Industrial equipment state early warning system and method based on multi-sensor cooperative monitoring

Abstract

The invention discloses a multi-sensor cooperative monitoring industrial equipment state early warning system and method, and particularly relates to the technical field of cooperative monitoring industrial equipment state early warning. Comprising the following steps: acquiring monitoring data acquired by multiple sensor nodes corresponding to target equipment in a monitoring area and a current forwarding binding relationship of the multiple sensor nodes in wireless networking, performing association organization on the monitoring data, and outputting a node association data set and a current binding relationship set; performing synchronous anomaly identification on the node associated data set, determining an abnormal node group with abnormal fluctuation in the same abnormal time period, extracting a shared forwarding binding relationship corresponding to the abnormal node group from the current binding relationship set, and outputting a to-be-judged abnormal fragment set; on the basis of identifying an abnormal node group in the same abnormal time period, controlled disturbance is executed on a shared forwarding binding relation, and abnormal attribution is judged according to abnormal form changes before and after disturbance.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment condition early warning technology based on collaborative monitoring, and more specifically, to an industrial equipment condition early warning system and method based on multi-sensor collaborative monitoring. Background Technology

[0002] In existing industrial equipment status early warning technologies, the mainstream approach in the industry mainly addresses the problem of timely detection of equipment anomalies and output of early warnings through the collection of results from multiple sensors. Typically, vibration, temperature, current, acoustic or pressure sensors are deployed around the target equipment, and the collection results from each node are uploaded to the gateway or central side through wireless networking. The source of the anomaly is then determined by combining threshold comparison, spatial clustering or equipment attribution rules. For example, in scenarios where multiple cabinet devices or fan arrays are operating in close proximity in a computer room, the structural spacing between adjacent devices is small, and vibration and heat propagation are significant. At the same time, there are also hard constraints on site, such as strong metal shielding, node sharing of relays, frequent wireless time slot competition, and the inability to significantly increase the sampling load. Under this constraint, the mainstream approach will consistently reveal an observable and verifiable defect: when multiple nodes in a certain area simultaneously exhibit abnormal fluctuations, although the system can detect the abnormal phenomenon, it cannot reliably determine whether the abnormality originates from the actual diffusion effect of a target device, the independent abnormality of adjacent devices, or the accompanying abnormality caused by the interference of the wireless network link. Consequently, the abnormal device object, abnormal scope, and abnormality type corresponding to the same abnormal event change repeatedly in different warning rounds, resulting in numerous warning contents but unclear attribution and insufficient credibility. The technical problem this application aims to solve is: how to distinguish the true attribution of multiple node synchronization anomalies in an industrial scenario where multiple devices are arranged in close proximity and use wireless networking for collaborative monitoring, so as to achieve industrial equipment status early warning with rich warning content and reliable attribution. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an industrial equipment status early warning system and method for multi-sensor collaborative monitoring. By identifying abnormal node groups within the same abnormal time period, controlling the perturbation of their shared forwarding binding relationship, and determining the attribution of the abnormality based on the changes in the abnormality morphology before and after the perturbation, the system solves the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for early warning of industrial equipment status based on multi-sensor collaborative monitoring, comprising: S1. Obtain the monitoring data collected by the multi-sensor nodes corresponding to each target device in the monitoring area and the current forwarding binding relationship of the multi-sensor nodes in the wireless network. Perform association organization on the monitoring data and output the node association dataset and the current binding relationship set. S2. Perform synchronous anomaly identification on the node association dataset, identify the abnormal node group that has abnormal fluctuations within the same abnormal time period, extract the shared forwarding binding relationship corresponding to the abnormal node group from the current binding relationship set, and output the set of abnormal segments to be judged. S3. For the set of abnormal segments to be determined, keep the installation position and collection object of the abnormal node group unchanged, perform controlled disgrouping or controlled merging operations on the shared forwarding binding relationship to form a disturbance forwarding binding relationship, and reacquire the monitoring data corresponding to the abnormal node group based on the disturbance forwarding binding relationship, and output the disturbance observation dataset. S4. Perform a correspondence comparison between the set of abnormal segments to be judged and the disturbance observation dataset to determine whether the abnormal fluctuations have migrated, reorganized or disappeared due to changes in the forwarding binding relationship, and output the abnormal migration judgment result. S5. Perform attribution determination on the abnormal migration judgment result. When the abnormal fluctuation maintains the same spatial attribution of the corresponding target device before and after the change of the forwarding binding relationship, it is determined to be a real abnormality of the target device. When the abnormal fluctuation migrates, reassembles or disappears with the change of the forwarding binding relationship, it is determined to be an abnormality accompanying the wireless networking, and the target device status warning information is output according to the attribution determination result.

[0005] In a preferred embodiment, S1 includes: S1-1. Obtain the node identifier, target device identifier, sensor type identifier, acquisition time, and current parent node identifier corresponding to each multi-sensor node. Classify each multi-sensor node according to the target device identifier, sensor type identifier, and current parent node identifier, and output the node classification result. S1-2. Sort the multi-sensor nodes of each type in the node classification results according to the acquisition time, and output the node association fragment and binding change fragment for the statistical data missing status and parent-child correspondence changes of the sorted multi-sensor nodes. S1-3. Write the node association fragments and binding change fragments according to the target device identifier to generate the node association dataset and the current binding relationship set.

[0006] In a preferred embodiment, S2 includes: S2-1. Read the target device identifier, sensor type identifier, acquisition time and monitoring data of each multi-sensor node in the node association dataset. Perform fluctuation comparison on the monitoring data of each multi-sensor node under continuous acquisition time, determine the abnormal start time and abnormal duration of each multi-sensor node, and output the abnormal period results of the node. S2-2. Perform overlapping identification on the results of abnormal node periods according to the duration of abnormal periods, identify multiple multi-sensor nodes that simultaneously exhibit abnormal fluctuations within the same abnormal period, generate abnormal node groups according to the target device identifier and the corresponding abnormal period, and output the abnormal node group results. S2-3. Read the parent-child binding relationship corresponding to the result of the abnormal node group in the current binding relationship set, perform shared parent node identification and continuous forwarding path tracing for each multi-sensor node in the same abnormal node group, extract the shared forwarding binding relationship corresponding to the abnormal node group, and generate a set of abnormal segments to be judged in combination with the corresponding abnormal time period.

[0007] In a preferred embodiment, the process of outputting the abnormal node group results in S2-2 further includes: S2-21. Read the target device identifier, abnormal start time, and abnormal end time of each multi-sensor node in the abnormal time period results. Construct the abnormal time period intervals corresponding to each multi-sensor node through the abnormal start time and abnormal end time. Perform overlapping duration calculation, boundary interval calculation, and interval inclusion relationship calculation on any two abnormal time period intervals. Determine the abnormal time period intervals with overlapping duration greater than zero and boundary intervals of zero as adjacent time period intervals. Determine the abnormal time period intervals with interval inclusion relationships as included time period intervals. Generate time period association results based on adjacent time period intervals and included time period intervals.

[0008] In a preferred embodiment, the process of outputting the abnormal node group results in S2-2 further includes: S2-22. The time period association results are aggregated within the group according to the target device identifier, and an abnormal node time period graph is constructed for the time period association results corresponding to each target device. Multi-sensor nodes with time period association results are identified as graph nodes, and the time period association results between any two graph nodes are identified as graph edges. Connected subgraph identification is performed on each abnormal node time period graph. Connected subgraphs containing only a single graph node are removed. For conflicting nodes that appear in multiple connected subgraphs at the same time, the total overlap time with the other graph nodes in each connected subgraph is calculated. Conflicting nodes are assigned to the connected subgraph with the higher value of the total overlap time. The abnormal node subgraph results are output. S2-23. Perform common abnormal time period calculation on each connected subgraph in the abnormal node subgraph result. Determine the later time among the abnormal start times of each multi-sensor node in each connected subgraph as the common start time, and determine the earlier time among the abnormal end times of each multi-sensor node in each connected subgraph as the common end time. When the common start time is earlier than the common end time, determine the multi-sensor nodes in the corresponding connected subgraph as the abnormal node group in the same abnormal time period, and write the target device identifier with the common start time and common end time to generate the abnormal node group result.

[0009] In a preferred embodiment, S3 includes: S3-1. Read the shared forwarding binding relationship, target device identifier and abnormal time period corresponding to each abnormal node group in the set of abnormal segments to be judged. Perform parent node affiliation statistics and path overlap statistics on each multi-sensor node in the shared forwarding binding relationship. Determine the multi-sensor nodes with the same parent node identifier as the same parent node set. Determine the multi-sensor nodes that pass through the same intermediate node in the forwarding path as the same path set. Output the binding aggregation result. S3-2. Determine the perturbation method for the binding aggregation result. When the number of multi-sensor nodes in the same parent node set is not less than two, perform a controlled splitting operation on the same parent node set, and switch at least one of the multi-sensor nodes to another parent node identifier that is not occupied by the same parent node set. When the multi-sensor nodes in the same path set belong to different parent node identifiers, perform a controlled merging operation on the same path set, and merge the multi-sensor nodes with the same forwarding path endpoint into the same parent node identifier to generate a perturbation forwarding binding relationship.

[0010] In a preferred embodiment, S3 further includes: S3-3. Write the disturbance forwarding binding relationship into each multi-sensor node corresponding to the abnormal node group. Perform synchronous acquisition on each multi-sensor node after writing in the subsequent acquisition period corresponding to the original abnormal period to obtain the monitoring data of each multi-sensor node under the disturbance forwarding binding relationship and generate a disturbance observation dataset. S3-4. Perform corresponding association on the disturbance observation dataset according to the target device identifier, multi-sensor node identifier, acquisition time and disturbance forwarding binding relationship. Write the changes in parent node identifier before and after the disturbance, the changes in forwarding path and the corresponding changes in monitoring data accordingly to generate the bound disturbance association fragment in the disturbance observation dataset.

