Data-driven air compressor control system and method

Abstract

This invention discloses a data-driven air compressor control system and method, relating to the field of air compressor control technology. It synchronously acquires the physical parameters of the air compressor and the equipment response data corresponding to each control command; based on historical normal operation data, it establishes equipment baseline response characteristics corresponding to different control commands; by comparing real-time response data with the baseline response characteristics, it identifies response drift, delay extension, feedback structure changes, and behavioral deviations under similar commands when the control logic remains unchanged, and outputs the dynamic evolution trajectory of the deviation characteristics. This invention, by establishing baseline response characteristics and comparing them with real-time response data, identifies early response deviations of the equipment, detects potential faults in advance, and ensures stable equipment operation.

Description

Technical Field

[0001] This invention belongs to the field of air compressor control technology, specifically a data-driven air compressor control system and method. Background Technology

[0002] An air compressor is a device that converts mechanical energy into gas pressure energy and is widely used in industries such as manufacturing, construction, and medicine.

[0003] During air compressor operation, traditional control methods struggle to accurately detect early potential faults. Due to long-term operation, issues such as structural fatigue, control link anomalies, or internal media degradation may occur. These problems often don't initially cause physical parameters to exceed limits, but they do cause subtle changes in the equipment's response, such as response drift or increased latency. Failure to identify these early signs and adjust control commands in a timely manner can lead to more serious faults, affecting the air compressor's normal operation and lifespan. Therefore, accurately identifying and controlling early equipment anomalies based on data is crucial for ensuring the stable and efficient operation of air compressors. Summary of the Invention

[0004] The purpose of this invention is to provide a data-driven air compressor control system and method, which solves the technical problem in the prior art that it is impossible to accurately identify early equipment anomalies and control them based on data.

[0005] Data-driven air compressor control methods include:

[0006] The physical parameters of the air compressor and the equipment response data corresponding to each control command are acquired synchronously.

[0007] Based on historical normal operation data, establish the equipment baseline response characteristics corresponding to different control commands;

[0008] By comparing real-time response data with baseline response characteristics, we can identify response drift, delay extension, feedback structure changes, and behavioral deviations under similar commands when the control logic remains unchanged, and output the dynamic evolution trajectory of the deviation characteristics.

[0009] When the behavior deviation exceeds the set range, the key turning point of the deviation evolution trajectory is used as the boundary to divide the transmission stage of parameter micro-fluctuation. Then, the influence relationship between parameters in each stage is checked to see if the transmission law of the historical degradation process under the same working conditions is reproduced. If the change order and intensity distribution of the influence relationship of parameters in each stage are consistent with the historical characteristics, even if the physical parameters are not exceeded, it is still judged as an early sign of structural fatigue, abnormal control link or deterioration of internal medium in the equipment.

[0010] Adjust control instructions based on risk assessment results, and feed the new instructions and response data back to the expected response model unit to update the baseline characteristics.

[0011] Furthermore, after determining that the device exhibits early signs, this also includes:

[0012] Obtain the parameter influence relationships in each transmission stage that are inconsistent with the historical characteristic change order, and mark them as abnormal action points;

[0013] Starting from the abnormal point of action, obtain the response time difference sequence of the parameter after the control command is issued, and compare it with the characteristics of the time difference sequence of the historical normal response.

[0014] If the interval fluctuation amplitude of a certain link in the time difference sequence exceeds the usual range of similar working conditions, extract the equipment operation sound pattern or vibration spectrum segment corresponding to that link and compare it with the sound pattern / spectrum of the same link in the early stage of historical degradation to make resonance peak comparison.

[0015] If the resonance peak shift does not reach the warning threshold, adjust the analysis dimension of the parameter interaction relationship and re-examine the progression order;

[0016] After each round of analysis, the number of abnormal action points is counted to see if it has decreased. If it has not decreased in two consecutive rounds, the typical deviation transmission path of the same type of equipment under the same degradation scenario is introduced as a reference, and the consistency between the current time difference sequence and the reference path is checked step by step.

[0017] Until the number of abnormal action points falls below a set threshold, the finally located deviation source and verification parameters are associated with the early sign judgment results and updated to the baseline feature library of the expected response model unit.

[0018] Furthermore, in the initial stage of the micro-fluctuation propagation phase of the partitioning parameters, it also includes:

[0019] Obtain parameter combinations that do not match the intensity distribution of historical degradation transmission patterns in each stage, and align the peaks of the control command curves corresponding to these parameters with the equipment response curves.

[0020] If, after alignment, the number of curve intersections exceeds the preset range for the same type of working condition in multiple consecutive fluctuation cycles, the vibration waveform of the equipment operation in the dense intersection area is extracted and compared with the vibration waveform of the normal state in the same stage by the trough spacing.

[0021] Adjust the sampling interval of the parameter combination according to the direction of the difference in the trough spacing, recheck the intensity distribution until it conforms to the historical characteristics, and archive the adjusted sampling interval and vibration waveform characteristics together.

[0022] Furthermore, the intermediate stage of verifying the influence of parameters also includes:

[0023] Find the group of parameters with the largest deviation from the historical evolution order from each stage, and calculate the interval between the instruction trigger time and the time when the response extreme value appears for the corresponding group.

[0024] If the dispersion of the interval value exceeds the normal range of similar operating conditions, extract the equipment operating temperature field distribution segment corresponding to the set of parameters and compare it with the temperature field segment of the normal state at the same stage to make hot spot migration path comparison.

[0025] Based on the direction of hotspot migration differences, the threshold for determining the order of parameter changes is adjusted until it conforms to historical characteristics. The adjusted threshold is then associated with and saved with the temperature field characteristics.

[0026] Furthermore, in the later stages of the transmission phase, it also includes:

[0027] Select the parameter interaction relationship that deviates most significantly from the historical pattern in each stage, and statistically analyze the ratio of instruction duration to response amplitude change corresponding to this relationship;

[0028] If the ratio fluctuation exceeds the common range of similar working conditions, extract the fluid pressure pulsation curve of the equipment during this stage and compare the overlap of the pulsation cycle with the normal curve under the same condition.

[0029] Based on the difference in period overlap, the analysis weights of parameter interaction relationships are adjusted until they conform to historical characteristics. The adjustment results are then associated with and stored in relation to pressure pulsation characteristics.

[0030] Furthermore, establishing equipment baseline response characteristics includes: grouping historical normal data according to control command type, extracting response parameter change curves within a preset time window after each command is issued; merging multiple curves of the same type of command to generate a baseline template containing response start threshold, steady-state fluctuation range, and peak occurrence time; and establishing baseline feature sub-libraries for different operating conditions, containing typical time constants and parameter correlation coefficients of command response under that operating condition.

[0031] Furthermore, the division of the parameter micro-fluctuation propagation stages includes: taking the second derivative of the deviation evolution trajectory and using the inflection point where the derivative changes from positive to negative as the stage node; recording the propagation path of the initial fluctuation parameters in the first stage; statistically analyzing the number of parameters involved in the fluctuation and their influence intensity in the second stage; and calculating the decay rate of the fluctuation amplitude of each parameter in the third stage.

