A method and system for processing vibration and impact data

Abstract

This invention relates to the field of industrial data processing technology, and discloses a method and system for processing vibration data and shock data. The method includes: a sensor determining a dynamic floating threshold based on neighbor vibration summaries; initiating a local consensus challenge when the shock exceeds the threshold; if the challenge fails, querying the survival status of physically non-adjacent witness sensors, and performing final arbitration and differentiated data routing based on the query results to assess the source reliability of the shock event. This invention, by establishing a pre-emptive source reliability arbitration process at the data source, transforms the basis for judging data reliability from passive trust in individual sensors to active verification of sensor community consensus and external witnesses. This not only suppresses false data caused by individual faults but also prevents the erroneous filtering of real shocks due to local cluster failures, avoiding the risks of data contamination and loss of critical information.

Description

Technical Field

[0001] This invention relates to a method and system for processing vibration data and impact data, belonging to the field of industrial data processing technology. Background Technology

[0002] Currently, in the application of health monitoring and fault diagnosis of large-scale equipment, the deployment of large-scale sensor networks to obtain high-density equipment status data is a common technical approach in the industry. The technical basis of this approach lies in the fact that the core of the system architecture design is to ensure the maximum acquisition of front-end data and the complete transmission to the back-end analysis platform, so as to support subsequent accurate diagnosis with comprehensive data. At the same time, there is an objective physical law about signal propagation: high-energy impact events originating from the actual physical process of the equipment have spatial correlation when their energy is conducted in the medium, and will be jointly sensed by a group of sensors in the vicinity of its physical location; while abnormal signals generated by the occasional failure of a single sensor or its data link are characterized by the isolation of individual information generation.

[0003] As the scale of sensor deployment expands, the mismatch between the aforementioned technical methods and objective physical laws creates an inherent limitation in system operation. In the initial stage of data reception, the system gives equal processing priority to all data uploaded by sensors before verification, failing to effectively distinguish the credibility of data sources. This architecture treats community signals carrying real physical information and isolated noise generated by a single fault point equally in terms of information value. This results in a large number of false impact signals generated by isolated sensor faults being received and processed by the system, increasing the computational load of the backend analysis system and affecting the accuracy of diagnostic results. To solve this problem, a common improvement is to enhance the data cleaning algorithm of the backend platform, attempting to filter false signals through more complex calculations. However, this approach places the data quality verification stage at the end of the data processing chain, causing the verification behavior to be separated from its source in time and space, increasing processing latency and resource consumption, and failing to solve the problem of data pollution occupying the frontend communication bandwidth.

[0004] Analysis reveals the following inherent limitations in existing technologies: 1. Data quality control is reactive, passively filtering data after it enters the system, lacking an active verification mechanism at the data source before formal acceptance; 2. System reliability relies on the individual health status of each sensor, lacking a method to leverage the collaborative relationships of sensor clusters to improve the overall robustness of the data acquisition system. These architectural limitations are also reflected in specific data processing control methods. Even in multi-sensor systems, inherent flaws in their control logic make it difficult to effectively distinguish between real physical events and individual sensor malfunctions. For example, the patent with authorization announcement number CN104657518B... A Chinese invention patent discloses a method and system for processing vibration and impact data. While this approach uses two independent sensors for vibration and impact, its core logic involves a backend microcontroller classifying the data after collection based on fixed, pre-set thresholds and time difference rules. The fundamental flaw of this approach is that it passively receives all data, assuming the data source is reliable and relying entirely on backend analysis based on fixed thresholds. It lacks a mechanism to verify the signal's reliability at the data generation source. It cannot distinguish between community-based impacts originating from the actual physical processes of the equipment and isolated impacts from individual sensor circuit failures, leading to extremely high false alarm rates and invalid data transmission, thus failing to address the root cause of the problem. Therefore, the technical problem this invention aims to solve is how to construct a lightweight pre-processing workflow that utilizes the location correlation of sensor clusters and the physical laws of signal propagation to effectively arbitrate the reliability of the data source before reporting, thereby suppressing the impact of low-quality data on big data analysis systems. Summary of the Invention

[0005] This invention provides a method for processing vibration and impact data. Its main purpose is to solve the problem that existing industrial big data systems, due to the lack of a credibility arbitration mechanism for front-end data sources, can cause false impact signals generated by a single sensor failure to directly affect the entire data system.

[0006] To achieve the above objectives, the present invention provides a method for processing vibration data and impact data, comprising the following steps: Step a: The first sensor in the sensor network periodically sends a proof of life containing the identity information of itself and one or more neighboring second sensors to one or more physically non-adjacent witness sensors. Step b: The first sensor determines a real-time noise baseline characterizing local environmental noise by calculating a weighted moving average of the energy values ​​of the vibration state summaries broadcast by one or more neighboring second sensors within a preset historical time window, based on the periodic vibration state summaries it has historically received from one or more neighboring second sensors. Step c: When the first sensor detects an impact event whose energy exceeds a floating threshold dynamically adjusted by a real-time noise baseline, a consensus challenge is initiated to one or more second sensors. Step d: If a consensus confirmation is received from at least one second sensor, the source credibility of the impact event is arbitrated as credible. Step e: If no consensus confirmation is received from any second sensor, a query is initiated to the witness sensor to inquire whether a survival certificate issued by the second sensor has been continuously received within a preset time window of five minutes. Based on the feedback obtained from the witness sensor, a final arbitration is conducted. The rules for the final arbitration are as follows: if the feedback indicates that the second sensor has lost contact, the source credibility of the impact event is arbitrated as trustworthy; if the feedback indicates that the second sensor is still alive, it is arbitrated as untrustworthy. Step f involves performing differentiated data routing on the raw data of the shock event based on whether the arbitration result is credible or uncredible.

[0007] Preferably, the calculation steps for the vibration state summary include: acquiring an original vibration signal sequence and calculating the cumulative sum of the squares of the amplitudes of the sampling points in the sequence to obtain an original energy value; directly mapping the calculated original energy value to an intermediate value after logarithmic compression through a lookup table preset in the local memory; and quantizing the intermediate value into an unsigned integer through a linear scaling transformation as the vibration state summary.

[0008] Preferably, the differentiated data routing includes: when the source confidence of the impact event is arbitrated as credible, authorizing the reporting of the raw waveform data of the impact event with a high-precision timestamp; and when the source confidence of the impact event is arbitrated as untrustworthy, suppressing the reporting of the raw data of the impact event and reporting a self-diagnostic log, the self-diagnostic log containing the event timestamp, sensor ID, and an event code marked as an isolated high-energy event.

[0009] Preferably, when the source credibility of the impact event is arbitrated as credible, the step of authorizing the reporting of the original data of the impact event includes authorizing the first sensor and the second sensor that makes a consensus confirmation to report the original data of the impact event.