[0011] In a preferred embodiment, S4 includes: S4-1. Read the target device identifier, common abnormal time period, abnormal node group composition and shared forwarding binding relationship corresponding to each abnormal node group in the abnormal segment set to be judged, as well as the disturbance forwarding binding relationship and disturbance monitoring data of the corresponding abnormal node group in the disturbance observation data set, perform abnormal re-identification on the disturbance monitoring data, and generate disturbance abnormal node group results. S4-2. Perform a corresponding comparison between the composition of the abnormal node group and the result of the disturbed abnormal node group, calculate the difference in abnormal node identifiers before and after the disturbance, the difference in the common abnormal time period, and the difference in the corresponding target device identifier. When an abnormal node exits before the disturbance and other nodes enter after the disturbance, it is determined as an abnormal migration. When the connection relationship of abnormal nodes splits or merges before and after the disturbance, it is determined as an abnormal reorganization. When no corresponding abnormal node group is identified after the disturbance, it is determined as an abnormal disappearance. Output the abnormal morphology judgment result. S4-3. Perform corresponding write operations on the abnormal form determination results and the shared forwarding binding relationship and the disturbance forwarding binding relationship to generate an abnormal migration determination result containing the target device identifier, abnormal node group identifier, abnormal form category and binding relationship change results.

[0012] In a preferred embodiment, S5 includes: S5-1. Read the target device identifier, abnormal form category, shared forwarding binding relationship and disturbance forwarding binding relationship corresponding to each abnormal node group in the abnormal migration judgment result. Statistically analyze the distribution of target device identifiers and changes in node composition of each multi-sensor node in each abnormal node group before and after the disturbance. When the target device identifiers corresponding to the multi-sensor nodes in the abnormal node group remain consistent before and after the disturbance and the abnormal form category does not include migration, reorganization and disappearance, the corresponding abnormal node group is determined as the real abnormality of the target device, and the real abnormality result is output. S5-2. For each abnormal node group that is not determined to be a real abnormality of the target device in the abnormal migration judgment result, perform network companion judgment based on the abnormal form category and the change result of forwarding binding relationship before and after the disturbance. When the abnormal form category includes migration, reorganization or disappearance, the corresponding abnormal node group is determined as a wireless network companion abnormality and the companion abnormality result is output. S5-3. Write the actual abnormal results and accompanying abnormal results according to the target device identifier, generate target device status warning information for actual abnormalities of the target device, generate network interference warning information for accompanying abnormalities of wireless networking, and output the corresponding target device status warning information.

[0013] An industrial equipment status early warning system based on multi-sensor collaborative monitoring, the system comprising a data association module, an anomaly grouping module, a disturbance binding module, a judgment module, and an early warning module: The data association module is used to obtain the monitoring data collected by the multi-sensor nodes corresponding to each target device in the monitoring area and the current forwarding binding relationship of the multi-sensor nodes in the wireless network. It performs association organization on the monitoring data and outputs the node association dataset and the current binding relationship set. The anomaly grouping module is used to perform synchronous anomaly identification on the node-associated dataset, identify the abnormal node groups that exhibit abnormal fluctuations within the same abnormal time period, extract the shared forwarding binding relationship corresponding to the abnormal node group from the current binding relationship set, and output the set of anomaly fragments to be judged. The binding disturbance module is used to maintain the installation position and collection object of the abnormal node group for the set of abnormal segments to be judged, perform controlled disgrouping or controlled merging operations on the shared forwarding binding relationship, form a disturbance forwarding binding relationship, and re-acquire the monitoring data corresponding to the abnormal node group based on the disturbance forwarding binding relationship, and output the disturbance observation dataset. The judgment module is used to perform a correspondence comparison between the set of abnormal segments to be judged and the disturbance observation dataset, to determine whether the abnormal fluctuations have migrated, reorganized or disappeared due to changes in the forwarding binding relationship, and output the abnormal migration judgment result. The early warning module is used to determine the attribution of abnormal migration judgment results. When the abnormal fluctuation maintains the same spatial attribution of the corresponding target device before and after the change of forwarding binding relationship, it is determined to be a real abnormality of the target device. When the abnormal fluctuation migrates, reassembles or disappears with the change of forwarding binding relationship, it is determined to be an abnormality accompanying the wireless networking, and the target device status early warning information is output according to the attribution determination result.

[0014] The technical effects and advantages of this invention are as follows: 1. This solution first forms an abnormal node group and its shared forwarding binding relationship, then performs controlled disgrouping or controlled merging operations on the shared forwarding binding relationship, and performs attribution determination in combination with the abnormal morphology changes before and after the disturbance. This can distinguish the true attribution of multiple node synchronous anomalies, thereby relatively improving the problem of repeated changes in the warning object, warning range and warning type in the scenario of multiple devices in close parallel. 2. Perform single-node abnormal time period identification, time period association construction, abnormal node time period graph connectivity merging and common abnormal time period calculation on the node association dataset. This can organize discrete abnormal records into abnormal node groups within the same abnormal time period, thereby relatively reducing the misjudgment of synchronization anomalies caused by isolated fluctuations of single nodes or misalignment of abnormal time periods. 3. Under the condition that the installation location and the object of collection remain unchanged, only the forwarding binding relationship of the abnormal node group is changed and the monitoring data is reacquired. This can place the impact of the device status and the impact of the network binding change in the same comparison framework, thereby providing verifiable disturbance observation basis for subsequent abnormal migration, abnormal reorganization and abnormal disappearance determination. 4. Perform corresponding comparisons on the composition of abnormal node groups before and after the disturbance, the common abnormal time period, and the target device identifier, and generate abnormal migration judgment results by combining the shared forwarding binding relationship and the disturbance forwarding binding relationship. This can relatively improve the consistency of the judgment of abnormal morphological changes, thereby providing a unified judgment basis for distinguishing between real anomalies and accompanying anomalies. 5. During the attribution determination phase, the distribution of target device identifiers, changes in node composition, and changes in binding relationships are statistically analyzed simultaneously. Target device status warning information and network interference warning information are generated separately. This enables the warning output to cover both device status changes and network interference, thereby relatively improving the completeness of the warning content and the directionality of subsequent investigations. Attached Figure Description

[0015] Figure 1 This is a flowchart outlining the method steps of the present invention; Figure 2 This is a schematic diagram of the system module structure of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Refer to the instruction manual appendix Figure 1-2 The multi-sensor collaborative monitoring method for early warning of industrial equipment status of the present invention includes: S1. Obtain the monitoring data collected by the multi-sensor nodes corresponding to each target device in the monitoring area and the current forwarding binding relationship of the multi-sensor nodes in the wireless network. Perform association organization on the monitoring data and output the node association dataset and the current binding relationship set. This implementation method is used to uniformly classify, time-series organize, and bind relationship organize the raw acquisition results of multiple sensor nodes within a monitoring area, thereby forming the node association dataset and current binding relationship set required for subsequent synchronous anomaly identification. Its working mechanism is as follows: first, the acquisition records scattered across various target devices, sensor types, and forwarding parent nodes are organized into node classification results with a unified attribution criterion; then, within each classification, the continuous acquisition sequence is restored according to the acquisition time, while simultaneously identifying acquisition gaps and parent-child binding changes; finally, the data association results and binding change results are written into the corresponding result sets according to the target device identifier, ensuring that subsequent readings can trace both continuous changes in monitoring data by target device and changes in wireless network status by node binding changes under the same target device. This implementation process includes the following steps: The purpose of S1-1 is to first establish a unified attribution standard for the data collected by multiple sensor nodes, so that data collection records under the same target device, the same sensor type, and with the same current parent node identifier are organized into the same classification unit, facilitating subsequent sequential organization and statistical binding of data by category. During execution, the input is the raw data collection records reported by each multi-sensor node. These records must include at least the node identifier, target device identifier, sensor type identifier, data collection time, monitoring data, and current parent node identifier. The processing actions include performing a field integrity check on each raw data collection record. The rules for this check are derived from the communication protocol. The segment definition and data acquisition record storage format constraints are as follows: when the node identifier, target device identifier, sensor type identifier, acquisition time, and current parent node identifier all exist, the original acquisition record is determined as a valid record. When there are missing fields, the record corresponding to the missing field is written into the supplementary record table, and the cached record corresponding to the same node identifier at the previous acquisition time is called to perform the completion judgment. If there is a field with the same name at the previous acquisition time and the field is a static field, the field value at the previous acquisition time is written into the supplementary record. If the missing field belongs to the acquisition time or the current parent node identifier, the record is marked as an invalid record and the classification process is stopped. For valid records, a classification key value is generated according to the target device identifier, sensor type identifier, and current parent node identifier. Original acquisition records with the same classification key value are grouped into the same node classification unit. The output is the node classification result. Each classification unit in the node classification result contains at least the target device identifier, sensor type identifier, current parent node identifier, the set of node identifiers belonging to the unit, and the corresponding original acquisition record sequence, which are then read by S1-2. In terms of exception handling, when multiple original acquisition records with different current parent node identifiers appear at the same acquisition time for the same node identifier, the record with the later acquisition time is taken as the current record, and the remaining records are written to the conflict record table to avoid duplicate classification at the same time. The purpose of S1-2 is to restore the continuous acquisition sequence within each node classification unit, and to identify acquisition gaps and parent-child binding changes under the continuous acquisition sequence, thereby forming the node association fragments required for the subsequent establishment of the node association dataset and the binding change fragments required for the establishment of the current binding relationship set. In specific execution, the input is the node classification result output by S1-1. The processing actions include arranging the original acquisition records in each node classification unit in ascending order of acquisition time. The comparison caliber of acquisition time is based on the value reported by the local clock of the acquisition terminal, and the time correction is performed with the unified clock correction value of the gateway. The correction rule comes from the gateway clock synchronization rule. After sorting, the interval is calculated for the acquisition time difference between two adjacent acquisition records, and the interval value is compared with the standard acquisition cycle corresponding to the sensor type. The standard acquisition cycle comes from the node acquisition configuration table. When the adjacent acquisition time difference is equal to one standard acquisition cycle, the two acquisition records are determined to be continuous acquisition records. When the adjacent acquisition time difference is greater than one standard acquisition cycle, the missing acquisition cycle segment is determined to be a data missing segment, and the missing start time, missing end time, corresponding node identifier, and corresponding target device identifier are written into the node association fragment. Subsequently, a time-by-time comparison is performed on the current parent node identifier corresponding to the same node identifier at continuous acquisition times. When the current parent node identifiers corresponding to adjacent acquisition times are the same, it is determined that the parent-child correspondence of the node remains unchanged within the corresponding time period. When the current parent node identifiers corresponding to adjacent acquisition times are different, the change time is determined as the change time of the parent-child correspondence. The parent node identifier before the change, the parent node identifier after the change, the corresponding node identifier, the corresponding target device identifier, and the change time are written into the binding change segment. The output is the node association segment and the binding change segment. The node association segment includes at least the continuous acquisition segment and the data missing segment. The binding change segment includes at least the parent node switching record and is available for S1-3 to read. In terms of anomaly handling, when there is only one original acquisition record in a certain classification unit, the single record is written into the node association segment separately, and the classification unit is marked as a low continuity unit so that its direct grouping priority is reduced when identifying synchronization anomalies later. The purpose of S1-3 is to organize the node-related segments and binding change segments into a unified result set according to the target device identifier, forming the node-related dataset and the current binding relationship set that can be directly called in subsequent steps. In specific execution, the input is the node-related segments and binding change segments output by S1-2. The processing actions include first performing group writing on the node-related segments according to the target device identifier, and under the same target device identifier, writing the continuous acquisition segments, data missing segments, node identifiers and acquisition time ranges corresponding to each sensor type identifier into the same node-related dataset record. Then, performing group writing on the binding change segments according to the same target device identifier, and under the same target device identifier, writing the node identifier, change time, parent node identifier before change and parent node identifier after change into the same current binding relationship set record. To ensure consistent subsequent reading order, internal indexes are generated for the node association dataset and the current binding relationship set during the writing process. These internal indexes include the target device identifier, node identifier, and the start time of data collection, facilitating the rapid extraction of corresponding node records by S2 based on the target device identifier and collection time. The output consists of the node association dataset and the current binding relationship set. The node association dataset is used to identify abnormal node groups within the same abnormal time period, while the current binding relationship set is used to extract the shared forwarding binding relationships corresponding to the abnormal node groups. Regarding anomaly handling, when no binding change fragment exists under a certain target device identifier, a stable binding flag corresponding to that target device identifier is written into the current binding relationship set. When only binding change fragments exist under a certain target device identifier without valid node association fragments, the target device identifier is written into the table of devices to be re-collected, allowing for priority re-collection in the next data collection round. Through the above implementation process, the original multi-sensor node acquisition results can be transformed from scattered records into structured results with unified attribution, continuous time sequence and binding change information. This allows subsequent synchronous anomaly identification to have both continuous monitoring data as input and corresponding wireless networking binding status as reference, thereby avoiding the problems of attribution confusion, time sequence breakage and binding status loss caused by directly facing the original discrete records in subsequent steps. In practical applications: Taking multiple fan cabinets deployed in parallel in a computer room as an example, each fan cabinet corresponds to a vibration node, a temperature node, and a current node. Each node uploads the collected results through wireless networking. During implementation, the node identifier, fan cabinet identifier, sensor type identifier, collection time, and current parent node identifier reported by each node are read first. Records of the same sensor type under the same fan cabinet and with the same current parent node identifier are grouped into the same classification unit. Then, the continuous collection order of each classification unit is restored according to the collection time. The collection gaps caused by wireless packet loss and the changes in parent nodes caused by relay switching are identified. Finally, the continuous collection segments and missing data segments are written into the node association dataset of the corresponding fan cabinet, and the parent node switching record is written into the current binding relationship set of the corresponding fan cabinet. Subsequently, abnormal node groups within the same abnormal time period can be directly identified based on these two types of results, and their shared forwarding binding relationships can be extracted.