[0032] Furthermore, following the dynamic evolution trajectory of the output deviation characteristics, it also includes:

[0033] Extract segments from the trajectory whose deviation increases exceed common values ​​for similar working conditions within several consecutive sampling periods, and align the start and end points of the instruction execution segment and the response change segment corresponding to the segment.

[0034] If the slope fluctuation of the response change segment after alignment exceeds the historical normal range, extract the equipment operation vibration waveform of that segment and compare the peak interval with the earlier waveform of the same type of deviation.

[0035] Based on the difference in peak intervals, the criteria for determining the increase in deviation are adjusted, and feature segments are re-extracted until they conform to historical fluctuation patterns. The adjusted criteria are then associated with the vibration waveform characteristics and archived.

[0036] Furthermore, the process of identifying behavioral deviations also includes:

[0037] Perform overlay analysis on parameter change curves of multiple consecutive behavioral deviations under the same type of command, and statistically analyze the area ratio of the overlapping region of the curves;

[0038] If the percentage is lower than the set value, the interval with the most significant deviation from each curve is selected and compared with the valley depth of the parameter curve of the same type in the early stage of degradation in history.

[0039] Based on the direction of the difference in valley depth, the identification threshold for behavioral deviation is corrected, and the analysis is re-overlaid until the overlapping area reaches the standard. The corrected threshold and curve features are then stored together.

[0040] Secondly, this application provides a data-driven air compressor control system, which includes:

[0041] The acquisition module synchronously acquires the physical parameters of the air compressor and the equipment response data corresponding to each control command;

[0042] A module is established to create equipment baseline response characteristics corresponding to different control commands based on historical normal operation data;

[0043] The identification module compares real-time response data with baseline response characteristics to identify response drift, delay extension, feedback structure changes, and behavioral deviations under similar instructions when the control logic remains unchanged, and outputs the dynamic evolution trajectory of the deviation characteristics.

[0044] When the behavior deviation exceeds the set range, the output module first divides the transmission stage of parameter micro-fluctuation by the key turning point of the deviation evolution trajectory, and then checks whether the influence relationship between parameters in each stage reproduces the transmission law of the historical degradation process under the same working conditions. If the change order and intensity distribution of the influence relationship of parameters in each stage are consistent with the historical characteristics, even if the physical parameters are not exceeded, it is still determined that the equipment has early signs of structural fatigue, control link abnormality or internal medium condition deterioration.

[0045] The feedback module adjusts control commands based on the risk assessment results and feeds the new commands and response data back to the expected response model unit to update the baseline characteristics.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] This invention establishes benchmark response characteristics and compares them with real-time response data to accurately identify early response deviations of the equipment. It also combines the transmission patterns of historical degradation processes to determine whether there are early signs of failure in the equipment and adjusts control commands accordingly. This enables intelligent and precise control of the air compressor, early detection of potential faults, and ensures stable equipment operation. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of an air compressor.

[0049] Figure 2 This is a schematic diagram of the system framework structure of the present invention. Detailed Implementation

[0050] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Please see Figure 2 This application provides a data-driven air compressor control method, applicable to... Figure 1 Air compressors, including:

[0052] The physical parameters of the air compressor and the equipment response data corresponding to each control command are acquired synchronously.

[0053] Based on historical normal operation data, establish the equipment baseline response characteristics corresponding to different control commands;

[0054] By comparing real-time response data with baseline response characteristics, we can identify response drift, delay extension, feedback structure changes, and behavioral deviations under similar commands when the control logic remains unchanged, and output the dynamic evolution trajectory of the deviation characteristics.

[0055] When the behavior deviation exceeds the set range, the key turning point of the deviation evolution trajectory is used as the boundary to divide the transmission stage of parameter micro-fluctuation. Then, the influence relationship between parameters in each stage is checked to see if the transmission law of the historical degradation process under the same working conditions is reproduced. If the change order and intensity distribution of the influence relationship of parameters in each stage are consistent with the historical characteristics, even if the physical parameters are not exceeded, it is still judged as an early sign of structural fatigue, abnormal control link or deterioration of internal medium in the equipment.

[0056] Adjust control instructions based on risk assessment results, and feed the new instructions and response data back to the expected response model unit to update the baseline characteristics.

[0057] Physical parameters refer to physical quantities that reflect the operating status of an air compressor, including temperature, pressure, and flow rate. These parameters can be collected in real time by devices such as temperature sensors, pressure sensors, and flow meters to monitor the basic operating status of the air compressor.

[0058] Among them, equipment response data refers to the status change data of the air compressor after receiving control commands, specifically including speed change rate, pressure regulation rate, power consumption change, etc., which can be recorded in real time by the air compressor's control system and used to analyze the equipment's response to control commands.

[0059] Among them, the baseline response characteristics refer to the typical response patterns of the air compressor under different control commands during its historical normal operation. Specifically, it can be established by using the sliding window statistical method (with the window size set to 100 sets of data) to extract features (such as mean, variance, and trend) from historical normal operation data, and is used as a reference standard to judge whether the real-time response is normal.

[0060] Among them, the dynamic evolution trajectory of the deviation feature refers to the path of the deviation generated by the real-time response data compared with the baseline response feature over time. Specifically, it can be obtained by continuously recording the deviation value and plotting the curve, which is used to intuitively show the development and changes of the equipment response deviation.

[0061] Among them, the transmission stage of parameter micro-fluctuations refers to dividing the transmission process of parameter micro-fluctuations between various systems of the equipment into different stages, with the key turning point of the deviation evolution trajectory as the boundary. Each stage corresponds to a specific parameter influence relationship and change characteristics, which is used to deeply analyze the laws of deviation generation and transmission.

[0062] Among them, the transmission law of the historical degradation process refers to the law of the changing order and intensity distribution of the influence relationship between various parameters when the air compressor has experienced failures such as structural fatigue, control link abnormality or internal media condition deterioration. It can be obtained by analyzing and summarizing historical failure data and is used to help judge whether there are early failure signs in the current equipment.

[0063] Among them, the expected response model unit refers to the model unit used to store and update the baseline response characteristics. The gradient descent method is used to continuously optimize the baseline characteristics based on new control commands and response data, so that they are more in line with the current actual operating state of the air compressor.

[0064] The innovation of this application lies in the fact that by establishing a benchmark response characteristic and comparing it with real-time response data, the early response deviation of the equipment can be accurately identified. Combined with the transmission law of the historical degradation process, it can be determined whether there are signs of early failure in the equipment, and the control commands can be adjusted accordingly. This realizes intelligent and precise control of the air compressor, detects potential faults in advance, and ensures stable operation of the equipment.

