Inflator pump air tightness detection method and system based on TPMS feedback and flow modeling

Abstract

The invention provides an inflator pump air tightness detection method and system based on TPMS feedback and flow modeling, and the method comprises the steps: obtaining a real-time pressure fluctuation signal and an instantaneous flow fluctuation waveform of an inflator pump gas conveying pipeline, and extracting a corresponding relation between a continuous change rule of a pressure gradient and a flow peak appearance moment; on the basis of the corresponding relation and a three-dimensional structure model of a sealed cavity in the inflator pump, pressure gradient spatial and temporal distribution and flow peak spatial and temporal distribution in the gas flowing process are reconstructed, the flowing track of gas in a pipeline is traced, and the spatial mapping relation between a flow abnormal fluctuation area and a pressure attenuation change area is recognized; and comparing the gas flow path with a gas flow path in a standard sealing state, positioning a key node of sealing failure in combination with a space mapping relationship, and determining the quantification level of the air tightness defect of the inflator pump according to the abnormal change degree of the key node and in combination with the distribution characteristics and mutual influence of the plurality of key nodes. According to the invention, the accuracy and reliability of air tightness detection are effectively improved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a method and system for detecting the air tightness of an air pump based on TPMS feedback and flow modeling. Background Technology

[0002] As a key device for gas delivery and pressure control, the airtightness of air pumps is a core indicator for evaluating product quality and operational safety, directly affecting gas delivery efficiency and system operational stability. Current airtightness testing technologies for air pumps traditionally rely on monitoring static parameters such as pressure decay and flow stability during the inflation process, or simply comparing pressure and flow values ​​at specific time points, combined with preset thresholds to determine airtightness compliance. These methods generally fail to adequately capture the dynamic and coordinated changes in pressure and flow during gas flow, easily leading to missed detections of early, minute leaks. Furthermore, when locating seal failures, traditional methods often rely on overall performance parameters to infer possible leak areas, lacking precise analysis of the spatial correlation between abnormal flow and pressure decay. This results in ambiguous seal failure location, affecting the accuracy and reliability of airtightness defect assessment and failing to meet the demands for high-precision, high-reliability airtightness testing. Summary of the Invention

[0003] This invention provides a method and system for detecting the air tightness of an air pump based on TPMS feedback and flow modeling.

[0004] In a first aspect, embodiments of the present invention provide a method for detecting the air tightness of an air pump based on TPMS feedback and flow modeling. The method includes: acquiring real-time pressure fluctuation signals from a tire pressure monitoring system during the operation of the air pump and instantaneous flow fluctuation waveforms from the gas delivery pipeline of the air pump. The real-time pressure fluctuation signals are marked with continuous timestamps, and the instantaneous flow fluctuation waveforms are time-aligned with the timestamps. A time-by-time correlation analysis is performed between the pressure change curve corresponding to the real-time pressure fluctuation signals and the instantaneous flow fluctuation waveforms to extract the correspondence between the continuous change pattern of the pressure gradient and the time of occurrence of the flow peak within the complete working cycle of the air pump. This correspondence characterizes the synchronous change characteristics of pressure and flow during gas flow. The correspondence is used in conjunction with the three-dimensional structural model of the internal sealed cavity of the air pump to reconstruct the spatiotemporal distribution of pressure gradient and flow peak during gas flow. This allows for tracing the gas flow trajectory in the pipeline and identifying the spatial mapping relationship between abnormal flow fluctuation areas and pressure decay change areas. This spatial mapping relationship characterizes the positional correlation between abnormal flow and pressure decay. By comparing the gas flow path with the gas flow path under standard sealing conditions, the path differences are analyzed segment by segment. Combined with the spatial mapping relationship, the key nodes of seal failure are located, which are the locations leading to abnormal flow and pressure decay. Based on the degree of abnormal change at these key nodes, and considering the distribution characteristics and mutual influence of multiple key nodes, the quantitative level of the airtightness defect of the air pump is determined.

[0005] Secondly, embodiments of the present invention provide a computer system, the computer device including a processor and a memory, the memory storing a computer program, the computer program being loaded and executed by the processor to implement the airtightness detection method for an air pump based on TPMS feedback and flow modeling as described above.

[0006] The embodiments of this application have the following beneficial effects: This invention acquires the real-time pressure fluctuation signal of the tire pressure monitoring system and the instantaneous flow fluctuation waveform of the gas delivery pipeline of the air pump, and performs time-by-time correlation analysis on the pressure change curve and the flow fluctuation waveform to extract the continuous change law of the pressure gradient and the correspondence between the time of the flow peak within the complete working cycle of the air pump. Through the deep coupling of pressure and flow in the time dimension, it can accurately capture the synchronous characteristics of the two in dynamic changes, avoiding the detection blind spots caused by using pressure or flow parameters in isolation. On this basis, the gas flow path is sorted out based on the correspondence and the spatial mapping relationship between the abnormal flow fluctuation area and the pressure decay change area is identified, realizing the dynamic correlation from the data level to the tracing of the flow trajectory at the physical level. Compared with the conventional overall performance judgment, it can more accurately locate the spatial correlation of abnormal areas. Furthermore, by comparing the standard flow path, the key nodes of sealing failure are located, and the distribution characteristics and mutual influence of multiple key nodes are combined to determine the quantitative level of air tightness defects. The interaction between nodes is considered, making the defect level assessment more consistent with the actual characteristics of the air pump pipeline system as a whole, effectively improving the accuracy and reliability of air tightness detection. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the application environment provided in the embodiments of this application; Figure 2 This is a flowchart illustrating the air tightness detection method for an air pump based on TPMS feedback and flow modeling provided in the embodiments of this application. Figure 3 This is a structural block diagram of the computer system provided in the embodiments of this application. Detailed Implementation

[0008] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0009] In some embodiments, the air tightness detection method for air pumps based on TPMS feedback and flow modeling provided in this application is applied to, for example... Figure 1 The application environment shown. For example, as... Figure 1 As shown, the application environment includes a tire pressure monitoring system 10 and a computer system 20, and the tire pressure monitoring system 10 and the computer system 20 are connected for data transmission.

[0010] One point that needs to be clarified is that the above Figure 2 The descriptions provided are merely exemplary and illustrative. In exemplary embodiments, the functions of the tire pressure monitoring system 10 and the computer system 20 can be flexibly configured and adjusted, and the embodiments of this application do not limit this.

[0011] Please refer to Figure 2 This document illustrates a flowchart of an airtightness testing method for an air pump based on TPMS feedback and flow modeling, provided in an embodiment of this application. The steps in this method can be derived from the above-described... Figure 1 The computer system 20 executes the method. The method may include the following steps: Step S100: Acquire the real-time pressure fluctuation signal of the tire pressure monitoring system during the operation of the air pump and the instantaneous flow fluctuation waveform of the air pump gas delivery pipeline. The real-time pressure fluctuation signal is marked with continuous timestamps, and the instantaneous flow fluctuation waveform is time-aligned with the timestamps.

[0012] A tire pressure monitoring system (TPMS) continuously monitors changes in tire pressure while the air pump is operating. The real-time pressure fluctuation signal reflects the dynamic changes in pressure over time during inflation. A timestamp assigns precise time information to each pressure data point, and the instantaneous flow rate fluctuation waveform in the air pump's gas delivery pipeline reflects the flow rate characteristics over time as the gas flows through the pipeline. Ensuring that the instantaneous flow rate fluctuation waveform is aligned with the timestamp is crucial for accurately correlating pressure changes with flow rate changes in time, thereby enabling a deeper exploration of the correlation between the two.

[0013] To acquire real-time pressure fluctuation signals, a pressure sensor can be installed inside the tire to detect minute changes in tire pressure and convert the pressure value into an electrical signal, which is then transmitted to a data acquisition system. The data acquisition system automatically adds a corresponding timestamp to each acquired pressure data point, forming a time-stamped real-time pressure fluctuation signal. For acquiring instantaneous flow rate fluctuation waveforms, a flow sensor can be installed on the gas delivery line of the air pump. The flow sensor measures the gas flow rate in the line in real time and outputs the flow data as a waveform. The data acquisition system synchronously records the flow waveform and its corresponding timestamp, ensuring temporal consistency with the real-time pressure fluctuation signal.

[0014] Step S200: Perform time-by-time correlation analysis between the pressure change curve corresponding to the real-time pressure fluctuation signal and the instantaneous flow fluctuation waveform, extract the continuous change law of the pressure gradient and the correspondence between the time of the flow peak within the complete working cycle of the air pump, and characterize the synchronous change characteristics of pressure and flow during the gas flow process.

[0015] As one implementation method, step S200 specifically includes the following steps S210~S260: Step S210: Perform time-series framing on the real-time pressure fluctuation signal, and divide the pressure change curve into continuous pressure data segments according to the time interval of the mechanical action cycle of the air pump. Each pressure data segment contains all pressure sampling values ​​within that time period and the start and end timestamps. The time interval is determined by the correlation between the air pump motor speed and the piston movement cycle.

[0016] In this step, the real-time pressure fluctuation signal is segmented into frames based on the mechanical operation cycle of the air pump. The mechanical operation of the air pump is mainly achieved by the piston movement driven by a motor. The motor speed determines the piston's movement frequency, and the piston's movement cycle directly affects the working rhythm of the air pump. By analyzing the correlation between the air pump motor speed and the piston's movement cycle, the segmentation time interval of the pressure change curve can be accurately determined.

[0017] Each pressure data segment contains all pressure sample values ​​within that time period, along with corresponding start and end timestamps. When the air pump motor operates at a certain speed, the piston moves according to a corresponding cycle. Based on the correlation between motor speed and piston cycle, a suitable time interval is determined, and the pressure change curve is divided according to this time interval. Each resulting pressure data segment has a clearly defined start and end time and contains complete pressure sample values ​​within that time period. In actual operation, the segmentation time interval can be accurately determined by monitoring the air pump motor speed sensor signal and piston position sensor signal. Then, a data processing algorithm is used to accurately segment the pressure change curve according to the determined time interval, generating continuous pressure data segments.

[0018] Step S220: Perform fluctuation feature extraction for each pressure data segment, and use sliding window trend recognition to distinguish the upward trend segment, downward trend segment and stable segment of pressure sample values ​​within the segment. Calculate the absolute value of the slope, the number of continuous sampling points and the fluctuation frequency of each trend segment to generate a pressure segment feature vector containing multi-dimensional trend parameters.

[0019] As one implementation method, step S220 specifically includes the following steps S221~S226: Step S221: Initialize the sliding window parameters. The window length is determined according to the proportion of the total number of sampling points in the pressure data segment. The window step size is a single sampling point. Traverse the pressure sampling value sequence within the pressure data segment.

[0020] A sliding window is a tool used for localized analysis of pressure data segments. The window length is determined by the proportion of the total number of sampling points in the pressure data segment. This allows the window size to be dynamically adjusted according to the actual length of the data segment to adapt to different data characteristics and analytical needs. Setting the window step size to a single sampling point means that the window moves forward by the distance of one sampling point each time, thus enabling detailed capture of subtle changes in the pressure data.