[0018] S2. Perform synchronous anomaly identification on the node association dataset, identify the abnormal node group that has abnormal fluctuations within the same abnormal time period, extract the shared forwarding binding relationship corresponding to the abnormal node group from the current binding relationship set, and output the set of abnormal segments to be judged. This implementation method, based on an existing node association dataset, further identifies the abnormal start time and duration of each multi-sensor node. Within the same target device range, it extracts abnormal node groups exhibiting common abnormal fluctuations within the same abnormal time period from multiple abnormal time periods. Then, it combines this with the current binding relationship set to extract the shared forwarding binding relationship corresponding to the abnormal node group, forming a set of abnormal segments to be judged for subsequent binding disturbance and abnormal migration determination. Its mechanism involves first comparing fluctuation changes in monitoring data collected continuously at a single node level to obtain node abnormal time periods with a locatable time range. Then, within the same target device range, it performs overlapping, inclusion, and connectivity merging processing on the abnormal time period intervals of each node, eliminating isolated nodes that cannot form synchronous abnormal relationships, and extracting abnormal node groups under common abnormal time periods. Finally, it maps the abnormal node groups to the parent-child binding relationship and continuous forwarding path in the current binding relationship, ensuring that subsequent steps read not discrete abnormal records, but structured abnormal segments that have already completed target device attribution, abnormal time period merging, and binding relationship extraction. This implementation process includes the following steps: The purpose of S2-1 is to first identify the abnormal start time and abnormal duration of each multi-sensor node from the node association dataset, providing a unified single-node abnormal time period input for subsequent abnormal time period overlap identification. Specifically, the input includes the target device identifier, sensor type identifier, node identifier, acquisition time, and monitoring data corresponding to each multi-sensor node in the node association dataset. The node association dataset is output by S1. The processing actions include extracting the monitoring data sequence of the same multi-sensor node at consecutive acquisition times according to the target device identifier and node identifier, and selecting the corresponding fluctuation comparison caliber based on the sensor type identifier. The comparison caliber is derived from the sensor type configuration table. For vibration monitoring data, the difference between adjacent acquisition times is compared with the average difference of multiple consecutive acquisition times; for temperature monitoring data, the difference between adjacent acquisition times is compared with the slope of multiple consecutive acquisition times; and for current monitoring data, the difference between adjacent acquisition times is compared with the average deviation of the same operating condition segment. The length of multiple consecutive acquisition times is derived from the analysis window length in the acquisition configuration table. If the acquisition configuration table is not configured separately, five consecutive acquisition times corresponding to the standard acquisition cycle of the sensor type are taken. Subsequently, a fluctuation status marker is generated for each acquisition moment. When two or more consecutive acquisition moments are marked as abnormal fluctuations, the first abnormal fluctuation acquisition moment is determined as the abnormal start moment, and reading continues in chronological order until two or more consecutive normal fluctuation acquisition moments appear. Then, the previous abnormal fluctuation acquisition moment is determined as the abnormal end moment, and the abnormal duration period is formed from the abnormal start moment to the abnormal end moment. The output is the node abnormal duration result. The node abnormal duration result must include at least the target device identifier, node identifier, sensor type identifier, abnormal start moment, and abnormal end moment, and is available for S2-2 and S2-21 to read. In terms of anomaly handling, when a multi-sensor node has a data missing segment within the analysis window, it is first determined whether the length of the missing segment is equal to one standard acquisition cycle. If it is equal to one standard acquisition cycle, the monitoring data of the two acquisition moments before and after the missing segment are linearly interpolated and then compared. If it is greater than one standard acquisition cycle, the missing segment is split into an abnormal duration period, and the node abnormal duration results before and after the missing segment are output separately. The purpose of S2-2 is to identify multiple multi-sensor nodes that synchronously exhibit abnormal fluctuations within the same abnormal time period from multiple single-node abnormal time periods, and to form abnormal node groups according to the target device identifier and common abnormal time period, providing grouped objects for subsequent extraction of shared forwarding binding relationships; in specific execution, the input is the node abnormal time period results output by S2-1; the processing actions include first grouping the node abnormal time period results according to the target device identifier, and then performing overlap identification on the abnormal time periods of each node within the same target device identifier range. The overlap identification does not directly use the intersection of arbitrary times as the grouping basis, but calls the processing results of S2-21 to S2-23 to complete the time period association, connectivity merging, and common abnormal time period solution; After obtaining the common abnormal period, the multi-sensor nodes within the common abnormal period are identified as the same abnormal node group, and a group identifier is generated for each abnormal node group. The output is the abnormal node group result, which includes at least the target device identifier, the abnormal node group identifier, the node identifier set, the start point of the common abnormal period, and the end point of the common abnormal period, and is available for S2-3 to read. In terms of anomaly handling, when only one multi-sensor node under the same target device identifier has an anomaly in a certain period, the result corresponding to that multi-sensor node is written to the single node anomaly record table, and no abnormal node group result is generated, so as to avoid directly outputting isolated node anomalies as synchronous anomalies. The purpose of S2-3 is to map the abnormal node group results to the current binding relationship set, extract the shared parent node and continuous forwarding path of each abnormal node group under the corresponding abnormal time period, and thus generate the set of abnormal segments to be judged required for subsequent binding disturbance steps. In specific execution, the input is the abnormal node group results output by S2-2 and the current binding relationship set output by S1. The processing actions include first extracting the parent-child binding correspondence of each multi-sensor node in the common abnormal time period from the current binding relationship set according to the target device identifier, node identifier set and common abnormal time period in the abnormal node group results, then reading the current parent node identifier corresponding to each multi-sensor node according to each acquisition time in the common abnormal time period, counting the parent node identifiers that are commonly corresponding to two or more multi-sensor nodes under the same acquisition time, and determining the parent node identifier as the shared parent node. Subsequently, the path record field of each multi-sensor node is traced from its own node identifier to the end point of the forwarding path from the path record field of the current binding relationship set. In this embodiment, the end point of the forwarding path is fixed as the gateway node identifier. The condition for continuous forwarding path identification is that the intermediate node sequence remains unchanged at adjacent acquisition times. When two or more multi-sensor nodes have overlapping intermediate node sequences at consecutive acquisition times during the common abnormal period, the corresponding intermediate node sequence is identified as a shared forwarding path. The output is a set of abnormal segments to be determined. The set of abnormal segments to be determined is written with at least the target device identifier, abnormal node group identifier, node identifier set, common abnormal period, shared parent node identifier, and shared forwarding path, and is available for S3 to read. In terms of abnormal handling, when an abnormal node group does not have a shared parent node but has a shared forwarding path during the common abnormal period, a set of abnormal segments to be determined is still generated and the shared parent node field is written as null. When neither the shared parent node nor the shared forwarding path exists, the abnormal node group is written to the non-shared abnormal segment table and does not enter the subsequent binding disturbance process. The purpose of S2-21 is to convert the single-node abnormal time period results into time period association results that can directly participate in time period connectivity merging, so that the generation of subsequent abnormal node groups is based on clear interval overlap and interval inclusion relationships. In specific execution, the input quantities are the target device identifier, abnormal start time, and abnormal end time corresponding to each multi-sensor node in the node abnormal time period results. The processing actions include constructing the abnormal time period intervals corresponding to each multi-sensor node through the abnormal start time and abnormal end time, and performing overlap duration calculation, boundary interval calculation, and interval inclusion relationship calculation on any two abnormal time period intervals under the same target device identifier. The overlap duration is obtained by subtracting the later time from the start time of the two abnormal time period intervals from the earlier time of the end time of the two abnormal time period intervals. If the result is greater than zero, it is recorded as the actual overlap duration; if the result is not greater than zero, it is recorded as zero. The boundary interval is obtained by subtracting the end time of the previous abnormal time interval from the start time of the subsequent abnormal time interval. If the result is zero, it means that the two intervals are connected end to end. The interval inclusion relationship is determined by comparing whether the start time and end time of one abnormal time interval are located inside another abnormal time interval. When the overlap duration is greater than zero and the boundary interval is zero, the corresponding two abnormal time intervals are determined as adjacent time intervals. When there is an interval inclusion relationship, the corresponding two abnormal time intervals are determined as time inclusion intervals, and the above relationship is written into the time association result. The output is the time association result, which at least includes the target device identifier, the previous node identifier, the next node identifier, the overlap duration, the boundary interval, and the association relationship type, and is available for S2-22 to read. In terms of anomaly handling, when multiple abnormal time intervals of the same multi-sensor node overlap, the abnormal time intervals within the multi-sensor node are first merged, and the merged intervals then participate in the generation of the time association result. The purpose of S2-22 is to construct an abnormal node time period graph based on the time period association results, and to group abnormal nodes with interrelationships under the same target device into the same abnormal node subgraph by identifying connected subgraphs and merging conflicting nodes. In specific execution, the input is the time period association results output by S2-21. The processing actions include performing intra-group aggregation according to the target device identifier. Under each target device identifier, multi-sensor nodes with time period association results are identified as graph nodes, and the time period association results between any two graph nodes are identified as graph edges, thereby constructing the abnormal node time period graph. Subsequently, connected subgraph identification is performed on the graphs of each abnormal node's time period. Connected subgraph identification adopts a graph traversal method, starting from any unvisited graph node and expanding along the graph edges one by one until no further expansion is possible to form a connected subgraph. Connected subgraphs containing only a single graph node are directly eliminated and not included in the abnormal node subgraph results. For conflicting nodes that appear in multiple connected subgraphs, the total overlap time between the conflicting node and the other graph nodes in each connected subgraph is calculated, and the conflicting node is assigned to the connected subgraph with the higher total overlap time value. When the total overlap time values ​​are the same, the connected subgraph with more graph nodes is given priority. If the number of graph nodes is still the same, the connected subgraph with the earlier start time of their common abnormal time period is assigned. The output is the abnormal node subgraph result, which includes at least the target device identifier, connected subgraph identifier, and graph node set, and is available for S2-23 to read. In terms of anomaly handling, when the merging of conflicting nodes causes the original connected subgraph to lose all graph nodes, the original connected subgraph is deleted and no longer participates in the common abnormal time period calculation. The purpose of S2-23 is to extract the common abnormal time period that can simultaneously cover all multi-sensor nodes within the connected subgraph from the abnormal node subgraph results, and to transform the connected subgraph into the final, directly callable abnormal node group results. Specifically, the input is the abnormal node subgraph results output by S2-22, along with the abnormal start and end times of the corresponding multi-sensor nodes. The processing actions include calculating the common abnormal time period for each connected subgraph, determining the later time among the abnormal start times corresponding to each multi-sensor node within the connected subgraph as the common start time, determining the earlier time among the abnormal end times corresponding to each multi-sensor node as the common end time, and then performing a sequential judgment on the common start and end times. If the common start time is earlier than the common end time, it is determined that all multi-sensor nodes within the connected subgraph are simultaneously in an abnormal fluctuation state between the common start and end times, and the multi-sensor nodes in the connected subgraph are identified as an abnormal node group within the same abnormal time period. Subsequently, the target device identifier, common start time, common end time, node identifier set, and abnormal node group identifier are written into the abnormal node group result; the output is the abnormal node group result, which is called by S2-2 and S2-3; in terms of exception handling, when the common start time is not earlier than the common end time, it means that there is no abnormal period covered by all nodes in the connected subgraph, so the connected subgraph is split back into the single node exception record corresponding to each graph node, and no abnormal node group result is generated; Through the above implementation process, the anomaly start time and duration can be stably identified at the single node level. Then, within the same target device, through time interval association, anomaly node time interval graph construction, connected subgraph identification, and common anomaly time interval calculation, anomaly node group results with clear common anomaly time intervals and clear node sets are generated. Finally, the shared parent node and shared forwarding path are extracted by combining the current binding relationship, thereby unifying the input required for subsequent binding disturbance and anomaly migration judgment into a set of anomaly fragments to be judged. After this processing, the subsequent steps do not read the scattered and difficult-to-compare single-node anomaly records, but rather structured anomaly objects that have completed time interval merging, node grouping, and binding relationship extraction. In practical applications: Taking the monitoring scenario of chemical pump sets as an example, vibration nodes, temperature nodes, and current nodes are deployed on the same target pump body. The gateway periodically reads the monitoring data of each node in the node association dataset. For vibration nodes, it calculates the amplitude difference between adjacent acquisition times and the mean difference of continuous acquisition windows. For temperature nodes, it calculates the temperature rise difference between adjacent acquisition times and the slope of continuous acquisition windows. For current nodes, it calculates the current difference between adjacent acquisition times and the deviation of the operating condition segment. This allows the gateway to determine the abnormal start time and abnormal end time of each node. Subsequently, an abnormal time period interval is constructed within the same pump body. The time period association relationship between vibration nodes and current nodes with an overlapping duration greater than zero is identified. Then, an abnormal node time period diagram is constructed and the common abnormal time period is solved. The vibration node and current node are identified as the same abnormal node group. Finally, the parent node identifier and forwarding path of the two nodes in the common abnormal time period are read from the current binding relationship set. If the two nodes share the same relay node and have the same intermediate node sequence in the path to the gateway, a set of abnormal segments to be judged, including the abnormal node group, the common abnormal time period, the shared parent node identifier, and the shared forwarding path, is generated for subsequent steps to perform controlled disgrouping or controlled merging operations.

