Vibration event early warning diffusion method based on edge calculation

Abstract

The invention discloses a vibration event early warning diffusion method based on edge calculation, and relates to the technical field of computer application, and the method comprises the following steps: constructing a topological phase coordinate under a unified time baseline, playing back a historical track of early warning diffusion based on the topological phase coordinate, recognizing a continuous path segment with a closed trend in the track, and carrying out the early warning diffusion. Outputting a loop suspicious region; a fingerprint code of early warning information is constructed on the basis of a loop suspicious area, a three-dimensional feature set containing time sequence features, path sequence features and phase sequence features is extracted, and a corresponding loop suspicion degree spectrum is generated. According to the method, through space-time modeling, path fingerprint construction, causal discrimination and duplicate removal mechanisms, accurate identification and repeated suppression of a vibration early warning information diffusion path are realized, dynamic threshold regulation and reverse intervention are combined, loop closed-loop treatment and topology self-healing are realized, and early warning accuracy, timeliness and system stability are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer application, in particular to a vibration event early warning diffusion method based on edge computing. BACKGROUND

[0002] The "vibration event early warning diffusion based on edge computing" refers to the large amount of real-time data generated in the vibration monitoring process, which is preferentially processed and intelligently analyzed in the edge computing node close to the vibration source, rather than being transmitted to the central server. When the system detects a possible abnormal or risk vibration event, the edge node will immediately generate an early warning information, and through a hierarchical, regional or node diffusion mechanism, the early warning signal will be quickly transmitted to the relevant equipment, control center or other edge nodes in a low-delay and high-efficiency manner, realizing multi-point synchronous perception and response. This method can not only reduce the delay and bandwidth pressure caused by data transmission, but also ensure that the early warning information quickly covers the target area in complex environments, thereby greatly improving the perception speed, early warning accuracy and timeliness of event disposal of vibration abnormalities.

[0003] The prior art has the following disadvantages: In the prior art, vibration event early warning diffusion usually relies on dynamic network topology between multiple nodes for information transmission. However, when part of the nodes cause instantaneous changes in the topology structure due to load adjustment or link switching, an implicit loop structure may be formed in the diffusion path. In this loop, the originally single early warning information will be continuously superimposed due to cyclic forwarding, producing an amplification effect similar to resonance. Since the prior art lacks identification and suppression mechanisms for this type of implicit loop, the monitoring center will misjudge the same early warning repeatedly diffused as multiple independent vibration risk events in the receiving process, thereby triggering excessive disposal measures, causing unnecessary resource consumption, and even possibly causing systemic scheduling confusion and secondary risks.

[0004] The above information disclosed in the background section is only intended to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a vibration event early warning diffusion method based on edge computing to solve the problems in the background art.

[0006] In order to achieve the above-mentioned purpose, the present application provides the following technical solution: a vibration event early warning diffusion method based on edge computing, comprising the following steps: S1, constructing a topology phase coordinate under a unified time baseline, replaying the history track of early warning diffusion based on the topology phase coordinate, identifying the continuous path segment with a closed trend in the track, and outputting the loop suspicious area; S2, on the basis of suspicious areas of the loop, construct fingerprint coding of early warning information, extract a three-dimensional feature set containing time series features, path series features and phase series features, and generate the corresponding loop suspicion spectrum; S3, based on the loop doubt spectrum, performs causal consistency judgment, strips out duplicate paths and locates the boundary region forming the loop, and constructs a one-way commitment chain to freeze the identified duplicate path segments; S4 constructs a cross-node deduplication proof chain under the constraint of a one-way commitment chain. It ensures path uniqueness by injecting one-time tokens, monotonic counters, and time watermarks, thereby eliminating the repeated propagation of early warning information. S5, based on the deduplicated diffusion path, deploys an adaptive threshold control mechanism to adjust the priority of the diffusion path according to the baseline residual, path density slope and diffusion arrival difference, to converge high-risk paths and stabilize the diffusion structure. S6 performs time-reversal diffusion control based on a stable diffusion structure. It constructs an anti-phase feedback window through phase-driven perturbation injection to absorb residual energy in the path and writes the correction result into the topology adjustment vector to achieve closed-loop suppression and adaptive correction of loop diffusion.

[0007] Preferably, step S1 includes: After establishing a unified time baseline and topological phase coordinates, the historical trajectory of the early warning propagation process is replayed based on the topological phase coordinates. The node forwarding events are sorted in ascending order of absolute time, and trajectory segments are constructed by combining spatial path numbers and phase offsets. In the constructed trajectory segments, identify the propagation delay, spatial offset direction, and phase change of continuous path jump segments, and screen trajectory segments with propagation delay below the average level, repeated changes in spatial offset direction, and phase value decreasing to near zero as suspected return segments; In the suspicious return segment, the included angle of the trajectory line segment, the spacing of the path number, the time interval, the amplitude of the phase fluctuation and the curvature change in three-dimensional space are calculated to identify the trajectory path point set with a closing trend; The time range, path number range, and phase change interval of the identified trajectory path point set are extracted, along with the path return probability, propagation convergence speed, and closure curvature amplitude range, and the output is the suspected loop area.

[0008] Preferably, step S2 includes: Based on the completed loop suspicious area, the timestamp information of the forwarding node is extracted, the time value is calculated and a time vector sequence is generated. The sliding window process is used to identify the time-dense segment to form time series features. Record the sending node number and receiving node number of each hop forwarding event to form a path number sequence, calculate the path hop sequence and node repetition rate, extract path return pattern fragments, and constitute path sequence features; The phase value is obtained by comparing the node time value with the path hop count, forming a phase sequence, and the phase convergence segment is identified to constitute the phase sequence feature; Synchronous mapping of time series features, path series features, and phase series features is performed to construct a three-dimensional feature set. The concentration of each time segment, the amplitude of path jump, and the degree of phase convergence are extracted using a sliding analysis window to generate a loop skepticism spectrum.