[0021] When initializing the sliding window parameters, the window length is first determined based on the total number of sampling points in the pressure data segment and a preset ratio. For example, if the preset ratio is a certain percentage and the pressure data segment has a certain number of sampling points, the window length is the number multiplied by the preset ratio. Then, the window step size is set to a single sampling point. Starting from the first sampling point of the pressure data segment, the window covers the corresponding number of sampling points, and subsequent analysis and processing of the pressure sampling values ​​within the window are performed. Afterward, the window moves forward one sampling point, covering a new set of sampling points, and the analysis continues until the entire pressure sampling value sequence within the pressure data segment has been traversed.

[0022] Step S222: Perform linear fitting on the pressure sampling values ​​within each window, and determine the trend type by the sign of the slope of the fitted line. If the slope is positive, it is marked as an upward trend window; if the slope is negative, it is marked as a downward trend window; if the absolute value of the slope is less than the trend threshold, it is marked as a stationary trend window.

[0023] When performing linear fitting, the least squares method can be used. For each pressure sample value within a window, it is treated as a data point, and the slope of the fitted line is calculated using the least squares method. Based on the sign of the slope and the comparison with a trend threshold, the trend type of the window is marked. If the slope is positive, it indicates that the pressure value shows an upward trend within the window, and it is marked as an upward trend window; if the slope is negative, it shows a downward trend, and it is marked as a downward trend window; if the absolute value of the slope is less than a pre-set trend threshold, it indicates that the pressure value changes little within the window and remains basically stable, and it is marked as a stable trend window.

[0024] Step S223: Merge windows with consecutive trends of the same type into trend segments. Merge upward trend windows into upward trend segments, downward trend windows into downward trend segments, and stable trend windows into stable trend segments. Record the starting sampling point index and ending sampling point index of each trend segment.

[0025] After labeling the trend type for each window, to more clearly describe the overall trend changes within the pressure data segment, consecutive windows with the same trend type are merged into a single trend segment. This integrates scattered window data into continuous data segments with clear trend characteristics. An upward trend segment is formed by merging consecutive upward trend windows, a downward trend segment by merging consecutive downward trend windows, and a stable trend segment by merging consecutive stable trend windows. Recording the starting and ending sampling point indices of each trend segment accurately defines the position and range of each trend segment within the pressure data segment, providing a basis for subsequent calculations of relevant parameters for the trend segment. For example, for a series of consecutive upward trend windows, they are merged into one upward trend segment, and the starting and ending sampling point indices of this upward trend segment are recorded. Similarly, the same merging operation is performed on downward trend windows and stable trend windows, and the corresponding starting and ending sampling point indices are recorded.

[0026] Step S224: Calculate the absolute value of the representative slope of each trend segment by taking the arithmetic mean of the absolute values ​​of the slopes of all fitted straight lines within the merged window. Calculate the number of sampling points contained in the trend segment based on the starting and ending sampling point indices as the number of continuous sampling points. Calculate the fluctuation frequency by dividing the number of extreme points of the pressure sampling values ​​within the trend segment by the number of continuous sampling points.

[0027] The absolute value of the slope comprehensively reflects the average rate of pressure change within a trend segment. By taking the arithmetic mean of the absolute values ​​of the slopes of all fitted straight lines within the merged window, an index representing the speed of pressure change in that trend segment can be obtained. The number of continuous sampling points refers to the number of pressure samples contained in the trend segment, reflecting the duration of the trend segment. The fluctuation frequency reflects the severity of pressure fluctuations within the trend segment, obtained by dividing the number of extreme pressure sampling points within the trend segment by the number of continuous sampling points.

[0028] For a trend segment, first sum the absolute values ​​of the slopes of the fitted lines within all windows of the trend segment, then divide by the number of windows to obtain the representative absolute value of the slope. Based on the start and end sampling point indices of the trend segment, calculate the difference between them and add one to obtain the number of sampling points contained in the trend segment, i.e., the number of continuous sampling points. Within the trend segment, count the number of extreme points of the pressure sampling values, divide this number by the number of continuous sampling points, and obtain the fluctuation frequency. For example, for an upward trend segment, calculating its representative absolute slope, number of continuous sampling points, and fluctuation frequency allows us to describe the pressure change characteristics of the upward trend segment from different perspectives.

[0029] Step S225: Arrange the absolute value of the slope of the upward trend segment, the number of continuous sampling points, the fluctuation frequency, the corresponding parameters of the downward trend segment, and the corresponding parameters of the stable segment in a fixed order to obtain the pressure segment feature vector containing multiple dimension parameters; where each dimension parameter is determined by the cross-index of the trend segment type and the parameter type.

[0030] To comprehensively describe the fluctuation characteristics of each pressure data segment, key parameters from different trend segments need to be integrated. The absolute values ​​of the slopes, the number of continuous sampling points, and the fluctuation frequency of the upward trend segment, downward trend segment, and stable segment are arranged in a fixed order to form a pressure segment feature vector containing multiple dimensions of parameters. By cross-indexing the trend segment type (upward trend segment, downward trend segment, stable segment) and the parameter type (absolute value of slope, number of continuous sampling points, fluctuation frequency), the position of each dimension parameter in the feature vector can be accurately determined. For example, the parameters can be arranged in the following order: absolute value of the slope of the upward trend segment, number of continuous sampling points of the upward trend segment, fluctuation frequency of the upward trend segment; absolute value of the slope of the downward trend segment, number of continuous sampling points of the downward trend segment, fluctuation frequency of the downward trend segment; absolute value of the slope of the stable segment, number of continuous sampling points of the stable segment, fluctuation frequency of the stable segment. For a pressure data segment, the relevant parameters for the upward trend segment, downward trend segment, and stable segment are calculated separately, and then these parameters are arranged into a vector in the above order.

[0031] Step S226: Normalize the parameters of each dimension in the feature vector of the pressure segment, and map the parameter values ​​to a preset numerical range. The normalization is based on the maximum and minimum values ​​of the parameters of the same type of pressure data segment in historical detection data, so that the feature vectors of different pressure data segments are comparable.

[0032] In this step, the parameters of each dimension in the pressure segment feature vector are normalized, mapping their values ​​to a preset numerical range, such as the common [0,1] range. The normalization is based on the maximum and minimum values ​​of the parameters for the same type of pressure data segment in historical testing data. Since the parameters of different pressure data segments may have different value ranges, normalization eliminates the influence of these range differences, allowing the feature vectors of different pressure data segments to be compared and analyzed on the same scale.

[0033] For each dimension parameter in the pressure segment feature vector, a normalization formula is used to transform it to a preset numerical range based on the maximum and minimum values ​​of that parameter in historical detection data. For example, for a parameter whose current value is a certain value, the maximum value of that parameter in historical detection data is a relatively large value, and the minimum value is a relatively small value. The parameter is normalized using the linear normalization formula: Normalized value = (current value - minimum value) / (maximum value - minimum value). This process is performed on all dimension parameters in the pressure segment feature vector, ultimately resulting in a normalized pressure segment feature vector where all dimension parameters fall within the preset numerical range.

[0034] Step S230: Perform morphological feature analysis on the instantaneous flow fluctuation waveform, identify the start point of the rising edge, the end point of the rising edge, the peak plateau segment, the start point of the falling edge, and the end point of the falling edge in the waveform through multi-scale edge detection, record the timestamp coordinates of each feature point and the waveform curvature value at the corresponding position, and generate a flow feature event sequence arranged in chronological order.

[0035] As one implementation method, step S230 specifically includes the following steps S231~S236: Step S231: Gaussian filtering is performed on the instantaneous flow fluctuation waveform, and multiple filtering parameters with different smoothness are set for multi-scale filtering to generate a set of flow waveforms with different smoothness; the filtering parameters are determined by the correlation between the waveform sampling frequency and the gas compression frequency of the air pump.

[0036] Convolution of a signal using a Gaussian function can effectively remove noise while preserving its main features. In this step, multi-scale Gaussian filtering is applied to the instantaneous flow fluctuation waveform. Setting multiple filtering parameters with different smoothing levels allows for waveform analysis at different scales, thereby obtaining more information about the waveform characteristics. The filtering parameter settings are closely related to the waveform sampling frequency and the gas compression frequency of the air pump. The waveform sampling frequency determines the data acquisition time interval, while the gas compression frequency of the air pump reflects the frequency of gas flow rate changes. By analyzing the correlation between the two, appropriate filtering parameters can be determined. For example, when the waveform sampling frequency is high and the air pump gas compression frequency is low, a smaller filtering parameter can be selected to retain more waveform details; conversely, a larger filtering parameter can be selected to enhance the smoothing effect. By applying Gaussian filtering to the instantaneous flow fluctuation waveform using different filtering parameters, multiple flow waveforms with different smoothing levels are obtained, forming a set of flow waveforms containing different smoothing levels.

[0037] Step S232: Calculate the second derivative of the smooth flow waveform at each scale to obtain the curvature change curve of the waveform. Identify potential edge points of the waveform by detecting the zero crossing point of the curvature change curve. Edge points at different scales are fused by non-maximum suppression to eliminate false edge points.

[0038] Calculating the second derivative of the smoothed flow waveform at each scale yields the waveform's curvature variation curve. This curve reflects the change in the waveform's bending degree at various points, with its zero-crossing points typically corresponding to edge points where significant changes in curvature occur. Since edge detection at different scales may generate false edge points, affecting the accuracy of subsequent feature recognition, non-maximum suppression (NMS) is needed to fuse and filter edge points at different scales. The basic idea of ​​NMS is to compare the curvature values ​​of each potential edge point at different scales. If the curvature value at a certain scale is not a local maximum, it is suppressed, retaining only true edge points. This eliminates the interference of false edge points and improves the accuracy of edge detection.

[0039] Step S233: Select the rising edge start point of the flow waveform from the merged edge points. When the curvature change curve changes from negative to positive and the waveform slope begins to be greater than the rising threshold, the corresponding sampling point is marked as the rising edge start point, and its timestamp coordinates are recorded.

[0040] A change in curvature curve from negative to positive indicates a shift in waveform curvature, from concave to convex, which usually signifies an impending change in the waveform's trend. Conversely, a waveform slope greater than a rising threshold indicates that the flow rate is beginning to rise rapidly. When both conditions are met, the corresponding sampling point is designated as the rising edge start point. Recording the timestamp coordinates of the rising edge start point allows for accurate determination of the moment when the flow rate begins to rise rapidly. For example, when filtering merged edge points, each edge point is traversed, and its corresponding curvature curve and waveform slope are examined. When an edge point's curvature curve changes from negative to positive, and the waveform slope at that point exceeds a pre-set rising threshold, the sampling point is marked as the rising edge start point, and its timestamp coordinates are recorded.

[0041] Step S234: Identify the rising edge endpoint. When the waveform slope after the rising edge start point reaches the maximum upward angle and then begins to decrease, the corresponding sampling point is marked as the rising edge endpoint, and its timestamp coordinates and waveform curvature value are recorded. The waveform segment between the rising edge start point and the rising edge endpoint is the rising edge.

[0042] Recording the timestamp coordinates and waveform curvature value of the rising edge's endpoint allows for a precise description of the rising edge's ending position and shape characteristics. The waveform segment between the rising edge's start and end points constitutes the rising edge, reflecting the entire process of flow rate rapidly increasing from its initial stage to its maximum value. By analyzing the rising edge's length, slope changes, and other characteristics, we can gain a deeper understanding of the flow rate's variation patterns and characteristics during the rising phase. For example, after determining the rising edge's start point, we continuously monitor the waveform slope from that point. When the waveform slope reaches its maximum value and begins to decrease, assuming this corresponds to a specific sampling point, we mark this sampling point as the rising edge's endpoint and record its timestamp coordinates and waveform curvature value. The waveform segment between the rising edge's start and end points constitutes the rising edge of the flow rate waveform.