[0019] S3. For the set of abnormal segments to be determined, keep the installation position and collection object of the abnormal node group unchanged, perform controlled disgrouping or controlled merging operations on the shared forwarding binding relationship to form a disturbance forwarding binding relationship, and reacquire the monitoring data corresponding to the abnormal node group based on the disturbance forwarding binding relationship, and output the disturbance observation dataset. This implementation method is used to perform controlled binding perturbation on the abnormal node groups already identified in the set of abnormal segments to be judged. Under the condition that the installation location and data acquisition object remain unchanged, only the forwarding binding relationship in the wireless network of the same abnormal node group is changed, thereby constructing perturbation observation conditions that can be used for subsequent anomaly migration judgment. Its working mechanism is as follows: First, the parent node affiliation and path overlap of each multi-sensor node are statistically analyzed from the shared forwarding binding relationship to form a binding aggregation result that supports the selection of perturbation mode. Then, based on the binding aggregation result, controlled disgrouping or controlled merging operations are performed on the sets of nodes with the same parent and the sets of nodes with the same path to generate a perturbation forwarding binding relationship. Subsequently, the corresponding monitoring data is re-acquired under the perturbation forwarding binding relationship, and the binding changes and monitoring data changes before and after the perturbation are uniformly written into the binding perturbation associated segment to ensure that subsequent steps can directly compare whether the abnormal fluctuations migrate, reorganize, or disappear with the change in the forwarding binding relationship. The implementation process includes the following steps: The purpose of S3-1 is to first extract the shared binding structure and shared path structure within the abnormal node group from the set of abnormal segments to be determined, so as to provide clear input for subsequent controllable disgrouping or controllable merging operations. In specific execution, the input quantities are the target device identifier, abnormal node group identifier, node identifier set, abnormal time period, shared parent node identifier, and shared forwarding path corresponding to each abnormal node group in the set of abnormal segments to be determined. The set of abnormal segments to be determined is output by S2. The processing actions include first extracting the parent node record of each multi-sensor node in the corresponding node identifier set according to the abnormal node group identifier, performing parent node attribution statistics on each multi-sensor node in the same abnormal node group, and grouping multi-sensor nodes with the same parent node identifier into the same set of parent nodes. Subsequently, the complete forwarding path sequence corresponding to each multi-sensor node during the abnormal period is extracted. The forwarding path sequence consists of node identifiers, intermediate node identifiers at each level, and the forwarding path endpoint. In this embodiment, the forwarding path endpoint is fixed as the gateway node identifier. Then, path overlap statistics are performed on the forwarding path sequences of any two multi-sensor nodes. The rule for path overlap statistics is to compare whether there are consecutive identical node segments in the intermediate node identifiers of the two forwarding path sequences. When the length of the consecutive identical node segment is not less than two intermediate nodes, the corresponding two multi-sensor nodes are determined as path overlap nodes and are included in the same path set. The output is binding. The aggregation results are bound to at least the target device identifier, abnormal node group identifier, same parent node set, same path set, and corresponding abnormal time period, and are available for S3-2 to read. In terms of anomaly handling, when a multi-sensor node has multiple parent node records within an abnormal time period, its current parent node identifier is determined by the parent node identifier that appears more frequently within the abnormal time period. When the number of occurrences is the same, its current parent node identifier is determined by the parent node identifier that is closer to the end time of the abnormal time period. When the forwarding path record of a multi-sensor node is incomplete, the multi-sensor node is written into the path missing node table and is not included in the same path set statistics. The purpose of S3-2 is to select the perturbation method corresponding to the current shared binding structure based on the binding aggregation result, and to generate a perturbation forwarding binding relationship without changing the installation location of the abnormal node group and the acquisition object. Specifically, the input is the binding aggregation result output by S3-1. The processing actions include first performing perturbation condition judgments on the same parent node set and the same path set respectively. When the number of multi-sensor nodes in the same parent node set is not less than two and all nodes in the same parent node set have available parent nodes outside the group, a controlled disgrouping operation is performed on the same parent node set. The source of the available parent nodes outside the group is the current binding relationship set with the target device. The parent node identifier corresponding to the backup identifier and not occupied by the same parent node set during the abnormal period is selected. If there are multiple available parent nodes outside the group, the switching target is determined by the parent node identifier with the smaller difference between the path hop count to the gateway node and the path hop count of the original parent node. If the difference in path hop count is the same, the switching target is determined by the parent node identifier with the fewer nodes attached to the parent node during the abnormal period. When performing controlled disgrouping operation, the multi-sensor node with the last node identifier in the same parent node set is switched to the determined available parent node identifier outside the group. The node identifier sorting rule is derived from the node identifier dictionary order to ensure that the disturbance selection caliber is fixed. Subsequently, a disturbance condition judgment is performed on the same path set. When the multi-sensor nodes in the same path set belong to different parent node identifiers and their corresponding forwarding path endpoints are the same, a controlled merging operation is performed on the same path set. The execution rule of the controlled merging operation is as follows: first, count the number of nodes corresponding to each parent node identifier in the same path set, determine the parent node identifier with more nodes as the retained parent node identifier, and switch the multi-sensor nodes under the remaining parent node identifiers to the retained parent node identifier. When the number of nodes is the same, determine the parent node identifier with fewer hops to the gateway node as the retained parent node identifier. When the same parent node set meets the controlled disgrouping operation conditions, the controlled disgrouping operation is performed first. When the same parent node set does not meet the controlled disgrouping operation conditions, the controlled disgrouping operation is performed first. When the conditions for splitting and grouping operations are met, and the set of nodes on the same path also meets the conditions for controlled grouping operations, the controlled grouping operation is executed. When neither condition is met, the original shared forwarding binding relationship is maintained, and the abnormal node group is written into the undisturbed segment table. The output is the disturbed forwarding binding relationship. The disturbed forwarding binding relationship must include at least the target device identifier, the abnormal node group identifier, the disturbance type, the parent node identifier before the disturbance, the parent node identifier after the disturbance, and the forwarding path after the disturbance, and is available for S3-3 to read. In terms of exception handling, when there is no available parent node outside the group, the controlled splitting operation is not executed. When any multi-sensor node cannot form a complete forwarding path to the gateway node after grouping, the controlled grouping operation is canceled and the original shared forwarding binding relationship is maintained. The purpose of S3-3 is to reacquire disturbance monitoring data of the same abnormal node group after the disturbance forwarding binding relationship has been determined, while keeping the collection objects of the abnormal node group unchanged, so as to provide post-disturbance observation results for subsequent abnormal migration judgment. In specific execution, the input is the disturbance forwarding binding relationship output by S3-2 and the abnormal time period and node identifier set of the corresponding abnormal node group in the set of abnormal segments to be judged. The processing actions include first writing the disturbance forwarding binding relationship into the network control table corresponding to each multi-sensor node, and issuing a binding update command to each multi-sensor node. After each multi-sensor node returns the binding update confirmation result, it is determined that the disturbance forwarding binding relationship has been written. Then, the subsequent collection period is determined. The starting point of the subsequent collection period is taken as the starting point of the first complete collection cycle after the binding update confirmation is completed. The length of the subsequent collection period is taken as the length of the original abnormal period. The length of the original abnormal period is obtained by subtracting the abnormal start time from the abnormal end time. During the subsequent acquisition period, synchronous acquisition is performed on each multi-sensor node corresponding to the abnormal node group. The scope of synchronous acquisition is that the acquisition time of each multi-sensor node falls within the same acquisition window. The acquisition window length is taken as one standard acquisition cycle corresponding to the sensor type with the longer standard acquisition cycle in the abnormal node group. The standard acquisition cycle is derived from the acquisition configuration table. For the monitoring data returned in each acquisition window, the target device identifier, node identifier, sensor type identifier, acquisition time, monitoring data, and current parent node identifier are recorded. The output is a disturbance observation dataset. The disturbance observation dataset must at least include the target device identifier, abnormal node group identifier, node identifier, acquisition time, monitoring data, and the parent node identifier after disturbance, and is available for S3-4 to read. In terms of anomaly handling, when a multi-sensor node does not return monitoring data in the first complete acquisition cycle after binding update, an acquisition command is resent to the multi-sensor node. If no monitoring data is returned after resentment, the multi-sensor node is marked as a disturbance acquisition failure node and written into the disturbance failure node table, but the disturbance observation data corresponding to the other multi-sensor nodes are retained. The purpose of S3-4 is to organize the binding change results, path change results, and monitoring data change results before and after the disturbance into the same associated segment, so that subsequent steps can directly read the structured binding disturbance information to perform corresponding comparisons. Specifically, the inputs are the disturbance observation dataset output by S3-3, the disturbance forwarding binding relationship output by S3-2, and the shared forwarding binding relationship and abnormal time period corresponding to the set of abnormal segments to be judged. The processing actions include first establishing the correspondence before and after the disturbance according to the target device identifier, abnormal node group identifier, and node identifier, then extracting the parent node identifier before the disturbance and the parent node identifier after the disturbance for each multi-sensor node to generate the parent node identifier change result; then extracting the forwarding path before the disturbance and the forwarding path after the disturbance, comparing the node identifiers at each level in the path step by step to generate the forwarding path change result. Next, the monitoring data before and after the disturbance are time-aligned according to the corresponding collection order of the abnormal period and the subsequent collection period. The time alignment rule is to take the first collection record in the original abnormal period and correspond it to the first collection record in the subsequent collection period, and then match them one by one. The difference calculation is performed on the monitoring data at each corresponding position to generate the monitoring data change result. Finally, the parent node identifier change result, forwarding path change result, and monitoring data change result are written according to the target device identifier, node identifier, and collection time to form a bound disturbance association fragment. The output is the bound disturbance association fragment in the disturbance observation dataset. The bound disturbance association fragment must at least include the target device identifier, abnormal node group identifier, node identifier, parent node identifier before the disturbance, parent node identifier after the disturbance, forwarding path before the disturbance, forwarding path after the disturbance, monitoring data before the disturbance, monitoring data after the disturbance, and monitoring data change result, and is available for S4 to read. In terms of anomaly handling, when the number of collection records before and after the disturbance is inconsistent, the corresponding writing is performed according to the number of collection records with less data. The remaining collection records that cannot be correspondingly completed are written separately into the uncorresponding collection record table and do not participate in the generation of the bound disturbance association fragment in this round. Through the above implementation process, the shared parent node structure and shared path structure of the abnormal node group can be extracted from the set of abnormal segments to be judged. Then, controlled disgrouping or controlled merging operations are performed according to fixed disturbance selection rules to form a disturbance forwarding binding relationship. Under the disturbance forwarding binding relationship, the corresponding monitoring data is reacquired. Finally, the binding changes, path changes and monitoring data changes before and after the disturbance are organized into a unified binding disturbance association segment. After this processing, the subsequent abnormal migration judgment steps no longer need to trace the network status and monitoring results before and after the disturbance. Instead, they can directly judge whether the abnormal fluctuation has migrated, reorganized or disappeared with the change of the forwarding binding relationship based on the binding disturbance association segment of the same abnormal node group. In practical applications: Taking two parallel circulating pumps as an example, the vibration node and current node corresponding to one of the circulating pumps form an abnormal node group during the same abnormal period. The two were originally connected to the same relay node and uploaded to the gateway along the same path. During implementation, the parent node affiliation and path overlap of the abnormal node group during the abnormal period are first counted. The vibration node and current node are assigned to the same parent node set. Then, another parent node identifier that is not occupied by the same parent node set and can reach the gateway is selected from the current binding relationship set. The current node with the last node identifier in the node identifier sorting is subjected to a controlled disgrouping operation to switch it to the other parent node identifier. Subsequently, starting from the beginning of the first complete acquisition cycle after the binding update is completed, synchronous acquisition is performed according to the original abnormal period length to obtain the monitoring data of the vibration node and current node under the disturbance forwarding binding relationship. Finally, the parent node identifier, forwarding path and corresponding monitoring data before and after the disturbance are written into the binding disturbance association fragment for subsequent steps to determine whether the abnormal node group has experienced abnormal migration, abnormal reorganization or abnormal disappearance due to binding disturbance.

