An edge-computing-based vibration motor operation data analysis method and system

Abstract

The present application relates to the technical field of vibration motor operation data analysis, in particular to a vibration motor operation data analysis method and system based on edge computing, which comprises the following steps: obtaining a structured data block sequence generated from a vibration signal, evaluating a complexity index according to an energy index, triggering a transient state marker, generating a mapping table by dividing priorities in combination with kinematic parameters, establishing a quota parameter and a scheduling instruction based on a computing power difference, extracting an amplitude warning from a high priority, and performing analysis on low priority features. In the present application, a multi-dimensional complexity dynamic evaluation mechanism is established to identify signal transient characteristics, and a frequency band division and computing capacity forward estimation is used to establish a real-time computing resource quota fine scheduling allocation criterion, ensuring that the processor computing power is limited to complete the core fault feature extraction first, avoiding task timeout and early hidden danger omission caused by computing peak stimulation, and ensuring the continuous and effective fault warning of the motor under variable frequency transient working conditions.

Description

Technical Field

[0001] This invention relates to the field of vibration motor operation data analysis technology, and in particular to a vibration motor operation data analysis method and system based on edge computing. Background Technology

[0002] The field of vibration motor operation data analysis technology covers the core aspects of multi-dimensional analysis of various physical parameters generated by vibration motors during operation. Its overall technical field systematically includes front-end physical quantity sensing, multi-source signal synchronous acquisition, communication link data backhaul, and multi-domain signal feature extraction. It mainly focuses on in-depth analysis of operating data such as motor acceleration, velocity, displacement, and stator temperature to characterize the motor's dynamic characteristics and health evolution process. Among them, traditional vibration motor operation data analysis methods refer to the technical aspects of vibration motor operation feature extraction and anomaly diagnosis.

[0003] Existing technologies mostly perform feature extraction with a fixed sampling rate. During the motor frequency conversion start-up phase, the vibration signal instantly transforms into a broadband non-stationary signal. The influx of broadband data causes instantaneous spikes in computing resources, which quickly saturates the computing power of edge nodes. The analysis task originally scheduled within a fixed time window is forced to be extended and marked as timed out. There is a lack of a mechanism to deal with transient high computational impact. The computing power saturation causes feature extraction to exceed the time boundary, resulting in the omission of early bearing fault features under transient operating conditions, causing the fault warning to fail during the motor start-up and shutdown phase. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a method and system for analyzing the operating data of a vibration motor based on edge computing.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for analyzing the operating data of a vibration motor based on edge computing, comprising the following steps: S1: Acquire the simulated vibration signals collected by the acceleration sensor installed on the bearing housing of the vibratory motor in the start-up state and the stable operation state, respectively. Convert the acquired simulated vibration signals into discrete data points for buffer storage. Combine the clock signal and the driver output frequency to extract effective data segments and splice them to generate a structured vibration data block sequence containing working condition labels. S2: Calculate the frequency migration velocity and frequency band energy dispersion index based on the structured vibration data block sequence, evaluate the comprehensive complexity index by combining the frequency migration velocity and the frequency band energy dispersion index, calculate the deviation between the comprehensive complexity index and the preset calculation budget threshold, and when the deviation meets the preset transient judgment condition, use the deviation as the complexity evaluation quantity and trigger the transient marking action to generate the start transient flag bit and the complexity evaluation quantity; S3: Calculate the theoretical fault center frequency based on the motor bearing kinematic parameters, the starting transient flag, and the complexity assessment quantity. Combine the frequency migration speed to divide the high-priority calculation frequency band and the low-priority accompanying frequency band, and integrate them to generate a priority frequency band mapping table. S4: Calculate the estimated value of the required computing power based on the priority frequency band mapping table, extract the computing power capacity difference by combining the number of available computing power clock cycles, generate resource quota parameters and downgrade processing coefficients based on the computing power capacity difference, and establish real-time computing resource allocation instructions and analysis boundary thresholds. S5: Based on the real-time computing resource allocation instructions and analysis boundary thresholds, extract the true amplitude of bearing fault characteristic frequencies from the high-priority computing frequency band to generate an instant warning message, and extract downgraded features from the low-priority accompanying frequency band to establish a transfer marker snapshot. Perform supplementary analysis based on the transfer marker snapshot and fuse the instant warning message to generate motor operating health status analysis data.

[0006] The present invention is improved in that the specific steps of S1 are as follows: S111: Acquire the simulated vibration signal collected by the acceleration sensor installed on the bearing housing of the vibration motor in the start-up state and the stable operation state, respectively; monitor the clock signal output by the hardware timer within the acquisition time period; convert the simulated vibration signal into discrete vibration data points through analog-to-digital conversion logic; and store the discrete vibration data points in the multi-level ring buffer of the edge node through the memory access controller to obtain the buffered discrete data points. S112: Divide the multi-level circular buffer recorded discrete data points based on the time window length, generate a global timestamp associated with the time window according to the clock signal, collect the driver output frequency to establish the working condition label corresponding to the time window, extract the divided discrete data points of the buffer storage according to the overlap ratio of adjacent time windows, generate effective data segments, and perform data splicing and combination on the effective data segments, the global timestamp and the working condition label in the order of the time axis to generate a structured vibration data block sequence.

[0007] The present invention is improved in that the specific steps of S2 are as follows: S211: Based on the working condition labels contained in the structured vibration data block sequence, extract the driver output frequency of the current time window and the previous time window, calculate the time difference between the driver output frequency and the current time window and the previous time window, and obtain the frequency migration speed. S212: The frequency distribution parameters of the effective data segments contained in the structured vibration data block sequence are calculated using the Fast Fourier Transform algorithm to obtain frequency domain distribution data. The frequency domain distribution data is divided into equal-width frequency band intervals along the frequency axis. The ratio between the signal energy contained in each equal-width frequency band interval and the full-band signal energy corresponding to the frequency domain distribution data is calculated to obtain the frequency band energy ratio coefficient. Each equal-width frequency band interval is regarded as a discrete information source, and the associated frequency band energy ratio coefficient is used as the occurrence probability parameter. The Shannon information entropy is calculated based on the occurrence probability parameter to obtain the frequency band energy dispersion index. S213: Obtain a preset velocity reference value and a preset energy dispersion upper limit value; calculate the ratio between the frequency migration velocity and the preset velocity reference value to obtain a velocity migration normalization coefficient; calculate the ratio between the frequency band energy dispersion index and the preset energy dispersion upper limit value to obtain an energy dispersion normalization coefficient; perform a weighted summation calculation on the velocity migration normalization coefficient and the energy dispersion normalization coefficient to obtain a comprehensive complexity index; obtain a preset calculation budget threshold; calculate the deviation between the comprehensive complexity index and the calculation budget threshold; compare the deviation with a preset transient judgment condition; when the deviation meets the preset transient judgment condition, use the deviation as a complexity evaluation quantity and trigger a transient marking action to generate a start transient flag bit and a complexity evaluation quantity.

[0008] The present invention is improved in that the process of obtaining the preset velocity reference value and the preset energy dispersion upper limit value is specifically as follows: Extract the historical frequency sequence and historical energy sequence of the vibratory motor; The maximum rate of change contained in the historical frequency sequence is set as a preset velocity benchmark value; The maximum information entropy value associated with the historical energy sequence is set to a preset upper limit value for energy dispersion. The process of weighted summation of the velocity migration normalization coefficient and the energy dispersion normalization coefficient is as follows: Obtain evaluation values ​​for frequency changes and energy fluctuations; Calculate the first ratio of the frequency change evaluation value within the sum of the frequency change evaluation value and the energy fluctuation evaluation value, and establish the first ratio as the first weighting coefficient; Calculate the second ratio of the energy fluctuation evaluation value within the sum of the frequency change evaluation value and the energy fluctuation evaluation value, and establish the second ratio as the second weighting coefficient; Calculate the first product between the velocity migration normalization coefficient and the first weighting coefficient, and calculate the second product between the energy dispersion normalization coefficient and the second weighting coefficient; Calculate the sum of the first product and the second product to obtain the overall complexity index; The process of obtaining a preset computational budget threshold, calculating the deviation between the comprehensive complexity index and the computational budget threshold, and comparing the deviation with a preset transient judgment condition specifically includes: Extract the quantified value of available computing clock cycles from the edge computing node, and set the quantified value of available computing clock cycles as the computing budget threshold; Obtain the fault tolerance ratio parameter and calculate the boundary product between the fault tolerance ratio parameter and the calculation budget threshold; The numerical range formed by the product of zero and the boundary is set as a preset transient determination condition; Calculate the deviation between the overall complexity index and the calculation budget threshold. Determine whether the deviation amount belongs to the numerical range corresponding to the preset transient determination condition.

