Spinning machine life cycle management system based on big data model

Abstract

The invention relates to the technical field of life cycle management, in particular to a spinning machine life cycle management system based on a big data model, which comprises a signal analysis module, a pass division module, a state recognition module, a cycle modeling module and a cycle optimization module. According to the method, spinning pressure signals and time sequences are synchronously collected and subjected to peak-valley scanning, pressure fluctuation data are formed, details of the machining process are continuously depicted, pass boundary judgment and section statistical processing are combined, pressure change rules are structurally presented, continuous fluctuation section combination is promoted through consistency verification, and the machining precision is improved. A clear mark is obtained for an implicit operation state, a state transition relation is deduced based on cross-pass occurrence frequency and continuous characteristics, an equipment behavior evolution path tends to be clear, a life cycle result is converted from static recording to dynamic analysis, and after deviation distribution verification and grading adjustment are synchronously introduced, a pressure abnormity trend can be identified in time. The maintenance rhythm and the real working condition are kept coordinated, and the management judgment basis is more coherent.

Description

Technical Field

[0001] This invention relates to the field of lifecycle management technology, and in particular to a spinning machine lifecycle management system based on a big data model. Background Technology

[0002] Lifecycle management technology refers to the technical fields of information recording, status tracking, process analysis, and management control carried out around the entire process of equipment or products from planning and design, manufacturing and installation, operation and use, maintenance and repair to decommissioning and scrapping. Its core matters include the division of the entire life cycle of the object, the continuous collection and storage of life cycle data, the collation and correlation of historical status and behavior information, the establishment of management rules and evaluation criteria, and the unified coordination of multi-stage management activities. It is widely used in scenarios such as industrial equipment management, production operation management, and asset management.

[0003] The traditional spinning machine lifecycle management system refers to a technical solution for the spinning machine, a metal plastic forming equipment, to manage basic parameters, operating time, processing tasks, fault conditions, and maintenance records separately through manual recording or simple information systems during its design and filing, commissioning, daily operation, maintenance, and disposal stages. It typically adopts methods such as manually filling in equipment ledgers, periodically summarizing operating data, arranging maintenance plans according to fixed cycles, and judging equipment status through experience rules to manage relevant information of the spinning machine in a decentralized manner at multiple lifecycle stages.

[0004] Traditional spinning machine lifecycle management relies on manual ledgers and fixed-period summaries. The recording of operating data is lagging and discrete, making it difficult to present the continuous pressure change characteristics during processing. Historical status is mostly archived in the form of results, and process fluctuations lack correlation and organization, resulting in insufficient identification of equipment behavior trajectory. Experience rules dominate status judgment, making it difficult to identify implicit evolutionary relationships when facing multi-pass processing. Abnormal trends of equipment are not easily detected in advance, and the matching degree between maintenance arrangements and actual wear is low. After long-term operation, maintenance rhythm is prone to imbalance, which further leads to a decline in equipment utilization efficiency, a single basis for management decisions, and limited lifecycle information coordination capabilities. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a spinning machine lifecycle management system based on a big data model. The technical solution is as follows: On the one hand, a spinning machine lifecycle management system based on a big data model is provided, which includes: The signal analysis module acquires the analog electrical signal and timestamp output by the swirling force sensor, performs analog-to-digital conversion and sampling order verification, analyzes the swirling force time series and scans for peak and valley points, generates peak and valley sequence data and transmits it to the pass division module. The pass division module obtains the processing pass number, performs pass boundary judgment on the spun pressure sampling points, divides the peak and valley sequence data of the same pass into pressure fluctuation segments, calculates the average time interval between adjacent peaks for each pressure fluctuation segment, generates segment statistics data and transmits it to the status recognition module. The state recognition module determines the consistency between the peak position type and the peak-valley alternation sequence of adjacent pressure fluctuation segments based on the segment statistics, combines and marks continuous pressure fluctuation segments that meet the consistency condition, generates hidden state segment data, and transmits it to the periodic modeling module. The cycle modeling module maps the hidden state segment data to the corresponding processing pass number, constructs a probability transition matrix based on the frequency and duration of the hidden state in adjacent passes, performs topological analysis on the evolution path between hidden states, and generates life cycle management results. The cycle optimization module, based on the life cycle management results, calls the time series data of the swirling pressure in the corresponding hidden state segment, performs consistency difference judgment and hierarchical adjustment processing on the deviation distribution between the pressure value and the segment mean in the hidden state segment, and generates life cycle management optimization results.

[0006] As a further embodiment of the present invention, the peak-valley sequence data includes peak timestamp, peak amplitude, valley time index, and valley amplitude; the segment statistics include pressure fluctuation segment number, average time interval between adjacent peaks, and segment duration; the hidden state segment data includes hidden state number, segment start number, segment end number, and segment coverage length; the lifecycle management results include hidden state transition probability, hidden state evolution path, and path duration; and the lifecycle management optimization results include segment pressure deviation level, segment pressure correction amount, and hidden state stability index.

[0007] As a further aspect of the present invention, the signal analysis module includes: The signal acquisition submodule acquires the analog electrical signal output by the swirling force sensor, synchronously obtains the corresponding timestamp information, performs sequential pairing and buffering of the electrical signal amplitude and timestamp, extracts the original swirling force record sequence, writes the acquisition time identifier and completes the integrity check, and generates swirling force time series data. The sampling verification submodule performs analog-to-digital conversion on the cyclone pressure time series data, rearranges the discrete amplitudes according to the preset sampling frequency, compares the differences between adjacent timestamps to determine whether the sampling interval threshold is met, performs sequential verification marking on abnormal interval positions, and obtains the cyclone pressure sampling sequence. The peak-valley scanning submodule calls the cyclone sampling sequence, performs differential calculation on the amplitude of adjacent sampling points, determines the location of local extrema based on the change of the differential sign, indexes and records sampling points that meet the peak value determination conditions and valley value determination conditions, integrates the extrema point information in chronological order, and generates peak-valley sequence data.

[0008] As a further aspect of the present invention, the sampling interval threshold is determined by performing a reciprocal operation on the rated sampling frequency to obtain a standard sampling period reference, and then combining the clock jitter tolerance and the data transmission delay error range, and superimposing the maximum permissible deviation value on the standard sampling period reference.