[0020] S4. Perform a correspondence comparison between the set of abnormal segments to be judged and the disturbance observation dataset to determine whether the abnormal fluctuations have migrated, reorganized or disappeared due to changes in the forwarding binding relationship, and output the abnormal migration judgment result. This implementation method is used to re-identify abnormal node groups in the monitoring data after the controlled binding disturbance has been completed and a disturbance observation dataset has been formed. It compares the composition of the abnormal node groups before and after the disturbance, the common abnormal time period, and the target device affiliation to determine whether the abnormal fluctuation has migrated, reorganized, or disappeared due to changes in the forwarding binding relationship, thus forming the abnormal migration determination result required for subsequent attribution determination. Its working mechanism is as follows: First, within the target device range corresponding to the original abnormal node group and the subsequent collection time period, the disturbance monitoring data is re-identified using the aforementioned abnormal identification criteria to obtain the disturbance abnormal node group results. Then, the composition of the abnormal node group before and after the disturbance is matched item by item, and a joint judgment is made on changes in the node set, changes in the common abnormal time period, and changes in the target device identifier. Abnormal migration, abnormal reorganization, or abnormal disappearance is determined according to a fixed morphological judgment order. Finally, the abnormal morphological category and the changes in the binding relationship before and after the disturbance are uniformly written into the abnormal migration determination result for subsequent steps to determine the attribution of real anomalies and wireless networking-related anomalies. This implementation process includes the following steps: The purpose of S4-1 is to re-perform anomaly identification on disturbance monitoring data after the disturbance forwarding binding relationship has been written and subsequent data acquisition has been completed, so that the post-disturbance anomaly state is comparable to the pre-disturbance anomaly state in terms of caliber, object, and time period. Specifically, the input includes the target device identifier, anomaly group identifier, common anomaly time period, anomaly group composition, and shared forwarding binding relationship for each anomaly node group in the set of anomaly segments to be judged, as well as the disturbance forwarding binding relationship and disturbance monitoring data for the corresponding anomaly node group in the disturbance observation dataset. The processing actions include first extracting the corresponding disturbance monitoring data according to the target device identifier and anomaly group identifier, and then... The original common abnormal period length is used as the disturbance identification period length. The disturbance monitoring data is sorted according to the node identifier and the collection time, and the anomaly re-identification is performed according to the sensor type identification caliber used in S2-1. Among them, vibration monitoring data is still compared with the difference between adjacent collection times and the average difference of multiple consecutive collection times. Temperature monitoring data is still compared with the difference between adjacent collection times and the slope of multiple consecutive collection times. Current monitoring data is still compared with the difference between adjacent collection times and the deviation of the average value of the same working condition segment. The analysis window length of multiple consecutive collection times is still taken from the corresponding configuration value in the collection configuration table to ensure that the basis for anomaly identification before and after the disturbance is consistent. After each multi-sensor node re-obtains the anomaly start and end times, the disturbed anomaly nodes are regrouped according to the same time period association, anomaly node time period graph construction, connected subgraph identification, and common anomaly time period resolution process as described in S2-21 to S2-23, generating a disturbed anomaly node group result. The output is the disturbed anomaly node group result, which includes at least the target device identifier, anomaly node group identifier, set of node identifiers after disturbance, start time of common anomaly time period after disturbance, end time of common anomaly time period after disturbance, and disturbance forwarding binding relation. The system is used for reading by S4-2; in terms of anomaly handling, when a multi-sensor node has missing records during the acquisition period after the disturbance, the length of the missing segment is first determined according to the missing handling rules in S2-1. When the length of the missing segment is equal to one standard acquisition period, linear interpolation is performed before the anomaly is identified. When the length of the missing segment is greater than one standard acquisition period, the missing segment is used as the time period breakpoint for splitting and identifying the results. When no anomaly is identified in all multi-sensor nodes corresponding to a certain anomaly node group, the anomaly node group is marked as an empty group in the disturbance anomaly node group results. The purpose of S4-2 is to classify changes in abnormal node groups before and after disturbances into one of the following categories: abnormal migration, abnormal reorganization, or abnormal disappearance, by using a fixed correspondence comparison caliber and a fixed morphological judgment order, thus avoiding the same change falling into multiple morphological categories simultaneously. In specific execution, the input quantities are the abnormal node group composition, common abnormal time period, and target device identifier corresponding to each abnormal node group in the set of abnormal segments to be judged, as well as the disturbance abnormal node group results output by S4-1. The processing actions include first establishing a one-to-one correspondence between abnormal node groups before and after disturbances. The rule for establishing the correspondence is that within the same target device identifier range, the disturbance abnormal node group with a larger number of nodes intersecting with the original abnormal node group is selected as the corresponding group. When the number of nodes intersecting is the same, the disturbance abnormal node group with a longer overlap between the common abnormal time period after disturbance and the original common abnormal time period is selected as the corresponding group. After the mapping is completed, the differences in abnormal node identifiers before and after the disturbance, the differences in common abnormal time periods, and the differences in target device identifiers are calculated. The differences in abnormal node identifiers are obtained by comparing the node identifier set of the original abnormal node group with the node identifier set after the disturbance to obtain the exit node set and the entry node set. The differences in common abnormal time periods are obtained by comparing the start and end points of the original common abnormal time period with the start and end points of the common abnormal time period after the disturbance to obtain the change in the start point and the change in the end point. The differences in target device identifiers are obtained by comparing whether the target device identifiers corresponding to the abnormal node groups before and after the disturbance are consistent to obtain the change in device affiliation. Then, a morphological determination is performed. The morphological determination order is fixed as follows: first determine if the abnormality has disappeared, then determine if the abnormality has migrated, and then determine if the abnormality has recombined. When no disturbance abnormal node group corresponding to the original abnormal node group is identified after the disturbance, or when the result of the disturbance abnormal node group is an empty group mark, it is determined that the abnormality has disappeared. An abnormal migration is defined as follows: when a corresponding abnormal node group exists, and at least one node in the original abnormal node group exits while at least one node outside the original abnormal node group enters, and the target device identifier corresponding to the entering node is inconsistent with the target device identifier corresponding to the exiting node; when a corresponding abnormal node group exists, and the node connection relationship between the original abnormal node group and the abnormal node group splits or merges, an abnormal reorganization is defined as follows: the source of the node connection relationship is fixed as the connectivity relationship within the abnormal node time period graph constructed in S2-22 and S4-1. Splitting refers to the original one connected subgraph corresponding to two or more connected subgraphs after the disturbance, and merging refers to the original two or more connected subgraphs corresponding to one connected subgraph after the disturbance; when the same When an abnormal node group meets both migration and reorganization conditions, it is prioritized as an abnormal migration. When it only meets the reorganization condition, it is determined as an abnormal reorganization. The output is the abnormal morphology determination result. The abnormal morphology determination result should at least include the target device identifier, abnormal node group identifier, exit node set, entry node set, common abnormal time period difference, target device identifier corresponding difference, and abnormal morphology category, and be available for S4-3 to read. In terms of abnormal handling, when there are abnormal node groups before and after the disturbance but a unique correspondence cannot be established, all candidate disturbance abnormal node groups are written into the disambiguation group table, and are re-compared in the order of node intersection number, time period overlap duration, and target device identifier consistency until a unique corresponding group is obtained before morphology determination is performed. The purpose of S4-3 is to write the corresponding changes in the binding relationship before and after the disturbance into a unified result, so that the subsequent attribution determination step can directly read the anomaly migration judgment result that has completed the morphology classification and has binding change information. In specific execution, the input is the anomaly morphology judgment result output by S4-2, as well as the shared forwarding binding relationship corresponding to the set of anomaly segments to be judged and the disturbance forwarding binding relationship corresponding to the disturbance observation dataset. The processing actions include first establishing the correspondence between the anomaly morphology judgment result and the binding relationship record according to the target device identifier and the anomaly node group identifier, then extracting the shared parent node identifier and shared forwarding path in the shared forwarding binding relationship, and the disturbed parent node identifier and disturbed forwarding path in the disturbance forwarding binding relationship, and generating the binding relationship change result. The binding relationship change result includes at least the parent node change mark, the path change mark and the disturbance type mark. The parent node change mark is obtained by comparing whether the parent node identifier before and after the disturbance is consistent. The path change mark is obtained by comparing whether the intermediate node sequence in the forwarding path before and after the disturbance is consistent. The disturbance type mark is directly taken from the controlled splitting or controlled merging operation result already written in S3-2. Subsequently, the abnormal form category, exit node set, entry node set, common abnormal time period difference, target device identifier corresponding difference, and binding relationship change result are written according to the target device identifier and abnormal node group identifier to generate an abnormal migration judgment result; the output is the abnormal migration judgment result, which at least writes the target device identifier, abnormal node group identifier, abnormal form category, shared parent node identifier, shared forwarding path, disturbed parent node identifier, disturbed forwarding path, parent node change flag, path change flag, and disturbance type flag, and is available for S5 to read; in terms of abnormal handling, when an abnormal form judgment result cannot match a complete disturbance forwarding binding relationship, it is written to the binding information missing table, and the abnormal form category field is retained so that subsequent steps can perform separate processing for this type of result as a binding information missing anomaly; Through the above implementation process, while maintaining consistency in anomaly identification criteria, abnormal node groups can be re-identified from the monitoring data after disturbance. Changes in node sets before and after disturbance, changes in common abnormal time periods, changes in target device affiliation, and changes in binding relationships can all be incorporated into the same judgment chain. This allows the post-disturbance manifestations of abnormal fluctuations to be clearly attributed to abnormal migration, abnormal reorganization, or abnormal disappearance, providing a direct basis for distinguishing between real anomalies of target devices and anomalies accompanying wireless networking. In practical applications: Taking two circulating pumps running in close parallel as an example, the original abnormal node group consists of the vibration node and the current node on the first circulating pump. The two are synchronously abnormal during the same common abnormal period and share the same relay node and part of the forwarding path. After the controlled disgrouping operation is performed in S3, the current node is switched to another parent node identifier and the monitoring data is re-collected. Then, in S4, the abnormality of the disturbance monitoring data is re-identified. If the original current node on the first circulating pump is no longer abnormal after the disturbance, but the temperature node on the second circulating pump enters the abnormal node group, the correspondence before and after the disturbance is first established based on the number of node intersections and the duration of time overlap. Then, the exit node set and the entry node set are compared. Combined with the result that the target device identifier corresponding to the entry node is inconsistent with the target device identifier corresponding to the exit node, the change is determined as an abnormal migration. Then, the abnormal migration result, the change result of the parent node identifier before and after the disturbance, and the change result of the forwarding path are written into the abnormal migration judgment result for subsequent steps to further determine whether the abnormality is an accompanying abnormality of the wireless network or a real abnormality of the target device.