[0010] Preferably, the method further includes: recording the arrival times of the raw data reported by the first sensor and the second sensor that has made a consensus confirmation; and using a time difference of arrival (TDOA) localization algorithm to determine the physical source location of the impact event based on the time difference of arrival calculated by the second sensor that has made a consensus confirmation relative to the arrival time of the first sensor, and the known physical locations of the first and second sensors.

[0011] Preferably, a floating threshold The calculation rules are as follows: in, As a real-time noise baseline, The method includes, for a pre-defined tolerance value, the step of arbitrating the source reliability of the impact event as unreliable based on feedback indicating that the second sensor is still alive, further comprising: acquiring a vibration state summary of one or more second sensors below a floating threshold during the impact event; determining a real-time attenuation gradient characterizing the energy decay characteristics in space based on the impact event energy of the first sensor and the vibration state summary energy of the second sensor; and arbitrating the source reliability of the impact event as reliable again by comparing the real-time attenuation gradient with a pre-defined theoretical attenuation model, wherein if there is a nonlinear deviation between the real-time attenuation gradient and the theoretical attenuation model.

[0012] Preferably, it also includes: aggregating self-diagnostic logs from multiple sensors in the cloud, and periodic vibration state summaries attached when the source credibility of an impact event is arbitrated as credible and authorized for reporting; generating one or more macroscopic statistical indicators characterizing the overall health status of the sensor network based on the aggregated self-diagnostic logs or vibration state summaries; and identifying a chronic, systemic change in device health status by analyzing the patterns of one or more macroscopic statistical indicators over time.

[0013] Preferably, the macroscopic statistical indicators include: the background noise baseline represented by the moving average of all vibration state summary energy values, the volatility index represented by the standard deviation of all vibration state summary energy values, and the information entropy in space and time calculated based on the sensor ID and timestamp generated by the self-diagnostic log, which is the fault entropy.

[0014] Preferably, the step of arbitrating the source credibility of the impact event as credible further includes: the first sensor determining a stability index of a communication link with each of the second sensors that made consensus confirmations based on the arrival time of the periodic heartbeat signals it historically received from the second sensors that made consensus confirmations; and weighting the consensus confirmations received from the second sensors based on the stability index to calculate a weighted total score; and comparing the weighted total score with a preset weighted confirmation threshold, wherein the source credibility of the impact event is arbitrated as credible only when the weighted total score is greater than the threshold.

[0015] A vibration data and shock data processing system, the system comprising: A sensor network, comprising a first sensor, one or more second sensors, and one or more witness sensors; wherein the first sensor includes: A communication module is configured to periodically send a proof of survival containing its own identity information and that of one or more second sensors to one or more witness sensors, initiate a consensus challenge to one or more second sensors when an impact event is detected, and initiate a query to the witness sensors when no consensus confirmation is received. A baseline determination module is configured to determine a real-time noise baseline characterizing local environmental noise based on a summary of periodic vibration states received from one or more second sensors in the past. An event monitoring module is configured to trigger a consensus challenge by the communication module when an impact event is detected whose energy exceeds a floating threshold dynamically adjusted by a real-time noise baseline. A source credibility arbitration module is configured to: arbitrate the source credibility of the shock event as credible if a consensus confirmation is received from at least one second sensor; and perform final arbitration based on feedback from witness sensors regarding the survival status of the second sensors if no consensus confirmation is received from any second sensor; and a data routing module is configured to perform differentiated data routing on the raw data of the shock event based on the result of the source credibility arbitration module arbitrating it as credible or untrustworthy.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention discloses a method for processing vibration and impact data. When the first sensor detects an impact event, instead of immediately reporting the raw data, it first initiates a consensus challenge to a nearby second sensor. The reliability of the event's source is arbitrated based on whether a consensus confirmation is received. Finally, differentiated data routing is performed on the raw data based on the arbitration result. This combination of technical features establishes a front-end data acceptance and arbitration process at the entry point of the industrial data system. It moves the data quality control node from the back-end analysis platform to the source of data generation. This transforms the data system's judgment of information reliability from passive trust in the physical state of a single sensor to active verification of the local consensus formed by the sensor cluster at the moment of the event. From the system architecture level, this avoids data contamination caused by a single sensor failure from entering the main data processing flow.

[0017] Furthermore, during the execution of the above process, the historical reception of periodic vibration state summaries from the adjacent second sensor by the first sensor is used to determine a real-time baseline characterizing local environmental noise. A floating threshold dynamically adjusted by this real-time baseline is then used to determine whether an impact event has been triggered. This approach reuses the periodic communication data used for consensus challenge as a sensing input for environmental state, coupling the impact event triggering judgment mechanism with the consensus arbitration mechanism. As a result, the system acquires an operating characteristic that adaptively adjusts its detection sensitivity according to changes in the local working environment. Under conditions of severe background noise, the system automatically raises the judgment threshold to filter environmental interference, while under quiet conditions, it automatically lowers the threshold to capture weak signals. This achieves a dynamic balance between detection sensitivity and anti-interference robustness in complex and ever-changing industrial environments.

[0018] Furthermore, in specific cases where the first sensor fails to receive any consensus confirmation, the arbitration step can further acquire a summary of the vibration state of one or more second sensors below the impact threshold during the impact event. Based on the impact data from the first sensor and these low-energy summaries, a real-time gradient of energy decay in space is determined. This real-time gradient is then compared with a preset theoretical decay model to finally arbitrate the source reliability of the impact event. This method of transforming consensus failure events into secondary in-depth investigation trigger conditions enables the system to not only process confirmed events but also analyze unconfirmed events. It utilizes the failure process of a first consensus interaction to reverse-examine the propagation characteristics of impact energy in the physical medium, thereby enabling the system to distinguish between two different fault modes: signal source failure and abnormal signal propagation path, providing a deeper physical insight for equipment diagnosis. Attached Figure Description