[0043] Step S235: Identify the peak plateau segment. After the rising edge ends, if the absolute value of the waveform slope is continuously less than the plateau threshold within a preset duration, confirm that the waveform segment is marked as a peak plateau segment, and record the start timestamp, end timestamp, and average waveform curvature value of the plateau segment.

[0044] The basis for identifying a peak plateau segment is to examine the change in waveform slope after the end of the rising edge. If the absolute value of the waveform slope remains less than a pre-set plateau threshold for a preset duration, it indicates that the flow rate changes little and remains relatively stable during this period. In this case, the waveform segment can be confirmed as a peak plateau segment.

[0045] Recording the start and end timestamps and average waveform curvature value of a peak plateau segment allows for an accurate description of the time range and shape characteristics of that segment. The start and end timestamps define the time interval of the peak plateau segment, while the average waveform curvature value reflects the overall curvature of the waveform. By analyzing the duration and average waveform curvature value of the peak plateau segment, the characteristics of the flow rate during the steady-state phase and its correlation with pressure can be understood. For example, timing begins after the end of the rising edge, and the absolute value of the waveform slope is monitored. If, within a preset duration, the absolute value of the waveform slope is consistently less than a plateau threshold, this segment is marked as a peak plateau segment, and the start and end timestamps and average waveform curvature value are recorded. The average waveform curvature value can be obtained by calculating the average of the curvature values ​​at various points on the waveform segment.

[0046] Step S236: Identify the starting point and ending point of the falling edge. After the peak plateau segment ends, when the waveform slope changes from zero to a negative value and the absolute value is greater than the falling threshold, it is marked as the starting point of the falling edge; when the absolute value of the waveform slope after the falling edge is less than the falling threshold, it is marked as the ending point of the falling edge. Record the timestamp coordinates and waveform curvature values ​​of both. Define the feature combination containing the starting point of the rising edge, the ending point of the rising edge, the peak plateau segment, the starting point of the falling edge, and the ending point of the falling edge as a flow characteristic event, and generate a flow characteristic event sequence by arranging them in the order of timestamps.

[0047] The start and end of the falling edge mark the critical time points when the flow rate begins to decrease and ends, respectively. The condition for determining the start of the falling edge is that after the peak plateau, when the waveform slope changes from zero to a negative value and its absolute value is greater than a pre-set falling edge threshold, it indicates that the flow rate has begun to decrease at a relatively fast rate; the corresponding sampling point at this time is the start of the falling edge. The condition for determining the end of the falling edge is that when the absolute value of the waveform slope after the falling edge is less than the falling edge threshold, it indicates that the flow rate has decreased very slowly and has basically reached a relatively stable value; the corresponding sampling point at this time is the end of the falling edge.

[0048] Recording the timestamp coordinates and waveform curvature values ​​of the fall edge start and end points allows for a precise description of the start and end positions and shape characteristics of the flow decline phase. Combining these characteristic points and segments—rise edge start, rise edge end, peak plateau segment, fall edge start, and fall edge end—is defined as a flow characteristic event. Arranging these flow characteristic events in timestamp order generates a flow characteristic event sequence.

[0049] For example, after the peak plateau segment ends, the change in waveform slope is continuously monitored. When the waveform slope changes from zero to a negative value and the absolute value is greater than the falling threshold, the sampling point is marked as the start of the falling edge, and its timestamp coordinates and waveform curvature value are recorded. Monitoring of the waveform slope continues, and when the absolute value of the waveform slope after the falling edge is less than the falling threshold, the sampling point is marked as the end of the falling edge, and its timestamp coordinates and waveform curvature value are recorded. Arranging these feature points and feature segments in chronological order yields a sequence of flow characteristic events.

[0050] Step S240: Perform dynamic window matching between the pressure data segment and the flow characteristic event sequence based on timestamps, set a matching window that includes forward time offset and backward time offset, so that each flow characteristic event falls within the time range of a unique pressure data segment, and generate a time-related data structure.

[0051] Dynamic window matching is an effective method for accurately correlated different data sequences over time. In this step, pressure data segments and flow characteristic event sequences are matched based on timestamps. Since there may be some time discrepancies between the pressure data segments and flow characteristic event sequences, a matching window containing forward and backward time offsets is needed to ensure that each flow characteristic event is accurately associated with a pressure data segment. The forward and backward time offsets can be adjusted according to the characteristics of the actual data and the required accuracy of the analysis. The purpose of the matching window is to expand the temporal range of the pressure data segment, making it more likely that flow characteristic events fall within the time range of the pressure data segment. By moving the matching window, each flow characteristic event is ensured to fall within the time range of a unique pressure data segment. The generated temporal correlation data structure records the temporal correlation between the pressure data segment and the flow characteristic events, providing a foundation for subsequent calculations of the correlation between the pressure segment feature vector and the flow characteristic events.

[0052] For example, a pressure data segment spans from one moment to another. A forward time offset and a backward time offset are both set to a predetermined time length. The matching window extends from the start time of the pressure data segment minus the backward time offset to the end time plus the forward time offset. For a flow characteristic event, its timestamp is checked to see if it falls within the matching window. If it does, the flow characteristic event is associated with the pressure data segment. This matching operation is performed on all pressure data segments and flow characteristic events, ultimately generating a time-related data structure that clearly shows the time correspondence between pressure data segments and flow characteristic events.

[0053] Step S250: Calculate the correlation weight between the pressure segment feature vector and the flow characteristic event based on the time-related data structure. By statistically analyzing the recurrence frequency of the pressure segment feature vector corresponding to the same flow characteristic event type, select strong correlation combinations with a frequency higher than the correlation threshold.

[0054] The correlation weight reflects the degree of correlation between pressure segment feature vectors and flow characteristic events. Using a time-related data structure, it's possible to clearly identify which flow characteristic events each pressure data segment is associated with. The frequency of recurrence of pressure segment feature vectors corresponding to the same flow characteristic event type is counted; that is, for a given flow characteristic event type, the number of times the same feature vector appears in the associated pressure segment feature vectors is counted. The correlation threshold is a pre-defined standard used to filter out strongly correlated combinations between pressure segment feature vectors and flow characteristic events. If the correlation frequency between a pressure segment feature vector and a certain flow characteristic event type is higher than the correlation threshold, a strong correlation is considered to exist between them.

[0055] Step S260: Extract the absolute value of the slope of the upward trend segment of the pressure data segment in the strongly correlated combination as the pressure gradient change index, associate it with the start timestamp of the peak plateau segment in the corresponding flow characteristic event, and obtain the correspondence between the continuous change law of the pressure gradient covering the complete working cycle of the air pump and the time when the flow peak occurs.

[0056] The starting timestamp of the peak plateau segment in the corresponding flow characteristic event is correlated, as the start time of the peak plateau segment usually corresponds to the moment when the flow reaches its maximum value. By correlating the pressure gradient change index with the time of the flow peak occurrence, the correspondence between the continuous change pattern of the pressure gradient covering the entire working cycle of the air pump and the time of the flow peak occurrence can be obtained. This correspondence clearly demonstrates the intrinsic relationship between the change of pressure gradient and the occurrence of the flow peak throughout the entire working process of the air pump.

[0057] For example, a strongly correlated combination includes pressure data segment C and flow characteristic event type D. The absolute value of the slope of the upward trend segment in pressure data segment C is a specific value, and the starting timestamp of the peak plateau segment in flow characteristic event type D is a specific moment. The absolute value of the slope of the upward trend segment is correlated with this starting timestamp. This operation is performed on all strongly correlated combinations, ultimately yielding a series of correspondences between pressure gradient change indicators and the time of flow peak occurrence. This allows us to summarize the continuous change pattern of pressure gradient covering the entire working cycle of the air pump and the correspondence between the time of flow peak occurrence.

[0058] Step S300: Based on the correspondence and the three-dimensional structural model of the sealed cavity inside the air pump, reconstruct the spatiotemporal distribution of pressure gradient and flow peak during the gas flow process, trace the flow trajectory of the gas in the pipeline, identify the spatial mapping relationship between the abnormal flow fluctuation area and the pressure decay change area, and the spatial mapping relationship characterizes the positional correlation between abnormal flow and pressure decay.

[0059] As one implementation method, step S300 specifically includes the following steps S310~S360: Step S310: Obtain a three-dimensional solid model of the sealed cavity inside the air pump, and divide the model space into cubic mesh units uniformly through three-dimensional meshing.

[0060] A three-dimensional solid model of the internal sealed cavity of an air pump can be accurately created using 3D modeling software. This model accurately describes the geometry, dimensions, and structural features of the internal sealed cavity. 3D mesh generation is a method of spatial discretizing the three-dimensional solid model. By uniformly dividing the model space into cubic mesh elements, the continuous three-dimensional space can be transformed into discrete mesh points, facilitating subsequent numerical calculations and analysis.

[0061] When performing 3D mesh generation, it's crucial to ensure that the mesh cells are uniformly divided, meaning each cubic mesh cell has the same size and shape. This guarantees consistent computational accuracy and complexity for each mesh cell in subsequent calculations. 3D mesh generation assigns a unique 3D coordinate index to each mesh cell, facilitating its location and identification. For example, in a cuboid-shaped sealed cavity model of an air pump, dividing the space at regular intervals along the three coordinate axes yields a series of identical cubic mesh cells. Each mesh cell can be represented by a 3D coordinate, such as (1,2,3) representing the mesh cell in row 1, column 2, and layer 3.

[0062] Step S320: Combining the continuous change law of pressure gradient in the correspondence with the pre-established spatial decay model of pressure gradient, assign a pressure gradient magnitude estimate to each three-dimensional spatial grid cell under each time stamp, and construct the spatiotemporal distribution data structure of pressure gradient.

[0063] The continuous variation of the pressure gradient describes how the pressure gradient changes over time, while the pre-established spatial attenuation model of the pressure gradient describes its attenuation characteristics as it propagates through space. Combining these two pieces of information, a pressure gradient magnitude estimate can be assigned to each 3D spatial grid cell at each time stamp.

[0064] Pressure gradient magnitude estimation refers to the estimated magnitude of the pressure gradient in a specific grid cell at a given time stamp. By performing such estimations for all time stamps and all grid cells, and arranging these estimates in order of time stamp and spatial coordinates, a spatiotemporal distribution data structure for pressure gradients is constructed. This data structure can comprehensively describe the temporal and spatial variations of the pressure gradient during gas flow.

[0065] As one implementation method, step S320 specifically includes the following steps S321~S326: Step S321: Obtain a three-dimensional solid model of the sealed cavity inside the air pump, and divide the model space into cubic mesh units uniformly through three-dimensional meshing; the side length of the mesh unit is determined by the correlation between the minimum feature size of the model and the preset number of meshes, and a unique three-dimensional coordinate index is assigned to each mesh unit.

[0066] The minimum feature size of the model is the size of the smallest geometric feature in the model, while the preset number of mesh elements is predetermined based on the required computational accuracy and the limitations of computational resources. By reasonably determining the side length of the mesh elements, computational efficiency can be improved while ensuring computational accuracy. If the side length of the mesh elements is too large, it may lead to inaccurate description of model details; if the side length is too small, it will increase the computational complexity and workload. A unique three-dimensional coordinate index is assigned to each mesh element to facilitate subsequent positioning and calculation of each mesh element. For example, in a cubic mesh partitioning, each mesh element can be represented by three integers to indicate its position in the three coordinate axes. Using the three-dimensional coordinate index, the required mesh elements can be found quickly and accurately, and related calculations and analyses can be performed on them.