[0009] Preferably, step S3 includes: Based on the path segments whose scores exceed the preset threshold in the loop suspicion spectrum, read the timestamp, path number, phase value and propagation direction of the nodes, determine the time inversion value, path overlap and phase back-off magnitude, identify causal abnormal jump points and confirm redundant closed loop propagation segments. Around the causal abnormal jump point in the redundant closed-loop propagation segment, extend the preceding and following nodes to form a propagation segment set, extract the path number reversal point, the time interval sudden increase point and the phase inflection point, and superimpose the results to obtain the boundary node point set, and locate the boundary start point and boundary end point. Based on the path segment between the starting point and the ending point of the boundary, a one-way commitment chain is constructed. A node number, timestamp, path number and freeze flag are added to each event node to freeze the propagation behavior within the path segment and block repeated propagation.

[0010] Preferably, step S4 includes: Based on the one-way commitment chain, the source node number, path start timestamp, path number sequence and phase value are extracted to generate path digest information, and a one-time token generated by a random number generator is embedded in the propagation information; A monotonic counter is introduced in each hop of the propagation process. The initial value is set to one. The counter is incremented by one each time the propagation is carried out and the difference is recorded. If the hop is discontinuous or the increment fails, it is judged as an abnormal propagation and forwarding is stopped. Between the generation of a time watermark at the starting node of the path and the insertion of key fields within the information, the receiving node extracts the watermark and compares it with the locally reconstructed watermark to verify the consistency of propagation. After each hop of propagation, a one-time token, a counter value, a receiving timestamp, a path number, and a time watermark are encapsulated to form an event unit, which is then appended to the event chain to achieve propagation path tracking. At the end of the propagation, a path status identifier is generated, which includes the path start and end numbers, hop count, cumulative delay and final check code, and written into the frozen path list to establish a dual-index structure to block duplicate paths.

[0011] Preferably, step S5 includes: A propagation baseline residual measurement model is constructed. By calculating the time difference between the actual propagation time and the standard propagation model, the residual vector is extracted and statistical analysis is performed to identify the rhythm disorder path. Based on the network spatial distribution of propagation paths, a path density evolution model is constructed and the density slope is calculated to identify highly clustered path segments with continuously increasing density and abnormal jump rates. By integrating the path residual and density slope analysis results, the hop-level diffusion arrival difference is extracted, a ternary feature fusion matrix is ​​constructed, hop segments are prioritized and propagation thresholds are dynamically set.

[0012] Preferably, during segment propagation, propagation is only allowed when the residual regression value is lower than the basic standard, the local path density does not exceed the set upper limit, and the diffusion reaches the differential within the steady-state range; otherwise, execution is delayed to ensure that the diffusion behavior is controlled in high-risk path segments.

[0013] Preferably, step S6 includes: Extract segment propagation events to construct a propagation behavior set, identify path segments with abnormal propagation rhythm, and calculate the disturbance intensity value; The intervention window is set based on the disturbance intensity value, and a phase-driven micro-perturbation event is injected to form a local negative feedback structure; Collect jump propagation indicators before and after intervention to construct perturbation correction vectors and identify the changing trends of response nodes and paths; The perturbation correction vector is uniformly mapped to the topology adjustment vector, and the path direction, rate and propagation priority are dynamically adjusted.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention achieves high-precision spatiotemporal modeling and closure trend identification of diffusion paths by constructing topological phase coordinates under a unified time baseline, thus locking potential loop regions at the source. It constructs a warning information fingerprint using a three-dimensional feature set to deeply mine the path evolution patterns. Furthermore, it uses loop suspicion spectrum and causal consistency discrimination to remove redundant paths and construct a freezing mechanism to prevent repeated loop propagation of information in the network. Subsequently, it introduces one-time tokens and time watermarking to establish cross-node deduplication chains, ensuring path uniqueness at the mechanism level. Based on this, it deploys a dynamic threshold control mechanism to dynamically adjust the diffusion rhythm according to residual characteristics and path density trends, quickly converging high-risk paths and maintaining propagation structure stability. Finally, it injects anti-phase perturbation signals through a time inversion mechanism, combined with topology adjustment vectors to achieve closed-loop suppression and topology self-healing. Overall, this method possesses high path identification accuracy, perturbation suppression capability, and network adaptive adjustment capability. It not only significantly improves the accuracy and timeliness of warning information dissemination and reduces false alarm rate and system resource consumption, but also enhances the stability and robustness of multi-point linkage response to vibration events in complex environments. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a flowchart of the vibration event early warning and diffusion method based on edge computing according to the present invention. Detailed Implementation

[0017] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0018] This invention provides, for example Figure 1 The vibration event early warning and dissemination method based on edge computing shown includes the following steps: S1, construct topological phase coordinates under a unified time baseline, replay the historical trajectory of early warning spread based on the topological phase coordinates, identify continuous path segments with a closing trend in the trajectory, and output suspicious loop areas; To identify loop-prone paths during the early warning diffusion process of vibration events, a processing method is proposed that constructs topological phase coordinates based on a unified time baseline and replays historical diffusion trajectories. Through the following steps, the method achieves accurate identification of path segments with a closing trend and outputs them as suspected loop areas: A unified time baseline is constructed, and a topological phase coordinate system is established based on this baseline. Specifically, a time reference node is first defined, which periodically broadcasts timestamps from a highly stable time source (e.g., a local clock controlled by a temperature-controlled crystal oscillator) to synchronize the local clocks of all edge nodes in the network. After synchronization is established, a precise absolute timestamp is assigned to each forwarding event in an actual early warning propagation path, thus forming a unified time baseline. Under this time baseline, a three-dimensional topological phase coordinate system is established, where: The first dimension is the time axis, which records the trigger time of each hop in the early warning spread, in milliseconds; The second dimension is the spatial path axis, which is numbered according to the forwarding direction of the physical connection between nodes. For example, it is numbered from 1 to N according to the geographical distribution from west to east. The third dimension is the phase evolution axis, which describes the phase shift of each hop forward relative to the original signal triggering time. Specifically, it is calculated by the ratio of the propagation delay of each hop to the current position of the spread wavefront.

