management system for circuit board equipment based on the IoT as described in any one of claims 1-4, characterized in that, Includes the following steps: Step 1: Collect system interaction information from the circuit board equipment; Step 2: Model dependency relationships based on system interaction information, generate dependency topology and form dependency parameter set; Step 3: Generate and update dependency impact metrics and dynamic thresholds based on the dependency parameter set, device environment parameters, historical behavior, and anomaly warnings; Step 4: Build a health prediction model based on historical and real-time data, and output health trends; Step 5: Generate and execute regulatory decisions based on dependency impact indicators, dynamic thresholds, and health trends; Step 6: Dynamically adjust the fault tolerance threshold and fault tolerance level, and update the fault tolerance strategy; Step 7: Write the dependency topology, dependency parameter set, control actions and interaction information into the knowledge graph, and trigger updates based on the graph structure offset and semantic inconsistency to complete the dynamic reconstruction and optimization of the dependency relationship.

Images