[0065] The working principle of this application is as follows: First, the system synchronously collects the physical parameters of the air compressor and the equipment response data corresponding to each control command, providing basic data for subsequent analysis. Second, it uses historical normal operation data to establish equipment baseline response characteristics corresponding to different control commands, serving as the basis for judging whether the equipment response is normal. Then, it compares the real-time response data with the baseline response characteristics to identify various response deviations and outputs the dynamic evolution trajectory of the deviation characteristics, promptly detecting abnormal changes in the equipment response. Next, when the behavior deviation exceeds the set range, it divides the transmission stage of parameter micro-fluctuations and checks whether the influence relationship between parameters in each stage conforms to the transmission law of the historical degradation process, thereby determining whether there are early signs of equipment failure. Even if the physical parameters do not exceed the limits, problems can be detected in time. Finally, it adjusts the control commands based on the risk assessment results and feeds the new commands and response data back to the expected response model unit to update the baseline characteristics, making the control more adaptable to the actual state of the equipment.

[0066] As a preferred embodiment, the data-driven air compressor control method of this application is specifically implemented as follows:

[0067] First, install temperature sensors, pressure sensors, flow meters, and other equipment on the air compressor to collect physical parameters in real time, such as cylinder temperature (range 80-120℃), exhaust pressure (range 0.7-0.8MPa), and exhaust flow rate (range 5-8m³ / min). At the same time, record equipment response data, such as speed change rate and pressure regulation rate, through the air compressor control system.

[0068] Based on historical normal operation data from the past year, a sliding window statistical method (100 window groups) was used to establish baseline response characteristics for different control commands. For example, for the loading command, the baseline response characteristics are: the time for the pressure to rise from 0.4 MPa to 0.7 MPa is 15 ± 2 s, and the speed change rate is stable at around 5% / s.

[0069] During real-time operation, after the loading command is issued, the real-time response data shows that the pressure rises from 0.4MPa to 0.7MPa in 20 seconds, and the speed change rate fluctuates greatly. After comparison with the baseline response characteristics, it is found that there is a lengthened response delay and behavioral deviation, and the dynamic evolution trajectory of the output deviation characteristics is identified.

[0070] Because the behavior deviation exceeded the set range (time deviation exceeding 5 seconds), the point where the pressure rise rate in the trajectory significantly slowed down was taken as the key turning point, and the transmission stage of parameter micro-fluctuations was divided. Verification revealed that the order and intensity distribution of the influence relationships between parameters such as temperature, pressure, and rotational speed in each stage were consistent with the transmission pattern of degradation processes caused by control link anomalies in the past. Therefore, it was determined that there were early signs of anomaly in the control link, even though the physical parameters were all within the normal range at this time.

[0071] Based on the judgment results, the execution parameters of the loading command are adjusted, such as appropriately increasing the initial loading power. The new command and its corresponding response data are fed back to the desired response model unit, and the baseline response characteristics corresponding to the loading command are updated using the gradient descent method, making subsequent adjustments more precise.

[0072] This implementation method helps to detect early signs of air compressor failure in advance, adjust control commands in a timely manner, effectively prevent the failure from developing further, and improve the operational reliability and service life of the air compressor.

[0073] In some embodiments of this application, after determining that the device has early signs of structural fatigue, control link abnormality, or internal media degradation, there is a problem in how to accurately locate the source of the deviation and refine the anomaly analysis to improve the reliability of the determination result.

[0074] To address this, this application further proposes to obtain the parameter influence relationships in each transmission stage that are inconsistent with the historical characteristic change sequence, and mark them as abnormal action points; starting from the abnormal action points, obtain the response time difference sequence of the parameter after the control command is issued, and compare it with the characteristics of the time difference sequence of the historical normal response; if the interval fluctuation amplitude of a certain link in the time difference sequence exceeds the normal range of similar working conditions, extract the corresponding equipment operation sound pattern or vibration spectrum segment, and compare it with the sound pattern / spectrum of the same link in the early stage of historical degradation for resonance peak comparison; if the resonance peak deviation does not reach the warning threshold, adjust the analysis dimension of the parameter action relationship and re-examine the change sequence; after each round of analysis, count whether the number of abnormal action points has decreased; if it has not decreased for two consecutive times, introduce the typical deviation transmission path of the same model of equipment in the same degradation scenario as a reference, and check the consistency between the current time difference sequence and the reference path for each link; until the number of abnormal action points is lower than the set threshold, associate the finally located deviation source link and verification parameters with the early sign judgment result, and update the benchmark feature library of the expected response model unit.

[0075] Specifically, after determining that the equipment exhibits early signs of degradation, the system first identifies anomalous points of action that do not conform to the historical sequence of characteristic changes. Starting from these points, it traces the response time difference sequence and identifies the abnormal fluctuation points by comparing them with normal time difference characteristics. Then, it combines this with a comparison of the resonance peaks of acoustic signatures or vibration spectra to verify whether the deviation reaches the warning level at the physical signal level. If the threshold is not reached, the analysis dimensions are adjusted to reduce the number of anomalous points of action. When it is difficult to reduce the number of anomalous points, the typical degradation path of the same model of equipment is used as a reference to accurately locate the source of the deviation. This multi-step, multi-dimensional analysis method can penetrate complex parameter correlations and accurately find the root cause of the problem, making subsequent control command adjustments more targeted. Especially in the early stages of equipment degradation, when parameter fluctuations are subtle and the correlations are complex, this method can effectively capture key anomalous links, preventing early hidden dangers from being overlooked and improving the safety and effectiveness of air compressor operation and control. Simultaneously, the source location results are updated to the benchmark feature library, allowing the model to continuously learn new anomalous patterns, enhancing the system's self-optimization capability and long-term adaptability.

[0076] In the technical solution of this application, the identification and marking of anomalous action points can be achieved in various ways. For example, association rule mining algorithms can be used to find nodes that are inconsistent with the historical evolution order from the parameter influence relationship; when obtaining and comparing the response time difference sequence, time series analysis methods, such as dynamic time warping (DTW), can be used to measure the similarity between the real-time time difference sequence and the historical normal sequence.

[0077] Resonance peak comparison can be achieved using spectral analysis tools, such as Fast Fourier Transform (FFT), to convert acoustic or vibration signals into a spectrum, and then extracting resonance peaks for comparison using peak detection algorithms. When adjusting the analytical dimensions of parameter relationships, dimensionality reduction algorithms such as Principal Component Analysis (PCA) can be used to focus on key parameters. When introducing typical deviation transmission paths of equipment of the same model as a reference, a path matching model can be established, such as path similarity calculation based on graph theory, or using graph neural networks (GNNs) in deep learning to learn path features and perform matching.

[0078] In addressing the problem of accurately analyzing early signs of degradation in air compressors, the technical solution of this application first identifies potential problem parameter correlations by pinpointing abnormal action points. This step effectively filters out key nodes that deviate from historical patterns from a complex parameter transmission network. For example, when an air compressor transitions from normal operation to a slight degradation stage, certain parameters (such as the correlation between air pressure regulation response time and motor speed) may deviate from their historical normal progression. These abnormal action points can be quickly identified through association rule mining.