[0019] All node forwarding events form a discrete set of points in this three-dimensional space, constituting the basic point matrix of the trajectory, providing a coordinate mapping basis for trajectory playback.

[0020] Based on the constructed topological phase coordinates, the early warning propagation trajectory is replayed hop by hop. First, all node events in each trajectory are arranged in ascending order according to absolute timestamps to form a complete forwarding sequence. Then, starting from the first initiating node, subsequent forwarding nodes are read hop by hop in chronological order, and their position in the spatial path axis and the offset value of the phase evolution axis are combined to form trajectory path segments. During the replay process, the following key indicators are recorded: (1) the propagation delay of each hop, i.e., the current node timestamp minus the previous hop timestamp; (2) the spatial path offset, i.e., the current hop path number minus the previous hop path number; (3) the phase change, i.e., the current node phase value minus the previous node phase value. If the spatial path offset direction of a certain trajectory is reversed more than twice in consecutive hops, and the propagation delay is less than 70% of the historical average, while the phase change gradually decreases to near zero, the trajectory segment is recorded as a "return suspicious segment". This step avoids the problem of path tracing breakage caused by unidirectional propagation of nodes and ensures the continuity of the trajectory and the rationality of the phase.

[0021] Identify continuous path segments with a closing trend in the trajectory. Further similarity calculations and spatial closing trend analysis are performed on the aforementioned suspicious return segments. Specifically, the set of line segments in the current trajectory segment in three-dimensional coordinate space is compared with the cosine angle of its historical trajectory segments. If three or more line segments have an angle less than 30 degrees, and the distance between the start and end spatial path numbers does not exceed 5 number units, and the time interval between the start and end does not exceed twice the historical average, then the trajectory segment is determined to have a path closing trend. The total phase change is further statistically analyzed. If the change is alternating between positive and negative, fluctuating around 0, and the difference in amplitude between three fluctuations is within 20%, it is considered to have a cyclic evolution tendency. Based on this, the curvature change of each node event in three-dimensional space is analyzed point by point. Those curvature values ​​that fluctuate periodically and are concentrated in a small range are used as the final confirmed path point set with a closing trend. This method does not rely on path templates or pre-set cyclic structure forms; it is entirely derived from actual propagation data, avoiding the poor adaptability and high false alarm problems caused by template matching using static path diagrams in existing technologies.

[0022] The output identification results are used as suspected loop areas. All node events in the path segments with closed trends are organized into a continuous sequence, and their time range, path number range and phase change interval are extracted as annotation content, with the following additional parameters: (1) Path return probability, defined as the overlap rate between the return path node and the previous node; (2) Propagation convergence speed, defined as the average time taken for the phase change to change from positive to negative or from negative to positive; (3) Closed curvature amplitude range, used to identify the stability of the trajectory shape. These data are packaged into a structured output and used as input data for early warning fingerprint coding, loop suspicion calculation and causal consistency judgment in subsequent steps. In this way, not only are the temporal, spatial and phase characteristics of suspected loop areas clarified, but also full and quantifiable auxiliary data are provided. It can also stably and accurately capture closed trend trajectories in dynamic network environments, providing a solid foundation for subsequent governance steps.

[0023] S2, on the basis of suspicious areas of the loop, construct fingerprint coding of early warning information, extract a three-dimensional feature set containing time series features, path series features and phase series features, and generate the corresponding loop suspicion spectrum; To identify the propagation characteristics of early warning information in suspected areas of the loop and to provide a quantitative basis for subsequent causal determination, a fingerprint code for the early warning information is constructed based on the forwarding information of all nodes in the suspected areas of the loop. The process consists of the following steps: Time-series features are extracted from the suspected loop area. In practice, event timestamps of all forwarding nodes within the area are extracted one by one, and the relative time value of each node is calculated in milliseconds using a unified time baseline. All relative time values ​​are sorted in ascending order to generate a time vector sequence, used to characterize the actual time sequence of information propagation. Subsequently, the time vector is subjected to adjacent time difference processing to obtain a time interval sequence, where each element represents the actual propagation delay between two hops. Next, a sliding window mechanism is used, scanning the time interval sequence in fixed-length units (e.g., five hops), and the average propagation delay, standard deviation, maximum interval, and minimum interval within each window are statistically analyzed to identify areas of sudden increases in propagation density. If the average propagation delay within three consecutive windows is less than 70% of the mean delay within the suspected loop area, and the standard deviation is less than 20% of the total sequence, then this segment is marked as a time-dense segment, indicating that frequent information round trips may occur in this segment. Finally, all time interval sequences and marked segments are retained as input features for the time dimension to construct a time-varying model of early warning propagation.

[0024] Extracting path sequence features. For each hop forwarding event, record the sending node number and receiving node number, constructing a path number sequence in chronological order to represent the propagation trajectory of information in the physical network. Simultaneously, calculate the path change value for each hop (current path number minus the previous hop path number) to obtain a path jump sequence, used to assess the trend of propagation direction changes. When a path jump sequence contains three or more consecutive negative values, and each jump value is less than one-fifth of the total range of node numbers, the path is considered to have a tendency to backtrack. Simultaneously, perform duplicate detection on the path number sequence, marking node numbers that appear more than twice and recording their positions to determine if a local loop has formed. For each path loop interval, extract the five hops before and after it as context, and construct a set of path loop pattern fragments for subsequent feature fusion. This step, by quantifying path jumps and loop behavior, makes spatial propagation features more identifiable, showing a significant advantage in capturing local loop behavior compared to traditional methods that only perform global comparisons of path numbers.