[0019] Figure 1This is a schematic diagram of the source credibility arbitration process based on community consensus and external witnessing in this invention; Figure 2 This is a graph showing the evolution trend of macroscopic indicators of the health status of the system of the present invention over time; Figure 3 This is a diagram of a hierarchical system architecture for edge-side collaborative processing and cloud-based analysis of a sensor network, as described in this invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0021] This invention discloses a method and system for processing vibration and shock data, applied to a steam turbine generator diagnostic monitoring and management system. The system mainly consists of a sensor network deployed at key measuring points of the steam turbine generator set, and a cloud platform for data aggregation and analysis. Within a local area, the sensor network includes a first sensor acting as the event trigger and arbitrator, one or more second sensors surrounding it forming a consensus community, and one or more witness sensors not physically adjacent to this community. The core of this method lies in constructing a data source reliability arbitration and data routing mechanism based on front-end consensus verification, executed at the edge of the sensor network before the data enters the cloud platform. In a specific application scenario, the sensor network is deployed in the high-pressure cylinder bearing housing area of ​​a 600MW supercritical steam turbine generator set. The first sensor and four second sensors together form a monitoring community, with physical spacing between them ranging from 0.5 meters to 1.5 meters. One sensor is located at the low pressure of the unit... The sensors in the cylinder pressure area are assigned by the system as witness sensors for the community to avoid community-wide communication interruptions caused by failures in the local power supply module or wireless communication gateway of the high-pressure cylinder area. All sensors are equipped with triaxial accelerometers with a range of ±50g and a sampling frequency set to 25.6kHz. They also integrate a microprocessor with integer arithmetic capabilities and a wireless communication module supporting low-power wide-area network communication protocols. After the system enters steady-state operation, the processing steps are initiated. As step a, the first sensor sends a data packet as a liveness certificate to the assigned witness sensor through its communication module every 60 seconds. The liveness certificate data structure contains three fields: the ID of the sensor that initiated the certificate, a list of IDs of all second sensors that it has listened to in the past 60 seconds, and a local timestamp that generated the certificate. After receiving the data packet, the witness sensor updates an internal state table to record the community's active status and the latest communication time.

[0022] To ensure a defined operational baseline for the method during deployment, the roles of each sensor within the system—namely, the first sensor acting as the arbitrator, the second sensor acting as the consensus community, and one or more witness sensors acting as external verification nodes—are pre-configured during the initial deployment phase and stored in the non-volatile storage units of each sensor node. The number of witness sensors deployed can be configured based on the system's reliability requirements: deploying one witness sensor is the minimum configuration for external verification; deploying multiple physically distributed witness sensors further enhances the robustness of the external verification mechanism. Simultaneously, to support subsequent impact source localization calculations based on time difference of arrival, all nodes in the sensor network achieve timestamp synchronization through a locally running lightweight time synchronization procedure based on the IEEE 1588 protocol. This procedure ensures that the local clock deviation between any two sensor nodes is continuously calibrated to within 5 microseconds. This time synchronization accuracy is crucial for subsequent... The hyperboloid positioning algorithm is then used to process the signals received by sensors several meters apart, achieving meter-level positioning accuracy. Simultaneously, to address the fluctuations in background vibration noise of the steam turbine generator under different loads, an impact detection threshold that can adaptively adjust with the environment is needed. Therefore, as step b, the baseline determination module inside the first sensor is configured to continuously monitor and receive periodic vibration state summaries from all surrounding second sensors. This vibration state summary is a value obtained by accumulating the squares of the vibration signal amplitudes collected by the sensors within a 100-millisecond time window, taking the logarithm, and quantizing it into an 8-bit unsigned integer. This calculation uses integer operations to reduce computational overhead. The baseline determination module maintains a queue of length 600 in its local memory to store the latest summary values ​​received from all second sensors in the past minute, and calculates a weighted moving average of these summary values ​​every second, thereby obtaining a real-time noise baseline that characterizes the current local environmental noise. The dynamic change of this baseline value reflects the fluctuations in the overall vibration level of the steam turbine generator caused by the adjustment of operating conditions.

[0023] The calculation procedure for the vibration state summary is set as a fixed multi-step process. The sensor first acquires a raw vibration signal sequence with a duration of 100 milliseconds, and then calculates the sum of the squares of the amplitudes of all sampling points in the sequence to obtain a raw energy value. This value ,in For the first The signal amplitude at each sampling point, where N is the total number of sampling points within a 100-millisecond time window; subsequently, the processor uses a lookup table pre-installed in local memory to retrieve the calculated raw energy value. The direct mapping is represented by a logarithmically compressed intermediate value. This lookup table pre-stores the correspondence between energy values ​​and logarithmic values ​​to avoid performing floating-point logarithmic operations on the microprocessor. Finally, the intermediate value is quantized into a one-bit unsigned integer ranging from 0 to 255 through a linear scaling transformation, serving as the final output vibrational state summary. During this transformation process, any value exceeding the mapping range is set to 255. When this condition is met, the system proceeds to step c. This detection condition is not based on a fixed impact threshold, but rather on a real-time noise baseline. Dynamically adjusted floating threshold Floating threshold The determination follows a standardized offline calibration procedure. During the initial commissioning phase of the unit, the real-time noise baseline of the unit is recorded under healthy conditions, across the entire operating range from no-load to full-load. Historical data was used, and a known energy was injected to simulate the impact signal of early bearing pitting, by gradually adjusting a value beyond the tolerance. This makes the floating threshold The impact signal can be detected under all operating conditions, while minimizing false triggering due to background noise, ultimately ensuring the detection of this signal. The floating threshold is stored in the non-volatile memory of the sensor node. The calculation rules are set as follows: in, For floating thresholds, As a real-time noise baseline, For a pre-defined tolerance value; for example, the real-time noise baseline measured by the system at a certain moment. The tolerance was set to 150 quantization units. If the value is 100 units, then the floating threshold at this moment is 250 units. When the event monitoring module inside the first sensor detects an impact event with an instantaneous vibration energy value of 260 units, the energy of the impact event exceeds the floating threshold dynamically adjusted by the real-time noise baseline. The event monitoring module will immediately trigger the communication module to initiate a consensus challenge to all its neighboring second sensors. The consensus challenge data packet contains the event ID, the trigger timestamp, and the impact energy summary of the triggering node itself. After receiving the challenge, the surrounding second sensors check the vibration state summary recorded within the same time window. If their energy values ​​also exceed their locally calculated floating thresholds, they reply with a consensus confirmation to the first sensor. The system's source reliability arbitration module then starts arbitration. In step d, if the first sensor receives a consensus confirmation from at least one second sensor within a 500-millisecond waiting window, the source reliability arbitration module directly arbitrates the source reliability of the impact event as reliable.