[0079] Starting from the point of abnormal action, the response time difference sequence is traced. For example, after the control command is issued, the time difference sequence of the oil-gas separator pressure rise is compared with the historical normal sequence. If the fluctuation amplitude of a certain link (such as the pressure sensor feedback delay) exceeds the range of similar operating conditions, the operating sound pattern (such as the vibration sound of the separator shell) or vibration spectrum of that link is extracted. After FFT transformation, it is compared with the spectrum of the same link in the early stage of historical degradation. If the resonance peak shift is small (not reaching the warning threshold), the analysis dimension is adjusted by PCA, focusing on the core parameters related to pressure feedback (such as sensor voltage and pipeline resistance), and the parameter change sequence is re-examined.

[0080] If the number of abnormal points remains unchanged after two rounds of analysis (e.g., there are still 3 abnormalities), then a typical deviation propagation path for the same type of equipment under oil-gas separator degradation scenarios (e.g., "motor load fluctuation -- separator pressure anomaly -- sensor feedback delay") is introduced, and the consistency between the current time difference sequence and this path is checked step by step. Assuming that the consistency between the current sequence and the typical path reaches 80%, the source of the deviation can be located as filter element blockage inside the separator. This result and the corresponding verification parameters (e.g., pressure difference before and after the filter element, vibration spectrum peak value) are then correlated with the early indication judgment results and updated to the benchmark feature library.

[0081] In this way, the air compressor control system can accurately locate the root cause of the problem in the early stage of degradation, making the adjustment of control commands more targeted (such as appropriately increasing the frequency of filter cleanliness monitoring). At the same time, the updated benchmark feature library can more accurately identify similar degradation signs, significantly improving the operational reliability and control intelligence level of the equipment.

[0082] In some embodiments of this application, when the parameter combination does not match the intensity distribution of the historical degradation transmission law in the early stage of dividing the parameter micro-fluctuation transmission stage, there are still challenges in how to effectively calibrate the accuracy of parameter analysis and reduce misjudgments caused by data sampling or waveform feature deviations.

[0083] To address this, this application further proposes obtaining parameter combinations that do not match the intensity distribution of historical degradation transmission patterns in each stage, aligning the peaks of the control command curves corresponding to these parameters with the equipment response curves; if, after alignment, the number of curve intersections exceeds the preset range for similar operating conditions in multiple consecutive fluctuation cycles, the equipment operating vibration waveform in the dense intersection area is extracted and compared with the trough spacing of the vibration waveform in the normal state of the same stage; according to the direction of the difference in trough spacing, the sampling interval of the parameter combination is adjusted, and the intensity distribution is re-checked until it conforms to historical characteristics, and the adjusted sampling interval and vibration waveform characteristics are archived together.

[0084] Specifically, in the initial stage of the parameter micro-fluctuation transmission phase, when it is found that the intensity distribution of a certain parameter combination (such as the intake pressure and exhaust flow of an air compressor) does not match the historical pattern, the system first aligns the peaks of the corresponding control command curves (such as the intake valve control signal) and response curves (such as the exhaust flow feedback signal) to eliminate phase deviation on the time axis. If, after alignment, the number of intersections between the two curves exceeds the preset threshold for the same operating condition within three consecutive fluctuation cycles (e.g., ≤2 intersections under normal conditions, 5 currently appearing), then the vibration waveform of the dense intersection interval (such as the air compressor cylinder vibration signal) is extracted and compared with the vibration waveform of the normal state in the same stage—for example, the valley spacing of the normal waveform is stable at 0.2 seconds, while the valley spacing of the current waveform fluctuates drastically between 0.15 and 0.25 seconds. Based on the direction of the difference in valley spacing (e.g., generally too small), the sampling interval of the parameter combination is adjusted (e.g., from the original 10ms to 8ms), the intensity distribution is recalculated until it matches the historical characteristics, and the 8ms sampling interval and fluctuation waveform characteristics are archived.

[0085] This method effectively solves the problem of misjudgment of intensity distribution caused by improper data acquisition frequency by reversely calibrating the sampling interval through fine comparison of waveform features. Especially in the early stage of slight fluctuations in equipment parameters, when signal characteristics are weak and susceptible to noise interference, this method can lock key deviations through the geometric features of peaks / troughs, ensuring the accuracy of the transmission stage division. At the same time, the archived sampling intervals and waveform features can serve as analysis templates for similar operating conditions, reducing the cost of repeated calibration, improving the system's sensitivity to parameter fluctuations and response speed, and laying a data foundation for the accurate determination of early signs of equipment degradation.

[0086] In the technical solution of this application, the peak alignment of the control command curve and the device response curve can be achieved in various ways. For example, a peak detection algorithm (such as threshold-based local maximum identification) can be used to locate the peak positions of the two curves, and then the time axis can be aligned by translating or stretching the curves.

[0087] The comparison of the trough spacing of vibration waveforms can be achieved with the help of signal processing tools, such as extracting the low-frequency components of the waveform through wavelet transform to highlight the trough characteristics, and then using Euclidean distance or Manhattan distance to calculate the difference in spacing with the normal waveform; the adjustment of the parameter combination sampling interval can be based on the statistical characteristics of the trough spacing difference, such as when the difference is systematically small, the sampling interval is shortened proportionally (e.g., if the difference amplitude is 20%, the sampling interval is shortened by 20%).

[0088] In addressing the accuracy issue of dividing the transmission stage of micro-fluctuations in parameters, the technical solution of this application first eliminates the time phase difference between the control command and the response curve through peak alignment, ensuring the validity of the basic data for parameter correlation analysis. For example, when the air compressor is under low load, there is a 0.5-second delay between the peak of the intake valve control command (maximum opening) and the peak of the exhaust flow response. If the analysis is performed directly without alignment, the intensity distribution will show a "weak correlation between command and response," which does not match the historical pattern (strong correlation). After locating the peak through peak detection, the response curve is shifted by 0.5 seconds to achieve alignment. At this point, the number of recalculated intersection points can truly reflect the dynamic correlation between the two.

[0089] If the number of crosspoints remains abnormal after alignment (e.g., exceeding the preset range), the physical deviation is identified by the trough spacing of the vibration waveform. For example, if a certain parameter combination has too many crosspoints after alignment, the vibration waveform of the air compressor crankcase is captured. It is found that the trough spacing is stable at 0.3 seconds under normal conditions, but the current waveform, due to slight bearing wear, has a trough spacing shortened to 0.25 seconds and fluctuates frequently. Based on this difference (smaller spacing), the sampling interval of this parameter combination is adjusted from 15ms to 12ms (a 20% reduction). After re-collecting data, the matching degree between the intensity distribution and the historical degradation pattern increases from 60% to 92%, and the 12ms sampling interval and the vibration waveform characteristics at the initial stage of wear are archived.