[0009] As a further aspect of the present invention, the track division module includes: The processing pass numbering submodule obtains the processing pass number based on the peak-valley sequence data, detects the timestamp sequence of the swirling pressure sampling points, reads the time difference between adjacent sampling points and makes a continuity judgment, and classifies the sampling points sequentially according to the trend of time difference changes to generate a pass numbering sequence. The segment division submodule judges the pass boundaries of the swirling pressure sampling points according to the pass numbering sequence, locates the start and end positions of the pass, continuously combines the peak-valley sequence data associated within the same pass, and completes the segment division according to the alternation relationship between peak and valley values ​​to generate a pressure fluctuation segment set. The segment statistics submodule calls the pressure fluctuation segment set, obtains the peak timestamp sequence for each pressure fluctuation segment, sums and normalizes the time differences between adjacent peaks to obtain the average time interval within the segment, summarizes the track number and corresponding average information, and generates segment statistics data.

[0010] As a further aspect of the present invention, the state recognition module includes: The peak location submodule detects the peak distribution location type of adjacent pressure fluctuation segments based on the segment statistics, obtains the peak timestamps and segment boundary indices of multiple segments, classifies the locations based on the relative order of the indices, and generates a peak location distribution vector. The alternating sequence submodule, based on the peak position distribution vector, calls the peak-valley time series of adjacent pressure fluctuation segments, compares the order of occurrence of peaks and valleys, counts the sequence consistency and sequence difference of adjacent segments, analyzes the sequence correlation between segments, and obtains the peak-valley alternation consistency degree. The segment combination submodule, based on the peak-valley alternation consistency, calls the adjacent pressure fluctuation segment index sequence, performs consistency judgment and combination selection on continuous segments, records and summarizes the mapping relationship between the start and end indices of the combined segments and the segment number, and generates hidden state segment data.

[0011] As a further aspect of the present invention, the periodic modeling module includes: The state alignment submodule extracts the start and end indexes of the segment based on the hidden state segment data and the corresponding processing turn number, monitors the consistency of hidden state numbers in adjacent turns, completes segment sequence alignment and number mapping adjustment according to the turn order, and generates a hidden state turn alignment sequence. The transfer statistics submodule, based on the hidden state track alignment sequence, calls the combination of hidden state numbers in adjacent tracks, calculates the number of occurrences of multiple number combinations in the track sequence and the cumulative length of the corresponding segment span and normalizes them, and applies a preset weight coefficient to calculate the transfer occurrence rate to obtain the hidden state transfer probability set. The path aggregation submodule compares the transition probability of each hidden state starting number with the hidden state transition probability set, determines the relationship items that exceed the transition probability threshold, connects them in the order of the path number to form a continuous state evolution path, records the number of paths covered by the path and the length of the state dwell segment, and generates life cycle management results.

[0012] As a further aspect of the present invention, the period optimization module includes: The data access submodule, based on the lifecycle management results, calls the corresponding hidden state segment's vortex pressure time series data, obtains the start and end time indices of multiple hidden state segments, collects continuous pressure sampling values ​​within the segment, and splices the sequences according to the time index order to generate a segment pressure sequence set. The deviation assessment submodule obtains pressure sample values ​​within multiple hidden state segments based on the segment pressure sequence set, calculates the arithmetic mean of pressure in the same segment, performs difference calculation on the multiple sample values ​​and the segment mean, and statistically analyzes the distribution frequency of the absolute value of the difference to obtain the pressure deviation distribution. The graded adjustment submodule calls the pressure deviation distribution, calculates the deviation amplitude and compares it with the preset consistency difference threshold, divides the deviation level sequence according to the consistency difference threshold, maps the pressure correction amount to the multi-level associated hidden state segments, summarizes the segment correction results, and generates life cycle management optimization results.

[0013] As a further aspect of the present invention, the consistency difference threshold is determined by obtaining the absolute values ​​of all pressure deviations within the same segment, sorting them by numerical value, extracting the median value of the deviation as a distribution center characterization quantity, and then combining it with the deviation dispersion range corresponding to the median value.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By synchronously acquiring counter-rotating pressure signals and time series data and conducting peak-valley scanning, pressure fluctuation data with sequential characteristics is generated, enabling continuous characterization of processing details. Combined with pass boundary determination and section statistical processing, the pressure change pattern is presented in a structured manner. Consistency verification facilitates the combination of continuous fluctuation sections, and implicit operating states are clearly marked. Based on the frequency and continuous characteristics of cross-pass occurrences, the state transition relationship is deduced, and the equipment behavior evolution path becomes clearer. The life cycle results shift from static recording to dynamic analysis. With the synchronous introduction of deviation distribution verification and graded adjustment, abnormal pressure trends can be identified in a timely manner, the maintenance rhythm is coordinated with the actual working conditions, and the management judgment basis is more consistent. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the signal analysis module in this invention; Figure 4 This is a flowchart of the pass division module in this invention; Figure 5 This is a flowchart of the state recognition module in this invention; Figure 6 This is a flowchart of the periodic modeling module in this invention; Figure 7 This is a flowchart of the period optimization module in this invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] This invention provides a spinning mill lifecycle management system based on a big data model, such as... Figure 1-2 The diagram shown illustrates a spinning machine lifecycle management system based on a big data model. This system includes: The signal analysis module acquires the analog electrical signal and timestamp output by the swirling force sensor, performs analog-to-digital conversion and sampling order verification, analyzes the swirling force time series and scans for peak and valley points, generates peak and valley sequence data and transmits it to the pass division module. The pass division module obtains the processing pass number, judges the pass boundary of the spun pressure sampling point, divides the peak and valley sequence data of the same pass into pressure fluctuation segments, calculates the average time interval between adjacent peaks for each pressure fluctuation segment, generates segment statistics data and transmits it to the status recognition module. The state recognition module determines the consistency of the peak position type and peak-valley alternation sequence of adjacent pressure fluctuation segments based on the segment statistics, combines and marks continuous pressure fluctuation segments that meet the consistency conditions, generates hidden state segment data, and transmits it to the periodic modeling module. The cycle modeling module maps the hidden state segment data to the corresponding processing stage number, constructs a probability transition matrix based on the frequency and duration of the hidden state in adjacent stages, and performs topological analysis on the evolution path between hidden states to generate life cycle management results. The cycle optimization module, based on the life cycle management results, calls the time series data of the swirling pressure in the corresponding hidden state segment, performs consistency difference judgment and hierarchical adjustment on the deviation distribution between the pressure value and the segment mean in the hidden state segment, and generates life cycle management optimization results.