[0024] In the process of arbitrating a credible link, a weighted processing procedure based on communication link stability can be introduced. The first sensor, based on the arrival times of periodic heartbeat signals it has historically received from the second sensor that has made consensus confirmations, calculates the standard deviation of these arrival time intervals to determine an index characterizing the stability of the point-to-point communication link, namely the link jitter index. A smaller standard deviation corresponds to a more stable link and a higher weight. When the source credibility arbitration module receives multiple consensus confirmations, it performs a weighted summation of each confirmation according to the weight of its source link. Only when the total weighted score exceeds a preset threshold is the event finally arbitrated as credible. This reduces the impact of confirmation information from unstable links caused by signal interference on the arbitration result. Impact; Stability index, namely link jitter index, is calculated as follows: The first sensor maintains a timestamp queue of maximum length 20 for each neighboring second sensor locally, storing the arrival timestamps of the most recent 20 received periodic heartbeat signals. The queue is updated each time a new heartbeat signal is received. The system is configured to calculate the difference between all valid adjacent timestamps in the queue in real time, resulting in a sequence of time intervals of variable length. To ensure statistical validity, this sequence must contain at least two time interval values. The link jitter index is then determined as the standard deviation of this time interval sequence; a smaller standard deviation indicates a more stable communication link. During weighted processing, each consensus-confirmed weight... It is set as its corresponding link jitter index The reciprocal of, that is The source credibility arbitration module calculates the weighted total score of all received consensus confirmations. It is then compared to a preset weighted confirmation threshold (e.g., 1.5 times the highest single-link weight), and only when... Only when the value exceeds this threshold is the event ultimately arbitrated as trustworthy. This procedure reduces the impact of consensus confirmations from links with poor signal quality on the final arbitration result. If no second sensor responds to the consensus challenge within a preset time, the system proceeds to step e. At this point, the source credibility arbitration module does not immediately determine the event as untrustworthy, but instead initiates a secondary arbitration process. It first sends an emergency query to the witness sensor pre-defined in step a, querying whether it has continuously received liveness certificates from the designated second sensor (e.g., IDs A, B, C) within the past five-minute preset time window. The five-minute duration setting is consistent with the liveness certificate in step a. The transmission period (60 seconds) is directly related, and this window duration is set to five times the transmission period to ensure that the system can tolerate several (e.g., four) consecutive packet losses due to transient network fluctuations. The witness sensor responds according to its internally maintained state table. The final arbitration rule here is set as follows: if the witness sensor feedback indicates that one or more second sensors are out of service due to partial power supply or communication failure, the system arbitrates the source credibility of this impact event as credible; conversely, if the witness sensor feedback indicates that all second sensors are still alive, the impact event is arbitrated as unreliable and determined to be an isolated signal perceived only by the first sensor itself.

[0025] In the step where feedback indicates the second sensor is still alive, thus arbitrating the source reliability of the impact event as unreliable, to further distinguish between sensor failure and abnormal signal propagation path in the physical medium, the system can perform a second arbitration of channel health. The first sensor will again request all unresponsive second sensors to obtain precise values ​​of the vibration state summaries recorded during the impact event but with energy below the floating threshold. Based on the impact energy value of the first sensor and the low-energy summary values ​​of each second sensor, combined with the known physical distance of each sensor, the system locally calculates a real-time attenuation gradient characterizing the attenuation characteristics of impact energy in space. By comparing this real-time attenuation gradient with a theoretical attenuation model pre-calibrated under device health conditions, if there is a nonlinear deviation between the two, such as energy attenuation in a certain direction far exceeding the theoretical value, the system reclassifies the nature of the event as a high-reliability physical event under channel anomaly, and considers this phenomenon to indicate an abnormality in the structure between the sensor and its neighbors. The theoretical attenuation model may contain defects such as cracks. To ensure that the model can adapt to structural changes that occur during long-term operation (e.g., due to normal wear and tear or component replacement), it is not statically fixed but designed to be linked with the macroscopic health status assessment mechanism of the cloud platform. The cloud platform is configured to continuously track macroscopic statistical indicators characterizing the overall health status of the sensor network, as described below. When the cloud platform identifies a chronic change in the health status of the device by analyzing the evolution pattern of the macroscopic statistical indicators, such as a permanent drift in the background noise baseline, the system will be triggered to perform a recalibration of the theoretical attenuation model. This will ensure that the model benchmark matches the current health status of the device, avoiding re-arbitration failure due to model obsolescence. Finally, as step f, the system's data routing module performs differentiated data routing on the original data of the impact event based on the final arbitration result of the source reliability arbitration module, determining whether it is reliable or unreliable. When the source reliability is arbitrated as unreliable, the data routing module will suppress 25% of the impact event.Instead of reporting raw waveform data at the full sampling rate of 6kHz, only one self-diagnostic log is reported. This log contains an event timestamp, sensor ID, and an event code marking an isolated high-energy event, thus cutting off data contamination at the source. When the source credibility is arbitrated as trustworthy, the data routing module authorizes the first sensor and all second sensors that have reached a consensus to report their synchronously collected raw waveform data with high-precision timestamps to the cloud platform. On the cloud platform, the system can also use this authorized data from multiple perspectives to perform physical source localization of the impact event. The system first records the arrival time of the raw data reported by the first sensor and the second sensors that have reached a consensus. Based on the difference between these arrival times synchronized through a high-precision network time protocol (Time Difference of Arrival, TDOA), and the known three-dimensional physical coordinates of each sensor at deployment, the system uses the Chan algorithm or hyperboloid localization algorithm to calculate and ultimately determine the physical source location of the impact event within the turbine generator set structure, providing spatial guidance for subsequent fault diagnosis and maintenance.

[0026] In a specific location calculation example, after a community-based shock event is arbitrated as credible, the cloud platform receives a data packet from the first sensor (physical coordinates [10, 10, 0]) at 14:30:05.125000 UTC, a data packet from the second sensor A (physical coordinates [10, 20, 0]) that made the consensus confirmation at 14:30:05.125800, and a data packet from the second sensor B (physical coordinates [20, 10, 5]) at 14:30:05.126100. The system first calculates the time difference of arrival, obtaining two independent time difference measurements: the time difference of arrival of the second sensor A relative to the first sensor is +0.8 ms, and the time difference of arrival of the second sensor B relative to the first sensor is +1.1 ms. Assuming that the shock wave propagates at a speed of 5000 m / s in the structural medium of the steam turbine generator set, the system constructs a hyperboloid equation with the corresponding two sensors as foci based on each time difference value. By solving for the intersection of these two hyperboloid equations with the geometric model of the unit structure (Z≥0), the physical source coordinates of the impact event can be obtained in reverse analysis, for example [13.5, 14.2, 2.1]. In addition, to detect chronic systemic failures, this invention also includes a process for performing condition assessment using system operation byproducts in the cloud. The cloud platform will aggregate self-diagnostic logs from all sensors in the sensor network over a long period of time, as well as periodic vibration state summaries attached when authorized to report. Based on these aggregated data, the system generates and continuously tracks three statistical indicators characterizing the health status of the entire turbine generator set: the background noise baseline represented by the moving average of the energy values ​​of all vibration state summaries, the volatility index represented by the standard deviation of the energy values ​​of all vibration state summaries, and the fault entropy represented by the spatial and temporal distribution uniformity of the self-diagnostic logs. By analyzing the patterns of these one or more statistical indicators over time, the system can identify a chronic, systemic change in equipment health status.