[0025] Phase evolution sequence features are extracted and fused with time, path, and phase features to construct a three-dimensional feature set. Specifically, the phase value of each hop node is obtained by dividing its relative time value by the number of path hops between that node and the source node, representing the relationship between its propagation rhythm and propagation depth. The phase values ​​of all nodes are arranged chronologically to form a phase sequence. Continuous difference calculations are performed on the phase sequence to identify segments where the phase values ​​of multiple consecutive nodes show a decreasing trend, especially segments containing more than three negative differences with each difference magnitude within 10%, which are marked as phase convergence segments. Furthermore, the time sequence, path number sequence, and phase sequence are aligned, and the three types of features are mapped into a three-dimensional matrix with the node event timestamp as the main axis. The first dimension of the matrix is ​​the time step sequence, recording the forwarding time of all nodes; the second dimension is the path number, recording the node's position in the topology; and the third dimension is the phase value, recording the corresponding phase state of the node. In the three-dimensional matrix, each cell represents the path and phase state of a node at a certain time step, forming a complete propagation feature mapping map. This three-dimensional feature set not only preserves the spatiotemporal information of the original propagation behavior, but also introduces propagation dynamics features through the phase dimension, providing high-dimensional data support for the next step of quantitative analysis.

[0026] A loop skepticism spectrum is constructed based on a three-dimensional feature set. A sliding analysis window is set within the three-dimensional feature set to scan node behavior features hop-by-hop, quantitatively scoring the temporal concentration, path jump amplitude, and phase convergence degree of each window. Specifically, temporal concentration is calculated using the standard deviation of the time interval within the window; a smaller standard deviation results in a higher score. Path jump amplitude is calculated using the root mean square value of the path jump sequence within the window; a smaller jump amplitude indicates a tendency for propagation to return, thus increasing the score. Phase convergence degree is determined by the fluctuation range of the phase sequence; if the phase value within the window continuously decreases from a high value and approaches zero, phase closure behavior is considered to exist, and the corresponding score also increases. After normalizing the scores of the three dimensions, a weighted composite score is synthesized, serving as the loop skepticism score for the current position of the window. Finally, all window scores in the entire suspicious region are arranged in chronological order and path number, plotted as a two-dimensional chromatogram, where the horizontal axis represents the time step, the vertical axis represents the path number, and the color level represents the intensity of the skepticism score. A higher color level value indicates that the path segment is more likely to constitute a loop, which is used in subsequent steps for causal consistency determination.

[0027] S3, based on the loop doubt spectrum, performs causal consistency judgment, strips out duplicate paths and locates the boundary region forming the loop, and constructs a one-way commitment chain to freeze the identified duplicate path segments; To identify and freeze recurring propagation paths caused by loop structures during vibration early warning propagation, causal consistency judgment is performed based on the generated loop suspicion spectrum to accurately isolate recurring paths, locate loop boundary regions, and freeze confirmed recurring propagation segments by establishing a one-way commitment chain. Specifically, the steps include: Based on the loop suspicion spectrum, suspected redundant paths are identified and causal consistency is determined. In the previous technical step, a three-dimensional feature set was constructed and a loop suspicion spectrum covering the entire suspicious area of ​​the loop was generated. The spectrum uses the time axis as the horizontal axis and the path number as the vertical axis, and the color intensity represents the loop suspicion score of the corresponding path segment. First, a suspicion score threshold is set on the spectrum, for example, 0.85. The entire spectrum is scanned and all spectrum segments that continuously exceed the threshold are extracted as a candidate set of high suspicion areas. For each high suspicion spectrum segment, the corresponding node forwarding event is read, including four key attributes such as the event's timestamp, path number, phase value, and propagation direction. For each event, the records of the previous hop and the two hops of the propagation node are traced back according to the propagation time order, and the following are calculated: (1) Time inversion value, that is, whether the time difference between the current event time and the previous hop event time is negative; (2) Path overlap, that is, whether the current path number is the same as or differs from the previous two hop numbers by only one number; (3) Phase backoff magnitude, that is, whether the current phase value is less than the previous hop phase value. If two of the above three indicators are met simultaneously, the event is marked as a "causal anomaly jump point." The judgment is then extended to the next three jumps. If there is an overall reversal of the time-path-phase trend, i.e., an overall propagation pattern of "later-arriving first, near retreating far advancing, and phase compression," then the path segment is confirmed as a "redundant closed-loop propagation segment." Unlike traditional methods that rely solely on the repetition of path numbers or timestamp sorting, this method comprehensively introduces phase dimension and directional trend analysis to construct a causal closed loop at the structural and dynamic behavior levels, effectively improving identification accuracy and robustness.

[0028] After identifying redundant propagation paths, the closed-loop path is carefully stripped and the loop boundary region is located. For the identified causal abnormal jump points, five jump node events are extended before and after the time series to form a complete propagation segment set. In this propagation segment, the path number change trend, time interval fluctuation and phase evolution gradient of each jump event are analyzed one by one. The specific judgment criteria are as follows: (1) If the path number has three consecutive positive jumps and then suddenly has two negative jumps and returns to the previous path number range, it is marked as a path reversal point; (2) If the time interval between adjacent events changes from decreasing to increasing by more than twice the original average, and there is still forwarding behavior after this increase segment, then this position is a potential diffusion return starting point; (3) If the phase value change rate changes from positive to negative three times in a row, and the change amplitude remains within 30% of the original phase evolution amplitude, it is marked as a phase inflection point. The feature points of the above three types are intersected and superimposed to obtain the node point set with the most boundary attributes. Within the node set, the two nodes with the strongest abrupt change in propagation direction and the most prominent time interval are selected as loop boundary points, denoted as the boundary start point and boundary end point, respectively. Using these as the boundary, all propagation paths between them are combined into a closed-loop path segment, providing a precise boundary range for subsequent freezing processing. This method differs from traditional topology matching or graph search methods; it does not rely on a global path graph or centralized path list, but rather adaptively identifies the propagation boundary through real-time data evolution trends. It possesses dynamic adaptability and event-driven characteristics, better reflecting the frequent node changes in actual deployment environments.