[0090] This method allows the system to calibrate data acquisition standards early in the transmission phase of minor parameter fluctuations, avoiding misjudgments of patterns due to improper sampling intervals and ensuring the accuracy of subsequent early sign detection. Simultaneously, the archived feature data can be used to train a predictive model for parameter sampling intervals. When similar operating conditions recur, the optimal sampling interval can be automatically invoked, significantly improving the system's analysis efficiency and response speed, and providing a more reliable foundation for refined control and fault early warning of air compressors.

[0091] This application further proposes to identify the group of parameters with the largest deviation from the historical change order from each stage combination, and to statistically analyze the interval between the instruction trigger time and the time when the response extreme value appears for the corresponding group; if the dispersion of the interval value exceeds the normal range of similar operating conditions, the equipment operating temperature field distribution segment corresponding to the group of parameters is extracted and compared with the temperature field segment of the normal state in the same stage for hot spot migration path comparison; according to the direction of hot spot migration difference, the judgment threshold of parameter change order is corrected until it meets the historical characteristics, and the corrected threshold is associated with and saved with the temperature field characteristics.

[0092] Specifically, when the deviation in the changing order of a certain parameter group (such as the cooling system flow rate and outlet temperature of an air compressor) is the largest, the dispersion of the time interval between command triggering and response extreme values ​​is first calculated. If the dispersion is too large (e.g., the standard deviation exceeds 0.3 seconds), the temperature field distribution of the corresponding time period (e.g., infrared thermal imaging of the radiator area) is retrieved and compared with the hot spot migration path under normal conditions (e.g., hot spots normally diffuse from the center to the edge, but currently show local clustering). Based on the direction of the migration difference (e.g., the clustering speed is faster than historical data), the judgment threshold for the changing order of this parameter group is adjusted (e.g., shortened from the original 1.2 seconds to 1.0 seconds) until it matches historical characteristics, and the threshold is associated with the temperature field clustering characteristics and archived.

[0093] This method provides a basis for correction when the temporal correlation of parameters is ambiguous, by leveraging the physical characteristics of the temperature field, effectively eliminating the limitations of single-time-dimensional analysis. Especially in the mid-stage of equipment degradation, where parameter interactions intensify, it can accurately pinpoint the criteria for determining the progression order, providing a reliable foundation for subsequent risk assessment. Simultaneously, the associated archive feature can accelerate the processing efficiency of similar problems, forming a closed-loop optimization mechanism.

[0094] At the implementation level, the dispersion of interval values ​​can be calculated using the coefficient of variation; the comparison of hotspot migration in the temperature field can be achieved using an image analysis algorithm based on contour matching; and threshold correction can be implemented by establishing a mapping model between hotspot migration speed and threshold. The steps are logically progressive: the dispersion of interval values ​​triggers temperature field analysis, the difference in hotspot migration determines the direction of threshold correction, and the final archived results feed back into the verification process, improving the overall analysis accuracy.

[0095] In some embodiments of this application, when dividing the parameter micro-fluctuation transmission stage, the accuracy of parameter analysis has been gradually improved through the early sampling interval calibration and the mid-term judgment threshold correction. However, in the later stage of the transmission stage, when the parameter interaction relationship deviates significantly from the historical pattern, it remains a challenge to optimize the weight allocation of the interaction relationship through an analysis dimension that is more in line with the physical characteristics of the device.

[0096] To address this, this application further proposes selecting the parameter interaction relationship with the most significant deviation from historical patterns from each stage, statistically analyzing the ratio of command duration to response amplitude change corresponding to this relationship; if the ratio fluctuation exceeds the common range of similar operating conditions, extracting the fluid pressure pulsation curve of the equipment operation in that stage, and comparing the pulsation cycle overlap with the normal curve in the same state; adjusting the analysis weight of the parameter interaction relationship according to the difference in cycle overlap until it conforms to historical characteristics, and storing the adjustment result in association with the pressure pulsation characteristics.

[0097] Specifically, in the later stage of the transmission phase, when the interaction relationship of a certain parameter (such as the opening of the air compressor's intake valve and the cylinder pressure) deviates most significantly from the historical pattern, the system first calculates the ratio of the duration of the command under this relationship (such as the valve opening for 1.5 seconds) to the change in response amplitude (such as the cylinder pressure rising by 0.3 MPa). If the fluctuation range of this ratio (such as the normal range of 0.2-0.25 MPa / second, and the current range of 0.15-0.3 MPa / second) exceeds the common range, the fluid pressure pulsation curve of the corresponding stage (such as the high-frequency pulsation signal collected by the cylinder pressure sensor) is extracted and compared with the normal curve in the same state to compare the overlap of the pulsation cycle. For example, the pulsation cycle of the normal curve is stable at 0.02 seconds, and the overlap reaches 90%, while the current curve fluctuates due to valve leakage, and the overlap drops to 60%. Based on the difference in period overlap (reduced by 30%), the analysis weight of the interaction relationship of this parameter is adjusted (e.g., increased from 0.3 to 0.5). After recalculation, the interaction relationship is made to conform to historical characteristics, and the adjusted weight is associated with and stored with the pulsation period disorder characteristics caused by leakage.

[0098] This method optimizes the weights of parameter interactions by leveraging the periodic characteristics of fluid pressure pulsations, effectively addressing parameter correlation deviations caused by fluid dynamic anomalies such as valve wear and pipeline blockage. Particularly in the later stages of the transmission phase, when the impact of physical degradation on parameter interactions intensifies, this method can accurately calibrate and analyze weights using pressure pulsations—a signal directly reflecting the internal state of the equipment—ensuring that the final result of the transmission phase division accurately reflects the actual degradation process.

[0099] In the technical solution of this application, the ratio calculation of the instruction duration and response amplitude change can be achieved by the sliding window method, and the fluctuation range of the ratio can be tracked in real time; the comparison of the overlap degree of fluid pressure pulsation cycles can be achieved by using spectrum analysis combined with dynamic time warping (DTW) algorithm, first extracting the cycle features through Fourier transform, and then calculating the similarity with the normal curve; the adjustment of the parameter interaction relationship analysis weight can be based on the mapping model between the difference in cycle overlap degree and the weight, such as using proportional adjustment (for every 10% decrease in overlap degree, the weight increases by 0.1).

[0100] In some embodiments of this application, a basic idea for establishing equipment reference response characteristics is proposed as a reference standard for identifying parameter deviations and determining equipment status. However, there is still room for improvement in how to make the reference characteristics more closely match the actual characteristics of different control command types and operating conditions, and avoid insufficient identification accuracy due to the excessive universality of the reference template.

[0101] In response, this application further proposes to group historical normal data according to the type of control command, extract the response parameter change curve within a preset time window after each group of commands is issued; merge multiple curves of the same type of command to generate a benchmark template containing the response start threshold, steady-state fluctuation range, and peak occurrence time; and establish benchmark feature sub-libraries for different operating conditions, containing the typical time constant and parameter correlation coefficient of the command response under that operating condition.