[0027] Example 1: This embodiment is a specific operational example in a particular industrial application scenario. In a 600MW supercritical steam turbine generator unit operating at full load continuously, the No. 2 bearing housing of the high-pressure cylinder is designated as the measuring point of the first sensor. Due to a transient electronic noise interference in its internal analog-to-digital conversion circuit, its processor calculates a high-energy impact signal within a 0.2-millisecond time window. The characteristics of this signal are similar in morphology to the fault signal of bearing outer ring spalling. At this moment of system operation, because the unit is under full load, the overall background vibration level is high. The baseline determination module inside the first sensor has determined the real-time noise baseline based on the continuously received periodic vibration state summary of the adjacent second sensor. The number of quantization units is dynamically increased to 180, and then calculated according to the calculation rules described in the specific implementation method. ,in Set to 100, this is the floating threshold used to trigger the impact event detection. The quantization unit was set to 280; the signal generated by electronic noise interference had an energy value calculated to be 310 units, which exceeded the floating threshold under the current operating conditions. Therefore, the event monitoring module determines that an impact event has been triggered and immediately initiates a consensus challenge to the four surrounding second sensors via the communication module. Within a preset time window of 500 milliseconds after issuing the consensus challenge, the first sensor does not receive a consensus confirmation from any of the second sensors. This is because the root cause of the impact event lies in the circuitry of the first sensor itself, and the impact energy has not propagated through the physical structure of the unit. Therefore, the vibration state summaries detected by all second sensors within this time window are all below their respective local floating thresholds. It should be noted that the floating threshold... The settings enable the system to capture signals that exceed normal fluctuations even in a noisy environment, and the subsequent consensus challenge steps provide cross-validation of the source of the captured signal.

[0028] Given the lack of consensus confirmation, the source reliability arbitration module of the first sensor immediately initiated a secondary arbitration process, sending a query to the witness sensor located in the low-pressure cylinder area regarding the survival status of the local community. Upon receiving feedback from the witness sensor that all second sensors had maintained normal communication over the past five minutes, the source reliability arbitration module, based on preset final arbitration rules, arbitrated the source reliability of this impact event as unreliable. This process, by introducing an independent external observer, provides a criterion for distinguishing between individual sensor failures and local sensor community communication interruptions. Finally, based on the arbitration result of unreliable source reliability, the data routing module performed differential suppression of reporting on the raw data of this impact event. Data routing: The original waveform data describing the false impact was retained locally, without consuming wireless communication bandwidth or entering the cloud platform's analysis system. Instead, a self-diagnostic log containing only a timestamp, sensor ID, and isolated high-energy event code was reported to the turbine generator diagnostic monitoring and management system. This process places the judgment of data credibility at the edge of data acquisition, preventing invalid data from entering subsequent processing stages. The system's maintenance terminal interface did not trigger an emergency alarm regarding the unit's No. 2 bearing, and the turbine generator unit continued to operate stably. Meanwhile, the equipment health management module of the diagnostic monitoring and management system automatically generated a low-priority inspection work order for the first sensor, prompting maintenance personnel to inspect or replace it during the next routine maintenance.

[0029] Example 2: To objectively verify the technical effectiveness of the method claimed in this invention in suppressing spurious impact signals caused by individual sensor malfunctions and in reducing invalid data transmission, this embodiment constructs a hardware-in-the-loop simulation test platform for turbine vibration signals. This platform consists of a multi-channel arbitrary waveform generator and an embedded computing system running data acquisition and processing logic. The arbitrary waveform generator is used to simulate the vibration signals received by a monitoring community of five sensors deployed in the bearing housing area of ​​a turbine generator set. Its functional specifications include the ability to generate background noise with an adjustable signal-to-noise ratio of -5dB to 20dB and a bandwidth of 0Hz to 50kHz, and to superimpose two types of impact event signals onto this background noise. Two independent test sets are set up: one set is a control using existing data acquisition methods. The control group consists of two groups: a control group and an experimental group, which employs the method claimed in this invention. In the control group, all five sensors use a fixed energy threshold for impact judgment, and the original waveform data is immediately reported once the signal exceeds the threshold. In the experimental group, the five sensors fully execute the entire processing flow of the specific implementation method, including real-time noise baseline determination, floating threshold judgment, consensus challenge, source credibility arbitration, and differentiated data routing. The impact event signals injected in the experiment are divided into two categories: the first category is a community impact event simulating physical impact, in which the signal is synchronously applied to all five sensor channels, but the energy is graded according to the propagation attenuation model; the second category is an isolated impact event simulating a single point failure of the sensor, in which the signal is applied to only one sensor channel in the community.

[0030] The core parameters of the experiment were set as follows: the total experiment duration was set to 3600 seconds to ensure that the number of various randomly injected events reached statistical stability while balancing the execution efficiency of the experiment; in each round of the experiment, the total number of injections of both community-based and isolated impact events was set to 1000, and their distribution on the time axis followed a Poisson distribution to simulate the randomness of events occurring in an industrial environment; the energy amplitude of the events was randomly distributed between 1.1 and 5 times the floating threshold to cover various situations from weak impacts to impacts; during the experiment, when the five sensors in the control group received an isolated impact event signal, they judged the event as a real impact because its energy exceeded a fixed judgment threshold and uploaded complete raw waveform data to the backend; the first sensor in the sample group of this invention received an isolated impact event signal and exceeded its locally calculated floating threshold. Afterwards, a consensus challenge was immediately initiated. However, since the other four second sensors did not receive the signal, no consensus confirmation was received within the 500-millisecond waiting window. The source credibility arbitration module then initiated a secondary arbitration and, after confirming with the witness sensor that the neighbor was alive, arbitrated the source credibility of the event as untrustworthy. Finally, the data routing module suppressed the reporting of the original data and only reported one self-diagnostic log. For community impact events, both groups were able to report data, but the sample group of this invention reported data collaboratively by multiple sensors after consensus confirmation. Table 1 shows the average statistical data recorded after 10 consecutive rounds of experiments, used to compare the system behavior of the two methods when handling different types of events.

[0031] Table 1: Data table of comparative test on system behavior.