[0067] Step S322: Extract all timestamps and their corresponding pressure gradient change values ​​from the correspondence, generate a pressure gradient time series, and supplement the pressure gradient values ​​of missing timestamps through time interpolation so that the timestamps are evenly distributed at fixed time intervals.

[0068] To extract all timestamps and their corresponding pressure gradient changes, the data is correlated with the continuous variation of the pressure gradient and the occurrence of peak flow rates. These values ​​are then arranged chronologically to generate a pressure gradient time series. Since the actual collected data may have uneven time intervals or missing timestamps, time interpolation is needed to supplement the missing pressure gradient values. Time interpolation is a method of estimating unknown data points from known data points. Linear interpolation can be used to estimate the pressure gradient value of missing timestamps based on the pressure gradient values ​​of two adjacent known timestamps. The principle of linear interpolation is that the change in pressure gradient values ​​between two adjacent known data points is linear. Time interpolation ensures that timestamps are evenly distributed at fixed time intervals, facilitating subsequent calculations and analysis.

[0069] For example, a pressure gradient time series contains pressure gradient values ​​P1 and P3 corresponding to timestamps t1 and t3, but lacks the pressure gradient value at timestamp t2. Using linear interpolation, the pressure gradient value P2 at time t2 is estimated based on the values ​​of t1, P1, and t3, P3. Assuming t2 lies between t1 and t3, and the ratio of t2-t1 to t3-t1 is a specific ratio, P2 can be obtained through linear calculation. This interpolation operation is performed on all missing timestamps, ultimately resulting in a pressure gradient time series with timestamps uniformly distributed at fixed time intervals.

[0070] Step S323: Obtain the pre-established spatial attenuation model of the pressure gradient, which defines the attenuation characteristics of the pressure gradient as it propagates along different pipe sections, joints and cavity structures.

[0071] The pre-established spatial attenuation model of the pressure gradient is a pre-designed model that considers the attenuation characteristics of the pressure gradient when gas flows through different structural components inside the air pump. Different pipe sections, joints, and cavity structures have different effects on the attenuation of the pressure gradient.

[0072] For example, in longer pipeline sections, the resistance to gas flow is greater, and the pressure gradient gradually decreases with increasing distance. At joints, the pressure gradient may abruptly change due to changes in gas flow direction and increased local resistance. In cavity structures, gas diffusion and mixing can alter the distribution of the pressure gradient. This model can be established using experimental data or numerical simulations. By analyzing large amounts of experimental data or numerically simulating gas flow processes, the influence of different structural components on pressure gradient attenuation can be determined. Integrating these principles into the model allows for accurate consideration of spatial attenuation of the pressure gradient when estimating the pressure gradient magnitude for each grid cell.

[0073] Step S324: For each timestamp, calculate the estimated pressure gradient of each three-dimensional spatial grid cell within its influence range based on the current pressure gradient value and the attenuation model.

[0074] For each timestamp, the current pressure gradient value is obtained from the pressure gradient time series. Based on a pre-established spatial attenuation model of the pressure gradient, considering the propagation of the pressure gradient within the air pump, the estimated pressure gradient values ​​for each 3D spatial grid cell within its influence range are calculated. The influence range of the pressure gradient refers to the spatial region where the pressure will change significantly under the influence of the pressure gradient. For each grid cell, the estimated pressure gradient value is calculated using the spatial attenuation model of the pressure gradient, taking into account factors such as its distance from the pressure source, the pipeline section it is located in, and the structure of joints or cavities. For example, grid cells closer to the pressure source and located in smooth pipeline sections may have larger estimated pressure gradient values; while grid cells farther from the pressure source and passing through multiple joints and curved pipeline sections will have relatively smaller estimated pressure gradient values. During the calculation process, the spatial attenuation model of the pressure gradient performs corresponding attenuation calculations on the current pressure gradient value based on the specific location of the grid cell and the surrounding structural information.

[0075] Step S325: For each grid cell, integrate all timestamps and all relevant pressure gradient sources at that location to generate pressure gradient estimates. Then, fuse these estimates using the rules provided by the attenuation model or based on the confidence level of the estimates to obtain the final pressure gradient magnitude estimate for that grid cell at each timestamp.

[0076] Each grid cell is influenced by multiple pressure gradient sources at different timestamps. These sources can be pressure changes caused by different operating parts of an air pump or gas flow at different locations. The algorithm integrates the pressure gradient estimates generated by all relevant pressure gradient sources at each timestamp within that grid cell. Fusion can be performed according to rules provided by the attenuation model, such as weighted summation based on distance and propagation path. Pressure gradient sources closer to the pressure source and with smooth propagation paths have a greater impact on the grid cell and are assigned higher weights; pressure gradient sources farther away and with complex propagation paths have a smaller impact and are assigned lower weights. Alternatively, fusion can be performed based on the confidence level of the estimates, with estimates having higher confidence levels assigned higher weights.

[0077] Step S326: Arrange the pressure gradient magnitude estimates of all three-dimensional spatial grid cells corresponding to each time stamp in the order of timestamps as the first dimension of the data structure, and use the three-dimensional spatial grid coordinates as the second to fourth dimensions of the data structure to complete the spatiotemporal distribution data structure of pressure gradient.

[0078] The final pressure gradient magnitude estimates for all 3D spatial grid cells at all timestamps are arranged in time-stamp order, serving as the first dimension of the data structure. The 3D spatial grid coordinates, representing the second to fourth dimensions, indicate the positions of the grid cells along the three coordinate axes. For example, in a case with multiple timestamps and a large number of grid cells, the spatiotemporal distribution data structure of the pressure gradient can be viewed as a four-dimensional array. The length of the first dimension equals the number of timestamps, and the lengths of the second to fourth dimensions correspond to the number of grid cells along the three coordinate axes, respectively. This method constructs a complete spatiotemporal distribution data structure for the pressure gradient, accurately describing the temporal and spatial variations of the pressure gradient during gas flow.

[0079] Step S330: Combining the peak traffic occurrence time in the corresponding relationship with the pre-established traffic propagation model, assign a traffic amplitude estimate to each three-dimensional spatial grid cell under each timestamp, and construct a spatiotemporal distribution data structure for peak traffic.

[0080] The correspondence between peak flow occurrence times describes the occurrence of peak flow at different times, while the pre-established flow propagation model describes the propagation law of flow within the air pump. Combining these two pieces of information, a flow amplitude estimate can be assigned to each three-dimensional spatial grid cell at each timestamp. A flow amplitude estimate refers to the estimated value of the flow amplitude of a specific grid cell at a given timestamp. By performing such estimates for all timestamps and all grid cells, and arranging these estimates in the order of timestamp and spatial coordinates, a spatiotemporal distribution data structure for peak flow is constructed. This data structure can comprehensively describe the temporal and spatial distribution of peak flow during gas flow.

[0081] When calculating the flow amplitude estimate for each grid cell, the peak flow information at the current timestamp is first obtained from the correspondence. Then, based on the pre-established flow propagation model, factors such as the distance between the grid cell and the flow source, and the pipeline structure are considered to calculate the flow amplitude estimate for that grid cell. For example, the flow propagation model may take into account the impact of pipeline resistance and branching on flow. For grid cells that are close to the flow source and have low pipeline resistance, their flow amplitude estimate may be larger; while for grid cells that are far from the flow source and pass through multiple branches and narrow pipelines, their flow amplitude estimate will be relatively smaller.

[0082] Step S340: Perform spatiotemporal correlation analysis on the pressure gradient spatiotemporal distribution data structure and the flow peak spatiotemporal distribution data structure, calculate the cooperative change intensity index at each spatiotemporal location, and generate spatiotemporal distribution data of cooperative change intensity.

[0083] In one implementation, step S340 specifically includes the following steps S341 to S346: Step S341: Determine the processing window for spatiotemporal correlation analysis. This window contains the current timestamp and a preset number of timestamps before and after in the time dimension, and contains the current three-dimensional spatial grid cell and its adjacent grid cells in the spatial dimension.

[0084] In terms of time, the processing window includes the current timestamp and a preset number of timestamps before and after it. The selection of the preset number needs to comprehensively consider the data's variation characteristics and the required accuracy of the analysis. If the preset number is too small, it may not be able to fully account for changes in pressure and traffic at surrounding time points; if the preset number is too large, it will increase the amount of computation and computational complexity.

[0085] In the spatial dimension, the processing window contains the current 3D spatial grid cell and its coordinate-adjacent grid cells. The definition of adjacent grid cells can be determined according to the specific meshing method. For example, in a cubic meshing, adjacent grid cells can be grid cells that are adjacent to the current grid cell in the three coordinate axis directions. By setting the processing window, the pressure and flow changes around the current spatiotemporal location can be taken into account when calculating the cooperative change intensity index, thus more accurately reflecting the cooperative change relationship between pressure gradient and flow amplitude.

[0086] Step S342: Assign weight distribution to the processing window, where the weight value is determined based on the spatiotemporal distance between each spatiotemporal position within the window and the center position. The closer the spatiotemporal distance, the greater the weight value. All weight values ​​are normalized.

[0087] To highlight the influence of different spatiotemporal locations within the processing window on the calculation of the collaborative change intensity index, a weight distribution needs to be assigned to the processing window. The weight values ​​are determined based on the spatiotemporal distance between each spatiotemporal location within the window and the center location. The closer the spatiotemporal distance, the stronger the correlation between that location and the current spatiotemporal location, and the larger the weight value.

[0088] Spatiotemporal distance can be calculated by comprehensively considering both temporal and spatial distances. For example, the Euclidean distance formula can be used to weight and combine temporal and spatial distances to obtain a comprehensive spatiotemporal distance. For each spatiotemporal location within the processing window, its spatiotemporal distance to the center location is calculated, and corresponding weight values ​​are assigned based on the distance magnitude.

[0089] All weight values ​​are normalized so that their sum is 1. This ensures that the weight allocation is reasonable when calculating the intensity of synergistic change, and that the calculation results are not biased due to the sum of the weight values ​​not being 1. Normalization can be achieved by dividing each weight value by the sum of all weight values.

[0090] Step S343: Traverse each spatiotemporal location of the pressure gradient spatiotemporal distribution data structure and the flow peak spatiotemporal distribution data structure, and extract the pressure gradient magnitude estimation data subset and the flow amplitude estimation data subset within the processing window corresponding to that location.

[0091] Each spatiotemporal location in the spatiotemporal distribution data structure of pressure gradient and the spatiotemporal distribution data structure of flow peak is traversed. For each spatiotemporal location, according to the definition of the processing window, the subset of pressure gradient magnitude estimation data and the subset of flow amplitude estimation data within the processing window corresponding to that location are extracted.

[0092] A data subset refers to the collection of estimated pressure gradient magnitude and flow amplitude values ​​for all spatiotemporal locations within the processing window. These data subsets will be used for subsequent calculations of the coordinated change intensity index. During the traversal, for each spatiotemporal location, the range of the processing window is determined based on its timestamp and spatial coordinates. Then, all data within the processing window are extracted from the spatiotemporal distribution data structures of pressure gradient and peak flow, forming the pressure gradient magnitude estimation data subset and the flow amplitude estimation data subset, respectively.