[0020] Peak-valley sequence data includes peak timestamp, peak amplitude, valley time index, and valley amplitude. Segment statistics include pressure fluctuation segment number, average time interval between adjacent peaks, and segment duration. Hidden state segment data includes hidden state number, segment start track number, segment end track number, and segment coverage length. Lifecycle management results include hidden state transition probability, hidden state evolution path, and path duration. Lifecycle management optimization results include segment pressure deviation level, segment pressure correction amount, and hidden state stability index.

[0021] Specifically, such as Figure 2 , 3 As shown, the signal analysis module includes: The signal acquisition submodule acquires the analog electrical signal output by the swirling force sensor, synchronously obtains the corresponding timestamp information, performs sequential pairing and buffering of the electrical signal amplitude and timestamp, extracts the original swirling force record sequence, writes the acquisition time identifier and completes the integrity check, and generates swirling force time series data. The analog signal input interface is physically connected to a piezoelectric thin-film sensor with a range of 0 to 1000 Newtons, outputting a continuous analog voltage of 0 to 5 volts. The clock synchronization unit uses a 64-bit hardware counter to capture the clock signal in real time with a 1-microsecond time resolution. Upon capturing the rising edge of a voltage change at the analog signal input interface, it immediately triggers a latch operation to obtain the timestamp information of the current moment. The signal pairing unit maps the captured analog voltage amplitude to the latched timestamp, constructing a key-value pair data packet containing the "original voltage amplitude" and "acquisition time," and pushes it into a dual-channel buffer memory. This buffer memory uses a first-in, first-out (FIFO) queue mechanism to temporarily store data packets in real time and retrieve them sequentially. The acquisition time identifier writing unit adds an absolute time tag based on Coordinated Universal Time (UTC) to the beginning of each record sequence. The integrity verification logic unit is configured to perform a signal validity check operation based on the amplitude range. Specifically, the integrity verification logic unit acquires the currently recorded raw voltage amplitude, the reference low voltage value corresponding to the lower limit of the sensor range, and the reference high voltage value corresponding to the upper limit of the sensor range. Then, this unit executes numerical comparison logic to determine whether the raw voltage amplitude is greater than or equal to the reference low voltage value and simultaneously less than or equal to the reference high voltage value. If both comparison conditions are met, the record is deemed complete and valid; if either condition is not met, an anomaly marking logic is triggered. For example, when the reference low voltage value corresponding to the lower limit of the sensor range is set to 0.05 volts, the reference high voltage value corresponding to the upper limit is set to 4.95 volts, and the currently acquired raw voltage amplitude is 2.5 volts, the unit determines that 2.5 volts is within the range of 0.05 volts to 4.95 volts, and thus marks the data as valid; conversely, if the acquired value is 5.1 volts, it determines that the data overflows and marks it as an anomaly, finally outputting the integrity-filtered rotational pressure time series data.

[0022] The sampling and verification submodule performs analog-to-digital conversion on the swirling pressure time series data, rearranges the discrete amplitudes according to the preset sampling frequency, compares the differences between adjacent timestamps to determine whether the sampling interval threshold is met, marks the abnormal interval positions with sequential verification, and obtains the swirling pressure sampling sequence. The system receives analog voltage amplitudes from the swirl pressure time-series data, converts them into digital signals through quantization logic, and further maps them to physical pressure values. This conversion logic performs the following operations: acquiring the original analog voltage value, the full-scale resolution value of the analog-to-digital converter, the physical range span value of the sensor, and the zero-point offset compensation value; first, dividing the original analog voltage value by the full-scale resolution value to obtain a scaling factor; then multiplying this scaling factor by the physical range span value to obtain a preliminary pressure value; finally, adding the preliminary pressure value to the zero-point offset compensation value to obtain the discrete amplitude of the swirl pressure. For example, when the original analog voltage value is 2.5 volts, the voltage reference corresponding to the full-scale resolution is 5 volts, the physical range span value is 1000 Newtons, and the zero-point offset compensation value is 0 Newtons, the calculated discrete amplitude is 500 Newtons. The discrete data rearrangement unit aligns the discrete amplitudes in a grid on the time axis according to a preset sampling frequency of 1000 Hz. The time interval comparator is equipped with strict timing verification logic to monitor the continuity of the data stream. The comparator acquires the timestamps of two adjacent sampling points, i.e., the current sampling time value and the previous sampling time value, and performs a subtraction operation to obtain the actual sampling interval value. Simultaneously, it acquires the standard sampling period value and the allowable jitter tolerance ratio. By multiplying the standard sampling period value by the jitter tolerance ratio and adding it back, the high threshold of the sampling interval is calculated. The comparator then compares the actual sampling interval value with the high threshold. If the actual sampling interval value is greater than the high threshold, it determines that packet loss or transmission delay has occurred, and the position is marked for sequence verification. For example, if the standard sampling period value is set to 1 millisecond (corresponding to 1000 Hz) and the jitter tolerance ratio is 0.1 (i.e., 10%), the calculated high threshold of the sampling interval is 1.1 milliseconds. If the actual difference between the current sampling time value and the previous sampling time value is 1.2 milliseconds, since 1.2 milliseconds is greater than 1.1 milliseconds, the position is marked as an abnormal interval position, and the final output is a time-washed spin compression sampling sequence.

[0023] The peak-valley scanning submodule calls the cyclone sampling sequence, performs differential calculation on the amplitude of adjacent sampling points, determines the location of local extrema based on the change of the differential sign, indexes and records sampling points that meet the peak value determination conditions and valley value determination conditions, integrates the extrema point information in chronological order, and generates peak-valley sequence data. The pressure sampling sequence is invoked, and a first-order difference operation is performed on the sampling points arranged in chronological order within the sequence. The operation logic is as follows: the pressure amplitude of the current sampling point is obtained and compared with the pressure amplitude of the previous adjacent sampling point. A subtraction operation is then performed, subtracting the pressure amplitude of the previous adjacent sampling point from the current sampling point's pressure amplitude to obtain the amplitude difference value at the current position. The extreme value determination logic unit is configured with sign change monitoring logic to identify waveform inflection points. The amplitude difference value at the current position is obtained and compared with the amplitude difference value at the previous position. The extreme value nature is determined by comparing their sign attributes (positive or negative). Specifically, if the amplitude difference value at the previous position is positive and the amplitude difference value at the current position is negative, it indicates that the value has undergone a process of rising to falling, and the current sampling point is determined to meet the peak value determination condition; conversely, if the amplitude difference value at the previous position is negative and the amplitude difference value at the current position is positive, it indicates that the value has undergone a process of falling to rising, and the current sampling point is determined to meet the valley value determination condition. In addition, to filter out minor fluctuations, a minimum peak-valley significance threshold is introduced. Only when the absolute value of the difference between the pressure amplitude of a sampling point that meets the peak or valley determination criteria and the local baseline exceeds this threshold is it confirmed as a valid extreme value. The index recorder extracts and records the position index, timestamp, and pressure amplitude of the confirmed valid extreme value points in the original sequence. The sequence integration unit interweaves and arranges all recorded peak and valley information in chronological order to generate the final peak-valley sequence data. As shown in Table 1, this table displays the cyclone pressure peak-valley characteristic data fragments generated after the above logical processing.