[0009] The present invention is improved in that the specific steps of S3 are as follows: S311: Obtain the stored pitch diameter, contact angle and number of rolling elements of the motor bearing, establish the kinematic parameters of the motor bearing, respond to the start transient flag, evaluate the instantaneous speed estimate based on the complexity evaluation quantity and the driver output frequency, substitute the instantaneous speed estimate and the kinematic parameters of the motor bearing into the kinematic equation of the motor bearing for calculation, and obtain the theoretical fault center frequency of the differentiated fault type. S312: Based on the frequency migration speed, calculate the frequency band coverage range extending to both sides of each theoretical fault center frequency, generate multiple dynamic characteristic frequency band intervals, perform boundary merging and frequency truncation processing on the multiple dynamic characteristic frequency band intervals, and establish the independent or merged intervals obtained after processing as high-priority calculation frequency bands. S313: Remove the coverage area corresponding to the high-priority calculation frequency band from the preset full-band analysis area to generate a remaining bandwidth area. Establish the remaining bandwidth area as a low-priority accompanying frequency band. Integrate the frequency boundary information within the high-priority calculation frequency band and the low-priority accompanying frequency band under the current time window to generate a priority frequency band mapping table.

[0010] The present invention is improved in that the process of calculating the frequency band coverage range extending to both sides of each theoretical fault center frequency based on the frequency migration speed, and generating multiple dynamic characteristic frequency band intervals, specifically includes: Obtain the length of a preset time window, extract the maximum tolerance value of frequency fluctuation associated with historical stable operation phases, calculate the product between the time length and the maximum tolerance value of frequency fluctuation, and obtain the benchmark expansion coefficient. Calculate the product between the frequency shift rate and the reference extension factor to obtain the one-sided frequency extension margin; The upper limit frequency threshold is obtained by summing the theoretical fault center frequency and the single-sided frequency extension margin. The difference between the theoretical fault center frequency and the single-sided frequency extension margin is calculated to obtain the lower limit frequency threshold. The continuous frequency range defined by the lower frequency threshold and the upper frequency threshold is set as the dynamic characteristic frequency band interval; The process of performing boundary merging and frequency truncation processing on multiple dynamic feature frequency bands and establishing the resulting independent or merged intervals as high-priority calculation frequency bands is as follows: Obtain the upper limit value of the Nyquist frequency corresponding to the analog-to-digital conversion action, and compare the upper limit frequency threshold contained in each of the dynamic characteristic frequency bands with the upper limit value of the Nyquist frequency. When the upper limit frequency threshold is greater than the upper limit value of the Nyquist frequency, the upper limit frequency boundary of the dynamic characteristic frequency band interval is replaced with the upper limit value of the Nyquist frequency to obtain the corresponding corrected characteristic frequency band interval. Determine whether there are overlapping frequency band regions among the multiple modified feature frequency band intervals; When the overlapping frequency band regions exist, extract the minimum lower limit frequency value and the maximum upper limit frequency value associated with the multiple overlapping modified feature frequency band intervals. Based on the minimum lower limit frequency value and the maximum upper limit frequency value, perform boundary merging processing on multiple overlapping modified characteristic frequency band intervals to establish a high-priority calculation frequency band.

[0011] The present invention is improved in that the specific steps of S4 are as follows: S411: Read the operating system's performance counters to obtain the available computing power clock cycles of the multi-core processor, parse the high-priority computing frequency band and the low-priority accompanying frequency band contained in the priority frequency band mapping table, calculate the number of multiply-accumulate operations and memory accesses based on the bandwidth and frequency resolution within the high-priority computing frequency band, obtain the clock cycles consumed per multiply-accumulate operation and the clock cycles consumed per memory access, calculate the product between the number of multiply-accumulate operations and the clock cycles consumed per multiply-accumulate operation to obtain the computing clock cycle requirement, calculate the product between the number of memory accesses and the clock cycles consumed per memory access to obtain the memory access clock cycle requirement, and calculate the sum of the computing clock cycle requirement and the memory access clock cycle requirement to obtain the estimated value of the required computing power requirement; S412: Compare the number of available computing power clock cycles with the estimated value of required computing power to obtain the computing power capacity difference, obtain computing power allocation constraint rules, determine the allocation ratio of the number of available computing power clock cycles among different priority computing frequency bands according to the computing power allocation constraint rules and the computing power capacity difference, divide the number of available computing power clock cycles according to the allocation ratio, and generate resource quota parameters. S413: Calculate the absolute value of the computing power capacity difference, calculate the ratio between the absolute value and the theoretical processing clock cycle number corresponding to the low priority accompanying frequency band, obtain the downgraded processing coefficient, combine the resource quota parameter and the downgraded processing coefficient, and establish a real-time computing resource allocation instruction and analysis boundary threshold.

[0012] The present invention is improved in that the specific steps of S5 are as follows: S511: Based on the resource quota parameters included in the real-time computing resource allocation instruction and the analysis boundary threshold, set the filter parameters inside the bandpass filter, separate the narrowband signal belonging to the high-priority computing frequency band from the effective data segments included in the structured vibration data block sequence through the bandpass filter, calculate the corresponding signal envelope by the Hilbert transform algorithm, calculate the frequency distribution parameters corresponding to the signal envelope by the fast Fourier transform algorithm, and generate the true amplitude of the bearing fault characteristic frequency. S512: Based on the real-time computing resource allocation instruction and the order reduction processing coefficient included in the analysis boundary threshold, perform power spectral density estimation numerical calculation on the associated data segments belonging to the low-priority accompanying frequency band in the effective data segments included in the structured vibration data block sequence, extract the corresponding energy distribution order reduction features, obtain the memory address pointer corresponding to the time window inside the multi-level ring buffer, fuse the energy distribution order reduction features and the memory address pointer, and establish a transition marker snapshot; S513: Obtain the adaptive safety threshold for operating conditions, calculate the excess deviation between the true amplitude of the bearing fault characteristic frequency and the adaptive safety threshold for operating conditions, determine whether the excess deviation exceeds a preset safety limit, and when the excess deviation exceeds the preset safety limit, extract the current fault status information and the excess deviation to perform early warning encapsulation to obtain an immediate early warning message. Monitor the computing load corresponding to the multi-core processor, and when the computing load is idle, read the associated discrete vibration data points according to the memory address pointer contained in the receiving mark snapshot. Calculate the associated discrete vibration data points using a spectrum analysis algorithm to obtain the corresponding supplementary feature vector. Calculate the association alignment and splicing result of the supplementary feature vector and the true amplitude of the bearing fault characteristic frequency on the time axis according to the global timestamp contained in the structured vibration data block sequence. Merge the association alignment and splicing result with the immediate early warning message to generate motor operating health status analysis data.