[0093] Step S344: Based on the subset of pressure gradient magnitude estimation data and the subset of flow amplitude estimation data, calculate a scalar value that characterizes the spatiotemporal correlation strength between the two, as an index of the intensity of coordinated change at the current spatiotemporal location.

[0094] To measure the degree of coordinated change between pressure gradient and flow amplitude at the current spatiotemporal location, a scalar value representing the strength of their spatiotemporal correlation needs to be calculated based on a subset of pressure gradient magnitude estimation data and a subset of flow amplitude estimation data, serving as an indicator of the intensity of coordinated change at the current spatiotemporal location.

[0095] This scalar value can be calculated using correlation analysis. Correlation analysis is a statistical method used to measure the degree of association between two variables. For the subsets of pressure gradient magnitude estimation data and flow amplitude estimation data, their correlation coefficients can be calculated. The correlation coefficient ranges from -1 to 1; the closer the absolute value is to 1, the stronger the association between the two variables; the closer the absolute value is to 0, the weaker the association.

[0096] A common method for calculating correlation coefficients is the Pearson correlation coefficient method. First, the subsets of pressure gradient magnitude estimation data and flow amplitude estimation data are standardized to ensure they have the same scale and range. Then, the covariance and standard deviation of the two data subsets are calculated, and the correlation coefficient is calculated using the Pearson correlation coefficient formula. The resulting correlation coefficient is an indicator of the intensity of coordinated change at the current spatiotemporal location.

[0097] For example, when using the Pearson correlation coefficient method to calculate pressure gradient magnitude estimation data subsets P=[P1,P2,...,Pn] and flow rate magnitude estimation data subsets Q=[Q1,Q2,...,Qn], P and Q are first standardized to eliminate data scale differences. The standardized data subsets more accurately reflect the relative changes between the data.

[0098] After standardization, their covariance is calculated, which reflects the coordinated changing trend of the two data subsets. If the covariance is positive, it indicates that the pressure gradient magnitude and flow amplitude change in roughly the same direction; if it is negative, they change in roughly opposite directions. Simultaneously, their respective standard deviations are calculated, reflecting the degree of dispersion of the data.

[0099] According to the Pearson correlation coefficient formula, dividing the covariance by the product of the two standard deviations yields an index of the intensity of covariance at the current spatiotemporal location. This index intuitively reflects the degree of correlation between pressure gradient and flow amplitude at that spatiotemporal location.

[0100] Step S345: Organize the co-change intensity indicators of all spatiotemporal locations according to the dimensional order of the pressure gradient spatiotemporal distribution data structure to generate co-change intensity spatiotemporal distribution data.

[0101] To comprehensively and systematically demonstrate the coordinated changes of pressure gradient and flow amplitude across the entire spatiotemporal range, it is necessary to rationally organize the intensity indicators of coordinated changes at all spatiotemporal locations. The spatiotemporal distribution data structure of pressure gradient has a clear dimensional order, including both temporal and spatial dimensions.

[0102] Following this dimensional order, the coordinated change intensity indices calculated for each spatiotemporal location are arranged sequentially. This ensures that the resulting spatiotemporal distribution data of coordinated change intensity maintains dimensional consistency with the spatiotemporal distribution data of pressure gradients, facilitating subsequent analysis and comparison.

[0103] For example, the spatiotemporal distribution data structure of pressure gradients arranges the pressure gradient magnitudes of the three-dimensional spatial grid cells at each time point in chronological order. Similarly, the cooperative change intensity index is arranged in the same temporal order, listing the cooperative change intensity indices of each spatial grid cell at each time point. In the spatial dimension, the coordinate order of the grid cells is also followed to ensure the orderliness and correspondence of the data.

[0104] This organizational approach creates a complete dataset of spatiotemporal distribution of coordinated change intensity, which clearly shows the coordinated change patterns of pressure and flow at different times and locations.

[0105] Step S346: Post-process the generated spatiotemporal distribution data of cooperative variation intensity to reduce the impact of estimation noise and random fluctuations.

[0106] Post-processing can employ Gaussian filtering, using a Gaussian kernel function to perform convolution on the data. The Gaussian kernel function has smoothing properties and can assign different weights based on the distance of each data point from the center point. Points closer to the center point receive greater weights, while those farther away receive smaller weights.

[0107] Gaussian filtering can effectively smooth out some local noise and random fluctuations while preserving the main characteristics of the data. Another filtering method is median filtering, which is a non-linear filtering method. Median filtering replaces each data point with the median of its neighborhood. This method is very effective in removing impulse noise and can make the data more stable.

[0108] When performing filtering, the filtering parameters should be selected appropriately based on the characteristics of the data and the needs of the analysis. For example, for Gaussian filtering, the standard deviation of the Gaussian kernel function needs to be determined; for median filtering, the size of the neighborhood needs to be determined. By setting appropriate filtering parameters, the best post-processing effect can be achieved, enabling the spatiotemporal distribution data of the coordinated change intensity to more accurately reflect the coordinated changes in pressure and flow.

[0109] Step S350: Based on the spatiotemporal distribution data of the coordinated change intensity and the preset threshold, identify the active trajectory that characterizes the gas flow, optimize and match it according to the preset pipeline topology diagram inside the air pump, and generate the gas flow path.

[0110] The spatiotemporal distribution data of the intensity of coordinated change reflects the degree of coordinated change of pressure gradient and flow amplitude at different times and spatial locations. The preset threshold is a pre-defined standard used to distinguish between active and inactive regions of gas flow.

[0111] When the intensity of coordinated change exceeds a preset threshold, it indicates a significant coordinated change in pressure and flow rate at that spatiotemporal location, likely representing an active region through which gas flows. By traversing the spatiotemporal distribution data of coordinated change intensity, all spatiotemporal locations where the intensity of coordinated change exceeds the preset threshold are identified. Connecting these locations allows for the preliminary identification of active trajectories representing gas flow. However, the initially identified active trajectories may contain discontinuities or inconsistencies. In such cases, optimization and matching based on the preset pipeline topology diagram within the air pump are necessary. This diagram details the possible flow paths of gas within the pump, including the connection relationships and directions of each pipeline segment. By comparing the initially identified active trajectories with the preset pipeline topology diagram, corrections and adjustments are made to any parts that do not conform to the topology.

[0112] Step S360: On the gas flow path, identify the sections where the cooperative change intensity index is significantly lower than the overall level as candidate regions for abnormal flow fluctuations. In the pressure gradient spatiotemporal distribution data, identify the sections where the pressure gradient magnitude is significantly lower than the overall level as candidate regions for pressure attenuation. Calculate the spatial intersection of the two candidate regions to generate a spatial mapping relationship.

[0113] Further analysis is conducted on the distribution of the coordinated change intensity index and pressure gradient modulus along the established gas flow path. The coordinated change intensity index reflects the degree of coordinated change in pressure and flow rate. When the coordinated change intensity index in certain sections is significantly lower than the overall level, it indicates that the pressure and flow rate changes in these sections may be abnormal, and these sections are likely areas of abnormal flow rate fluctuations.

[0114] By statistically analyzing the coordinated change intensity index at various locations along the gas flow path, a standard is established to distinguish between normal and abnormal regions. When the coordinated change intensity index of a certain segment is lower than this standard, it is identified as a candidate region for abnormal flow fluctuations. Simultaneously, the pressure gradient magnitude at each spatial location is analyzed from the pressure gradient spatiotemporal distribution data. The pressure gradient magnitude reflects the rate of change of pressure in space and time. When the pressure gradient magnitude of certain segments is significantly lower than the overall level, it indicates that the pressure drop in these segments is more significant, and they are likely regions of pressure decay.

[0115] Similarly, through statistical analysis, a standard is determined to distinguish between normal and abnormal pressure gradients. When the pressure gradient magnitude of a certain section is lower than this standard, it is identified as a candidate region for pressure attenuation. The spatial intersection of the candidate regions for abnormal flow fluctuations and pressure attenuation is calculated. The spatial intersection indicates that both abnormal flow fluctuations and pressure attenuation phenomena exist within these regions.

[0116] Step S400: Compare the gas flow path with the gas flow path under standard sealing conditions, analyze the differences between the two paths segment by segment, and locate the key nodes of seal failure by combining spatial mapping relationship. The key nodes are the locations that cause abnormal flow and pressure decay.

[0117] As one implementation method, step S400 specifically includes the following steps S410~S460: Step S410: Obtain a gas flow path dataset under standard sealed conditions. The gas flow path dataset contains a sequence of standard pipeline segments arranged in the topological order of the air pump pipeline system. Each standard pipeline segment contains a standard spatial coordinate range, a standard inner wall roughness distribution, a standard cooperative variation intensity curve, and a standard pressure gradient threshold range.

[0118] Each standard pipeline segment contains several important pieces of information. The standard spatial coordinate range clearly defines the specific spatial location of this pipeline segment within the air pump, which helps to accurately determine whether the current gas flow path has passed through the correct spatial location in subsequent comparisons. The standard inner wall roughness distribution reflects the smoothness of the pipeline segment's inner wall. Inner wall roughness affects gas flow resistance, and thus affects gas flow rate and pressure distribution. The standard co-variation intensity curve describes the co-variation pattern of pressure and flow rate within this pipeline segment under standard sealing conditions. It is a curve that varies over time. By comparing the co-variation intensity curve of the current pipeline segment with the standard curve, it is possible to determine whether there are any abnormal co-variations. The standard pressure gradient threshold range specifies the normal range of pressure gradient variation within this pipeline segment. When the actual pressure gradient exceeds this range, it may indicate a sealing problem.

[0119] Step S420: Decompose the gas flow path of the current air pump into a current pipeline segment sequence corresponding to the standard pipeline segment sequence, establish a one-to-one correspondence through the pipeline segment identification number, and match the number and spatial location of the decomposed current pipeline segments with the standard pipeline segment sequence to generate pipeline segment comparison index data.

[0120] To accurately compare the current gas flow path with the standard gas flow path, the current gas flow path of the charging pump needs to be decomposed. The goal of this decomposition is to transform it into a current pipeline segment sequence corresponding to the standard pipeline segment sequence. By assigning a unique pipeline segment identifier to each pipeline segment, a one-to-one correspondence between the current pipeline segment and the standard pipeline segment can be established. During the decomposition process, it is crucial to ensure that the number and spatial location of the decomposed current pipeline segments match the standard pipeline segment sequence. For example, if the standard pipeline segment sequence contains a specific number of pipeline segments, and each segment has a defined spatial location, then the number and spatial location of the decomposed current gas flow path should also correspond accordingly. In this way, pipeline segment comparison index data is generated.

[0121] Step S430: For each corresponding pipeline segment pair in the pipeline segment comparison index data, calculate the spatial deviation between the trajectory point set of the pipeline segment in the current gas flow path and the preset spatial coordinate range of the standard pipeline segment, the deviation between the inner wall roughness distribution of the current pipeline segment and the standard inner wall roughness distribution, the similarity between the cooperative change intensity curve of the current pipeline segment and the standard cooperative change intensity curve, and the percentage of time during which the pressure gradient value of the current pipeline segment exceeds the standard pressure gradient threshold range.