[0029] Based on the loop boundary path segment, a one-way commitment chain is constructed and repeated propagation behavior is frozen. After completing the identification of the closed loop path segment boundary, all propagation events within the path segment are ordered, with timestamp as the primary order and path number as the secondary order, to construct a strictly unidirectional event sequence. A commitment chain attribute label is attached to each event, which contains four items: (1) the node number corresponding to the event; (2) the timestamp of the event (in milliseconds); (3) the path number where the event is located; and (4) the freeze flag. The freeze flag is set to an immutable boolean type, and its value is set to true to indicate that the event has been frozen and cannot continue to participate in the propagation forwarding operation. During the construction process, each event is linked to its successor event in a one-way linked list and the predecessor pointer is recorded. The direction of this one-way commitment chain is kept consistent with the actual direction of information propagation, and forward referencing or skipping links are not allowed to ensure the causal integrity and traceability in the chain structure. After the freeze is implemented, all events in the commitment chain are considered non-repeatable nodes. If the same warning content is received again during subsequent propagation, it will be compared with the unique identifier of the event and the commitment status. If the same commitment item already exists, it will be automatically blocked and no further propagation logic will be triggered. This freeze mechanism not only blocks the duplicate information flow in the identified closed-loop path, but also achieves the general elimination of redundant events across the entire network through a distributed structure, possessing global binding force and event-level consistency.

[0030] S4 constructs a cross-node deduplication proof chain under the constraint of a one-way commitment chain. It ensures path uniqueness by injecting one-time tokens, monotonic counters, and time watermarks, thereby eliminating the repeated propagation of early warning information. To achieve unique identification of the early warning information propagation path and complete blocking of repeated propagation, a cross-node deduplication proof chain is constructed based on the existing one-way commitment chain. This chain employs a triple verification mechanism, incorporating one-time tokens, a monotonic counter, and a time watermark, effectively ensuring the traceability and non-replicability of the entire propagation path. The specific implementation process is as follows: Based on the established one-way commitment chain, a one-time token is generated for each valid early warning propagation path. Specifically, at the starting node of the path, basic propagation parameters such as the source node number, path start timestamp, path number sequence, and node reception phase value are extracted and combined to form a unique path digest. Based on this digest, a 128-bit random bit sequence is generated by calling a random number generator based on physical entropy sources, and this sequence is appended to the early warning propagation information as a one-time token. This token is used only once during the path's lifecycle and remains unchanged across all subsequent hop nodes. Each propagation node, upon receiving the information, must compare the token. If it finds that it has already received and processed a propagation request with the same token identifier, it is considered a duplicate propagation request, and propagation is immediately terminated. By embedding a one-time token, the path can be uniquely identified from the source, preventing information from being copied and propagated repeatedly on other paths. Compared to traditional duplicate detection methods based on node identifiers or simple hash comparisons, this method has higher resistance to collisions and greater unpredictability.

[0031] A monotonic counter is introduced during each hop of information propagation to strictly define the forward propagation direction and node increment order of the path. This counter is initially set to 1 at the propagation starting node. After each valid hop, the counter value is incremented by one in the next hop node, and the difference between the current count and the previous hop count is recorded. If the count value of a hop is not strictly greater than the previous hop, or if a non-continuous jump occurs (e.g., from 5 to 7), it indicates abnormal transition behavior in the path. In this case, the node will refuse propagation and record the hop path as a count consistency error. Furthermore, each hop node binds its current count value, current path number, and reception time to form a snapshot of the propagation event. This snapshot data is carried as an additional field in the information for downstream verification. This continuously monotonically increasing and non-jumping structure effectively prevents circumvention of existing path freezing mechanisms through node reordering or path insertion, making it particularly suitable for scenarios with frequent additions and deletions of edge nodes in dynamic topology environments.

[0032] A time watermark is embedded in the early warning dissemination information to verify the authenticity of the dissemination path and the consistency of the content. Specifically, the physical timestamp of the current node is extracted at the starting node of the path. This timestamp, combined with a one-time token, path number sequence, and node phase value, generates a 64-bit time feature sequence using a specific encoding method. This sequence is not stored centrally in the information header but is interleaved and embedded between different fields in the main body of the dissemination information, forming an invisible identifier with anti-segmentation capabilities. After receiving the information, the receiving node extracts the time watermark from a fixed offset position and compares it with its reconstructed local watermark value. If any eight consecutive bits in the watermark are inconsistent, it indicates that the information has been copied, tampered with, or disseminated via a non-original path by intermediate nodes, and subsequent forwarding should be terminated immediately. Compared with traditional message digest verification methods, the time watermark integrates dynamic parameters of time, path, and node status, possessing strong anti-tampering and path-binding properties. While ensuring the uniqueness of the information, it significantly reduces the possibility of watermark forgery.

[0033] An event chain is constructed during each propagation process to achieve full path tracking and deduplication proof of propagation behavior. After each successful forwarding, the current node encapsulates the key information of the hop event, including a one-time token, current count value, receiving timestamp, path number, and time watermark, into an event unit structure and inserts it into the tail of the existing event chain in the propagation information. As a dynamic structure, the event chain grows gradually with each hop of propagation. Each subsequent receiving node appends the current hop event unit to the tail of the chain after completing the verification operation, forming an ordered event sequence. In subsequent path verification or network diagnosis, if the event chain in a path contains reversed event node positions, duplicate path numbers, or abnormal time watermarks, it can be accurately determined that the path has duplicate propagation, path tampering, or abnormal hop behavior, thereby executing an abort operation and adding the relevant nodes to the propagation blacklist. Compared with traditional fixed path marking methods, this chain structure has complete historical tracking capabilities and propagation trajectory reconstruction capabilities, providing strong support for behavioral proof of propagation information.