[0024] Table 1. Characteristics of peak-valley sequences in cyclone pressure monitoring

[0025] As shown in Table 1, each row of data represents an identified and recorded extreme event, clearly listing the exact time of the event, the physical quantity value, and its waveform characteristics, thus verifying the logical effectiveness of the peak-valley scanning submodule in extracting features from dynamic pressure signals.

[0026] Specifically, such as Figure 2 , 4 As shown, the track division module includes: The pass numbering submodule obtains the processing pass number based on peak-valley sequence data, detects the timestamp sequence of the spun pressure sampling points, reads the time difference between adjacent sampling points and makes a continuity judgment, classifies the sampling points in order according to the trend of time difference change, and generates a pass numbering sequence. The system directly reads the acquisition time information corresponding to all extreme points from the peak-valley sequence data in index order to construct a one-dimensional timestamp vector. The adjacent time difference calculator traverses this one-dimensional timestamp vector, obtains the timestamp value of the current extreme point and the timestamp value of the previous extreme point, and performs a subtraction operation, subtracting the previous timestamp value from the current timestamp value to obtain the time difference between adjacent sampling points. The pass interval determination logic unit has a preset pass switching determination threshold. This threshold is set based on the time consumed by the zeroing or tool changing action between different passes of the spinning machine, and is usually set to 1.5 to 2 times the processing cycle interval. The calculated time difference between adjacent sampling points is compared with the pass switching determination threshold: if the time difference between adjacent sampling points is less than or equal to the pass switching determination threshold, it is determined that the current point belongs to the same processing pass, and the current pass number remains unchanged; if the time difference between adjacent sampling points is greater than the pass switching determination threshold, it is determined that a pass switching has occurred, triggering a counter to increment by one and generating a new pass number. Based on the judgment result, the sequence classification marker writes the corresponding pass number into the attribute field of each sampling point, generating a pass number sequence containing pass assignment information. For example, if the pass switching judgment threshold is set to 3000 milliseconds (i.e., 3 seconds), and the timestamp of the previous extreme point is read as 15000 milliseconds, and the timestamp of the current extreme point is 15100 milliseconds, the difference is calculated to be 100 milliseconds. Since 100 milliseconds is less than 3000 milliseconds, it is classified as the current pass. If the timestamp of the next extreme point is 20000 milliseconds, the difference from the previous point is 4900 milliseconds. Since 4900 milliseconds is greater than 3000 milliseconds, it is determined to enter the next pass, and the pass number automatically increments. The advantage of this operation logic is that by setting a dynamic time threshold based on the physical action time, it can accurately and automatically identify and segment non-continuous machining processes in a continuous data stream without relying on the command signals of an external machine tool controller. The segment division submodule determines the segment boundary of the cyclone pressure sampling points according to the segment number sequence, locates the start and end positions of the segment, continuously combines the associated peak and valley sequence data within the same segment, and completes the segment division based on the alternation relationship between peak and valley values ​​to generate a pressure fluctuation segment set. Based on the pass number sequence, the index positions of the first and last occurrences of each independent pass number in the sequence are retrieved and identified as the pass start and end points, respectively. The validity filtering logic unit is configured to eliminate idle stroke data, obtain the pressure amplitude of each peak point within the pass, and compare it with a preset effective contact pressure benchmark. This benchmark is based on the minimum forming force setting of the spinning process, for example, set to 50 Newtons. If the pressure amplitude of a peak point is less than the effective contact pressure benchmark, it is determined as a non-contact idle point and eliminated; if it is greater than or equal to the benchmark, it is retained as a valid processing point. The waveform combiner continuously scans the retained valid processing points, combining continuous and complete fluctuation cycles into candidate waveform groups based on the alternating pattern of peaks and valleys on the time axis (i.e., the order of "peak-valley-peak" or "valley-peak-valley"). The segmentation generator segments the data based on the number of consecutive waveform groups. When the number of consecutive valid waveform groups reaches the preset minimum segment length (e.g., 10 complete fluctuation cycles), it is defined as an independent pressure fluctuation segment and assigned a unique identifier, generating a pressure fluctuation segment set. For example, in the data stream with track number 1, the pressure amplitude of data points with indices 100 to 200 is all above 500 Newtons and exhibits continuous peak-valley alternation, thus dividing it into "track 1 - segment 1"; while the data points with indices 201 to 250, although belonging to track 1, have a pressure amplitude of only 10 Newtons (less than the 50 Newton benchmark), and are determined to be an empty travel interval and not generated as a fluctuation segment. This logic, through threshold filtering of physical pressure and integrity constraints of waveform topology, ensures that the generated segment set only contains valid processing data that produces actual plastic deformation.

[0027] The section statistics submodule calls the pressure fluctuation section set, obtains the peak timestamp sequence for each pressure fluctuation section, sums and normalizes the time difference between adjacent peaks to obtain the average time interval within the section, summarizes the track number and corresponding average information, and generates section statistics data. The algorithm iterates through each independent segment in the pressure fluctuation segment set, extracting the timestamps of all peak points within that segment to form a peak timestamp subsequence. A differential accumulation operation is performed on the peak timestamp subsequence. Specifically, starting from the second timestamp of the sequence, the current peak timestamp is obtained sequentially, and the previous peak timestamp is subtracted to obtain the single-cycle time interval. All single-cycle time intervals within the segment are then summed to obtain the total time interval value. The normalization unit obtains the total number of peaks within the segment, subtracts 1 from the total number of peaks to obtain the number of interval items, and then performs a division operation, dividing the total time interval value by the number of interval items to calculate the average time interval within the segment. The statistical data assembler associates and stores the track number, segment identifier, and the calculated average time interval within the segment, generating the segment statistical data as shown in Table 2. For example, a segment contains 5 peak points with timestamps of 1000 milliseconds, 1100 milliseconds, 1205 milliseconds, 1300 milliseconds, and 1400 milliseconds. The cumulative summation unit calculates adjacent differences of 100 milliseconds, 105 milliseconds, 95 milliseconds, and 100 milliseconds, summing them to obtain a total time interval of 400 milliseconds. The normalization unit subtracts 1 from the total number of peak values ​​(5-1=4), resulting in 4 interval terms. Then, it calculates 400 / 4=100, yielding an average time interval of 100 milliseconds within the segment. The advantage of this operational logic is that by averaging multiple time intervals, it effectively smooths out instantaneous errors caused by sensor sampling jitter or localized minute speed variations, accurately quantifying the overall temporal rhythm characteristics of the processing segment.