[0102] Specifically, when establishing baseline response characteristics, the system first groups historical normal data according to control command type (such as "load command," "unload command," and "pressure regulation command" for air compressors). For each command group, it extracts the response parameter change curves (such as changes in parameters like pressure, temperature, and speed over time) within 5 seconds of issuance. For example, the system merges 100 curves from the "load command" group (by calculating the mean and standard deviation of each curve at the same time point) to generate a baseline template for that command: response initiation threshold (e.g., a pressure increase of 0.1 MPa is considered initiation), steady-state fluctuation range (e.g., pressure stabilizes at 0.8 ± 0.02 MPa), and peak occurrence time (e.g., pressure peak is reached 2.3 seconds after loading). Simultaneously, sub-libraries are established according to operating conditions (e.g., light load, full load, low temperature environment, high temperature environment). Each sub-library contains typical time constants for that operating condition (e.g., the time constant for pressure to reach steady state at full load is 1.8 seconds) and parameter correlation coefficients (e.g., the correlation coefficient between load current and exhaust pressure is 0.92).

[0103] This method, through two-layer subdivision, upgrades the benchmark features from a single template to a multi-dimensional, scenario-based reference system. Especially in scenarios where the equipment response pattern changes significantly with commands and operating conditions (such as the pressure rise rate of an air compressor under full load and light load can differ by up to 30%), it can effectively avoid misjudgment of normal deviations or omission of abnormal deviations caused by benchmark generalization.

[0104] In the technical solution of this application, the extraction of response parameter change curves can be achieved through a sliding time window, the window size of which is set according to the instruction execution cycle (e.g., 3 seconds for short instructions and 10 seconds for long instructions); curve fusion can be achieved using statistical methods (e.g., the mean curve within ±3 standard deviations) or clustering algorithms (e.g., K-means for extracting typical curves); the features of the benchmark template (start-up threshold, peak time, etc.) can be automatically identified through feature point detection algorithms. The division of the operating condition sub-library can be based on the clustering results of operating condition parameters (e.g., load rate, ambient temperature, intake pressure), the time constant can be calculated by fitting the response curve using a first-order inertial model, and the parameter correlation coefficient can be analyzed using Pearson correlation coefficient or partial correlation coefficient.

[0105] In some embodiments of this application, dividing the parameter micro-fluctuation transmission stages is a key step in analyzing equipment state changes and identifying early signs of degradation. However, there are challenges in how to extract the core features of each stage in a targeted manner, avoiding overly general stage divisions or feature extraction that lacks specificity.

[0106] In response, this application further proposes to calculate the second derivative of the deviation evolution trajectory and take the inflection point where the derivative changes from positive to negative as the stage node; the first stage records the propagation path of the initial fluctuation parameters, the second stage counts the number of parameters involved in the fluctuation and the intensity of their influence, and the third stage calculates the decay rate of the fluctuation amplitude of each parameter.

[0107] Specifically, when dividing the transmission stages of parameter micro-fluctuations, the system first calculates the second derivative of the deviation evolution trajectory (such as the curve of air compressor pressure deviation changing with time). The second derivative reflects the increasing or decreasing trend of the deviation change rate. When the derivative changes from positive to negative, it indicates that the rate of increase of deviation fluctuation begins to slow down, and this inflection point is taken as the stage node. For example, the second derivative of a certain deviation trajectory shows inflection points from positive to negative at 10 seconds and 25 seconds, thus dividing the transmission process into three stages: 0-10 seconds (stage 1), 10-25 seconds (stage 2), and after 25 seconds (stage 3).

[0108] The first stage focuses on the source and propagation of the initial fluctuations, recording how the first parameter to fluctuate (such as intake pressure) affects related parameters (such as flow rate and temperature), forming a propagation path map (such as "intake pressure-flow rate-cylinder temperature"). The second stage focuses on analyzing the degree of fluctuation diffusion, counting the number of newly added parameters involved in the fluctuation (such as increasing from 2 to 5), and quantifying the role of each parameter through the influence intensity coefficient (such as the percentage increase in the deviation of other parameters caused by the fluctuation of a certain parameter). The third stage focuses on the convergence trend of the fluctuations, calculating the decay rate of the fluctuation amplitude of each parameter over time (such as the rate at which the pressure fluctuation amplitude decreases from 0.2MPa to 0.05MPa is 0.03MPa / second).

[0109] This method divides the process into stages based on the mathematical characteristics of parameter fluctuations (inflection points of the second derivative), closely linking each stage node to the intrinsic dynamic changes of the fluctuations, rather than simply dividing the time frame. The feature extraction for each stage has its own focus, comprehensively presenting the entire process of micro-fluctuations from their origin to their intensification and eventual convergence. Especially in the early stages of equipment degradation, where parameter fluctuations are subtle and the transmission paths are complex, this method can accurately capture the stage-specific characteristics of the fluctuations, providing a clear analytical framework for identifying abnormal transmission patterns.

[0110] In the technical solution of this application, the second derivative of the deviation evolution trajectory can be calculated by numerical differentiation algorithm (such as finite difference method), and the inflection point identification can be combined with threshold judgment (the inflection point is confirmed when the second derivative changes from positive to negative for 3 consecutive sampling points); the propagation path record of the first stage can adopt a directed graph model, with nodes representing parameters and directed edges representing influence relationships; the influence intensity statistics of the second stage can be calculated by the correlation degree between parameters using Pearson correlation coefficient or mutual information entropy; the decay rate calculation of the third stage can be achieved by linearly fitting the curve of fluctuation amplitude changing with time, and the slope is the decay rate.

[0111] In some embodiments of this application, the dynamic evolution trajectory of the output deviation features provides basic data for identifying device anomalies. However, segments in the trajectory with abnormal deviation increases often contain crucial early degradation information. A challenge lies in accurately extracting these segments and calibrating the criteria for determining deviation increases to avoid feature omissions or misjudgments due to inappropriate judgment scales.

[0112] In response, this application further proposes to extract segments from the trajectory whose deviation increases exceed common values ​​for similar operating conditions within several consecutive sampling periods, align the start and end points of the corresponding instruction execution segment and response change segment; if the slope fluctuation amplitude of the aligned response change segment exceeds the historical normal range, extract the equipment operation vibration waveform of that interval and compare the peak interval with the early waveform of the same type of deviation; adjust the judgment scale of deviation increase according to the difference in peak interval, re-extract feature segments until they conform to the historical fluctuation pattern, and archive the adjusted scale with the vibration waveform features.