[0034] After token verification, counter verification, time watermark comparison, and event chain construction, a propagation path status identifier is generated and written to the frozen path list to prevent the path from being reused in the future. The path status identifier includes the path start number, path end number, total hop count, cumulative propagation delay, event chain integrity status, maximum counter value, and final watermark verification code. This identifier is generated at the propagation end node and written to the frozen list after being bound to the path's unique identifier. The frozen list uses a dual-index structure based on the path number and token value to ensure efficient path deduplication detection. When any future node receives a new propagation message, it must first perform a matching search in the frozen list. If a path status identifier is found to match an already frozen path, it is immediately identified as an illegal duplicate path, and propagation is terminated. Through the bidirectional indexing mechanism of the frozen list, the efficiency of path hit detection can be effectively improved, reducing network congestion and resource waste caused by duplicate propagation among large-scale edge nodes.

[0035] S5, based on the deduplicated diffusion path, deploys an adaptive threshold control mechanism to adjust the priority of the diffusion path according to the baseline residual, path density slope and diffusion arrival difference, to converge high-risk paths and stabilize the diffusion structure. To suppress high-risk paths in vibration early warning propagation, an adaptive threshold control mechanism is deployed based on deduplicated propagation. This mechanism comprehensively utilizes baseline residual analysis, path density slope assessment, and differential quantization results of propagation arrival to dynamically adjust the propagation priority of each path, achieving propagation convergence control and structural stability optimization. Specifically, the following steps are included: A residual measurement model based on the propagation baseline is constructed to quantify the deviation between the actual propagation behavior of a path and the ideal model. In the preliminary steps, all propagation paths have been verified for integrity through one-time token verification, monotonic counter verification, and time watermark comparison, and duplicate paths have been frozen and can no longer propagate. Based on this, the propagation data of each remaining valid path is used as input to construct a standard propagation model according to a unified time baseline. This standard model uses the trigger time of the path's starting point as the time origin and the path hop count multiplied by the theoretical hop propagation time as the reference duration to construct an ideal propagation curve. Subsequently, the propagation time of each hop for each path is sampled, and the time difference between the actual propagation time and the standard model is calculated to form a set of residual vectors. Statistical analysis is performed on this residual vector, including the residual mean, range, standard deviation, and continuous fluctuation amplitude, thereby forming a quantitative description of the propagation deviation of the path. If the residual mean of a path deviates significantly from the zero baseline and the standard deviation exceeds 30% of the total number of hops of the entire path, the path is considered to have a disordered propagation rhythm and should be included in the list of priority control paths. Compared to traditional methods that only average path delays, this step constructs a hop-by-hop residual analysis system to achieve fine-grained monitoring of the propagation process, effectively exposing propagation paths with unstable rhythms and abnormal jumps.

[0036] A path density evolution model is constructed based on the spatial distribution of network nodes, and the density slope is calculated to characterize the path aggregation trend. Based on residual analysis, the distribution of paths in the network space within each propagation time period is modeled. Specifically, the physical coordinates of nodes are extracted into three-dimensional coordinate values, and a point cloud set of path distributions for different time periods is constructed using propagation time as a hierarchical index. The propagation path is divided into consecutive hops according to time segments, and the number of paths contained in a unit spatial volume within each hop is counted to form a path density function. Subsequently, the first-order time derivative of the path density function is processed to obtain the density slope function, which is used to measure whether there is a rapid aggregation trend in the path during propagation. When the density slope in a certain region is greater than zero in two consecutive time periods, and the hop rate exceeds 1.5 times the global average growth rate, the paths in that region are marked as high-aggregation path segments. Such paths usually indicate potential bottleneck areas or high-load communication node aggregation, which are highly likely to cause information resonance or link congestion. Based on this, the priority of the hops corresponding to the paths in that region is downgraded, their propagation timing is delayed, or their diffusion order is adjusted to avoid them from conflicting with other paths in the same time window. This method is not only more dynamic and responsive than traditional density assessment based on path number ranking, but also directly reflects potential structural congestion risks through density change trends, thus improving the foresight of regulatory actions.

[0037] By fusing path residual results and density slope judgment, an adaptive diffusion threshold field is deployed to achieve a dynamic priority adjustment control mechanism at the hop level. For each effective propagation path, the arrival time sequence of each hop is extracted, and the time difference between consecutive hops is calculated to form an inter-hop diffusion arrival difference sequence. This difference sequence is then normalized and combined with the residual vector and density slope sequence to construct a ternary feature fusion matrix. Scoring is performed on a hop-by-hop basis, with scoring dimensions including deviation from the time arrival rate, local path aggregation risk index, and stability index of propagation phase change. For hops with scores exceeding 1.2 times the average, a diffusion threshold enhancement action is implemented: during propagation of this hop, propagation is only allowed if the residual regression value of the previous hop is lower than the baseline standard, the local density of the spatial region where the hop is located does not exceed the set upper limit, and the diffusion time difference value is within the steady-state range. If any condition is not met, the hop enters a waiting state, delaying propagation until the conditions are met to avoid disturbing other paths. This adaptive threshold mechanism does not employ a static fixed-value strategy, but continuously updates the judgment criteria throughout the path's lifecycle, thus achieving dynamic and adjustable path behavior control.

[0038] S6 performs time-reversal diffusion control on the basis of a stable diffusion structure. It constructs an anti-phase feedback window by injecting phase-driven perturbation to absorb residual energy in the path and writes the correction result into the topology adjustment vector to achieve closed-loop suppression and adaptive correction of loop diffusion. To completely suppress the residual propagation of vibration early warning information in the loop path and enhance the self-healing capability of the topology, a time-reversal-based diffusion control mechanism is proposed based on the construction of a stable diffusion structure. This mechanism achieves directional elimination of residual energy and dynamic optimization of topological relationships through phase-driven perturbation injection, feedback absorption, vector correction, and structure write-back operations. Specifically, the following steps are included: After adaptive threshold adjustment, the network propagation paths have formed a structure with stable priority ordering and hop rhythm control. Based on this structure, propagation events of each effective path hop are collected, and five key elements are extracted: timestamp, path number, propagation direction, phase value, and counter value, constructing a hop propagation behavior set. Sequential analysis of the hops identifies path segments with compressed propagation delays, continuously reversing phase values, and no increase in counter values ​​within the time series. These hops typically exhibit the following characteristics: a phase that gradually increases and then suddenly decreases, or decreases and then increases; a propagation direction that jumps back and forth; and a time difference reduced to less than 60% of the standard hop delay. These hops are selected as suspected residual perturbation propagation paths, with key segments exhibiting resonant feedback trends during propagation marked. Perturbation intensity values ​​are calculated based on three indicators: phase fluctuation amplitude, time compression ratio, and path repetition rate. Path segments with intensity values ​​higher than 1.5 times the system average are listed as intervention candidate targets, providing a basis for subsequent micro-perturbation injection.