[0122] In one implementation, step S430 specifically includes the following steps S431 to S434: Step S431: Obtain the three-dimensional spatial boundary defined by the preset spatial coordinate range of the standard pipeline section, obtain the three-dimensional spatial coordinate set of all trajectory points corresponding to the current pipeline section in the gas flow path, calculate the shortest distance from each point in the trajectory point set to the three-dimensional spatial boundary of the standard pipeline section, and calculate the statistical measure as the spatial deviation based on the shortest distance of all trajectory points. The statistical measure includes the average distance, the maximum distance, or the proportion of points whose distance exceeds the threshold.

[0123] A predefined spatial coordinate range for a standard pipeline segment establishes a clear three-dimensional spatial boundary, which determines the standard spatial position of the pipeline segment within the air pump. The set of three-dimensional spatial coordinates for all trajectory points corresponding to the current pipeline segment along the gas flow path is obtained; these trajectory points reflect the actual spatial position of the current pipeline segment. For each point in the trajectory point set, its shortest distance to the three-dimensional spatial boundary of the standard pipeline segment is calculated. This shortest distance reflects the degree of deviation of the point from the standard spatial position. Based on the shortest distances of all trajectory points, different statistical measures can be calculated as the spatial deviation.

[0124] The average distance is obtained by summing the shortest distances of all trajectory points and then dividing by the number of trajectory points. It reflects the average deviation of the entire trajectory point set from the standard spatial position. The maximum distance is the largest of the shortest distances of all trajectory points, reflecting the point in the trajectory point set that deviates the farthest from the standard spatial position. The proportion of points with distances exceeding a threshold is obtained by counting the number of trajectory point sets whose shortest distances exceed a pre-set threshold and then dividing by the total number of trajectory points. This proportion reflects how many trajectory points have seriously deviated from the standard spatial position. Through these statistics, the spatial deviation of the current pipeline segment from the standard pipeline segment can be comprehensively assessed.

[0125] Step S432: Obtain the inner wall roughness distribution data of the current pipeline section, obtain the three-dimensional point cloud data of the inner wall of the pipeline section through laser scanning, sample the roughness value at equal intervals along the axis of the pipeline section, generate the current roughness sequence, read the standard roughness sequence of the standard pipeline section, calculate the sum of the absolute differences between the corresponding positions of the two sequences, and divide by the sequence length to obtain the degree of deviation.

[0126] Roughness values ​​are sampled at equal intervals along the pipeline section's axis. These sampled roughness values ​​are then arranged sequentially to generate the current roughness sequence. This sequence reflects the distribution of roughness on the inner wall of the current pipeline section. Simultaneously, the standard roughness sequence of a standard pipeline section is read, representing the distribution of roughness on the inner wall of that section under standard sealing conditions. The sum of the absolute differences between corresponding positions in the two sequences is calculated. This involves subtracting the roughness values ​​at the same position in the current roughness sequence and the standard roughness sequence, taking the absolute values, and then summing them. This sum is then divided by the sequence length. The result indicates the degree of deviation between the current pipeline section's inner wall roughness distribution and the standard inner wall roughness distribution. A larger deviation indicates a greater difference between the current pipeline section's inner wall roughness and the standard condition, which may have a significant impact on gas flow.

[0127] Step S433: Extract the coordinated change intensity curve of the current pipeline segment and the standard coordinated change intensity curve of the standard pipeline segment. Adjust the time axis of the two curves through the dynamic time warping algorithm, calculate the sum of the differences between corresponding points of the aligned curves, and use the reciprocal of the sum of the differences as the similarity index.

[0128] The coordinated change intensity curve reflects the coordinated changes in pressure and flow within a pipeline segment. After extracting the coordinated change intensity curve of the current pipeline segment and the standard coordinated change intensity curve of a standard pipeline segment, a dynamic time warping algorithm is needed to adjust them due to potential time axis offsets. The dynamic time warping algorithm can find the optimal alignment between the two curves, taking into account potential stretching and distortion of the time series. This algorithm adjusts the time axes of the two curves to ensure better temporal correspondence. After adjustment, the sum of the differences between corresponding points on the aligned curves is calculated. That is, for the coordinated change intensity values ​​at the same time point on the two curves, the absolute values ​​are subtracted and then summed. The reciprocal of this sum of differences is used as a similarity index. A higher similarity index indicates a greater similarity between the coordinated change intensity curve of the current pipeline segment and the standard curve, and a closer approximation of the coordinated changes in pressure and flow within that pipeline segment to the standard state; conversely, a lower similarity index indicates a significant difference and potential sealing issues.

[0129] Step S434: Extract the pressure gradient magnitude of the current pipeline segment at all timestamps from the pressure gradient spatiotemporal distribution data structure to form a pressure gradient time series; read the standard pressure gradient threshold range of the standard pipeline segment, count the number of timestamps in the pressure gradient time series whose values ​​are greater than the upper limit of the threshold or less than the lower limit of the threshold, and divide by the total number of timestamps to obtain the time proportion.

[0130] The pressure gradient spatiotemporal distribution data structure records the pressure gradient magnitude values ​​at different times and spatial locations within the entire air pump. The pressure gradient magnitude values ​​for the current pipeline segment at all timestamps are extracted from this data structure and arranged chronologically to form a pressure gradient time series. The standard pressure gradient threshold range for a standard pipeline segment defines the normal range of pressure gradient variation under standard sealing conditions, including an upper and lower threshold. The number of timestamps in the pressure gradient time series whose values ​​are greater than the upper threshold or less than the lower threshold is counted. These timestamps correspond to pressure gradient values ​​outside the normal range, potentially indicating a sealing problem. Dividing the number of timestamps exceeding the threshold range by the total number of timestamps yields the percentage of time the current pipeline segment's pressure gradient value exceeds the standard pressure gradient threshold range. A larger percentage indicates a more severe pressure gradient anomaly in the pipeline segment, and is more likely to indicate a sealing failure.

[0131] Step S440: Combine spatial deviation, deviation degree, similarity degree and time proportion into a pipeline segment difference feature combination, and filter out the key difference features that have a significant impact on sealing failure through feature selection processing to generate a key difference feature vector.

[0132] As one implementation method, step S440 specifically includes the following steps S441~S445: Step S441: The spatial deviation, deviation degree, similarity degree and time proportion are standardized and transformed respectively. The spatial deviation is standardized as the difference between the actual spatial deviation and the minimum spatial deviation divided by the difference between the maximum spatial deviation and the minimum spatial deviation. The deviation degree is standardized as the difference between the actual deviation degree and the minimum deviation degree divided by the difference between the maximum deviation degree and the minimum deviation degree. The similarity degree is standardized as the difference between the actual similarity degree and the minimum similarity degree divided by the difference between the maximum similarity degree and the minimum similarity degree. The time proportion is standardized as the difference between the actual time proportion and the minimum time proportion divided by the difference between the maximum time proportion and the minimum time proportion.

[0133] Since features such as spatial deviation, degree of deviation, similarity, and time proportion may have different value ranges, they need to be standardized to ensure fair treatment of each feature in subsequent analyses. The purpose of standardization is to map the value of each feature to a uniform range, typically [0,1]. For spatial deviation, the actual spatial deviation is subtracted from the minimum spatial deviation recorded in historical data, and the difference is then divided by the difference between the maximum and minimum spatial deviations. This standardizes the spatial deviation to the [0,1] range. The same method is used to standardize degree of deviation, similarity, and time proportion. The actual degree of deviation, similarity, and time proportion are subtracted from their respective minimum values, and then divided by the difference between the maximum and minimum values. Through this standardization, each feature is represented on the same scale, avoiding the impact of different feature value ranges on subsequent analyses.

[0134] Step S442: Combine the standardized spatial deviation, deviation degree, similarity degree, and time proportion in sequence to form a pipeline segment difference feature combination.

[0135] After standardizing the spatial deviation, degree of deviation, degree of similarity, and time proportion, these characteristics are combined in a specific order to form a combination of pipeline segment difference features. This combination is an ordered set of features, each of which has undergone standardization and possesses the same scale and comparability.

[0136] Combining these features in sequence ensures that the arrangement of features remains consistent throughout subsequent analysis and processing, facilitating uniform operations and comparisons. For example, when building a machine learning model, the input feature vectors need to have a fixed order so that the model can correctly identify and process each feature.

[0137] Step S443: Collect normal pipeline section samples and failed pipeline section samples from historical inspection data. Each sample contains a corresponding combination of pipeline section difference features and a failure label.

[0138] To train the subsequent seal failure probability assessment model, it is necessary to collect samples of normal pipeline sections and failed pipeline sections from historical inspection data. Normal pipeline section samples refer to pipeline section samples that have proven to have good sealing performance in historical inspections, while failed pipeline section samples refer to pipeline section samples that have sealing problems.

[0139] Each sample contains a corresponding combination of pipeline segment difference features and a failure label. The combination of pipeline segment difference features, generated in the previous steps, includes key features such as spatial deviation, degree of deviation, similarity, and time proportion. The failure label is a binary label used to indicate whether the pipeline segment has failed to seal, typically using "0" to represent normal and "1" to represent failure.

[0140] By collecting a large number of normal and failed pipeline section samples, a rich dataset can be constructed. This dataset will be used to train and validate the seal failure probability assessment model, thereby improving the accuracy and reliability of the model.

[0141] Step S444: Evaluate the feature importance of the pipeline segment difference feature combination, calculate the contribution weight of each feature to the failure label prediction, and determine the contribution weight by the change in prediction accuracy of the feature subset in the classification process.

[0142] To further identify the most important features for predicting seal failure, a feature importance assessment is needed for combinations of pipeline segment differences. The purpose of the feature importance assessment is to determine the contribution of each feature in predicting seal failure labels.

[0143] To calculate the contribution weight of each feature to the prediction of the failure label, we can use the method of the change in prediction accuracy of a feature subset during classification. Specifically, we first perform classification using all features to obtain an initial prediction accuracy. Then, we remove one feature at a time and perform classification again to obtain the prediction accuracy after removing that feature.

[0144] Calculate the change in prediction accuracy after removing each feature. The larger the change, the greater the impact of that feature on prediction accuracy, and the higher its contribution weight.

[0145] Step S445: Sort by contribution weight from high to low, select the top number of features by weight as key difference features, extract the standardized parameters of the corresponding positions in the pipeline segment difference feature combination, and sort them by weight to form a key difference feature vector.

[0146] Based on the contribution weight of each feature calculated earlier, the features are sorted from highest to lowest contribution weight. Then, a predetermined number of features with the highest weights are selected as key difference features. The predetermined number can be determined based on specific analysis needs and model complexity. Standardized parameters for corresponding locations are extracted from the combination of pipeline segment difference features; these parameters are standardized feature values. These standardized parameters are arranged according to their contribution weights to form a key difference feature vector.

[0147] Step S450: Construct a pipeline segment failure probability assessment data model based on key difference feature vectors, train the data model using historical failed pipeline segment samples, output the failure probability value of each current pipeline segment, and mark pipeline segments with failure probability values ​​higher than the probability threshold as suspected sealing failure pipeline segments.

[0148] A data model for assessing pipeline section failure probability can be constructed based on key difference feature vectors. Common machine learning models, such as logistic regression, support vector machines, or neural networks, can be employed. This data model is then trained using historical failed pipeline section samples. These samples contain a large amount of data on pipeline sections with known sealing failure scenarios. Through this data, the model can learn the patterns and characteristics of key difference features during sealing failure.