[0113] Specifically, after outputting the dynamic evolution trajectory, the system first filters out segments from the trajectory where the deviation increase (e.g., air compressor pressure deviation increases from 0.05MPa to 0.15MPa) exceeds the common value for similar operating conditions (e.g., common increase ≤ 0.08MPa) within three consecutive sampling periods. The system then aligns the start and end points of the corresponding instruction execution segment (e.g., 0-2 seconds of the "pressure increase instruction") with the response change segment (e.g., 1-3 seconds of pressure increasing from 0.8MPa to 0.95MPa) to eliminate time offset. If the slope fluctuation amplitude of the aligned response change segment (e.g., the normal slope is stable at 0.07MPa / second, but currently fluctuates wildly between 0.05-0.09MPa / second) exceeds the historical range, the system extracts the vibration waveform of the corresponding interval (e.g., cylinder vibration signal) and compares the peak interval with the normal waveform of the same type of deviation earlier—for example, the normal waveform peak interval is 0.1 seconds, but the current interval has changed to 0.08-0.12 seconds due to component loosening. Based on the difference in peak intervals (a 40% increase in fluctuation amplitude), the judgment scale for the deviation increase is adjusted (e.g., the common value is relaxed from 0.08MPa to 0.1MPa). After re-extracting the segments, they are made to conform to historical patterns, and the adjusted scale is associated with the vibration waveform of the loosening characteristics and archived.

[0114] This method effectively addresses complex scenarios that are difficult to handle with single numerical analysis by using the physical characteristics of the vibration waveform to reverse-calibrate the criteria for determining the magnitude of the deviation. In particular, when the magnitude of the deviation is near the critical value, the difference in the interval between vibration wave peaks can provide a more sensitive criterion, which helps to avoid abnormal omissions caused by rigid scale.

[0115] In the technical solution of this application, the alignment of the instruction execution segment and the response change segment can be achieved using the Dynamic Time Warping (DTW) algorithm, allowing for a certain time distortion to achieve optimal matching; the slope fluctuation amplitude of the response change segment can be calculated by calculating the standard deviation of the slopes of adjacent sampling points; the comparison of the peak intervals of the vibration waveform can be achieved by combining peak detection and sliding window statistics to quantify the mean and fluctuation range of the intervals. The adjustment of the judgment scale can establish a mapping relationship between the peak interval difference and the scale correction amount (e.g., for every 10% increase in interval fluctuation, the scale is relaxed by 5%), ensuring that the adjusted scale is consistent with the physical signal characteristics.

[0116] In some embodiments of this application, identifying behavioral deviations under similar instructions is an important step in determining whether a device is abnormal. However, when the consistency of parameter change curves for multiple consecutive deviations is low, it is difficult to accurately define the criteria for determining deviations based solely on single curve analysis, which may lead to insufficient reliability of the identification results.

[0117] To address this, this application further proposes to perform overlay analysis on parameter change curves of consecutive behavioral deviations under the same type of command, and to statistically analyze the area ratio of the overlapping region of the curves; if the ratio is lower than a set value, the interval with the most significant deviation among each curve is extracted and compared with the valley depth of the parameter curve of the same type in the early stage of degradation in history; according to the direction of the difference in valley depth, the identification threshold of behavioral deviation is corrected, and the overlay analysis is repeated until the overlapping area reaches the standard, and the corrected threshold and curve features are stored together.

[0118] Specifically, when identifying behavioral deviations, the system first overlays the parameter change curves (e.g., pressure drop curves) of five consecutive behavioral deviations under the same command (e.g., the "unload command" of an air compressor), and calculates the area ratio of the overlapping region (normally ≥60%, currently only 45%). If the ratio is lower than the set value, the most significant deviation interval in each curve (e.g., the abnormal pressure drop segment) is extracted and compared with the valley depth of the parameter curves from the early stages of similar degradation in history (e.g., the initial stage of valve leakage). For example, the valley depth of the historical curve is 0.3 MPa, while the valley depths of the current five curves are scattered between 0.2 and 0.4 MPa. Based on the direction of the depth difference (overall 10% shallower), the behavioral deviation identification threshold is adjusted (e.g., the pressure deviation threshold is lowered from 0.2 MPa to 0.18 MPa). After re-overlay analysis, the overlapping area ratio is increased to 65%, and the adjusted threshold is stored together with the valley characteristics of this batch of curves.

[0119] This method, through the identification and comparison of multiple curves, helps to overcome the randomness problem of single curve analysis. Especially in the early stage of degradation, when the parameter deviation characteristics have not yet stabilized, it can capture hidden common deviation patterns by correcting the convergent and dispersed curve characteristics through threshold correction.

[0120] In the technical solution of this application, the curve overlay analysis can generate an average curve using a point-by-point weighted average method, and the consistency can be quantified by calculating the percentage of overlapping area between each curve and the average curve; the valley depth comparison can extract depth parameters by combining extreme value detection algorithms (such as local minimum value identification), and use variance analysis to measure the difference with historical curves; the correction of the identification threshold can establish a linear mapping relationship between the valley depth difference and the threshold adjustment amount (such as lowering the threshold by 8% if the depth is 10% shallower), ensuring that the corrected threshold matches the actual deviation characteristics.

[0121] Secondly, this invention also proposes a data-driven air compressor control system, which includes the following steps:

[0122] The acquisition module synchronously acquires the physical parameters of the air compressor and the equipment response data corresponding to each control command;

[0123] A module is established to create equipment baseline response characteristics corresponding to different control commands based on historical normal operation data;

[0124] The identification module compares real-time response data with baseline response characteristics to identify response drift, delay extension, feedback structure changes, and behavioral deviations under similar instructions when the control logic remains unchanged, and outputs the dynamic evolution trajectory of the deviation characteristics.

[0125] When the behavior deviation exceeds the set range, the output module first divides the transmission stage of parameter micro-fluctuation by the key turning point of the deviation evolution trajectory, and then checks whether the influence relationship between parameters in each stage reproduces the transmission law of the historical degradation process under the same working conditions. If the change order and intensity distribution of the influence relationship of parameters in each stage are consistent with the historical characteristics, even if the physical parameters are not exceeded, it is still determined that the equipment has early signs of structural fatigue, control link abnormality or internal medium condition deterioration.

[0126] The feedback module adjusts control commands based on the risk assessment results and feeds the new commands and response data back to the expected response model unit to update the baseline characteristics.