[0039] For each path segment marked as having high perturbation intensity, two hops are selected forward and two backward from its maximum perturbation point, forming a five-hop intervention window as the perturbation injection zone. Within this zone, a backpropagation event group is constructed. Each backpropagation event includes the following parameters: anti-phase value (same amplitude as the original event's phase value but opposite sign), time offset (offset relative to the center point of the intervention window), inverted path number (symmetrically mapped to the original path number), fixed counter identifier (to prevent propagation misjudgment), and a "not participating in the chain" mark. The backpropagation events are embedded into the intervention window path, forming a cross-over superposition structure with the original forward propagation events. When propagation is executed, the backpropagation events and the original events cancel each other out on the phase axis, the total propagation phase value tends to stabilize, and the propagation directionality difference significantly decreases in the hop segment statistics. This perturbation injection forms a local negative feedback structure that can absorb and neutralize residual perturbation energy without interrupting normal propagation behavior, preventing it from being transmitted to subsequent hop segments. Compared with existing delay-blocking path suppression methods, this structure can gently intervene in minor disturbances, avoid triggering full-path circuit breakers, and has a higher ability to ensure propagation continuity.

[0040] In the path segment after perturbation injection, the actual propagation data of the three hops before and after the perturbation intervention are recorded, and three dynamic indicators are extracted: hop propagation time difference, phase difference, and path jump amplitude. Comparing the data before and after the intervention, if the time difference fluctuation decreases by more than 40%, the absolute value of the phase difference decreases by more than 50%, and the path jump amplitude stabilizes within ±1, the intervention in that path segment is deemed effective. The three changed data are constructed into difference vectors and combined into a ternary perturbation correction vector, recording the path number to which the vector belongs, the intervention center node number, the start and end times of the intervention time window, and the point of maximum change amplitude. Among multiple effective correction vectors, high-frequency intervention response nodes (i.e., nodes that appear multiple times in the correction path) are extracted as the basis for identifying unstable areas in the local network state. Unlike traditional path performance evaluation based solely on mean delay or arrival rate, this step extracts perturbation correction vectors through targeted path segment responses, achieving directional quantification of local propagation state changes and ensuring that subsequent topology adjustments have a behavioral feedback basis.

[0041] After all disturbance correction vectors are aggregated, they are uniformly converted into topology adjustment vectors, which include: node pairs (representing the original propagation connections), directional correction values ​​(±1 indicates that the path needs to be reversed or maintained), propagation rate adjustment weights (calculated based on the proportion of time difference change), and path phase balance values ​​(calculated from the phase change amplitude). The topology adjustment vectors are mapped to the network path mapping table, constraining future propagation behavior along the same path: if the directional correction is negative, the initial path direction will be soft-redirected, and the priority of adjacent path selection will be reduced; if the rate adjustment weight is greater than 1.2, the hop segment waiting time threshold will be increased by 20%; if the phase balance value fluctuation remains high, delayed propagation will be set to avoid triggering disturbances that have not fully decayed. Throughout the entire network propagation cycle, the topology adjustment vectors will be continuously iterated and updated, ultimately moving the main propagation path away from the loop structure formed by the disturbance path segments and constructing a new stable propagation backbone.

[0042] This invention achieves high-precision spatiotemporal modeling and closure trend identification of diffusion paths by constructing topological phase coordinates under a unified time baseline, thus locking potential loop regions at the source. It constructs a warning information fingerprint using a three-dimensional feature set to deeply mine the path evolution patterns. Furthermore, it uses loop suspicion spectrum and causal consistency discrimination to remove redundant paths and construct a freezing mechanism to prevent repeated loop propagation of information in the network. Subsequently, it introduces one-time tokens and time watermarking to establish cross-node deduplication chains, ensuring path uniqueness at the mechanism level. Based on this, it deploys a dynamic threshold control mechanism to dynamically adjust the diffusion rhythm according to residual characteristics and path density trends, quickly converging high-risk paths and maintaining propagation structure stability. Finally, it injects anti-phase perturbation signals through a time inversion mechanism, combined with topology adjustment vectors to achieve closed-loop suppression and topology self-healing. Overall, this method possesses high path identification accuracy, perturbation suppression capability, and network adaptive adjustment capability. It not only significantly improves the accuracy and timeliness of warning information dissemination and reduces false alarm rate and system resource consumption, but also enhances the stability and robustness of multi-point linkage response to vibration events in complex environments.

[0043] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A vibration event early warning and diffusion method based on edge computing, characterized in that, Includes the following steps: S1, construct topological phase coordinates under a unified time baseline, replay the historical trajectory of early warning spread based on the topological phase coordinates, identify continuous path segments with a closing trend in the trajectory, and output suspicious loop areas; S2, on the basis of suspicious areas of the loop, construct fingerprint coding of early warning information, extract a three-dimensional feature set containing time series features, path series features and phase series features, and generate the corresponding loop suspicion spectrum; S3, based on the loop doubt spectrum, performs causal consistency judgment, strips out duplicate paths and locates the boundary region forming the loop, and constructs a one-way commitment chain to freeze the identified duplicate path segments; S4 constructs a cross-node deduplication proof chain under the constraint of a one-way commitment chain. It ensures path uniqueness by injecting one-time tokens, monotonic counters, and time watermarks, thereby eliminating the repeated propagation of early warning information. S5, based on the deduplicated diffusion path, deploys an adaptive threshold control mechanism to adjust the priority of the diffusion path according to the baseline residual, path density slope and diffusion arrival difference, to converge high-risk paths and stabilize the diffusion structure. S6 performs time-reversal diffusion control on the basis of a stable diffusion structure. It constructs an anti-phase feedback window by injecting phase-driven perturbations to absorb residual energy in the path and writes the correction results into the topology adjustment vector.