[0028] Table 2 Statistical Parameters of Spinning Process Sections

[0029] As shown in Table 2, the table details the key statistical indicators after processing by the above modules. The "mean time interval" column directly reflects the spinning pressure fluctuation period in a specific processing pass and section. The stability of this value (e.g., 100.0 milliseconds for the S-01 section of pass 001) directly corresponds to the matching accuracy between the spinning wheel speed and the workpiece speed, verifying the effectiveness of the section statistics submodule in quantifying the processing rhythm.

[0030] Specifically, such as Figure 2 , 5 As shown, the status recognition module includes: The peak location submodule detects the peak distribution location type of adjacent pressure fluctuation segments based on segment statistics, obtains peak timestamps and segment boundary indices for multiple segments, classifies the locations based on the relative order of the indices, and generates peak location distribution vectors. The peak occurrence timestamp for each pressure fluctuation segment is read from the segment statistics data; the boundary retrieval unit simultaneously obtains the start and end timestamps of the segment. The relative position calculator performs normalized position calculations, specifically: first, it obtains the segment end and start timestamps and performs subtraction to obtain the segment duration; simultaneously, it obtains the peak occurrence timestamp and start timestamp and performs subtraction to obtain the peak offset; then, it divides the peak offset by the segment duration to obtain a relative position ratio value rounded to two decimal places. The position classification logic unit has three preset position distribution intervals: the first interval (0.00 to 0.33), the middle interval (greater than 0.33 and less than or equal to 0.66), and the last interval (greater than 0.66 and less than or equal to 1.00). The calculated relative position ratio value is mapped to the above intervals: if the ratio value falls into the first interval, the position type is marked as "Type I"; if it falls into the middle interval, it is marked as "Type II"; if it falls into the last interval, it is marked as "Type III". All position type tags are combined according to the segment arrangement order to generate a peak position distribution vector. For example, a segment starts at 1000 milliseconds and ends at 2000 milliseconds, with its peak occurring at 1250 milliseconds. The calculator calculates the segment duration as 1000 milliseconds, the peak offset as 250 milliseconds, and the relative position ratio as 0.25. Since 0.25 falls between 0.00 and 0.33, the classification logic unit classifies it as "Type I". If the relative position ratio of subsequent adjacent segments is 0.80, it is classified as "Type III", ultimately generating a distribution vector in the form of [Type I, Type III, ...].

[0031] The alternating sequence submodule, based on the peak position distribution vector, calls the peak and valley time series of adjacent pressure fluctuation segments, compares the order of occurrence of peaks and valleys, counts the sequence consistency and sequence difference of adjacent segments, analyzes the sequence correlation between segments, and obtains the peak-valley alternation consistency degree. Based on the generated peak position distribution vector, the peak timestamp and valley timestamp within two adjacent pressure fluctuation segments (defined as preceding and succeeding segments) are obtained. The sequence comparison logic unit performs Boolean judgment on the peak-valley timing within each segment: if the peak timestamp within a segment is less than the valley timestamp, the segment is determined to be in a "peak-before-valley" pattern (denoted as Pattern 1); if the peak timestamp is greater than the valley timestamp, it is determined to be in a "valley-before-peak" pattern (denoted as Pattern 0). The sequence counter compares the patterns of the preceding and succeeding segments: if both patterns are the same (both 1 or both 0), the sequence consistency count is incremented; if the patterns are different, the sequence difference count is incremented. The consistency measurement calculator counts the total number of comparisons within a preset sliding window (e.g., 5 consecutive segments), divides the sequence consistency count by the total number of comparisons, and obtains the peak-valley alternation consistency. For example, in a sliding window containing 5 segments, there are 4 sets of adjacent comparison relationships (segments 1-2, 2-3, 3-4, 4-5). If segments 1, 2, 3, and 4 all follow a "peak before valley" pattern, while segment 5 follows a "valley before peak" pattern, then the comparison results for segments 1-2, 2-3, and 3-4 are all consistent (a total of 3 times), while the comparison result for segment 4-5 is different (a total of 1 time). The calculator divides the sequential consistency count of 3 by the total number of comparisons of 4, resulting in a peak-valley alternation consistency score of 0.75 for this window.

[0032] The segment combination submodule, based on the consistency of peak-valley alternation, calls the index sequence of adjacent pressure fluctuation segments, performs consistency judgment and combination selection on continuous segments, records and summarizes the mapping relationship between the start and end indices of the combined segments and the segment numbers, and generates hidden state segment data. Based on the generated peak-valley alternation consistency, a preset combination judgment threshold is used, for example, set to 1.0 (i.e., requiring absolute consistency). The pattern comparison results of adjacent pressure fluctuation segments are called. If the patterns of two adjacent segments are completely consistent, it is determined that the continuous combination condition is met. The segment fusion execution unit performs a physical index merging operation on multiple consecutive segments that meet the combination condition: extracting the starting index of the first segment in this group of consecutive segments as the new starting position after merging, and extracting the ending index of the last segment as the new ending position after merging, thereby eliminating intermediate boundaries and constructing a long-period hidden state segment spanning multiple original segments. The hidden state mapping recorder assigns a unique hidden state number to the newly generated hidden state segment and establishes an association mapping between the original segment number list and the hidden state number, summarizing and generating hidden state segment data. As shown in Table 3, three consecutive segments numbered S-01, S-02, and S-03 were detected. Their peak-valley alternation pattern is "before peak and after valley", and the peak position type is "I", which meets the consistency merging condition. The segment fusion execution unit reads the starting index 1000 of S-01 and the ending index 4000 of S-03, and merges them into a hidden state segment H-01. S-04, due to its alternating mode mutation to "valley before peak," does not participate in the merging and is processed independently.