[0149] After training, the key difference feature vectors for each current pipeline segment are input into the trained model, which then outputs the failure probability value for that segment. A probability threshold is a pre-defined standard; when the failure probability value of a pipeline segment exceeds this threshold, the segment is marked as a suspected seal failure. This marking helps to quickly locate pipeline segments that may have sealing problems, providing guidance for further inspection and repair.

[0150] Step S460: Query the set of spatial coordinates of the abnormal flow fluctuation area and the pressure attenuation change area in the spatial mapping relationship, calculate the spatial overlap volume between the spatial coordinate range of the suspected sealing failure pipeline section and the spatial coordinate set of the abnormal area, and locate the two end connection nodes of the suspected sealing failure pipeline section with the largest overlap volume as the key nodes of sealing failure.

[0151] The spatial mapping relationship records the set of spatial coordinates for areas of abnormal flow fluctuations and pressure attenuation changes; these areas are potential sealing problems. For each pipeline segment suspected of sealing failure, the spatial overlap volume between its spatial coordinate range and the set of spatial coordinates of the abnormal areas is calculated.

[0152] The spatial overlap volume reflects the degree of spatial overlap between the suspected sealing failure pipeline section and the abnormal area. The larger the overlap volume, the stronger the correlation between the pipeline section and the abnormal area, and the more likely it is to be the location of the sealing failure.

[0153] By comparing the spatial overlap volumes of all suspected leaky pipe sections, the section with the largest overlap volume is identified. The connection nodes at both ends of this section are then identified as the critical points for seal failure. This is because the connection nodes at both ends of a pipe section are often potential locations for gas leaks; when this section has significant spatial overlap with an abnormal area, its connection nodes are very likely to be the critical locations leading to seal failure.

[0154] Step S500: Based on the degree of abnormal changes in key nodes, combined with the distribution characteristics and mutual influence of multiple key nodes, determine the quantitative level of the airtightness defect of the air pump.

[0155] As one implementation method, step S500 specifically includes the following steps S510~S560: Step S510: Extract the abnormal change parameter set of each node from the key node, including the pressure gradient decay rate at the node, the rate of change of flow fluctuation amplitude, the proportion of abnormal duration and the deviation of coordinated change intensity, and arrange the four parameters in a fixed order as the key node abnormal feature vector.

[0156] As one implementation method, step S510 specifically includes the following steps S511~S516: Step S511: Locate the corresponding spatial grid cell of each key node in the pressure gradient spatiotemporal distribution data structure, extract the pressure gradient magnitude of the grid cell under all timestamps, and form a node pressure gradient sequence.

[0157] The pressure gradient spatiotemporal distribution data structure records the pressure gradient magnitudes at different timestamps for various spatial locations inside the air pump. To obtain the pressure gradient information of key nodes, it is necessary to locate the corresponding spatial grid cell for each key node within this data structure.

[0158] Each critical node has a defined spatial location inside the air pump, and its corresponding spatial grid cell can be found based on this location. The pressure gradient magnitude of this grid cell at all timestamps is extracted, and these magnitudes are arranged in chronological order to form a node pressure gradient sequence.

[0159] The node pressure gradient sequence can reflect the change of pressure gradient at the key node over time, providing a data basis for subsequent calculation of pressure gradient decay rate.

[0160] Step S512: Perform linear trend fitting on the nodal pressure gradient sequence, and calculate the pressure gradient decay rate at the node by the slope of the fitted line. When the slope is negative, the absolute value of the slope is the pressure gradient decay rate. When the slope is non-negative, the pressure gradient decay rate is zero.

[0161] Linear trend fitting of the nodal pressure gradient sequence aims to identify its trend over time. Linear trend fitting can be achieved using the least squares method, by finding a straight line that minimizes the sum of squared errors between the line and the data points in the nodal pressure gradient sequence.

[0162] The slope of the fitted line reflects the rate of change of the pressure gradient over time. When the slope is negative, it indicates that the pressure gradient decreases over time, and the absolute value of the slope is the pressure gradient decay rate at the node. If the slope is non-negative, it indicates that the pressure gradient has no decreasing trend, and the pressure gradient decay rate is zero.

[0163] Step S513: Locate the spatial grid cell corresponding to the key node in the spatiotemporal distribution data structure of the peak flow, extract the flow amplitude of the cell under all timestamps, identify the maximum and minimum points in the flow amplitude sequence, calculate the average value of the difference between adjacent maximum and minimum values, and then divide it by the average value of the corresponding time interval to obtain the rate of change of flow fluctuation amplitude.

[0164] The spatiotemporal distribution data structure for peak flow records the flow amplitude at different timestamps at various spatial locations inside the air pump. Key nodes are located within the corresponding spatial grid cells of this data structure, and the flow amplitude of each cell at all timestamps is extracted to form a flow amplitude sequence. Maximum and minimum points in the flow amplitude sequence are identified, reflecting the peaks and troughs of flow fluctuations. The difference between adjacent maximum and minimum values ​​is calculated to obtain the change in flow amplitude within each fluctuation cycle. The average of these differences is then calculated, reflecting the average change in flow fluctuation amplitude. Simultaneously, the average time interval between adjacent maximum and minimum values ​​is calculated, reflecting the average time interval of flow fluctuations. Dividing the average change in flow fluctuation amplitude by the average time interval yields the rate of change of flow fluctuation amplitude. A larger rate of change indicates more severe flow fluctuations, potentially indicating a sealing problem.

[0165] Step S514: Extract the sequence of coordinated change intensity indicators for key nodes corresponding to spatial grid cells from the spatiotemporal data structure of coordinated change intensity. Count the number of timestamps in the sequence whose indicator values ​​are lower than the corresponding time values ​​of the standard coordinated change intensity curve. Divide this number by the total number of timestamps to obtain the percentage of abnormal duration.

[0166] The spatiotemporal data structure for coordinated change intensity records the coordinated change intensity indices at different timestamps for various spatial locations within the air pump. The coordinated change intensity index sequence for the spatial grid cells corresponding to key nodes is extracted; this sequence reflects the changes in pressure and flow rate coordination at the key node over time. The standard coordinated change intensity curve represents the coordinated change pattern of pressure and flow rate at this node under standard sealing conditions. The number of timestamps in the coordinated change intensity index sequence whose index values ​​are lower than the corresponding values ​​of the standard coordinated change intensity curve is counted; these timestamps indicate anomalies in the coordinated change at that node. Dividing the number of anomalous timestamps by the total number of timestamps yields the percentage of anomalous duration. A larger percentage of anomalous duration indicates a longer duration of anomalous coordinated change at that node, and a greater likelihood of a sealing problem.

[0167] Step S515: Calculate the sum of the differences between the corresponding points of the coordinated change intensity index sequence at the key nodes and the standard coordinated change intensity curve, and divide the sum by the average value of the standard curve to obtain the deviation of the coordinated change intensity.

[0168] To measure the deviation of the coordinated change at critical nodes from the standard condition, the sum of the differences between corresponding points in the coordinated change intensity index sequence at the critical node and the standard coordinated change intensity curve is calculated. For the values ​​at the same time point in the coordinated change intensity index sequence and the standard curve, the absolute values ​​are subtracted and then summed to obtain the sum of the differences between corresponding points. Dividing this sum of differences by the average value of the standard curve yields the deviation of the coordinated change intensity. A larger deviation indicates a greater difference between the coordinated change pattern at that node and the standard condition, and a greater likelihood of a sealing problem.

[0169] Step S516: Arrange the calculated pressure gradient decay rate, flow fluctuation amplitude change rate, abnormal duration ratio, and cooperative change intensity deviation in order to obtain the key node abnormal feature vector.

[0170] The pressure gradient decay rate, flow fluctuation amplitude change rate, abnormal duration ratio, and deviation of coordinated change intensity calculated earlier are arranged in a fixed order to form a key node abnormal feature vector.

[0171] Step S520: Group the abnormal feature vectors of all key nodes by feature similarity. By setting spatial distance threshold and feature similarity threshold, key nodes with similar feature vectors and adjacent spatial locations are clustered into the same node cluster. Each node cluster represents a group of interconnected key nodes.

[0172] The abnormal feature vectors of all key nodes are grouped by feature similarity. The purpose is to group key nodes with similar abnormal features and spatial proximity together. Spatial distance thresholds and feature similarity thresholds are set to determine the criteria for grouping.

[0173] The spatial distance threshold is used to determine whether two key nodes are spatially adjacent. If the spatial distance between two nodes is less than the spatial distance threshold, they are considered spatially adjacent. The feature similarity threshold is used to determine whether the abnormal feature vectors of two key nodes are similar. Some similarity calculation methods can be used, such as Euclidean distance or cosine similarity.

[0174] If the similarity of the anomalous feature vectors of two key nodes is greater than the feature similarity threshold, and they are spatially adjacent, they are clustered into the same node cluster. Each node cluster represents a group of interconnected key nodes that may be affected by the same sealing problem, or whose sealing problems interact with each other.

[0175] Step S530: Calculate the average value of the intra-cluster anomaly feature vector of each node cluster as the overall anomaly level of the node cluster. At the same time, calculate the spatial coordinates of each key node in the cluster and determine the spatial clustering characteristics of the node cluster by the density of the coordinate distribution.

[0176] For each node cluster, the average value of the anomalous feature vectors of all key nodes within the cluster is calculated. The anomalous feature vectors contain information such as the pressure gradient decay rate, the rate of change of flow fluctuation amplitude, the proportion of anomalous duration, and the deviation of the intensity of coordinated changes. Calculating the average value yields the overall anomalous level of the node cluster.

[0177] The overall anomaly level reflects the combined anomaly status of all key nodes within the node cluster, providing a quantitative indicator for subsequent assessment of the cluster's impact on the airtightness of the air pump. Simultaneously, calculating the spatial coordinates of each key node within the cluster and analyzing the density of their distribution allows for the determination of the cluster's spatial aggregation characteristics.

[0178] If the spatial coordinates of the critical nodes within a node cluster are densely distributed, it indicates that the node cluster represents a relatively concentrated area of ​​sealing failure; if the distribution is relatively dispersed, it may indicate the existence of multiple relatively independent small sealing problems.

[0179] Step S540: Construct a key node association network. The network nodes are key nodes, and the edge weight is the product of the similarity between the abnormal feature vectors of two nodes and the inverse of the spatial distance. The higher the similarity and the closer the distance between the nodes, the greater the edge weight.

[0180] A key node relationship network is constructed to more intuitively represent the relationships between key nodes. Each node in the network is a key node, and each key node is represented by a single node in the network.

[0181] Edge weights represent the strength of the association between two key nodes. The edge weight is calculated by multiplying the similarity of the anomaly feature vectors of the two nodes by the inverse of their spatial distance. The similarity of the anomaly feature vectors of the two nodes can be determined using the similarity calculation method mentioned earlier; the higher the similarity, the more similar the anomalies of the two nodes, and they may be affected by the same sealing problem. The inverse of the spatial distance represents the spatial proximity of the two nodes; the closer the distance, the larger the inverse of the spatial distance. Multiplying the similarity by the inverse of the spatial distance yields the edge weight. Nodes with higher similarity and closer distances have larger edge weights, indicating a stronger association between them.