[0127] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A data-driven air compressor control method, characterized in that, include: The physical parameters of the air compressor and the equipment response data corresponding to each control command are acquired synchronously. Based on historical normal operation data, establish the equipment baseline response characteristics corresponding to different control commands; By comparing real-time response data with baseline response characteristics, we can identify response drift, delay extension, feedback structure changes, and behavioral deviations under similar commands when the control logic remains unchanged, and output the dynamic evolution trajectory of the deviation characteristics. When the behavior deviation exceeds the set range, the key turning point of the deviation evolution trajectory is used as the boundary to divide the transmission stage of parameter micro-fluctuation. Then, the influence relationship between parameters in each stage is checked to see if the transmission law of the historical degradation process under the same working conditions is reproduced. If the change order and intensity distribution of the influence relationship of parameters in each stage are consistent with the historical characteristics, even if the physical parameters are not exceeded, it is still judged as an early sign of structural fatigue, abnormal control link or deterioration of internal medium in the equipment. Adjust control instructions based on risk assessment results, and feed the new instructions and response data back to the expected response model unit to update the baseline characteristics; After determining that the device shows early signs, the following is also included: Obtain the parameter influence relationships in each transmission stage that are inconsistent with the historical characteristic change order, and mark them as abnormal action points; Starting from the abnormal point of action, obtain the response time difference sequence of the parameter after the control command is issued, and compare it with the characteristics of the time difference sequence of the historical normal response. If the interval fluctuation amplitude of a certain link in the time difference sequence exceeds the usual range of similar working conditions, extract the equipment operation sound pattern or vibration spectrum segment corresponding to that link and compare it with the sound pattern / spectrum of the same link in the early stage of historical degradation to make resonance peak comparison. If the resonance peak shift does not reach the warning threshold, adjust the analysis dimension of the parameter interaction relationship and re-examine the progression order; After each round of analysis, the number of abnormal action points is counted to see if it has decreased. If it has not decreased in two consecutive rounds, the typical deviation transmission path of the same type of equipment under the same degradation scenario is introduced as a reference, and the consistency between the current time difference sequence and the reference path is checked step by step. Until the number of abnormal action points falls below a set threshold, the finally located deviation source and verification parameters are associated with the early sign judgment results and updated to the baseline feature library of the expected response model unit.

2. The data-driven air compressor control method according to claim 1, characterized in that, In the initial stage of the parameter micro-fluctuation propagation phase, it also includes: Obtain parameter combinations that do not match the intensity distribution of historical degradation transmission patterns in each stage, and align the peaks of the control command curves corresponding to these parameters with the equipment response curves. If, after alignment, the number of curve intersections exceeds the preset range for the same type of working condition in multiple consecutive fluctuation cycles, the vibration waveform of the equipment operation in the dense intersection area is extracted and compared with the vibration waveform of the normal state in the same stage by the trough spacing. Adjust the sampling interval of the parameter combination according to the direction of the difference in the trough spacing, recheck the intensity distribution until it conforms to the historical characteristics, and archive the adjusted sampling interval and vibration waveform characteristics together.

3. The data-driven air compressor control method according to claim 1, characterized in that, The intermediate stage of verifying the influence of parameters also includes: Find the group of parameters with the largest deviation from the historical evolution order from each stage, and calculate the interval between the instruction trigger time and the time when the response extreme value appears for the corresponding group. If the dispersion of the interval value exceeds the normal range of similar operating conditions, extract the equipment operating temperature field distribution segment corresponding to the set of parameters and compare it with the temperature field segment of the normal state at the same stage to make hot spot migration path comparison. Based on the direction of hotspot migration differences, the threshold for determining the order of parameter changes is adjusted until it conforms to historical characteristics. The adjusted threshold is then associated with and saved with the temperature field characteristics.

4. The data-driven air compressor control method according to claim 1, characterized in that, The later stages of the transmission phase also include: Select the parameter interaction relationship that deviates most significantly from the historical pattern in each stage, and statistically analyze the ratio of instruction duration to response amplitude change corresponding to this relationship; If the ratio fluctuation exceeds the common range of similar working conditions, extract the fluid pressure pulsation curve of the equipment during this stage and compare the overlap of the pulsation cycle with the normal curve under the same condition. Based on the difference in period overlap, the analysis weights of parameter interaction relationships are adjusted until they conform to historical characteristics. The adjustment results are then associated with and stored in relation to pressure pulsation characteristics.

5. The data-driven air compressor control method according to claim 1, characterized in that, Establishing equipment baseline response characteristics includes: Historical normal data are grouped according to control command type, and the response parameter change curves within a preset time window after each command is issued are extracted. By integrating multiple curves of the same type of instruction, a benchmark template is generated that includes the response start threshold, steady-state fluctuation range, and peak occurrence time. A baseline feature sub-library is established for different operating conditions, which includes the typical time constant and parameter correlation coefficient of the command response under that operating condition.

6. The data-driven air compressor control method according to claim 1, characterized in that, The transmission stages of parameter micro-fluctuations include: Calculate the second derivative of the deviation evolution trajectory and take the inflection point where the derivative changes from positive to negative as the stage node; The first stage records the propagation path of the initial fluctuation parameters; the second stage counts the number of parameters involved in the fluctuation and their influence intensity; and the third stage calculates the decay rate of the fluctuation amplitude of each parameter.

7. The data-driven air compressor control method according to claim 1, characterized in that, Following the dynamic evolution trajectory of the output bias characteristics, the following is also included: Extract segments from the trajectory whose deviation increases exceed common values ​​for similar working conditions within several consecutive sampling periods, and align the start and end points of the instruction execution segment and the response change segment corresponding to the segment. If the slope fluctuation amplitude of the aligned response change segment exceeds the historical normal range, then the interval in the response change segment where the slope fluctuation amplitude exceeds the historical normal range is taken as the target interval, and the equipment operation vibration waveform corresponding to the target interval is extracted and compared with the peak interval of the early waveform of the same type of deviation. Based on the difference in peak intervals, the criteria for determining the increase in deviation are adjusted, and feature segments are re-extracted until they conform to historical fluctuation patterns. The adjusted criteria are then associated with the vibration waveform characteristics and archived.

8. The data-driven air compressor control method according to claim 1, characterized in that, The process of identifying behavioral deviations also includes: Perform overlay analysis on parameter change curves of multiple consecutive behavioral deviations under the same type of command, and statistically analyze the area ratio of the overlapping region of the curves; If the percentage is lower than the set value, the interval with the most significant deviation from each curve is selected and compared with the valley depth of the parameter curve of the same type in the early stage of degradation in history. Based on the direction of the difference in valley depth, the identification threshold for behavioral deviation is corrected, and the analysis is re-overlaid until the overlapping area reaches the standard. The corrected threshold and curve features are then stored together.

9. A data-driven air compressor control system, applicable to the data-driven air compressor control method according to any one of claims 1 to 8, characterized in that, The system includes: The acquisition module synchronously acquires the physical parameters of the air compressor and the equipment response data corresponding to each control command; A module is established to create equipment baseline response characteristics corresponding to different control commands based on historical normal operation data; The identification module compares real-time response data with baseline response characteristics to identify response drift, delay extension, feedback structure changes, and behavioral deviations under similar instructions when the control logic remains unchanged, and outputs the dynamic evolution trajectory of the deviation characteristics. When the behavior deviation exceeds the set range, the output module first divides the transmission stage of parameter micro-fluctuation by the key turning point of the deviation evolution trajectory, and then checks whether the influence relationship between parameters in each stage reproduces the transmission law of the historical degradation process under the same working conditions. If the change order and intensity distribution of the influence relationship of parameters in each stage are consistent with the historical characteristics, even if the physical parameters are not exceeded, it is still determined that the equipment has early signs of structural fatigue, abnormal control link or deterioration of internal medium. The feedback module adjusts control commands based on the risk assessment results and feeds the new commands and response data back to the expected response model unit to update the baseline characteristics.

Images