[0032]

[0033] Referring to Table 1, the data shows that the number of system alarms in the control group was the same as the total number of the two types of events it received, including 1000 false alarms caused by isolated impact events. In the present invention's sample group, only 2 out of 1000 isolated impact events were ultimately arbitrated as credible and triggered system alarms, achieving a false alarm suppression rate of 99.8%. The data in the "Number of Arbitrated as Credible" column presents the intermediate processing results of the source credibility arbitration module. It should be noted that when processing community-based impacts, the present invention's sample group had a higher total data traffic than the control group due to multiple sensor reports, but its data traffic was close to zero when processing isolated impacts. Considering both types of events, the total data traffic of the present invention's sample group was lower than that of the control group. The experimental results confirm that the method claimed in this invention can distinguish between community-based impacts originating from physical processes and isolated impacts originating from single-point faults at the front end of data acquisition, reducing the proportion of false alarms caused by isolated impacts and reducing the consumption of network bandwidth and backend storage and computing resources due to invalid signal reports.

[0034] To more systematically and quantitatively verify the specific technical effects of the method claimed in this invention in solving the aforementioned technical problems, and to make a direct comparison with the prior art, the following comparative examples and test cases are provided.

[0035] Comparative Example 1: This comparative example aims to simulate and verify the technical effect of a conventional data processing method in the prior art that uses a fixed threshold for judging impact events. To ensure the fairness and effectiveness of the comparison, this comparative example uses the same turbine vibration signal hardware-in-the-loop simulation test platform as described in Embodiment 2 of this invention. All hardware configurations, background noise signal models, and energy distribution ranges injected by impact events are consistent with Embodiment 2. The only essential difference between the method described in this comparative example and the method claimed in this invention is that this comparative example does not include any source credibility arbitration process based on community consensus and external witnesses executed at the edge of the sensor network. Instead, it adopts an independent reporting processing mechanism based on a fixed energy threshold. In this processing mechanism, each sensor in the simulation platform is configured to operate independently, and its impact event judgment is based on a fixed energy threshold. The value was set to 250 quantization units. This value was determined based on the 99.9th percentile of the historical background noise statistics of the steam turbine generator set within the range of 50% to 80% of the rated load. The aim was to effectively capture real physical impact signals under most operating conditions, while avoiding false triggering caused by normal background noise fluctuations to a certain extent. When the instantaneous value of vibration energy detected by any sensor exceeds this fixed threshold, the sensor is configured to immediately judge the event as a real impact and directly report the data packet containing the original waveform data before and after the impact event to the diagnostic monitoring and management system on the cloud platform. The cloud platform will then perform subsequent analysis and alarm. The test process is as follows: During the test duration of 3600 seconds, 1000 independent isolated impact events, designed to simulate transient electronic noise interference or data link anomalies of the sensor itself, were randomly injected into a sensor channel of the simulation platform.

[0036] Table 2: Comparative Example 1 Experimental Data Table.

[0037]

[0038] The experimental results show that when faced with 1000 isolated impact events, because the energy value of each impact exceeded the preset fixed threshold of 250 quantization units, the tested sensors all judged them as real impact events and reported the raw data. Ultimately, the cloud platform's diagnostic monitoring and management system received 1000 independent impact event reports and triggered 1000 system alarms accordingly. This process generated a total of 488.3 megabytes (MB) of invalid data traffic. Subsequent analysis confirmed that all 1000 alarms were caused by simulated single-point faults in the sensors. The false alarms, not originating from actual physical changes in the turbine generator set, constitute 100% false alarms. These numerous false alarms and invalid data increase the computational load and storage overhead of the backend analysis system and severely interfere with the fault diagnosis of maintenance personnel. The results of this comparative experiment confirm that the conventional technical approach based on fixed thresholds and independent reporting, in its system architecture design principle, is indeed unable to effectively distinguish between community-based impacts originating from real physical processes and isolated impacts originating from individual sensor failures, thus leading to an extremely high false alarm rate and waste of network resources.

[0039] Example 3: This embodiment combines Figures 1 to 3 A description of a method and system for processing vibration and shock data, such as... Figure 1 As shown, when a first sensor, acting as the arbitrator, detects an impact event, it does not directly report the data. Instead, it initiates a consensus challenge to one or more second sensors, which are its neighboring communities. Simultaneously, the first sensor periodically sends a survival certificate to a physically non-adjacent witness sensor. The source credibility arbitration module performs final arbitration on the source credibility of the impact event based on whether it receives consensus confirmation from the second sensor, or, if no confirmation is received, based on feedback from the witness sensor regarding the survival status of the second sensor. Based on the arbitration result, it performs differentiated data routing to the cloud platform. That is, when the arbitration is credible, the first and second sensors jointly report the original waveform data, while when the arbitration is untrustworthy, only the first sensor reports a self-diagnostic log.

[0040] like Figure 2As shown, the horizontal axis is the time axis in days, ranging from day 1 to day 30. The left vertical axis is the background noise baseline measured in quantified units, and the right vertical axis is the volatility index. The figure contains three curves: the background noise baseline represented by a solid line, the volatility index represented by a long dashed line, and the fault entropy represented by a dotted line. The data pattern shown reflects that the background noise baseline shows a fluctuating upward trend over time, while the volatility index shows an overall downward trend after an initial increase. The fault entropy remains in a relatively stable range for most of the time, but at a specific time point, such as around day 21, a significant sharp drop occurs.

[0041] like Figure 3 As shown, the architecture consists of a sensor network layer, a first sensor processing layer, and a cloud platform layer from bottom to top. The sensor network layer physically deploys a first sensor S1 as the main body for event monitoring and consensus initiation, multiple second sensors S2, S3, S4, and S5 as the consensus community, and a witness sensor W1 as an external verification node. The first sensor processing layer logically integrates a communication module, a baseline determination module, an event monitoring module, a source credibility arbitration module, and a data routing module to perform a complete floating threshold calculation and consensus challenge processing flow. The cloud platform layer is responsible for the final data aggregation and analysis, location calculation, fault diagnosis and alarm management of the trusted data after front-end arbitration, and generating macroscopic statistical indicators that characterize the overall health status of the system.

[0042] Example 4: This embodiment aims to supplement the explanation of the establishment process of key parameters and models; in the initial commissioning stage of deploying the method of this invention for the first time in a specific steam turbine generator unit, it determines the operating parameters of the sensor network to adapt to the specific unit structure and operating environment, especially for calculating the floating threshold. Exceeding tolerance value The theoretical attenuation model used for re-arbitration is determined according to an engineering calibration procedure. The initial state of this calibration procedure is defined as follows: the object of action is a sensor community that has been installed, containing one first sensor and four second sensors, and the sensor spacing and installation position are fixed; the enabling environment is the complete operating condition change process of the steam turbine generator set in a fault-free state, according to the predetermined commissioning plan, from start-up, warm-up, no-load operation to full-load operation, and a hammer with an output energy range of 0.1 joules to 5 joules and an accuracy of 0.05 joules, used to apply a controlled impact.