[0033] Table 3 Mapping Table of Peak-Valley Alternation and Section Combination Hidden States

[0034] As shown in Table 3, the table details the intermediate parameters and final combination results generated based on the peak position and alternating sequence logic. By merging S-01 to S-03, the three short-cycle fluctuations that were originally mechanically separated are restored into a complete H-01 steady-state processing process, verifying the segment combination submodule's ability to identify and reconstruct continuous and consistent action sequences.

[0035] Specifically, such as Figure 2 , 6 As shown, the periodic modeling module includes: The state alignment submodule extracts the start and end indices of each segment based on the hidden state segment data and the corresponding processing turn number, monitors the consistency of hidden state numbers in adjacent turns, completes segment sequence alignment and number mapping adjustment according to the turn order, and generates hidden state turn alignment sequence. The generated hidden state segment data is invoked, and the start and end ranges of each hidden state segment are traversed to perform interval overlap detection. Specifically, for any given processing track window, all hidden state segments that overlap with it on the time axis are retrieved. If a hidden state segment is completely contained within the track window, its hidden state number is retained. If a hidden state segment crosses the boundary of the track window, the state cutting aligner performs a truncation operation, that is, based on the boundary value of the track window, the hidden state segment crossing the boundary is forcibly divided into a part belonging to the current track and a part belonging to the next track, and its index range is updated accordingly, ensuring that the boundary of the hidden state is strictly limited to the track boundary. The sequence recombination unit rearranges the processed hidden state segments according to the ascending order of the track numbers, generating a hidden state track alignment sequence. For example, the starting index of track number 005 is 10000, and the ending index is 20000. The detected hidden state H-01 ranges from 9000 to 12000, and the hidden state H-02 ranges from 12000 to 25000. The state cutting aligner truncates H-01 at index 10000, retaining only the portion from 10000 to 12000 and assigning it to track number 005; it truncates H-02 at index 20000, retaining the portion from 12000 to 20000 and assigning it to track number 005. Finally, track number 005 is mapped to an ordered sequence of states containing [H-01, H-02]. The advantage of this process is that, through forced physical boundary alignment, it solves the problem of misalignment between continuous monitoring data and discrete processing steps in the time dimension, providing a precise data foundation for subsequent process-based statistical analysis.

[0036] The transfer statistics submodule, based on the hidden state track alignment sequence, calls the combination of hidden state numbers in adjacent tracks, calculates the number of occurrences of multiple number combinations in the track sequence and the cumulative length of the corresponding segment span and normalizes them, and applies preset weight coefficients to calculate the transfer occurrence rate to obtain the hidden state transfer probability set. The hidden state sequence is traversed, and a sliding window mechanism is used to identify all adjacent hidden state pairs, extracting the "current state - successor state" binary relationship. A frequency accumulation counter counts the total number of occurrences of each specific binary pair in the entire sequence, and also counts the total number of occurrences of the "current state" as the starting point. A span-weighted calculator introduces a time dimension weight. The specific calculation logic is as follows: obtain the duration of the "current state" in each occurrence of the binary pair, accumulate the duration of the "current state" in all cases where the binary pair occurs to obtain the sum of the specific transition durations; at the same time, accumulate the duration of the "current state" in all cases to obtain the sum of the global state durations. The probability generation unit performs a weighted transition rate calculation. First, it divides the number of occurrences of a specific pair by the total number of occurrences of the "current state" to obtain the frequency probability. Then, it divides the sum of specific transition durations by the sum of global state durations to obtain the duration probability. Finally, it multiplies the frequency probability and duration probability by preset weighting coefficients (e.g., frequency weight is 0.4, duration weight is 0.6) and sums them to calculate the final hidden state transition probability. For example, hidden state H-01 occurs 50 times, and the total global state duration is 50,000 milliseconds. The case where H-01 is followed by H-02 occurs 40 times, with a cumulative duration of 45,000 milliseconds for H-01 in these 40 occurrences. First, it calculates 40 / 50 = 0.8; then it calculates 45,000 / 50,000 = 0.9. Subsequently, 0.8 * 0.4 = 0.32 and 0.9 * 0.6 = 0.54 are added together to obtain a combined transfer probability of 0.86 for the transition from H-01 to H-02. The advantage of this operational logic is that by introducing time-weighted calculations, it can effectively distinguish between transient interference and steady-state processing trends, ensuring that high-frequency but short-term noise fluctuations will not mislead the judgment of the transfer pattern.

[0037] The path aggregation submodule compares the transition probability of each hidden state starting number with the transition probability based on the hidden state transition probability set, determines the relationship items that exceed the transition probability threshold, connects them in the order of the path number to form a continuous state evolution path, records the number of paths covered by the path and the length of the state dwell segment, and generates life cycle management results. Based on the generated set of hidden state transition probabilities, all calculated hidden state transition probabilities are traversed, and the probability value of each transition relationship is numerically compared with a transition probability judgment threshold. If the transition probability of a pair of hidden states (such as H-01 to H-02) is greater than or equal to the threshold, the transition relationship is determined to be a strongly correlated evolution path and is retained; if it is less than the threshold, it is regarded as a random discrete event and is discarded. The path link builder connects all the retained strongly correlated transition relationships end to end according to the state flow direction, constructing a continuous state evolution path with a directed acyclic graph structure. The lifecycle analyzer tracks the range of processing stages covered by this path, counts how many consecutive stages the evolution path remains stable, and accumulates the duration of all state nodes on the path to generate lifecycle management results. As shown in Table 4, the calculated transition probability from H-01 to H-02 is 0.86, which is greater than the threshold of 0.85. Therefore, the main evolutionary path from "rough machining steady state" to "semi-finishing steady state" is established. The transition probability from H-01 to H-03 is only 0.14, which is judged as an invalid bifurcation path and automatically ignored. Experimental data shows that after applying this path convergence logic, the prediction lead for the defect trend in spinning reaching three processing passes is achieved, effectively avoiding batch scrap accidents.

[0038] Table 4 Hidden State Transition Probability and Path Convergence Determination Table

[0039] As shown in Table 4, the table details the entire process data from transfer statistics to path decision-making. The "Comprehensive Transfer Probability" column, based on the weighted calculation results of the aforementioned frequency and duration, directly determines the direction of the "path construction action." The table clearly demonstrates how, from a complex network of state transfer relationships, quantitative calculations are used to select the main paths representing the primary processing flow (H-01→H-02→H-05→H-09), verifying the logical effectiveness of the path convergence submodule in extracting key process features.