[0013] The present invention is improved in that the process of obtaining the adaptive safety threshold for operating conditions, calculating the excess deviation between the true amplitude of the bearing fault characteristic frequency and the adaptive safety threshold for operating conditions, and determining whether the excess deviation exceeds the preset safety limit is specifically as follows: Extract historical vibration amplitude parameters of the vibratory motor under normal operating conditions and correlated with the current working conditions; Calculate the average value of the historical vibration amplitude parameters, calculate the sum of the average value and the tolerance constant, and set the sum as the working condition adaptive safety threshold. Calculate the absolute value of the difference between the true amplitude of the bearing fault characteristic frequency and the adaptive safety threshold of the operating condition to obtain the excess deviation; Obtain the maximum allowable deviation ratio parameter corresponding to the equipment safety specification, calculate the boundary product value between the maximum allowable deviation ratio parameter and the working condition adaptive safety threshold, and set the boundary product value as a preset safety limit. Compare the value of the excess deviation with the value of the preset safety limit to determine whether the excess deviation is greater than the preset safety limit; The process of calculating the corresponding supplementary feature vector by using a spectrum analysis algorithm on the associated discrete vibration data points, and calculating the correlation and alignment of the supplementary feature vector with the true amplitude of the bearing fault characteristic frequency on the time axis based on the global timestamp contained in the structured vibration data block sequence is as follows: Perform a discrete Fourier transform on the discrete vibration data points to obtain the full-band amplitude spectrum; Extract the discrete amplitude peak sequence retained within the full-band amplitude spectrum, and set the discrete amplitude peak sequence as a supplementary feature vector; Extract the global timestamp contained in the structured vibration data block sequence, and locate the target time series position that perfectly matches the global timestamp value within the historical feature database; The supplementary feature vector is inserted into the target time series position, and a multidimensional array merging operation is performed with the true amplitude of the bearing fault feature frequency at the target time series position to generate an associated aligned splicing result.

[0014] A vibration motor operation data analysis system based on edge computing, wherein the edge computing-based vibration motor operation data analysis system is used to implement the aforementioned edge computing-based vibration motor operation data analysis method, the system comprising: The data preparation module acquires the simulated vibration signals collected by the acceleration sensor installed on the bearing housing of the vibratory motor in the start-up state and the stable operation state, respectively. It converts the acquired simulated vibration signals into discrete data points for buffer storage, extracts effective data segments by combining the clock signal and the driver output frequency, and splices them to generate a structured vibration data block sequence containing working condition labels. The transient identification module calculates the frequency migration velocity and frequency band energy dispersion index based on the structured vibration data block sequence, evaluates the comprehensive complexity index by combining the frequency migration velocity and the frequency band energy dispersion index, calculates the deviation between the comprehensive complexity index and the preset calculation budget threshold, and when the deviation meets the preset transient judgment condition, uses the deviation as the complexity evaluation quantity and triggers the transient marking action to generate a transient start flag bit and a complexity evaluation quantity. The frequency band planning module calculates the theoretical fault center frequency based on the motor bearing kinematic parameters, the start-up transient flag, and the complexity evaluation quantity. It then divides the frequency band into high-priority calculation frequency bands and low-priority accompanying frequency bands based on the frequency migration speed, and integrates them to generate a priority frequency band mapping table. The computing power scheduling module calculates the estimated value of the required computing power based on the priority frequency band mapping table, extracts the computing power capacity difference by combining the number of available computing power clock cycles, generates resource quota parameters and downgrade processing coefficients based on the computing power capacity difference, and establishes real-time computing resource allocation instructions and analysis boundary thresholds. The feature execution module, based on the real-time computing resource allocation instructions and analysis boundary thresholds, extracts the true amplitude of bearing fault characteristic frequencies from the high-priority computing frequency band to generate an instant warning message, and extracts downgraded features from the low-priority accompanying frequency band to establish a transfer marker snapshot. Based on the transfer marker snapshot, it performs supplementary analysis and merges the instant warning message to generate motor operating health status analysis data.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a structured data block sequence is generated by splicing vibration signals, calculating frequency migration speed and frequency band energy dispersion to assess the comprehensive complexity index, extracting calculation deviation and triggering transient marking actions, establishing the fault center frequency in conjunction with bearing parameters, dividing high and low priority frequency band mapping relationships based on migration speed, estimating computing power requirements and capacity differences to generate resource quota parameters and downgrade processing coefficients, establishing resource allocation boundaries accordingly, extracting fault features from high priority frequency bands to generate real-time warnings, and performing supplementary analysis on low priority features. By utilizing dynamic frequency band planning and resource scheduling allocation, the omission of features induced by computing power saturation is avoided, ensuring the reliability of warnings under transient impacts. Attached Figure Description

[0016] Figure 1 This is the main flowchart for the vibration motor operation data analysis of the present invention; Figure 2 This is a flowchart of the vibration data preparation and transient identification process of the present invention; Figure 3 This is a flowchart of the priority frequency band planning and computing power scheduling process of the present invention; Figure 4 This is a flowchart of the feature extraction, early warning, and supplementary analysis process of the present invention. Detailed Implementation

[0017] 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 merely illustrative and not intended to limit the invention.