[0021] S5. Perform attribution determination on the abnormal migration judgment result. When the abnormal fluctuation maintains the same spatial attribution of the corresponding target device before and after the change of the forwarding binding relationship, it is determined to be a real abnormality of the target device. When the abnormal fluctuation migrates, reassembles or disappears with the change of the forwarding binding relationship, it is determined to be an abnormality accompanying the wireless networking, and the target device status warning information is output according to the attribution determination result. This implementation method is used to determine the affiliation of each abnormal node group based on the anomaly migration judgment results, and to generate target device status warning information and network interference warning information respectively. Its working mechanism is as follows: First, based on the anomaly type, target device identifier distribution, changes in node composition, and changes in binding relationships, it judges whether the physical affiliation of the abnormal node group remains consistent before and after the disturbance. Abnormal node groups that still maintain the same target device spatial affiliation and have not migrated, reorganized, or disappeared are identified as having genuine target device anomalies. Then, for the remaining abnormal node groups, a network-related judgment is performed based on the anomaly type and changes in binding relationships. Abnormal node groups whose anomalies are affected by binding disturbances are identified as having accompanying wireless network anomalies. Finally, the two types of results are written into the corresponding warning results according to the target device identifier. This implementation process includes the following steps: The purpose of S5-1 is to first screen out the abnormal node groups with stable physical affiliation from the abnormal migration judgment results and identify them as the real abnormalities of the target device. In specific execution, the input quantities are the target device identifier, abnormal morphology category, exit node set, entry node set, shared forwarding binding relationship, disturbance forwarding binding relationship, and the difference in the target device identifier corresponding to each abnormal node group in the abnormal migration judgment results. The processing actions include first extracting the target device identifiers corresponding to each multi-sensor node before and after the disturbance according to the abnormal node group identifier, statistically analyzing the target device identifier distribution of the node set before the disturbance and the target device identifier distribution of the node set after the disturbance, and then reading the exit node set and the entry node set to calculate the node composition change results. Then, a true anomaly determination is performed. The conditions for true anomaly determination are: the distribution of target device identifiers before and after the disturbance is consistent; the exit node set is empty; the entry node set is empty; the corresponding differences of target device identifiers are consistent; and the anomaly type does not include migration, reorganization, and disappearance. When the above conditions are met simultaneously, the corresponding abnormal node group is determined as a true anomaly of the target device, and the target device identifier, abnormal node group identifier, anomaly type, shared forwarding binding relationship, and disturbance forwarding binding relationship are written into the true anomaly result. The output is the true anomaly result, which is read by S5-3. In terms of anomaly handling, when the distribution of target device identifiers before and after the disturbance is consistent but the exit node set or the entry node set is not empty, it is not determined as a true anomaly of the target device, and the process is transferred to S5-2 for further determination. The purpose of S5-2 is to perform network-related judgment on abnormal node groups that are not identified as true anomalies of the target device, and to identify abnormal node groups affected by binding disturbances as wireless network-related anomalies. Specifically, the input quantities are the anomaly type, shared forwarding binding relationship, disturbed forwarding binding relationship, parent node change flag, path change flag, exiting node set, entering node set, and target device identifier corresponding to each abnormal node group that did not enter the true anomaly result in the anomaly migration judgment results. The processing actions include first reading the parent node change results and path change results between the shared forwarding binding relationship and the disturbed forwarding binding relationship to form a binding relationship change result, and then performing network-related judgment based on the anomaly type and binding relationship change result. The conditions for network-related judgment are that the anomaly type includes migration, reorganization, or disappearance, and the parent node change flag or path change flag is changed. When the above conditions are met, the corresponding abnormal node group is identified as a wireless networking accompanying anomaly, and the target device identifier, abnormal node group identifier, abnormal mode category, binding relationship change result, exit node set, and entry node set are written into the accompanying anomaly result. When the abnormal mode category does not include migration, reorganization, and disappearance, or when the abnormal mode category includes migration, reorganization, or disappearance but the parent node change marker and path change marker are unchanged, the corresponding abnormal node group is written into the anomaly table to be reviewed for subsequent manual review or further judgment in the next round of disturbance collection. The output is the accompanying anomaly result and is read by S5-3. In terms of anomaly handling, when the shared forwarding binding relationship or the disturbance forwarding binding relationship is missing, the abnormal node group is directly written into the anomaly table to be reviewed, and the wireless networking accompanying anomaly determination is not performed. The purpose of S5-3 is to organize the actual anomaly results and accompanying anomaly results into directly output early warning information according to the target device identifier. Specifically, the inputs are the actual anomaly results output by S5-1 and the accompanying anomaly results output by S5-2. The processing actions include first writing the attribution of the actual anomaly results according to the target device identifier to generate target device status early warning information, which at least includes the target device identifier, anomaly node group identifier, actual anomaly marker, anomaly type, and corresponding node identifier set; then writing the attribution of the accompanying anomaly results according to the target device identifier to generate network interference early warning information, which at least includes the target device identifier, anomaly node group identifier, accompanying anomaly marker, anomaly type, parent node change marker, and path change marker. Subsequently, target device status warning information and network interference warning information under the same target device identifier are aggregated and output. When only target device status warning information exists for the same target device, the corresponding target device status warning information is output. When only network interference warning information exists for the same target device, the corresponding network interference warning information is output. When both types of warning information exist for the same target device, both types of warning information are output separately and a concurrent warning flag is written into the output record. The output quantity is the corresponding target device status warning information. In terms of anomaly handling, when neither the actual anomaly result nor the accompanying anomaly result exists, the corresponding target device identifier is written into the no-warning-result table. Through the above implementation process, the abnormal migration judgment results can be further divided into two categories: the actual abnormality of the target device and the abnormality accompanying the wireless networking. Corresponding warning information is generated according to the target device identifier, thereby avoiding the direct misjudgment of abnormal morphological changes caused by binding disturbances as abnormalities of the device itself. In practical applications: Taking two wind turbines deployed in parallel as an example, if the vibration nodes and current nodes of the first wind turbine remain unchanged before and after the disturbance, the node structure remains unchanged, and the abnormal morphology category does not include migration, reorganization, or disappearance, then the abnormal node group is identified as a real abnormality of the first wind turbine, and a target equipment status warning information for the first wind turbine is generated; if the abnormal node group corresponding to the second wind turbine experiences node exit and other target equipment nodes entering after the disturbance, and both the parent node change marker and the path change marker are changed, and the abnormal morphology category is migration, then the abnormal node group is identified as a wireless networking accompanying abnormality, and a networking interference warning information for the second wind turbine is generated.