[0043] The first step in the calibration procedure is to determine the tolerance value. In this step, as the unit experiences the transition from no-load to full-load, all sensors within the system record a summary of the vibration status reported by their respective neighbors and calculate the local real-time noise baseline. However, it does not trigger any consensus challenges. This process continues to collect data until the various operating conditions are met. A stable statistical distribution was formed, thus obtaining a full-condition background noise database for the unit under healthy conditions. Subsequently, with the unit operating at 50% rated load, a 0.2 joule impact was applied using a hammer at a distance of 0.5 meters from the first sensor. This energy level was predefined as the lower limit of the fault signal that the system needed to capture. The system recorded the energy value detected by the first sensor during this impact. and the real-time noise baseline at that moment. ; The initial value is set by setting it to To determine, if It consists of 245 quantization units. If it is 135 units, then The initial value was set to 110; the system then used this... The value was retrospectively verified against the full-condition background noise database, and statistics were compiled on this value. Set the number of false triggers caused by background noise fluctuations. If this number exceeds the upper limit of once per hour, increase the number in increments of 5 units. The value is repeatedly verified until the number of false triggers falls below the upper limit, and the final result is obtained. The value is embedded into all sensors in this sensor community.

[0044] The second step in the calibration procedure is to establish a theoretical attenuation model. This step utilizes the impact data from the first step. When an impact of known energy is applied near the first sensor, the system not only records the response of the first sensor but also simultaneously records the precise value of the vibration state summary received by all second sensors within the community at the same time. Given that the three-dimensional coordinates of each sensor are known, the system can calculate the physical distance from the impact point to each second sensor. By conducting hundreds of impact tests at different energy levels and locations, the system obtained a dataset containing the impact source energy, the response energy of each sensor, and the propagation distance. Based on this dataset, the system uses the least squares method to fit an exponential function describing the relationship between energy attenuation and distance for each first-sensor-second-sensor pair, serving as the theoretical attenuation model for this path. An exemplary function form is as follows: ;in, In order to receive energy, As the source of energy, For distance, coefficient and This characterizes the attenuation characteristics of the specific propagation path; the model parameters of all paths are jointly stored in the non-volatile memory of the first sensor for subsequent use in the re-arbitration step; by executing the above calibration procedure, the key operating parameters of each sensor community are set based on the specific physical environment and equipment conditions of the deployment, providing a reproducible judgment benchmark for the judgment and arbitration of impact events during subsequent online operation.

[0045] Example 5: This embodiment aims to supplement the explanation of the response mechanism of the method in the specific implementation when facing a key boundary condition, namely, community silence. In a steam turbine generator set where the system of the present invention is deployed, due to a failure of the power supply module in a local area, all second sensors in a certain sensor community collectively enter a communication interruption state. At this time, the operating state and processing logic of a first sensor that is still working normally at the edge of the community are as follows: When the first sensor detects that the energy exceeds its local floating threshold... In the event of an impact event, the first sensor, following a pre-defined procedure, immediately initiates a consensus challenge to all its neighboring second sensors. Since all second sensors are in a communication interruption state, within a 500-millisecond waiting window, the first sensor receives no response, neither consensus confirmation nor rejection. At this point, the system does not directly equate this lack of response to an isolated impact event, i.e., a situation where other sensors are functioning normally but have not detected the impact. Given the absence of any response from the local community, the first sensor's source reliability arbitration module triggers a secondary arbitration process, initiating a query to a physically non-adjacent witness sensor regarding whether it has continuously received survival certificates from second sensors within a pre-defined time window. After querying its internally maintained state table, the witness sensor reports that it has not received any survival certificates from any second sensor in the community within the past five minutes. Based on this feedback, the first sensor's source reliability arbitration module, according to the final arbitration rules, determines the current situation as an impact event due to a collective loss of community contact and arbitrates the source reliability of this impact event as reliable.

[0046] Ultimately, based on the credible arbitration result of the source information, the data routing module authorizes the first sensor to report the raw waveform data of this impact event to the cloud platform. It should be noted that, in addition to the raw vibration data, the reported data packet also includes a status code to indicate that this report was triggered based on the special arbitration logic of community disconnection. After receiving this data, the cloud platform will not only treat the impact event itself as a high-priority alarm, but also immediately generate a higher-level systemic fault alarm for the sensor community infrastructure, such as local power supply or network communication, based on the status code. This ensures that in the extreme case of a collective failure of the local sensor network, a real physical impact signal will not be incorrectly suppressed, and the root cause of the fault can be accurately located.

[0047] Example 6: This embodiment aims to supplement the cloud-based analysis procedure for identifying chronic, systemic changes in equipment health status using aggregated data. On the cloud platform side of the steam turbine generator diagnostic monitoring and management system, the system employs a pattern recognition algorithm based on deterministic rules to quantify and analyze the evolution of macroscopic statistical indicators of the overall sensor network status over time. The initial state of this algorithm is defined as follows: the object of operation is a database that aggregates all self-diagnostic logs reported by all sensors of a complete steam turbine generator set over the past 365 days, as well as historical vibration state summaries attached when authorized to report impact events; the enabling environment is a server backend with statistical computing and time series analysis capabilities, whose processing power is sufficient to support daily calculations on the aforementioned database. The task and algorithm execution flow are set as a daily timed automated task. The first step is data preprocessing and index calculation. The system first extracts the energy values ​​of all vibration state summaries from the past 90 days and removes outliers with energy values ​​in the top 1% caused by impact events to obtain a background noise dataset reflecting the normal operating status of the equipment. Based on this dataset, the system calculates the background noise baseline, which is the 30-day weighted moving average of the energy values ​​in the dataset, and the volatility index, which is the 30-day weighted moving standard deviation of the energy values ​​in the dataset. At the same time, the system counts all self-diagnostic logs from the past 90 days and calculates their information entropy in space (sensor community) and time (week) based on the sensor ID and timestamp of the occurrence, as a fault entropy index.

[0048] The second step of the algorithm is pattern determination, which employs quantitative comparison rules. Specifically, the rule for determining a slow warming pattern is as follows: the system performs a linear regression analysis on the background noise baseline data points for the past 30 days. If the slope of the regression line is greater than a preset baseline drift threshold, and this condition is met for three consecutive calculation cycles, then the slow warming pattern is determined to be triggered. The baseline drift threshold is determined based on statistical analysis of the background noise baseline slope before similar faults occurred in the past for this type of unit and is set to 0.05 quantization units per day. The rule for determining an epidemic outbreak pattern is as follows: if the fault entropy value calculated this week is lower than... If the moving average of the fault entropy over the past four weeks reaches 50%, the epidemic outbreak mode is considered to have been triggered. In a specific operational instance, the system, through the above procedure, determined that the slow temperature rise mode was triggered, and its output was a structured diagnostic suggestion, which stated: Over the past 30 days, the unit background noise baseline has shown a continuous upward trend with a slope of 0.08 quantization units per day. The correlation between this trend and the changes in unit power load is less than 0.2. It is recommended that structural components that may cause systemic damping changes be inspected during the next shutdown maintenance. This suggestion provides maintenance personnel with objective data and troubleshooting directions for decision-making.