[0040] Specifically, such as Figure 2 , 7 As shown, the cycle optimization module includes: The data access submodule, based on the lifecycle management results, calls the corresponding hidden state segment's vortex pressure time series data, obtains the start and end time indices of multiple hidden state segments, collects continuous pressure sampling values ​​within the segment, and splices the sequences according to the time index order to generate a segment pressure sequence set. The system receives the generated lifecycle management results and identifies key latent state numbers marked as steady-state evolution paths, such as latent state H-02. The time index retrieval unit, based on this latent state number, traverses the historical processing database to search all processing passes containing this state, extracting the physical start and end timestamps of each corresponding segment. For example, latent state H-02 is identified to occur in the 12500ms to 15500ms interval of pass 5 and the 13000ms to 16000ms interval of pass 6. The data reading interface directly calls the cyclone pressure time series records stored in the database to read the historical cyclone pressure values ​​within the aforementioned time windows. The sequence splicing unit extracts the pressure value arrays within each pass sequentially according to the natural ascending order of the pass numbers and performs a data merging operation to generate a segment pressure sequence set containing the full lifecycle characteristics of this latent state. For example, 3000 sampling points were extracted in the 5th pass and 3000 sampling points were extracted in the 6th pass. The stitching unit merged them into a pressure sampling value vector with a length of 6000. This operation process ensures that subsequent analysis can accurately capture the material rheological properties under specific process conditions based on a sufficient sample size, and shields against invalid data interference during the idle stroke of the equipment.

[0041] The deviation assessment submodule obtains pressure sample values ​​within multiple hidden state sections based on the section pressure sequence set, calculates the arithmetic mean of pressure in the same section, performs difference calculation on the multiple sample values ​​and the section mean, and statistically analyzes the frequency distribution of the absolute value of the difference to obtain the pressure deviation distribution. The system reads all 6000 pressure samples from the pressure sequence set of the hidden state segment, performs an accumulation operation to obtain the total pressure, and then divides the total pressure by the total number of sampling points to calculate the arithmetic mean of the pressure in that hidden state segment. For example, if the total pressure of the 6000 points is 900,000 Newtons, the calculated arithmetic mean is 150 Newtons. The difference operation logic unit iterates through each pressure sample in the sequence, subtracts the calculated arithmetic mean from each instantaneous pressure value, and obtains an instantaneous deviation value with a positive or negative sign. For example, for a sampling point with a reading of 158 Newtons, a deviation of +8 Newtons is calculated; for a sampling point with a reading of 145 Newtons, a deviation of -5 Newtons is calculated. The absolute value converter converts all signed deviation values ​​into non-negative absolute differences to quantify the fluctuation amplitude. The frequency distribution statistician has several preset deviation statistical intervals, such as 0 to 2 Newtons, 2 to 5 Newtons, and above 5 Newtons. This unit scans and categorizes all absolute differences, counting the number of sampling points falling into each interval. For example, the statistics show that 5400 points fall in the 0 to 2 Newton interval, 500 points fall in the 2 to 5 Newton interval, and 100 points exceed 5 Newtons. This pressure deviation distribution directly reflects the stability of the processing, with the proportion of high deviation frequencies directly related to quality indicators such as workpiece surface waviness.

[0042] The graded adjustment submodule calls the pressure deviation distribution, calculates the deviation amplitude and compares it with the preset consistency difference threshold, divides the deviation level sequence according to the consistency difference threshold, maps the pressure correction amount to the multi-level associated hidden state segments, summarizes the segment correction results, and generates life cycle management optimization results. The pressure deviation distribution is invoked, using a preset consistency difference threshold. This threshold is set based on the allowable spinning pressure range for plastic deformation of the metallic material. For example, the first-level threshold is set to 5 Newtons, representing a small, permissible fluctuation; the second-level threshold is set to 10 Newtons, representing a significant fluctuation requiring intervention. These thresholds are derived from spinning failure experiments, which show that when pressure fluctuations exceed 10 Newtons, the rate of microcrack formation within the material increases sharply. The deviation level classification unit calculates the 95th quantile of the pressure deviation distribution and compares this value with the aforementioned thresholds. For example, if the 95th quantile of the statistical data is at 8 Newtons, falling between the first-level threshold of 5 Newtons and the second-level threshold of 10 Newtons, the current hidden state's deviation level is determined to be "Level II Concern." The correction mapper invokes a preset expert strategy library to match the corresponding pressure correction amount according to the deviation level. For the "Level II Concern" level, the strategy library specifies that a base spinning pressure of 1.5 Newtons should be added in the next processing cycle to suppress vibration. The results aggregation unit associates the calculated correction value with the corresponding hidden state number H-02 and the relevant processing passes to generate the final lifecycle management optimization results. As shown in Table 5, instructions for maintaining, fine-tuning, or resetting are given for the deviation performance of different hidden states. Experimental verification shows that after applying this hierarchical adjustment strategy, the wall thickness consistency error of the spun workpiece is reduced and the product qualification rate is improved, demonstrating the significant advantages of data-driven closed-loop control logic in the field of precision forming.

[0043] Table 5. Mapping Table for Implicit Pressure Deviation Assessment and Grading Correction

[0044] As shown in Table 5, the table details the pressure statistics and corresponding control decisions for each key hidden state. Through targeted compensation for highly volatile states such as H-02 and H-05, potential molding defect risks were successfully eliminated, validating the core role of the deviation assessment and graded adjustment submodule in achieving adaptive optimization of process parameters.

[0045] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A spinning machine lifecycle management system based on a big data model, characterized in that, The system includes: The signal analysis module acquires the analog electrical signal and timestamp output by the swirling force sensor, performs analog-to-digital conversion and sampling order verification, analyzes the swirling force time series and scans for peak and valley points, generates peak and valley sequence data and transmits it to the pass division module. The pass division module obtains the processing pass number, performs pass boundary judgment on the spun pressure sampling points, divides the peak and valley sequence data of the same pass into pressure fluctuation segments, calculates the average time interval between adjacent peaks for each pressure fluctuation segment, generates segment statistics data and transmits it to the status recognition module. The state recognition module determines the consistency between the peak position type and the peak-valley alternation sequence of adjacent pressure fluctuation segments based on the segment statistics, combines and marks continuous pressure fluctuation segments that meet the consistency condition, generates hidden state segment data, and transmits it to the periodic modeling module. The cycle modeling module maps the hidden state segment data to the corresponding processing stage number, constructs a probability transition matrix based on the frequency and duration of the hidden state in adjacent stages, performs topological analysis on the evolution path between hidden states, and generates life cycle management results.