[0022] Furthermore, the present invention also includes an industrial equipment status early warning system for multi-sensor collaborative monitoring, the system comprising a data association module, an anomaly grouping module, a disturbance binding module, a judgment module, and an early warning module: The data association module is used to obtain the monitoring data collected by the multi-sensor nodes corresponding to each target device in the monitoring area and the current forwarding binding relationship of the multi-sensor nodes in the wireless network. It performs association organization on the monitoring data and outputs the node association dataset and the current binding relationship set. The anomaly grouping module is used to perform synchronous anomaly identification on the node-associated dataset, identify the abnormal node groups that exhibit abnormal fluctuations within the same abnormal time period, extract the shared forwarding binding relationship corresponding to the abnormal node group from the current binding relationship set, and output the set of anomaly fragments to be judged. The binding disturbance module is used to maintain the installation position and collection object of the abnormal node group for the set of abnormal segments to be judged, perform controlled disgrouping or controlled merging operations on the shared forwarding binding relationship, form a disturbance forwarding binding relationship, and re-acquire the monitoring data corresponding to the abnormal node group based on the disturbance forwarding binding relationship, and output the disturbance observation dataset. The judgment module is used to perform a correspondence comparison between the set of abnormal segments to be judged and the disturbance observation dataset, to determine whether the abnormal fluctuations have migrated, reorganized or disappeared due to changes in the forwarding binding relationship, and output the abnormal migration judgment result. The early warning module is used to determine the attribution of abnormal migration judgment results. When the abnormal fluctuation maintains the same spatial attribution of the corresponding target device before and after the change of forwarding binding relationship, it is determined to be a real abnormality of the target device. When the abnormal fluctuation migrates, reassembles or disappears with the change of forwarding binding relationship, it is determined to be an abnormality accompanying the wireless networking, and the target device status early warning information is output according to the attribution determination result.

[0023] Working principle: This solution first clarifies the multi-sensor data and wireless networking relationships of each device, then identifies which sensors simultaneously exhibit abnormalities, forming an abnormal node group. Next, without changing the sensor positions, it actively alters the forwarding binding relationships of these abnormal nodes and re-collects data. Finally, it compares whether the abnormality changes along with the binding relationship before and after the disturbance. If the abnormality consistently falls on the same device, it's more likely a device-specific malfunction. If the abnormality migrates, reassembles, or disappears with the change in binding relationship, it's more likely an accompanying abnormality brought about by the wireless network. This allows for the differentiation between genuine device abnormalities and false network abnormalities. For example, in a server room, two coolant pumps located very close to each other are equipped with vibration, current, and temperature sensors. The system first detects that the vibration and current nodes on one of the coolant pumps are simultaneously abnormal, and that they are using the same wireless forwarding path. Then, the system temporarily switches one of the nodes to another forwarding path and checks whether the abnormality still exists on the original device. If the abnormality is still on this coolant pump, it indicates a device status warning; if the abnormality disappears or moves to another node, it indicates a network interference warning. After seeing the results, maintenance personnel can directly determine whether to check the device first or the wireless network first.

[0024] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for early warning of industrial equipment status based on multi-sensor collaborative monitoring, characterized in that, include: S1. Obtain the monitoring data collected by the multi-sensor nodes corresponding to each target device in the monitoring area and the current forwarding binding relationship of the multi-sensor nodes in the wireless network. Perform association organization on the monitoring data and output the node association dataset and the current binding relationship set. S2. Perform synchronous anomaly identification on the node association dataset, identify the abnormal node group that has abnormal fluctuations within the same abnormal time period, extract the shared forwarding binding relationship corresponding to the abnormal node group from the current binding relationship set, and output the set of abnormal segments to be judged. S3. For the set of abnormal segments to be determined, keep the installation position and collection object of the abnormal node group unchanged, perform controlled disgrouping or controlled merging operations on the shared forwarding binding relationship to form a disturbance forwarding binding relationship, and reacquire the monitoring data corresponding to the abnormal node group based on the disturbance forwarding binding relationship, and output the disturbance observation dataset. S4. Perform a correspondence comparison between the set of abnormal segments to be judged and the disturbance observation dataset to determine whether the abnormal fluctuations have migrated, reorganized or disappeared due to changes in the forwarding binding relationship, and output the abnormal migration judgment result. S5. Perform attribution determination on the abnormal migration judgment result. When the abnormal fluctuation maintains the same spatial attribution of the corresponding target device before and after the change of the forwarding binding relationship, it is determined to be a real abnormality of the target device. When abnormal fluctuations migrate, reorganize, or disappear due to changes in forwarding binding relationships, they are identified as wireless networking-related anomalies, and target device status warning information is output based on the attribution determination result.