[0018] Please see Figures 1 to 4 This embodiment provides a method for analyzing the operating data of a vibration motor based on edge computing. In practical applications, for example, during the operation of a vibration motor driven by a frequency converter and receiving the vibration signal of the bearing housing by an edge node, an acceleration sensor is attached to the measuring point of the bearing housing, and a hardware timer, memory access controller, and multi-core processor form a link around the edge node. The frequency output by the driver enters the time window label, including the following steps: S1: Acquire simulated vibration signals from the accelerometer mounted on the bearing housing of the vibratory motor during both startup and stable operation. Convert the acquired simulated vibration signals into discrete data points for buffer storage. Extract valid data segments by combining the clock signal and the driver output frequency, and concatenate them to generate a structured vibration data block sequence containing operating condition labels. The structured vibration data block sequence is a set of data blocks with time windows as the carrier, carrying valid data segments, global timestamps, operating condition labels, and cache address association fields. Valid data segments are the set of data points retained after sampling continuity verification, timestamp alignment verification, duplicate point removal, saturation distortion elimination, and noise status verification. S111: Acquires analog vibration signals from an accelerometer mounted on the bearing housing of the vibratory motor during both startup and stable operation. Monitors the clock signal output by the hardware timer within the acquisition period. Converts the analog vibration signals into discrete vibration data points using analog-to-digital conversion logic. These discrete data points are then stored in a multi-level circular buffer at the edge node via a memory access controller, resulting in buffered discrete data points. The analog vibration signal output by the accelerometer first enters the acquisition status verification path. The edge node writes sensor disconnection, output saturation, sampling jump, and interface interruption statuses into the acquisition status field. The analog-to-digital conversion logic generates discrete vibration data points for normal states and retains an anomaly identifier and buffer placeholder for abnormal states, ensuring continuous indexing of subsequent time windows. The hardware timer provides the acquisition timing reference. When the memory access controller writes to the multi-level circular buffer, it synchronously records the acquisition sequence number, clock source identifier, buffer level identifier, and write confirmation status. The multi-level circular buffer includes a current window write area, an adjacent window overlap interception retention area, and a supplementary analysis traceability area, all connected in series via memory address pointers. When the device is cold-started, the factory baseline field and the first valid acquisition window are used together as the initial buffer basis. After the continuous windows are formed, they inherit the acquisition status and cache address of the previous valid window. S112: Discrete data points are buffered and stored in a multi-level circular buffer based on the time window length. A global timestamp associated with the time window is generated according to the clock signal. The driver output frequency is collected to establish a corresponding operating condition label for the time window. Based on the overlap ratio of adjacent time windows, the buffered and stored discrete data points are truncated to generate valid data segments. The valid data segments, global timestamps, and operating condition labels are then combined in chronological order to generate a structured vibration data block sequence. The time window length is derived from the window field in the edge node configuration area. This field is written by the device acquisition strategy and updated when the sampling strategy or driver communication configuration changes. The global timestamp is generated jointly by the clock signal of the hardware timer and the window boundary record. The driver output frequency is written to the operating condition label field via the industrial communication interface. The operating condition label also carries a status identifier indicating whether the system is in startup or in stable operation. Before being combined, valid data segments undergo integrity checks. Missing points, duplicate points, and data points with inconsistent formats are written to their corresponding status fields and enter the exception buffer. After the splicing is completed, the structured vibration data block sequence is written into the real-time analysis queue and outputs window-level data blocks, working condition labels, global timestamps and cache address pointers to the transient identification step. S2: Based on the structured vibration data block sequence, calculate the frequency migration speed and band energy dispersion index. Combine the frequency migration speed and band energy dispersion index to evaluate the comprehensive complexity index. Calculate the deviation between the comprehensive complexity index and the preset computational budget threshold. When the deviation meets the preset transient judgment condition, use the deviation as the complexity evaluation quantity and trigger a transient marking action, generating a startup transient flag and a complexity evaluation quantity. The frequency migration speed is a window-level field formed by the trend of the driver output frequency along the time axis in adjacent time windows. The band energy dispersion index is a dimensionless field formed by the dispersion of the energy distribution of effective data segments in the frequency domain. The comprehensive complexity index is used to describe the edge computing pressure caused by the frequency change and frequency domain energy distribution during the startup phase. The computational budget threshold is derived from the quantified field of available computational clock cycles, and the preset transient judgment condition is derived from the transient rule field. S211: Based on the operating condition tags contained in the structured vibration data block sequence, extract the driver output frequency of the current time window and the previous time window, calculate the time difference between the driver output frequency and the current time window and the previous time window, and obtain the frequency migration speed. The driver output frequency of the current time window is provided by the operating condition tags written in S112. The driver output frequency of the previous time window is preferentially read from the previous valid structured vibration data block in the real-time analysis queue. When the device has just started and the previous window is insufficient, the output frequency status most recently confirmed by the driver control interface is used as the initial previous field and marked as the startup alternative source. The frequency migration speed generation process first performs a source verification on the timestamps of the current time window and the previous time window, and then reads the frequency change direction and change rate status according to the time sequence. If the driver output frequency is missing or the communication interface returns an abnormality, the current window enters the frequency migration pending confirmation state. The edge nodes retain the previous valid frequency and the current vibration data block. After the driver frequency field is completed, the frequency migration speed field is rewritten. S212: The frequency distribution parameters of the effective data segments contained in the structured vibration data block sequence are calculated using the Fast Fourier Transform (FFT) algorithm to obtain frequency domain distribution data. This frequency domain distribution data is then divided into equal-width frequency bands along the frequency axis. The ratio between the signal energy contained in each equal-width frequency band and the corresponding full-band signal energy of the frequency domain distribution data is calculated to obtain the frequency band energy ratio coefficient. Each equal-width frequency band is considered a discrete information source, and the associated frequency band energy ratio coefficient is used as the occurrence probability parameter. Based on the occurrence probability parameter, the Shannon information entropy is calculated to obtain the frequency band energy dispersion index. Before the effective data segments enter the frequency domain processing, the window boundary, sampling status, and cache address fields are retained. The FFT algorithm reads discrete vibration data points that have passed integrity verification. The time positions with missing status fields are processed according to the adjacent effective status continuation rule within the window, and the missing source identifier is retained in the frequency domain distribution data. The equal-width frequency bands are generated by the frequency band division rule field, and the frequency band boundaries are constrained by the upper frequency domain limit formed by the sampling links. The frequency band energy ratio coefficient is used to represent the occupancy status of the frequency band relative to the total energy of the current window frequency domain. In this scheme, the Shannon information entropy is used as the generation rule for the frequency band energy dispersion index, outputting the state field of concentrated or dispersed energy in the current window frequency domain for S213 to call. S213: Obtain the preset speed reference value and the preset energy dispersion upper limit value; calculate the ratio between the frequency migration speed and the preset speed reference value to obtain the speed migration normalization coefficient; calculate the ratio between the frequency band energy dispersion index and the preset energy dispersion upper limit value to obtain the energy dispersion normalization coefficient; perform a weighted summation of the speed migration normalization coefficient and the energy dispersion normalization coefficient to obtain the comprehensive complexity index; obtain the preset calculation budget threshold; calculate the deviation between the comprehensive complexity index and the calculation budget threshold; compare the deviation with the preset transient judgment condition; when the deviation meets the preset transient judgment condition, use the deviation as the complexity evaluation quantity and trigger the transient marking action, generating a start transient flag bit and a complexity evaluation quantity. The preset speed reference value comes from the frequency change boundary field recorded in the historical frequency sequence of the vibration motor; when historical data is insufficient, the frequency change permission field in the driver parameter table is read as the start alternative source. The preset energy dispersion upper limit value comes from the energy dispersion boundary field associated with the historical energy sequence and is refreshed with maintenance calibration or operating condition strategy updates. The speed migration normalization coefficient and energy dispersion normalization coefficient are used to convert states from different sources into the same evaluation caliber. Frequency change evaluation values ​​are derived from the driver operating condition evaluation field, and energy fluctuation evaluation values ​​are derived from the historical energy fluctuation evaluation field. These two values ​​are verified to form frequency-side weighting coefficients and energy-side weighting coefficients. The weighting field is stored in the complexity evaluation rule table and updated by equipment commissioning records, historical operating condition boundaries, and maintenance strategy constraints. The calculation budget threshold is formed by quantifying the available computing clock cycles, and the fault tolerance ratio parameter is stored in the resource scheduling configuration table to form transient judgment boundaries. When the deviation falls into the judgment interval corresponding to the transient rule field, a transient start flag is generated and written to the current window as a complexity evaluation quantity; when it does not fall into the judgment interval, the current window retains the non-transient state and transmits the frequency migration speed and band energy dispersion indicators. S3: Calculate the theoretical fault center frequency based on the motor bearing kinematic parameters, starting transient flags, and complexity assessment values. Combine this with the frequency migration speed to divide the frequency band into high-priority calculation bands and low-priority accompanying bands, and integrate them to generate a priority frequency band mapping table. The motor bearing kinematic parameters are a set of parameters consisting of bearing pitch diameter, contact angle, number of rolling elements, and bearing structure configuration fields. These parameters are stored in the equipment parameter area and bound to the vibration motor equipment identifier. The theoretical fault center frequency is the frequency center field corresponding to the differentiated bearing fault type under the current speed estimation state. The priority frequency band mapping table records the high-priority calculation frequency band, low-priority accompanying frequency band, frequency band boundaries, fault type identifier, window timestamp, and processing priority status. S311: Obtain the stored pitch diameter, contact angle, and number of rolling elements of the motor bearing, establish the motor bearing kinematic parameters, respond to the start-up transient flag, evaluate the instantaneous speed estimate based on the complexity assessment quantity and the driver output frequency, substitute the instantaneous speed estimate and the motor bearing kinematic parameters into the motor bearing kinematic equation for calculation, and obtain the theoretical fault center frequency of differentiated fault types. Bearing structural parameters are written to the equipment parameter area after equipment commissioning or bearing replacement. Before edge nodes read, they perform consistency checks on the equipment identifier and parameter version; if parameters are missing or versions are inconsistent, the current window enters the bearing parameter pending confirmation state, fault center frequency updates are paused, and the frequency band mapping table of the original stable window continues to be used as a temporary analysis boundary. When the start-up transient flag is valid, the complexity assessment quantity is used as the speed estimation correction field during the start-up phase and enters the instantaneous speed estimation process. The driver output frequency provides the drive-side operating condition source for the current window. Edge nodes generate theoretical fault center frequency fields corresponding to fault types such as outer ring, inner ring, rolling elements, and cage according to bearing kinematic rules and write them to the fault center frequency cache. S312: Based on the frequency migration speed, calculate the frequency band coverage extending to both sides of each theoretical fault center frequency, generating multiple dynamic characteristic frequency band intervals. Perform boundary merging and frequency truncation processing on these dynamic characteristic frequency band intervals, and establish the resulting independent or merged intervals as high-priority calculation frequency bands. The preset time window length comes from the window field used in S112, and the maximum tolerance value of frequency fluctuations associated with historical stable operation phases comes from the frequency fluctuation boundary field. Together, they form the baseline expansion coefficient. After the frequency migration speed enters the expansion rule, a unilateral frequency extension margin is generated. This margin is used to form coverage boundaries on both sides of each theoretical fault center frequency, ensuring that fault characteristic drift caused by frequency changes during the startup phase is included in the dynamic characteristic frequency band intervals. After the dynamic characteristic frequency band intervals are formed, the edge nodes read the Nyquist frequency upper limit value corresponding to the analog-to-digital conversion action. This upper limit value comes from the sampling link configuration and analog-to-digital conversion configuration confirmation records. When the upper limit boundary of a dynamic characteristic frequency band exceeds the frequency limit, it is replaced with the frequency boundary allowed by the sampling link and the truncation status is recorded; when there are overlapping frequency bands between intervals, the boundaries of the overlapping intervals are extracted to form a merged high-priority calculation frequency band, and the correspondence between fault types is retained. S313: Remove the coverage area corresponding to the high-priority calculation frequency band from the preset full-band analysis area to generate a remaining bandwidth area. Establish the remaining bandwidth area as a low-priority accompanying frequency band. Integrate the frequency boundary information within the high-priority calculation frequency band and the low-priority accompanying frequency band under the current time window to generate a priority frequency band mapping table. The full-band analysis area is limited by the sampled link's processable frequency domain range field. The remaining bandwidth area after removal retains its original frequency order and boundary continuity and is written into the low-priority accompanying frequency band field. The priority frequency band mapping table includes window timestamp, operating condition label, frequency band boundary, fault type identifier, boundary truncation status, interval merging status, and resource scheduling entry field. When no processable remaining bandwidth area exists after removal, the low-priority accompanying frequency band field is written to an idle state; when the high-priority calculation frequency band is not generated due to parameter anomalies, the current window uses the previous valid mapping table. S4: Calculate the estimated required computing power based on the priority frequency band mapping table, extract the computing power capacity difference by combining it with the available computing power clock cycles, generate resource quota parameters and downgrading processing coefficients based on the computing power capacity difference, and establish real-time computing resource allocation instructions and analysis boundary thresholds. The estimated required computing power is an estimated field formed by the processing resources required for high-priority computing frequency bands to complete filtering, envelope analysis, and frequency domain calculations. The available computing power clock cycles are provided by the operating system performance counter. The computing power capacity difference is used to represent the margin or gap status of the available computing resources of the current edge node relative to the required demand. The resource quota parameters record the share of processing resources that different priority frequency bands can occupy, and the downgrading processing coefficients record the simplified analysis intensity of the low-priority accompanying frequency bands. S411: Reads the operating system's performance counters to obtain the available computing power clock cycles of the multi-core processor. It parses the high-priority computing frequency bands and low-priority accompanying frequency bands contained in the priority frequency band mapping table. Based on the bandwidth and frequency resolution within the high-priority computing frequency bands, it calculates the number of multiply-accumulate operations and memory accesses, obtaining the clock cycles consumed per multiply-accumulate operation and the clock cycles consumed per memory access. It calculates the product of the number of multiply-accumulate operations and the clock cycles consumed per multiply-accumulate operation to obtain the computing clock cycle requirement. It also calculates the product of the number of memory accesses and the clock cycles consumed per memory access to obtain the memory access clock cycle requirement. Finally, it calculates the sum of the computing clock cycle requirement and the memory access clock cycle requirement to obtain the estimated required computing power. Before reading the performance counters, it verifies access permissions and counter refresh status. If the permission verification fails, it uses the available computing power field from the previous confirmation cycle and writes a permission pending recovery flag. If the counter returns an exception, it enters a conservative scheduling state. The high-priority computing frequency bands retain the lowest executable resource entry point, while the low-priority accompanying frequency bands wait for S413 to generate a downgrade processing coefficient. The bandwidth and frequency resolution are derived from the priority band mapping table and the analysis boundary configuration field. The number of multiply-accumulate operations represents the basic computational scale required for filtering and envelope spectrum processing. The number of memory accesses represents the memory access scale formed by reading effective data segments, writing intermediate spectrum, and writing back the true amplitude. The number of clock cycles consumed per multiply-accumulate operation and the number of clock cycles consumed per memory access are derived from the processor calibration table. The computation clock cycle requirement and the memory access clock cycle requirement are combined to form the estimated value of the required computing power and then passed to S412. S412: Compare the available computing power clock cycles with the estimated required computing power to obtain the computing power capacity difference. Obtain the computing power allocation constraint rules. Based on these rules and the capacity difference, determine the allocation ratio of available computing power clock cycles among different priority computing frequency bands. Divide the available computing power clock cycles according to the allocation ratio and generate resource quota parameters. The computing power allocation constraint rules are stored in the resource scheduling configuration table, including priority retention requirements for high-priority computing frequency bands, downgrade requirements for low-priority accompanying frequency bands, preemption order for real-time early warning tasks, and delayed execution entry points for supplementary analysis tasks. When the computing power capacity difference is in a surplus state, high-priority computing frequency bands receive complete analysis resources, and low-priority accompanying frequency bands receive supplementary analysis preparation resources. When the computing power capacity difference is in a deficit state, high-priority computing frequency bands maintain the true amplitude extraction link, and low-priority accompanying frequency bands switch to a downgraded processing link. Resource quota parameters consist of fields such as window timestamp, frequency band priority, task type, available computing power source, scheduling status, and confirmation flag, and are written into the real-time computing resource allocation instruction; when the processing thread fails to confirm, the scheduling retry flag is triggered, high-priority computing frequency bands are bound to conservative resource entry points, and low-priority accompanying frequency bands retain the snapshot entry point of the acceptance flag; S413: Calculate the absolute value of the computing power capacity difference, and then calculate the ratio between this absolute value and the theoretical processing clock cycles corresponding to the low-priority accompanying frequency band to obtain the downgraded processing coefficient. Combine the resource quota parameters and the downgraded processing coefficient to establish the real-time computing resource allocation instruction and analysis boundary threshold. The theoretical processing clock cycles are derived from the bandwidth boundary of the low-priority accompanying frequency band, the preset spectrum analysis task type, and the processor calibration table. The downgraded processing coefficient represents the processing intensity of the low-priority accompanying frequency band in preserving energy distribution characteristics under the current resource state. After this coefficient is written into the downgrade rule field, it controls the low-priority accompanying frequency band to perform full spectrum analysis, power spectral density estimation, or inherit the marker snapshot generation. The analysis boundary threshold is generated jointly by the resource quota parameters, the downgraded processing coefficient, and the priority frequency band mapping table. The content includes the high-priority filtering boundary, the low-priority downgrade boundary, the warning trigger amplitude boundary field, and the supplementary analysis start boundary field. The real-time computing resource allocation instruction and the analysis boundary threshold are sent to S5 together. S5: Based on real-time computing resource allocation instructions and analysis boundary thresholds, the system extracts the true amplitude of bearing fault characteristic frequencies from high-priority computing frequency bands to generate immediate early warning messages. It also extracts reduced-order features from low-priority accompanying frequency bands to establish a transition marker snapshot. Supplementary analysis is performed based on the transition marker snapshot, and the immediate early warning messages are fused to generate motor operating health status analysis data. The true amplitude of bearing fault characteristic frequencies is the envelope spectrum amplitude field within the high-priority computing frequency band corresponding to the theoretical fault center frequency. The immediate early warning message is encapsulated by fault status information, over-limit deviation, window timestamp, and device identifier. The transition marker snapshot is a delayed analysis index object of the low-priority accompanying frequency band, containing energy distribution reduced-order features, memory address pointers, window timestamps, and reduced-order processing status. S511: Based on the resource quota parameters included in the real-time computing resource allocation instructions and analysis boundary thresholds, the filter parameters inside the bandpass filter are set. The bandpass filter separates the narrowband signal belonging to the high-priority computing frequency band from the effective data segments contained in the structured vibration data block sequence. The Hilbert transform algorithm is used to calculate the corresponding signal envelope of the narrowband signal. The frequency distribution parameters corresponding to the signal envelope are calculated using the fast Fourier transform algorithm to generate the true amplitude of the bearing fault characteristic frequency. The bandpass filter parameters are derived from the high-priority computing frequency band boundary, resource quota parameters, and analysis boundary thresholds. Before filter processing, the cache address and integrity identifier of the effective data segment are read to confirm that the timestamp of the current window matches the priority frequency band mapping table. If the timestamps are inconsistent, the current filtering task stops and a window mismatch status is written. After the narrowband signal enters the Hilbert transform algorithm, a signal envelope is formed, which retains the source of the high-priority frequency band and the fault type identifier. The Fast Fourier Transform algorithm performs frequency domain processing on the signal envelope. Edge nodes read the true amplitude field within the frequency position corresponding to the theoretical fault center frequency and its dynamic characteristic frequency band, and then pass the true amplitude field to S513. S512: Based on the real-time computing resource allocation instructions and the reduction processing coefficients included in the analysis boundary thresholds, power spectral density estimation is performed on the associated data segments belonging to low-priority accompanying frequency bands in the effective data segments contained in the structured vibration data block sequence. The corresponding energy distribution reduction features are extracted, and the memory address pointers corresponding to the time window within the multi-level circular buffer are obtained. The energy distribution reduction features and memory address pointers are fused to establish a transition snapshot. The reduction processing coefficients control the processing depth of the low-priority accompanying frequency bands. When resources are sufficient, a relatively complete energy distribution description is retained; when resources are limited, reduction features that indicate energy location and energy change direction are retained. The associated data segments are indexed from the effective data segments by the low-priority accompanying frequency band boundaries in the priority frequency band mapping table. The indexing process retains the window timestamp and cache address. The power spectral density estimation numerical calculation forms the energy distribution reduction features, which are simplified description fields of energy distribution along frequency positions within the low-priority accompanying frequency band. The memory address pointers come from the multi-level circular buffer written in S111. The transition snapshot binds the energy distribution reduction features with the memory address pointers and writes them to the supplementary analysis queue. S513: Obtain the adaptive safety threshold for operating conditions, calculate the over-limit deviation between the true amplitude of the bearing fault characteristic frequency and the adaptive safety threshold, determine whether the over-limit deviation exceeds the preset safety limit, and when the over-limit deviation exceeds the preset safety limit, extract the current fault status information and the over-limit deviation to perform early warning encapsulation, obtain an immediate early warning message, monitor the computing load corresponding to the multi-core processor, and when the computing load is idle, read the associated discrete vibration data points according to the memory address pointer contained in the snapshot, calculate the corresponding supplementary feature vector through the spectrum analysis algorithm, calculate the association alignment and splicing result of the supplementary feature vector and the true amplitude of the bearing fault characteristic frequency on the time axis according to the global timestamp contained in the structured vibration data block sequence, fuse the association alignment and splicing result with the immediate early warning message, and generate motor operating health status analysis data. The adaptive safety threshold for operating conditions is generated by the historical vibration amplitude parameters associated with the current operating condition when the vibration motor is in normal operating condition. The historical vibration amplitude parameters come from the window records marked as normal state in the high-priority feature cache and are bound to the operating condition label formed by the driver output frequency. The tolerance constant is stored in the equipment safety specification configuration table. The edge node merges the average state of historical vibration amplitude parameters with the tolerance constant field to form the safety threshold field corresponding to the current operating condition. When historical records are insufficient, the initial safety threshold field in the equipment safety specification configuration table is read, and the current window is marked as a historical state to be supplemented. The over-limit deviation is used to represent the deviation of the true amplitude of the bearing fault characteristic frequency from the operating condition adaptive safety threshold. The preset safety limit is formed by the maximum allowable deviation ratio parameter in the equipment safety specification and the operating condition adaptive safety threshold. The maximum allowable deviation ratio parameter is stored as a configuration field in the safety rule table. The edge node first confirms that the true amplitude field, the operating condition adaptive safety threshold field, and the preset safety limit field belong to the same window and the same operating condition label, and then performs a size comparison. When the over-limit deviation exceeds the preset safety limit, the fault status information is formed by combining the fault type identifier, frequency band source, window timestamp, and over-limit deviation status, and encapsulated into an immediate early warning message. When the computing load is idle, the edge node retrieves the received marked snapshot from the supplementary analysis queue, reads back the discrete vibration data points in the multi-level circular buffer according to the memory address pointer, and verifies the cache validity. When the cache address is overwritten or expired, the snapshot is marked as invalid and retains the reduced-order energy distribution characteristics. When the cache is valid, the spectrum analysis algorithm generates a full-band amplitude spectrum and extracts the discrete amplitude peak sequence from the full-band amplitude spectrum as a supplementary feature vector. The global timestamp is used to locate the target time series position in the historical feature database. After the supplementary feature vector is inserted into the target time series position, a multi-dimensional array merging operation is performed with the true amplitude of the bearing fault characteristic frequency at the same target time series position to generate an associated, aligned, and stitched result.The associated alignment and splicing results are fused with the real-time early warning messages to form motor operation health status analysis data. This data includes equipment identifier, window timestamp, operating condition label, start-up transient flag, complexity assessment quantity, frequency band mapping status, true amplitude field, supplementary features, early warning message status and anomaly handling identifier, and is written to the health status cache and the upper-level communication message queue.