2. The spinning machine lifecycle management system based on a big data model according to claim 1, characterized in that, The peak-valley sequence data includes peak timestamp, peak amplitude, valley time index, and valley amplitude. The segment statistics include pressure fluctuation segment number, average time interval between adjacent peaks, and segment duration. The hidden state segment data includes hidden state number, segment start number, segment end number, and segment coverage length. The lifecycle management results include hidden state transition probability, hidden state evolution path, and path duration.

3. The spinning machine lifecycle management system based on a big data model according to claim 1, characterized in that, The signal analysis module includes: The signal acquisition submodule acquires the analog electrical signal output by the spun force sensor, performs analog-to-digital conversion on the electrical signal to obtain discrete amplitude, sequentially pairs the discrete amplitude with the timestamp, extracts the original spun force record sequence, writes the acquisition time identifier, completes the integrity check, and generates spun force time series data. The sampling verification submodule acquires sampling point information based on the swirling force time series data, rearranges discrete amplitudes according to a preset sampling frequency, compares the differences between adjacent timestamps to determine whether the sampling interval threshold is met, marks abnormal interval positions with sequential verification, and acquires the swirling force sampling sequence. The peak-valley scanning submodule calls the cyclone sampling sequence, performs differential calculation on the amplitude of adjacent sampling points, determines the location of local extrema based on the change of the differential sign, indexes and records sampling points that meet the peak value determination conditions and valley value determination conditions, integrates the extrema point information in chronological order, and generates peak-valley sequence data.

4. The spinning machine lifecycle management system based on a big data model according to claim 3, characterized in that, The sampling interval threshold is determined by taking the reciprocal of the rated sampling frequency to obtain a standard sampling period reference, and then superimposing the maximum allowable deviation value on the standard sampling period reference, in conjunction with the clock jitter tolerance and the data transmission delay error range.

5. The spinning machine lifecycle management system based on a big data model according to claim 1, characterized in that, The track division module includes: The processing pass numbering submodule obtains the processing pass number based on the peak-valley sequence data, detects the timestamp sequence of the swirling pressure sampling points, reads the time difference between adjacent sampling points and makes a continuity judgment, and classifies the sampling points sequentially according to the trend of time difference changes to generate a pass numbering sequence. The segment division submodule judges the pass boundaries of the swirling pressure sampling points according to the pass numbering sequence, locates the start and end positions of the pass, continuously combines the peak-valley sequence data associated within the same pass, and completes the segment division according to the alternation relationship between peak and valley values ​​to generate a pressure fluctuation segment set. The segment statistics submodule calls the pressure fluctuation segment set, obtains the peak timestamp sequence for each pressure fluctuation segment, sums and normalizes the time differences between adjacent peaks to obtain the average time interval within the segment, summarizes the track number and corresponding average information, and generates segment statistics data.

6. The spinning machine lifecycle management system based on a big data model according to claim 1, characterized in that, The status recognition module includes: The peak location submodule detects the peak distribution location type of adjacent pressure fluctuation segments based on the segment statistics, obtains the peak timestamps and segment boundary indices of multiple segments, classifies the locations based on the relative order of the indices, and generates a peak location distribution vector. The alternating sequence submodule, based on the peak position distribution vector, calls the peak-valley time series of adjacent pressure fluctuation segments, compares the order of occurrence of peaks and valleys, counts the sequence consistency and sequence difference of adjacent segments, analyzes the sequence correlation between segments, and obtains the peak-valley alternation consistency degree. The segment combination submodule, based on the peak-valley alternation consistency, calls the adjacent pressure fluctuation segment index sequence, performs consistency judgment and combination selection on continuous segments, records and summarizes the mapping relationship between the start and end indices of the combined segments and the segment number, and generates hidden state segment data.

7. The spinning machine lifecycle management system based on a big data model according to claim 1, characterized in that, The periodic modeling module includes: The state alignment submodule extracts the start and end indexes of the segment based on the hidden state segment data and the corresponding processing turn number, monitors the consistency of hidden state numbers in adjacent turns, completes segment sequence alignment and number mapping adjustment according to the turn order, and generates a hidden state turn alignment sequence. The transfer statistics submodule, based on the hidden state track alignment sequence, calls the combination of hidden state numbers in adjacent tracks, calculates the number of occurrences of multiple number combinations in the track sequence and the cumulative length of the corresponding segment span and normalizes them, and applies a preset weight coefficient to calculate the transfer occurrence rate to obtain the hidden state transfer probability set. The path aggregation submodule compares the transition probability of each hidden state starting number with the hidden state transition probability set, determines the relationship items that exceed the transition probability threshold, connects them in the order of the path number to form a continuous state evolution path, records the number of paths covered by the path and the length of the state dwell segment, and generates life cycle management results.

8. The spinning machine lifecycle management system based on a big data model according to claim 1, characterized in that, The system also includes: The cycle optimization module, based on the life cycle management results, calls the time series data of the swirling pressure in the corresponding hidden state segment, performs consistency difference judgment and hierarchical adjustment processing on the deviation distribution between the pressure value and the segment mean in the hidden state segment, and generates life cycle management optimization results. The lifecycle management optimization results include the section pressure deviation level, section pressure correction amount, and hidden state stability index.

9. The spinning machine lifecycle management system based on a big data model according to claim 8, characterized in that, The cycle optimization module includes: The data access submodule, based on the lifecycle management results, calls the corresponding hidden state segment's vortex pressure time series data, obtains the start and end time indices of multiple hidden state segments, collects continuous pressure sampling values ​​within the segment, and splices the sequences according to the time index order to generate a segment pressure sequence set. The deviation assessment submodule obtains pressure sample values ​​within multiple hidden state segments based on the segment pressure sequence set, calculates the arithmetic mean of pressure in the same segment, performs difference calculation on the multiple sample values ​​and the segment mean, and statistically analyzes the distribution frequency of the absolute value of the difference to obtain the pressure deviation distribution. The graded adjustment submodule calls the pressure deviation distribution, calculates the deviation amplitude and compares it with the preset consistency difference threshold, divides the deviation level sequence according to the consistency difference threshold, maps the pressure correction amount to the multi-level associated hidden state segments, summarizes the segment correction results, and generates life cycle management optimization results.

10. The spinning machine lifecycle management system based on a big data model according to claim 9, characterized in that, The consistency difference threshold is determined by obtaining the absolute values ​​of all pressure deviations within the same segment, sorting them by numerical value, extracting the median value of the deviation as a distribution center characterization quantity, and then combining it with the deviation dispersion range corresponding to the median value.

Images