[0049] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for processing vibration data and impact data, characterized in that, Includes the following steps: The first sensor in the sensor network periodically sends a proof of life containing the identity information of itself and one or more neighboring second sensors to one or more physically non-adjacent witness sensors. The first sensor determines a real-time noise baseline characterizing local environmental noise by calculating a weighted moving average of the energy values ​​of the vibration state summaries broadcast by one or more neighboring second sensors within a preset historical time window, based on the periodic vibration state summaries it has received in the past. When the first sensor detects an impact event whose energy exceeds a floating threshold dynamically adjusted by a real-time noise baseline, a consensus challenge is initiated to one or more second sensors. If a consensus confirmation is received from at least one second sensor, the source credibility of the impact event is arbitrated as credible. If no consensus confirmation is received from any second sensor, a query is initiated to the witness sensor to inquire whether a survival certificate issued by the second sensor has been continuously received within a preset time window of five minutes. A final arbitration is then conducted based on the feedback obtained from the witness sensor. The rules for the final arbitration are as follows: if the feedback indicates that the second sensor has lost contact, the source credibility of the impact event is arbitrated as credible; if the feedback indicates that the second sensor is still alive, it is arbitrated as uncredible. Differentiated data routing is performed on the raw data of the impact event based on whether the arbitration result is credible or uncredible.

2. The method for processing vibration data and impact data according to claim 1, characterized in that, The calculation steps for the vibration state summary include: A raw vibration signal sequence is acquired, and the sum of the squares of the amplitudes at the sampling points in the sequence is calculated to obtain a raw energy value. The calculated raw energy value is directly mapped to a log-compressed intermediate value using a lookup table pre-installed in local memory. The intermediate values ​​are quantized into an unsigned integer through a linear scaling transformation, serving as a summary of the vibration state.

3. The method for processing vibration data and impact data according to claim 1, characterized in that, The differentiated data routing includes: When the source credibility of an impact event is arbitrated as credible, the original waveform data of the impact event with high-precision timestamps is authorized to be reported. When the source credibility of an impact event is arbitrated as unreliable, the reporting of the raw data of the impact event is suppressed, and a self-diagnostic log is reported. The self-diagnostic log contains the event timestamp, sensor ID, and an event code marked as an isolated high-energy event.

4. The method for processing vibration data and impact data according to claim 1, characterized in that, When the source of an impact event is deemed credible by arbitration, the step of authorizing the reporting of the original data of the impact event includes authorizing the first sensor and the second sensor that has reached a consensus to report the original data of the impact event.

5. The method for processing vibration data and impact data according to claim 4, characterized in that, Also includes: Record the arrival time of the raw data reported by the first sensor and the second sensor that made the consensus confirmation; Based on the time difference of arrival calculated from the arrival time of the second sensor relative to the arrival time of the first sensor, and the known physical locations of the first and second sensors, the physical source location of the impact event is determined using a time difference of arrival localization algorithm.

6. The method for processing vibration data and impact data according to claim 1, characterized in that, The floating threshold The calculation rules are as follows: in, As a real-time noise baseline, For a pre-defined tolerance value; The step of arbitrating the source reliability of the impact event as unreliable, based on feedback indicating that the second sensor is still alive, also includes: Obtain a summary of vibration status below a floating threshold from one or more second sensors during an impact event; Based on the impact event energy from the first sensor and the vibration state summary energy from the second sensor, a real-time decay gradient characterizing the energy decay characteristics in space is determined. Furthermore, by comparing the real-time decay gradient with a preset theoretical decay model, if there is a nonlinear deviation between the real-time decay gradient and the theoretical decay model, the source credibility of the impact event is re-arbitrated as credible.

7. The method for processing vibration data and impact data according to claim 3, characterized in that, Also includes: The system aggregates self-diagnostic logs from multiple sensors in the cloud, along with a summary of periodic vibration states attached when the source credibility of an impact event is arbitrated as credible and authorized for reporting. Based on aggregated self-diagnostic logs or vibration status summaries, one or more macroscopic statistical indicators characterizing the overall health status of the sensor network are generated. And by analyzing the patterns of one or more macro-statistics over time, a chronic, systemic change in equipment health can be identified.

8. The method for processing vibration data and impact data according to claim 7, characterized in that, Macroscopic statistical indicators include: the background noise baseline represented by the moving average of all vibration state summary energy values, the volatility index represented by the standard deviation of all vibration state summary energy values, and the information entropy in space and time calculated based on the sensor ID and timestamp generated by the self-diagnostic log, which is the fault entropy.

9. The method for processing vibration data and impact data according to claim 4, characterized in that, The steps to arbitrate the credibility of sources regarding a shock event further include: The first sensor determines a stability index of a communication link with each of the second sensors that have made consensus confirmations based on the arrival time of the periodic heartbeat signals it has historically received from the second sensors that have made consensus confirmations. And based on stability metrics, the consensus confirmations received from the second sensor are weighted to calculate a weighted total score; The weighted total score is compared with a preset weighted confirmation threshold, wherein the source credibility of the shock event is arbitrated as credible only when the weighted total score is greater than the threshold.

10. A vibration data and impact data processing system, used to implement the vibration data and impact data processing method of claim 1, characterized in that, The system includes: A sensor network, comprising a first sensor, one or more second sensors, and one or more witness sensors; wherein the first sensor includes: A communication module is configured to periodically send a proof of survival containing its own identity information and that of one or more second sensors to one or more witness sensors, initiate a consensus challenge to one or more second sensors when an impact event is detected, and initiate a query to the witness sensors when no consensus confirmation is received. A baseline determination module is configured to determine a real-time noise baseline characterizing local environmental noise based on a summary of periodic vibration states received from one or more second sensors in the past. An event monitoring module is configured to trigger a consensus challenge by the communication module when an impact event is detected whose energy exceeds a floating threshold dynamically adjusted by a real-time noise baseline. A source credibility arbitration module is configured to: arbitrate the source credibility of the shock event as credible if a consensus confirmation is received from at least one second sensor; and perform final arbitration based on feedback from witness sensors regarding the survival status of the second sensors if no consensus confirmation is received from any second sensor; and a data routing module is configured to perform differentiated data routing on the raw data of the shock event based on the result of the source credibility arbitration module arbitrating it as credible or untrustworthy.

Images