[0019] The edge computing-based vibration motor operation data analysis system implemented in conjunction with this method includes a data preparation module, a transient identification module, a frequency band planning module, a computing power scheduling module, and a feature execution module. The data preparation module handles tasks S1, including data acquisition, analog-to-digital conversion, cache writing, time window stitching, and operating condition label generation, outputting a structured vibration data block sequence. The transient identification module handles tasks S2, including frequency migration speed, frequency band energy dispersion, comprehensive complexity index, and generation of start-up transient flags, and outputs complexity assessment and frequency migration speed to the frequency band planning module. The frequency band planning module handles tasks S3, including reading bearing kinematic parameters, generating theoretical fault center frequencies and priority frequency band mapping tables, and passing frequency band boundaries to the computing power scheduling module. The computing power scheduling module handles tasks S4, including reading available computing power, estimating required computing power, generating resource quota parameters, and generating reduction-order processing coefficients, and sends real-time computing resource allocation instructions and analysis boundary thresholds to the feature execution module. The feature execution module handles tasks S5, including extracting true amplitude, reducing order snapshots, supplementary analysis, early warning message fusion, and generating health status analysis data. The modules maintain data correspondence through window timestamps, device identifiers, operating condition tags, and cache address pointers, and abnormal states are transmitted along the same data link through the corresponding status fields.