2. The vibration event early warning and dissemination method based on edge computing according to claim 1, characterized in that, Step S1 includes: After establishing a unified time baseline and topological phase coordinates, the historical trajectory of the early warning propagation process is replayed based on the topological phase coordinates. The node forwarding events are sorted in ascending order of absolute time, and trajectory segments are constructed by combining spatial path numbers and phase offsets. In the constructed trajectory segments, identify the propagation delay, spatial offset direction, and phase change of continuous path jump segments, and screen trajectory segments with propagation delay below the average level, repeated changes in spatial offset direction, and phase value decreasing to near zero as suspected return segments; In the suspicious return segment, the included angle of the trajectory line segment, the spacing of the path number, the time interval, the amplitude of the phase fluctuation and the curvature change in three-dimensional space are calculated to identify the trajectory path point set with a closing trend; The time range, path number range, and phase change interval of the identified trajectory path point set are extracted, along with the path return probability, propagation convergence speed, and closure curvature amplitude range, and the output is the suspected loop area.

3. The vibration event early warning and diffusion method based on edge computing according to claim 1, characterized in that, Step S2 includes: Based on the completed loop suspicious area, the timestamp information of the forwarding node is extracted, the time value is calculated and a time vector sequence is generated. The sliding window process is used to identify the time-dense segment to form time series features. Record the sending node number and receiving node number of each hop forwarding event to form a path number sequence, calculate the path hop sequence and node repetition rate, extract path return pattern fragments, and constitute path sequence features; The phase value is obtained by comparing the node time value with the path hop count, forming a phase sequence, and the phase convergence segment is identified to constitute the phase sequence feature; Synchronous mapping of time series features, path series features, and phase series features is performed to construct a three-dimensional feature set. The concentration of each time segment, the amplitude of path jump, and the degree of phase convergence are extracted using a sliding analysis window to generate a loop skepticism spectrum.

4. The vibration event early warning and diffusion method based on edge computing according to claim 1, characterized in that, Step S3 includes: Based on the path segments whose scores exceed the preset threshold in the loop suspicion spectrum, read the timestamp, path number, phase value and propagation direction of the nodes, determine the time inversion value, path overlap and phase back-off magnitude, identify causal abnormal jump points and confirm redundant closed loop propagation segments. Around the causal abnormal jump point in the redundant closed-loop propagation segment, extend the preceding and following nodes to form a propagation segment set, extract the path number reversal point, the time interval sudden increase point and the phase inflection point, and superimpose the results to obtain the boundary node point set, and locate the boundary start point and boundary end point. Based on the path segment between the starting point and the ending point of the boundary, a one-way commitment chain is constructed. A node number, timestamp, path number and freeze flag are added to each event node to freeze the propagation behavior within the path segment and block repeated propagation.

5. The vibration event early warning and diffusion method based on edge computing according to claim 1, characterized in that, Step S4 includes: Based on the one-way commitment chain, the source node number, path start timestamp, path number sequence and phase value are extracted to generate path digest information, and a one-time token generated by a random number generator is embedded in the propagation information; A monotonic counter is introduced in each hop of the propagation process. The initial value is set to one. The counter is incremented by one each time the propagation is carried out and the difference is recorded. If the hop is discontinuous or the increment fails, it is judged as an abnormal propagation and forwarding is stopped. Between the generation of a time watermark at the starting node of the path and the insertion of key fields within the information, the receiving node extracts the watermark and compares it with the locally reconstructed watermark to verify the consistency of propagation. After each hop of propagation, a one-time token, a counter value, a receiving timestamp, a path number, and a time watermark are encapsulated to form an event unit, which is then appended to the event chain to achieve propagation path tracking. At the end of the propagation, a path status identifier is generated and written into the frozen path list to establish a dual-index structure to block duplicate paths.

6. The vibration event early warning and diffusion method based on edge computing according to claim 1, characterized in that, Step S5 includes: A propagation baseline residual measurement model is constructed. By calculating the time difference between the actual propagation time and the standard propagation model, the residual vector is extracted and statistical analysis is performed to identify the rhythm disorder path. Based on the network spatial distribution of propagation paths, a path density evolution model is constructed and the density slope is calculated to identify highly clustered path segments with continuously increasing density and abnormal jump rates. By integrating the path residual and density slope analysis results, the hop-level diffusion arrival difference is extracted, a ternary feature fusion matrix is ​​constructed, hop segments are prioritized and propagation thresholds are dynamically set.

7. The vibration event early warning and diffusion method based on edge computing according to claim 6, characterized in that, During segment propagation, propagation is only allowed when the residual regression value is lower than the baseline standard, the local path density does not exceed the set upper limit, and the diffusion reaches the differential within the steady-state range. Otherwise, execution is delayed to ensure that the diffusion behavior is controlled in high-risk path segments.

8. The vibration event early warning and diffusion method based on edge computing according to claim 1, characterized in that, Step S6 includes: Extract segment propagation events to construct a propagation behavior set, identify path segments with abnormal propagation rhythm, and calculate the disturbance intensity value; The intervention window is set based on the disturbance intensity value, and a phase-driven micro-perturbation event is injected to form a local negative feedback structure; Collect jump propagation indicators before and after intervention to construct perturbation correction vectors and identify the changing trends of response nodes and paths; The perturbation correction vector is uniformly mapped to the topology adjustment vector, and the path direction, rate and propagation priority are dynamically adjusted.

Images