2. The industrial equipment status early warning method for multi-sensor collaborative monitoring according to claim 1, characterized in that: S1 includes: S1-1. Obtain the node identifier, target device identifier, sensor type identifier, acquisition time, and current parent node identifier corresponding to each multi-sensor node. Classify each multi-sensor node according to the target device identifier, sensor type identifier, and current parent node identifier, and output the node classification result. S1-2. Sort the multi-sensor nodes of each type in the node classification results according to the acquisition time, and output the node association fragment and binding change fragment for the statistical data missing status and parent-child correspondence changes of the sorted multi-sensor nodes. S1-3. Write the node association fragments and binding change fragments according to the target device identifier to generate the node association dataset and the current binding relationship set.

3. The industrial equipment status early warning method based on multi-sensor collaborative monitoring according to claim 2, characterized in that: S2 includes: S2-1. Read the target device identifier, sensor type identifier, acquisition time and monitoring data of each multi-sensor node in the node association dataset. Perform fluctuation comparison on the monitoring data of each multi-sensor node under continuous acquisition time, determine the abnormal start time and abnormal duration of each multi-sensor node, and output the abnormal period results of the node. S2-2. Perform overlapping identification on the results of abnormal node periods according to the duration of abnormal periods, identify multiple multi-sensor nodes that simultaneously exhibit abnormal fluctuations within the same abnormal period, generate abnormal node groups according to the target device identifier and the corresponding abnormal period, and output the abnormal node group results. S2-3. Read the parent-child binding relationship corresponding to the result of the abnormal node group in the current binding relationship set, perform shared parent node identification and continuous forwarding path tracing for each multi-sensor node in the same abnormal node group, extract the shared forwarding binding relationship corresponding to the abnormal node group, and generate a set of abnormal segments to be judged in combination with the corresponding abnormal time period.

4. The industrial equipment status early warning method based on multi-sensor collaborative monitoring according to claim 3, characterized in that: The process of outputting the results of the abnormal node group in S2-2 also includes: S2-21. Read the target device identifier, abnormal start time, and abnormal end time of each multi-sensor node in the abnormal time period results. Construct the abnormal time period intervals corresponding to each multi-sensor node through the abnormal start time and abnormal end time. Perform overlapping duration calculation, boundary interval calculation, and interval inclusion relationship calculation on any two abnormal time period intervals. Determine the abnormal time period intervals with overlapping duration greater than zero and boundary intervals of zero as adjacent time period intervals. Determine the abnormal time period intervals with interval inclusion relationships as included time period intervals. Generate time period association results based on adjacent time period intervals and included time period intervals.

5. The industrial equipment status early warning method for multi-sensor collaborative monitoring according to claim 4, characterized in that: The process of outputting the results of the abnormal node group in S2-2 also includes: S2-22. The time period association results are aggregated within the group according to the target device identifier, and an abnormal node time period graph is constructed for the time period association results corresponding to each target device. Multi-sensor nodes with time period association results are identified as graph nodes, and the time period association results between any two graph nodes are identified as graph edges. Connected subgraph identification is performed on each abnormal node time period graph. Connected subgraphs containing only a single graph node are removed. For conflicting nodes that appear in multiple connected subgraphs at the same time, the total overlap time with the other graph nodes in each connected subgraph is calculated. Conflicting nodes are assigned to the connected subgraph with the higher value of the total overlap time. The abnormal node subgraph results are output. S2-23. Perform common abnormal time period calculation on each connected subgraph in the abnormal node subgraph result. Determine the later time among the abnormal start times of each multi-sensor node in each connected subgraph as the common start time, and determine the earlier time among the abnormal end times of each multi-sensor node in each connected subgraph as the common end time. When the common start time is earlier than the common end time, determine the multi-sensor nodes in the corresponding connected subgraph as the abnormal node group in the same abnormal time period, and write the target device identifier with the common start time and common end time to generate the abnormal node group result.

6. The industrial equipment status early warning method for multi-sensor collaborative monitoring according to claim 5, characterized in that: S3 includes: S3-1. Read the shared forwarding binding relationship, target device identifier and abnormal time period corresponding to each abnormal node group in the set of abnormal segments to be judged. Perform parent node affiliation statistics and path overlap statistics on each multi-sensor node in the shared forwarding binding relationship. Determine the multi-sensor nodes with the same parent node identifier as the same parent node set. Determine the multi-sensor nodes that pass through the same intermediate node in the forwarding path as the same path set. Output the binding aggregation result. S3-2. Determine the perturbation method for the binding aggregation result. When the number of multi-sensor nodes in the same parent node set is not less than two, perform a controlled splitting operation on the same parent node set, and switch at least one of the multi-sensor nodes to another parent node identifier that is not occupied by the same parent node set. When the multi-sensor nodes in the same path set belong to different parent node identifiers, perform a controlled merging operation on the same path set, and merge the multi-sensor nodes with the same forwarding path endpoint into the same parent node identifier to generate a perturbation forwarding binding relationship.

7. The industrial equipment status early warning method based on multi-sensor collaborative monitoring according to claim 6, characterized in that: S3 also includes: S3-3. Write the disturbance forwarding binding relationship into each multi-sensor node corresponding to the abnormal node group. Perform synchronous acquisition on each multi-sensor node after writing in the subsequent acquisition period corresponding to the original abnormal period to obtain the monitoring data of each multi-sensor node under the disturbance forwarding binding relationship and generate a disturbance observation dataset. S3-4. Perform corresponding association on the disturbance observation dataset according to the target device identifier, multi-sensor node identifier, acquisition time and disturbance forwarding binding relationship. Write the changes in parent node identifier before and after the disturbance, the changes in forwarding path and the corresponding changes in monitoring data accordingly to generate the bound disturbance association fragment in the disturbance observation dataset.

8. The industrial equipment status early warning method based on multi-sensor collaborative monitoring according to claim 7, characterized in that: S4 includes: S4-1. Read the target device identifier, common abnormal time period, abnormal node group composition and shared forwarding binding relationship corresponding to each abnormal node group in the abnormal segment set to be judged, as well as the disturbance forwarding binding relationship and disturbance monitoring data of the corresponding abnormal node group in the disturbance observation data set, perform abnormal re-identification on the disturbance monitoring data, and generate disturbance abnormal node group results. S4-2. Perform a corresponding comparison between the composition of the abnormal node group and the result of the disturbed abnormal node group, calculate the difference in abnormal node identifiers before and after the disturbance, the difference in the common abnormal time period, and the difference in the corresponding target device identifier. When an abnormal node exits before the disturbance and other nodes enter after the disturbance, it is determined as an abnormal migration. When the connection relationship of abnormal nodes splits or merges before and after the disturbance, it is determined as an abnormal reorganization. When no corresponding abnormal node group is identified after the disturbance, it is determined as an abnormal disappearance. Output the abnormal morphology judgment result. S4-3. Perform corresponding write operations on the abnormal form determination results and the shared forwarding binding relationship and the disturbance forwarding binding relationship to generate an abnormal migration determination result containing the target device identifier, abnormal node group identifier, abnormal form category and binding relationship change results.

9. The industrial equipment status early warning method based on multi-sensor collaborative monitoring according to claim 8, characterized in that: S5 includes: S5-1. Read the target device identifier, abnormal form category, shared forwarding binding relationship and disturbance forwarding binding relationship corresponding to each abnormal node group in the abnormal migration judgment result. Statistically analyze the distribution of target device identifiers and changes in node composition of each multi-sensor node in each abnormal node group before and after the disturbance. When the target device identifiers corresponding to the multi-sensor nodes in the abnormal node group remain consistent before and after the disturbance and the abnormal form category does not include migration, reorganization and disappearance, the corresponding abnormal node group is determined as the real abnormality of the target device, and the real abnormality result is output. S5-2. For each abnormal node group that is not determined to be a real abnormality of the target device in the abnormal migration judgment result, perform network companion judgment based on the abnormal form category and the change result of forwarding binding relationship before and after the disturbance. When the abnormal form category includes migration, reorganization or disappearance, the corresponding abnormal node group is determined as a wireless network companion abnormality and the companion abnormality result is output. S5-3. Write the actual abnormal results and accompanying abnormal results according to the target device identifier, generate target device status warning information for actual abnormalities of the target device, generate network interference warning information for accompanying abnormalities of wireless networking, and output the corresponding target device status warning information.

10. A multi-sensor collaborative monitoring industrial equipment status early warning system, used to implement the multi-sensor collaborative monitoring industrial equipment status early warning method according to any one of claims 1-9, the system comprising a data association module, an anomaly grouping module, a disturbance binding module, a judgment module, and an early warning module, characterized in that: The data association module is used to obtain the monitoring data collected by the multi-sensor nodes corresponding to each target device in the monitoring area and the current forwarding binding relationship of the multi-sensor nodes in the wireless network. It performs association organization on the monitoring data and outputs the node association dataset and the current binding relationship set. The anomaly grouping module is used to perform synchronous anomaly identification on the node-associated dataset, identify the abnormal node groups that exhibit abnormal fluctuations within the same abnormal time period, extract the shared forwarding binding relationship corresponding to the abnormal node group from the current binding relationship set, and output the set of anomaly fragments to be judged. The binding disturbance module is used to maintain the installation position and collection object of the abnormal node group for the set of abnormal segments to be judged, perform controlled disgrouping or controlled merging operations on the shared forwarding binding relationship, form a disturbance forwarding binding relationship, and re-acquire the monitoring data corresponding to the abnormal node group based on the disturbance forwarding binding relationship, and output the disturbance observation dataset. The judgment module is used to perform a correspondence comparison between the set of abnormal segments to be judged and the disturbance observation dataset, to determine whether the abnormal fluctuations have migrated, reorganized or disappeared due to changes in the forwarding binding relationship, and output the abnormal migration judgment result. The early warning module is used to determine the attribution of abnormal migration judgment results. When the abnormal fluctuation maintains the same spatial attribution of the corresponding target device before and after the change of forwarding binding relationship, it is determined to be a real abnormality of the target device. When abnormal fluctuations migrate, reorganize, or disappear due to changes in forwarding binding relationships, they are identified as wireless networking-related anomalies, and target device status warning information is output based on the attribution determination result.

Images