[0020] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for analyzing the operating data of a vibration motor based on edge computing, characterized in that, Includes the following steps: S1: Acquire the simulated vibration signals collected by the acceleration sensor installed on the bearing housing of the vibratory motor in the start-up state and the stable operation state, respectively. Convert the acquired simulated vibration signals into discrete data points for buffer storage. Combine the clock signal and the driver output frequency to extract effective data segments and splice them to generate a structured vibration data block sequence containing working condition labels. S2: Calculate the frequency migration velocity and frequency band energy dispersion index based on the structured vibration data block sequence, evaluate the comprehensive complexity index by combining the frequency migration velocity and the frequency band energy dispersion index, calculate the deviation between the comprehensive complexity index and the preset calculation budget threshold, and when the deviation meets the preset transient judgment condition, use the deviation as the complexity evaluation quantity and trigger the transient marking action to generate the start transient flag bit and the complexity evaluation quantity; S3: Calculate the theoretical fault center frequency based on the motor bearing kinematic parameters, the starting transient flag, and the complexity assessment quantity. Combine the frequency migration speed to divide the high-priority calculation frequency band and the low-priority accompanying frequency band, and integrate them to generate a priority frequency band mapping table. S4: Calculate the estimated value of the required computing power based on the priority frequency band mapping table, extract the computing power capacity difference by combining the number of available computing power clock cycles, generate resource quota parameters and downgrade processing coefficients based on the computing power capacity difference, and establish real-time computing resource allocation instructions and analysis boundary thresholds. S5: Based on the real-time computing resource allocation instructions and analysis boundary thresholds, extract the true amplitude of bearing fault characteristic frequencies from the high-priority computing frequency band to generate an instant warning message, and extract downgraded features from the low-priority accompanying frequency band to establish a transfer marker snapshot. Perform supplementary analysis based on the transfer marker snapshot and fuse the instant warning message to generate motor operating health status analysis data.

2. The method for analyzing vibration motor operation data based on edge computing according to claim 1, characterized in that, The specific steps of S1 are as follows: S111: Acquire the simulated vibration signal collected by the acceleration sensor installed on the bearing housing of the vibration motor in the start-up state and the stable operation state, respectively; monitor the clock signal output by the hardware timer within the acquisition time period; convert the simulated vibration signal into discrete vibration data points through analog-to-digital conversion logic; and store the discrete vibration data points in the multi-level ring buffer of the edge node through the memory access controller to obtain the buffered discrete data points. S112: Divide the multi-level circular buffer recorded discrete data points based on the time window length, generate a global timestamp associated with the time window according to the clock signal, collect the driver output frequency to establish the working condition label corresponding to the time window, extract the divided discrete data points of the buffer storage according to the overlap ratio of adjacent time windows, generate effective data segments, and perform data splicing and combination on the effective data segments, the global timestamp and the working condition label in the order of the time axis to generate a structured vibration data block sequence.

3. The method for analyzing vibration motor operation data based on edge computing according to claim 2, characterized in that, The specific steps of S2 are as follows: S211: Based on the working condition labels contained in the structured vibration data block sequence, extract the driver output frequency of the current time window and the previous time window, calculate the time difference between the driver output frequency and the current time window and the previous time window, and obtain the frequency migration speed. S212: The frequency distribution parameters of the effective data segments contained in the structured vibration data block sequence are calculated using the Fast Fourier Transform algorithm to obtain frequency domain distribution data. The frequency domain distribution data is divided into equal-width frequency band intervals along the frequency axis. The ratio between the signal energy contained in each equal-width frequency band interval and the full-band signal energy corresponding to the frequency domain distribution data is calculated to obtain the frequency band energy ratio coefficient. Each equal-width frequency band interval is regarded as a discrete information source, and the associated frequency band energy ratio coefficient is used as the occurrence probability parameter. The Shannon information entropy is calculated based on the occurrence probability parameter to obtain the frequency band energy dispersion index. S213: Obtain a preset velocity reference value and a preset energy dispersion upper limit value; calculate the ratio between the frequency migration velocity and the preset velocity reference value to obtain a velocity migration normalization coefficient; calculate the ratio between the frequency band energy dispersion index and the preset energy dispersion upper limit value to obtain an energy dispersion normalization coefficient; perform a weighted summation calculation on the velocity migration normalization coefficient and the energy dispersion normalization coefficient to obtain a comprehensive complexity index; obtain a preset calculation budget threshold; calculate the deviation between the comprehensive complexity index and the calculation budget threshold; compare the deviation with a preset transient judgment condition; when the deviation meets the preset transient judgment condition, use the deviation as a complexity evaluation quantity and trigger a transient marking action to generate a start transient flag bit and a complexity evaluation quantity.

4. The method for analyzing vibration motor operation data based on edge computing according to claim 3, characterized in that, The process of obtaining the preset velocity reference value and the preset energy dispersion upper limit value is as follows: Extract the historical frequency sequence and historical energy sequence of the vibratory motor; The maximum rate of change contained in the historical frequency sequence is set as a preset velocity benchmark value; The maximum information entropy value associated with the historical energy sequence is set to a preset upper limit value for energy dispersion. The process of weighted summation of the velocity migration normalization coefficient and the energy dispersion normalization coefficient is as follows: Obtain evaluation values ​​for frequency changes and energy fluctuations; Calculate the first ratio of the frequency change evaluation value within the sum of the frequency change evaluation value and the energy fluctuation evaluation value, and establish the first ratio as the first weighting coefficient; Calculate the second ratio of the energy fluctuation evaluation value within the sum of the frequency change evaluation value and the energy fluctuation evaluation value, and establish the second ratio as the second weighting coefficient; Calculate the first product between the velocity migration normalization coefficient and the first weighting coefficient, and calculate the second product between the energy dispersion normalization coefficient and the second weighting coefficient; Calculate the sum of the first product and the second product to obtain the overall complexity index; The process of obtaining a preset computational budget threshold, calculating the deviation between the comprehensive complexity index and the computational budget threshold, and comparing the deviation with a preset transient judgment condition specifically includes: Extract the quantified value of available computing clock cycles from the edge computing node, and set the quantified value of available computing clock cycles as the computing budget threshold; Obtain the fault tolerance ratio parameter and calculate the boundary product between the fault tolerance ratio parameter and the calculation budget threshold; The numerical range formed by the product of zero and the boundary is set as a preset transient determination condition; Calculate the deviation between the overall complexity index and the calculation budget threshold. Determine whether the deviation amount belongs to the numerical range corresponding to the preset transient determination condition.

5. The method for analyzing vibration motor operation data based on edge computing according to claim 3, characterized in that, The specific steps of S3 are as follows: S311: Obtain the stored pitch diameter, contact angle and number of rolling elements of the motor bearing, establish the kinematic parameters of the motor bearing, respond to the start transient flag, evaluate the instantaneous speed estimate based on the complexity evaluation quantity and the driver output frequency, substitute the instantaneous speed estimate and the kinematic parameters of the motor bearing into the kinematic equation of the motor bearing for calculation, and obtain the theoretical fault center frequency of the differentiated fault type. S312: Based on the frequency migration speed, calculate the frequency band coverage range extending to both sides of each theoretical fault center frequency, generate multiple dynamic characteristic frequency band intervals, perform boundary merging and frequency truncation processing on the multiple dynamic characteristic frequency band intervals, and establish the independent or merged intervals obtained after processing as high-priority calculation frequency bands. S313: Remove the coverage area corresponding to the high-priority calculation frequency band from the preset full-band analysis area to generate a remaining bandwidth area. Establish the remaining bandwidth area as a low-priority accompanying frequency band. Integrate the frequency boundary information within the high-priority calculation frequency band and the low-priority accompanying frequency band under the current time window to generate a priority frequency band mapping table.

6. The method for analyzing vibration motor operation data based on edge computing according to claim 5, characterized in that, The process of calculating the frequency band coverage range extending to both sides of each theoretical fault center frequency based on the frequency migration speed, and generating multiple dynamic characteristic frequency band intervals, specifically involves: Obtain the length of a preset time window, extract the maximum tolerance value of frequency fluctuation associated with historical stable operation phases, calculate the product between the time length and the maximum tolerance value of frequency fluctuation, and obtain the benchmark expansion coefficient. Calculate the product between the frequency shift rate and the reference extension factor to obtain the one-sided frequency extension margin; The upper limit frequency threshold is obtained by summing the theoretical fault center frequency and the single-sided frequency extension margin. The difference between the theoretical fault center frequency and the single-sided frequency extension margin is calculated to obtain the lower limit frequency threshold. The continuous frequency range defined by the lower frequency threshold and the upper frequency threshold is set as the dynamic characteristic frequency band interval; The process of performing boundary merging and frequency truncation processing on multiple dynamic feature frequency bands and establishing the resulting independent or merged intervals as high-priority calculation frequency bands is as follows: Obtain the upper limit value of the Nyquist frequency corresponding to the analog-to-digital conversion action, and compare the upper limit frequency threshold contained in each of the dynamic characteristic frequency bands with the upper limit value of the Nyquist frequency. When the upper limit frequency threshold is greater than the upper limit value of the Nyquist frequency, the upper limit frequency boundary of the dynamic characteristic frequency band interval is replaced with the upper limit value of the Nyquist frequency to obtain the corresponding corrected characteristic frequency band interval. Determine whether there are overlapping frequency band regions among the multiple modified feature frequency band intervals; When the overlapping frequency band regions exist, extract the minimum lower limit frequency value and the maximum upper limit frequency value associated with the multiple overlapping modified feature frequency band intervals. Based on the minimum lower limit frequency value and the maximum upper limit frequency value, perform boundary merging processing on multiple overlapping modified characteristic frequency band intervals to establish a high-priority calculation frequency band.

7. The method for analyzing vibration motor operation data based on edge computing according to claim 5, characterized in that, The specific steps of S4 are as follows: S411: Read the operating system's performance counters to obtain the available computing power clock cycles of the multi-core processor, parse the high-priority computing frequency band and the low-priority accompanying frequency band contained in the priority frequency band mapping table, calculate the number of multiply-accumulate operations and memory accesses based on the bandwidth and frequency resolution within the high-priority computing frequency band, obtain the clock cycles consumed per multiply-accumulate operation and the clock cycles consumed per memory access, calculate the product between the number of multiply-accumulate operations and the clock cycles consumed per multiply-accumulate operation to obtain the computing clock cycle requirement, calculate the product between the number of memory accesses and the clock cycles consumed per memory access to obtain the memory access clock cycle requirement, and calculate the sum of the computing clock cycle requirement and the memory access clock cycle requirement to obtain the estimated value of the required computing power requirement; S412: Compare the number of available computing power clock cycles with the estimated value of required computing power to obtain the computing power capacity difference, obtain computing power allocation constraint rules, determine the allocation ratio of the number of available computing power clock cycles among different priority computing frequency bands according to the computing power allocation constraint rules and the computing power capacity difference, divide the number of available computing power clock cycles according to the allocation ratio, and generate resource quota parameters. S413: Calculate the absolute value of the computing power capacity difference, calculate the ratio between the absolute value and the theoretical processing clock cycle number corresponding to the low priority accompanying frequency band, obtain the downgraded processing coefficient, combine the resource quota parameter and the downgraded processing coefficient, and establish a real-time computing resource allocation instruction and analysis boundary threshold.

8. The method for analyzing vibration motor operation data based on edge computing according to claim 7, characterized in that, The specific steps of S5 are as follows: S511: Based on the resource quota parameters included in the real-time computing resource allocation instruction and the analysis boundary threshold, set the filter parameters inside the bandpass filter, separate the narrowband signal belonging to the high-priority computing frequency band from the effective data segments included in the structured vibration data block sequence through the bandpass filter, calculate the corresponding signal envelope by the Hilbert transform algorithm, calculate the frequency distribution parameters corresponding to the signal envelope by the fast Fourier transform algorithm, and generate the true amplitude of the bearing fault characteristic frequency. S512: Based on the real-time computing resource allocation instruction and the order reduction processing coefficient included in the analysis boundary threshold, perform power spectral density estimation numerical calculation on the associated data segments belonging to the low-priority accompanying frequency band in the effective data segments included in the structured vibration data block sequence, extract the corresponding energy distribution order reduction features, obtain the memory address pointer corresponding to the time window inside the multi-level ring buffer, fuse the energy distribution order reduction features and the memory address pointer, and establish a transition marker snapshot; S513: Obtain the adaptive safety threshold for operating conditions, calculate the excess deviation between the true amplitude of the bearing fault characteristic frequency and the adaptive safety threshold for operating conditions, determine whether the excess deviation exceeds a preset safety limit, and when the excess deviation exceeds the preset safety limit, extract the current fault status information and the excess deviation to perform early warning encapsulation to obtain an immediate early warning message. Monitor the computing load corresponding to the multi-core processor, and when the computing load is idle, read the associated discrete vibration data points according to the memory address pointer contained in the receiving mark snapshot. Calculate the associated discrete vibration data points using a spectrum analysis algorithm to obtain the corresponding supplementary feature vector. Calculate the association alignment and splicing result of the supplementary feature vector and the true amplitude of the bearing fault characteristic frequency on the time axis according to the global timestamp contained in the structured vibration data block sequence. Merge the association alignment and splicing result with the immediate early warning message to generate motor operating health status analysis data.

9. The method for analyzing vibration motor operation data based on edge computing according to claim 8, characterized in that, The process of obtaining the adaptive safety threshold for operating conditions, calculating the excess deviation between the true amplitude of the bearing fault characteristic frequency and the adaptive safety threshold for operating conditions, and determining whether the excess deviation exceeds the preset safety limit is specifically as follows: Extract historical vibration amplitude parameters of the vibratory motor under normal operating conditions and correlated with the current working conditions; Calculate the average value of the historical vibration amplitude parameters, calculate the sum of the average value and the tolerance constant, and set the sum as the working condition adaptive safety threshold. Calculate the absolute value of the difference between the true amplitude of the bearing fault characteristic frequency and the adaptive safety threshold of the operating condition to obtain the excess deviation; Obtain the maximum allowable deviation ratio parameter corresponding to the equipment safety specification, calculate the boundary product value between the maximum allowable deviation ratio parameter and the working condition adaptive safety threshold, and set the boundary product value as a preset safety limit. Compare the value of the excess deviation with the value of the preset safety limit to determine whether the excess deviation is greater than the preset safety limit; The process of calculating the corresponding supplementary feature vector by using a spectrum analysis algorithm on the associated discrete vibration data points, and calculating the correlation and alignment of the supplementary feature vector with the true amplitude of the bearing fault characteristic frequency on the time axis based on the global timestamp contained in the structured vibration data block sequence is as follows: Perform a discrete Fourier transform on the discrete vibration data points to obtain the full-band amplitude spectrum; Extract the discrete amplitude peak sequence retained within the full-band amplitude spectrum, and set the discrete amplitude peak sequence as a supplementary feature vector; Extract the global timestamp contained in the structured vibration data block sequence, and locate the target time series position that perfectly matches the global timestamp value within the historical feature database; The supplementary feature vector is inserted into the target time series position, and a multidimensional array merging operation is performed with the true amplitude of the bearing fault feature frequency at the target time series position to generate an associated aligned splicing result.

10. A vibration motor operation data analysis system based on edge computing, characterized in that, The system is used to implement the edge computing-based vibration motor operation data analysis method as described in any one of claims 1-9, and the system includes: The data preparation module acquires the simulated vibration signals collected by the acceleration sensor installed on the bearing housing of the vibratory motor in the start-up state and the stable operation state, respectively. It converts the acquired simulated vibration signals into discrete data points for buffer storage, extracts effective data segments by combining the clock signal and the driver output frequency, and splices them to generate a structured vibration data block sequence containing working condition labels. The transient identification module calculates the frequency migration velocity and frequency band energy dispersion index based on the structured vibration data block sequence, evaluates the comprehensive complexity index by combining the frequency migration velocity and the frequency band energy dispersion index, calculates the deviation between the comprehensive complexity index and the preset calculation budget threshold, and when the deviation meets the preset transient judgment condition, uses the deviation as the complexity evaluation quantity and triggers the transient marking action to generate a transient start flag bit and a complexity evaluation quantity. The frequency band planning module calculates the theoretical fault center frequency based on the motor bearing kinematic parameters, the start-up transient flag, and the complexity evaluation quantity. It then divides the frequency band into high-priority calculation frequency bands and low-priority accompanying frequency bands based on the frequency migration speed, and integrates them to generate a priority frequency band mapping table. The computing power scheduling module calculates the estimated value of the required computing power based on the priority frequency band mapping table, extracts the computing power capacity difference by combining the number of available computing power clock cycles, generates resource quota parameters and downgrade processing coefficients based on the computing power capacity difference, and establishes real-time computing resource allocation instructions and analysis boundary thresholds. The feature execution module, based on the real-time computing resource allocation instructions and analysis boundary thresholds, extracts the true amplitude of bearing fault characteristic frequencies from the high-priority computing frequency band to generate an instant warning message, and extracts downgraded features from the low-priority accompanying frequency band to establish a transfer marker snapshot. Based on the transfer marker snapshot, it performs supplementary analysis and merges the instant warning message to generate motor operating health status analysis data.

Images