[0182] Step S550: Perform dynamic evolution simulation on the key node association network. Initially, the abnormal feature vector element values ​​of each key node are used as the initial energy values ​​of the nodes. Energy transfer is iterated according to the edge weights. In each iteration, the node energy value is updated to the weighted sum of its remaining energy and the received energy. The number of iterations is determined according to the working cycle of the air pump.

[0183] As one implementation method, step S550 specifically includes the following steps S551~S556: Step S551: Initialize the energy value of each node in the key node association network, and use the weighted sum of the four element values ​​of the node abnormal feature vector as the initial energy value. The weight is determined according to the degree of influence of each abnormal parameter on the sealing failure.

[0184] To begin the dynamic evolution simulation of the critical node network, the energy values ​​of each node need to be initialized. The initial energy values ​​are obtained by weighted summing of the four elements of the node's anomaly feature vector: pressure gradient decay rate, flow fluctuation amplitude change rate, anomaly duration percentage, and cooperative change intensity deviation.

[0185] The weights are determined based on the degree of influence of each abnormal parameter on the seal failure. For example, if the pressure gradient decay rate has a significant impact on the seal failure, it can be assigned a higher weight; if the rate of change of flow fluctuation amplitude has a relatively small impact, it can be assigned a lower weight. By setting the weights appropriately, the overall degree of abnormality at each node can be more accurately reflected, providing reasonable initial conditions for subsequent energy transfer iterations.

[0186] Step S552: Obtain the energy transfer attenuation coefficient, which is positively correlated with the length of the pipeline segment of the gas flow path between nodes. The longer the pipeline segment, the greater the attenuation during the energy transfer process.

[0187] The energy transfer attenuation coefficient describes the attenuation of energy as it is transferred from one node to another in a critical node interconnected network. This coefficient is positively correlated with the length of the gas flow path between nodes. When the gas flow path between nodes is long, the gas encounters more resistance and losses during flow, leading to greater attenuation during energy transfer. Therefore, the energy transfer attenuation coefficient increases with the increase of the path length. The energy transfer attenuation coefficient can be obtained through experimental measurement or theoretical calculation. Experimental measurement involves measuring the energy transfer attenuation at different path lengths in an actual air pump, thus obtaining the relationship between the energy transfer attenuation coefficient and the path length. Theoretical calculation can be based on the physical principles of gas flow, establishing a mathematical model to calculate the energy transfer attenuation coefficient.

[0188] Step S553: ​​Perform energy transfer iterations according to the timestamp sequence of the air pump's working cycle. In each iteration, each node multiplies its current energy value by the edge weight and attenuation coefficient, then transfers energy to the connected downstream nodes and receives energy from the upstream nodes.

[0189] In the dynamic evolution simulation, energy transfer is iterated according to the timestamp sequence of the air pump's working cycle. Each iteration represents a time step, simulating the energy transfer between key nodes within that time step.

[0190] In each iteration, each node multiplies its current energy value by its edge weight and a decay coefficient. The edge weight represents the strength of the connection between the node and its downstream nodes, and the decay coefficient represents the energy loss during transmission. The product is then used as the energy passed to the downstream nodes.

[0191] Simultaneously, each node receives energy transmitted from its upstream nodes. This energy, transmitted from upstream nodes, is processed with edge weights and attenuation coefficients. In this way, dynamic energy transfer and mutual influence between key nodes are achieved.

[0192] Step S554: The node energy update method is that the new energy value is equal to the weighted sum of its own remaining energy and the received energy. The ratio of its own remaining energy and the ratio of the received energy are determined according to the structural characteristics of the node, and the sum of the two ratios is one.

[0193] A node's new energy value is equal to the weighted sum of its remaining energy and the energy it received. Its remaining energy refers to the energy the node did not pass on in this iteration, while the received energy refers to the energy passed on from upstream nodes.

[0194] The proportion of a node's own remaining energy and the proportion of energy it receives are determined based on the node's structural characteristics. These characteristics include the node's importance and the number of edges it connects to. For example, if a node is a critical component in an air pump, its own remaining energy proportion may be set higher to ensure its stability; if a node has many connected edges, its received energy proportion may be relatively large because it receives more influence from other nodes.

[0195] It is necessary to ensure that the sum of the remaining energy ratio and the received energy ratio is equal to one in order to guarantee the conservation of node energy and make the simulation results more reasonable.

[0196] Step S555: In each iteration, if the node energy value exceeds the preset energy threshold, the node is marked as entering a deep failure state. In subsequent iterations, the energy transfer weight of nodes in the deep failure state is increased by the corresponding proportion.

[0197] In each iteration of the dynamic evolution simulation, a preset energy threshold is set. The energy threshold is a pre-defined standard used to determine whether a node has entered a deep failure state.

[0198] When a node's energy value exceeds a preset energy threshold, it indicates that the node's anomaly has become very severe, and the node is marked as entering a deep failure state. In subsequent iterations, the energy transfer weight of a node in a deep failure state increases proportionally. Increasing the energy transfer weight means that the node's influence on other connected nodes will increase, allowing for a more accurate simulation of the propagation and expansion of sealing failure problems.

[0199] For example, if a node enters a deep failure state, its energy transfer weight increases by a certain proportion, causing it to transfer more energy to downstream nodes. This may make downstream nodes more prone to abnormal situations, further reflecting the chain reaction of sealing failure problems.

[0200] Step S556: Set the number of iterations to the number of timestamps included in the complete working cycle of the air pump. After all iterations are completed, record the final energy value of each node as the node's abnormal contribution.

[0201] The number of iterations is set to the number of timestamps included in the complete working cycle of the air pump. This allows us to simulate the dynamic propagation and evolution of abnormal situations between key nodes during a complete working process of the air pump.

[0202] After all iterations are completed, the final energy value of each node is recorded. The final energy value reflects the combined result of the node's abnormal impact on other nodes and its own abnormal condition throughout the entire working cycle of the air pump. The final energy value is used as the node's abnormal contribution.

[0203] Step S560: After the simulation, extract the final energy value of each node as the node's abnormal contribution. Add up the abnormal contributions of all nodes to obtain the total abnormal energy value of the system. Compare the total abnormal energy value with the preset level threshold to determine the quantitative level of the air tightness defect of the air pump.

[0204] After the simulation, the final energy value of each node is extracted as the node anomaly contribution. The node anomaly contribution reflects the degree to which that node contributes to the airtightness defect of the air pump throughout the entire dynamic evolution simulation. The anomaly contributions of all nodes are summed to obtain the total anomaly energy value of the system. The total anomaly energy value of the system integrates the anomalies of all key nodes and can comprehensively reflect the overall airtightness defect of the air pump.

[0205] The preset level thresholds are a series of pre-defined standards used to classify different levels of airtightness defects. The total abnormal energy value of the system is compared with the preset level thresholds. If the total abnormal energy value of the system is lower than a certain level threshold, the airtightness defect quantification level of the air pump is determined to be the level corresponding to that level.

[0206] Please refer to Figure 3 This diagram illustrates the structural block diagram of a computer system 20 provided in one embodiment of this application. This computer system can be used to implement the functions of the aforementioned airtightness detection method for an air pump based on TPMS feedback and flow modeling. Specifically, the computer system 20 includes a central processing unit 21, a system memory 24 including a random access memory 22 and a read-only memory 23, and a system bus 25 connecting the system memory 24 and the central processing unit 21. The computer system 20 also includes a basic input / output system 26 to facilitate information transfer between various devices within the computer, and a mass storage device 27 for storing the operating system 271.

[0207] The memory includes a computer program stored in the memory and configured to be executed by one or more processors to implement the above-described method for detecting the air tightness of an air pump based on TPMS feedback and flow modeling.

Claims

1. A method for detecting the air tightness of an air pump based on TPMS feedback and flow modeling, characterized in that, The method includes: acquiring real-time pressure fluctuation signals and instantaneous flow fluctuation waveforms of the air pump's gas delivery pipeline from a tire pressure monitoring system during the operation of the air pump; wherein the real-time pressure fluctuation signals are marked with continuous timestamps, and the instantaneous flow fluctuation waveforms are time-aligned with the timestamps; performing time-by-time correlation analysis between the pressure change curve corresponding to the real-time pressure fluctuation signals and the instantaneous flow fluctuation waveforms to extract the correspondence between the continuous change law of the pressure gradient and the time of occurrence of the flow peak within the complete working cycle of the air pump, wherein the correspondence characterizes the synchronous change characteristics of pressure and flow during gas flow; and based on the correspondence and the three-dimensional structure of the internal sealed cavity of the air pump... The model reconstructs the spatiotemporal distribution of pressure gradient and peak flow rate during gas flow, traces the gas flow trajectory in the pipeline, and identifies the spatial mapping relationship between abnormal flow fluctuation areas and pressure decay change areas. This spatial mapping relationship characterizes the locational correlation between abnormal flow and pressure decay. The model compares the gas flow path with the gas flow path under standard sealing conditions, analyzes the path differences segment by segment, and locates the key nodes of sealing failure based on the spatial mapping relationship. These key nodes are the locations that cause abnormal flow and pressure decay. Based on the degree of abnormal change of the key nodes, combined with the distribution characteristics and mutual influence of multiple key nodes, the quantitative level of the airtightness defect of the air pump is determined.

2. The method according to claim 1, characterized in that, The step of performing a time-by-time correlation analysis between the pressure change curve corresponding to the real-time pressure fluctuation signal and the instantaneous flow fluctuation waveform to extract the correspondence between the continuous change law of the pressure gradient and the time of occurrence of the flow peak within the complete working cycle of the air pump includes: performing time-series framing of the real-time pressure fluctuation signal, dividing the pressure change curve into continuous pressure data segments according to the time interval of the air pump's mechanical action cycle, with each pressure data segment containing all pressure sample values ​​and start and end timestamps within that time interval, the time interval being determined by the correlation between the air pump motor speed and the piston motion cycle; performing fluctuation feature extraction on each pressure data segment, distinguishing the upward trend segment, downward trend segment, and stable segment of the pressure sample values ​​within the segment through sliding window trend recognition, calculating the absolute value of the slope, the number of continuous sampling points, and the fluctuation frequency of each trend segment, and generating a pressure segment feature vector containing multi-dimensional trend parameters; and performing morphological feature analysis on the instantaneous flow fluctuation waveform, identifying the starting point of the flow rising edge and the upper edge of the waveform through multi-scale edge detection. The system records the timestamp coordinates and waveform curvature values ​​of each feature point, including the rising edge endpoint, peak plateau segment, falling edge start point, and falling edge endpoint, generating a flow characteristic event sequence arranged chronologically. The pressure data segment and the flow characteristic event sequence are then dynamically window-matched based on timestamps. A matching window containing forward and backward time offsets is set to ensure each flow characteristic event falls within the time range of a unique pressure data segment, generating a time-related data structure. Based on this time-related data structure, the correlation weight between the pressure segment feature vector and the flow characteristic event is calculated. By statistically analyzing the recurrence frequency of pressure segment feature vectors corresponding to the same flow characteristic event type, strongly correlated combinations with frequencies higher than the correlation threshold are selected. The absolute value of the slope of the rising trend segment of the pressure data segment in the strongly correlated combination is extracted as a pressure gradient change index. This index is then correlated with the starting timestamp of the peak plateau segment in the corresponding flow characteristic event to obtain the correspondence between the continuous change pattern of the pressure gradient covering the complete working cycle of the air pump and the occurrence time of the flow peak.

3. The method according to claim 2, characterized in that, The process involves extracting fluctuation features for each pressure data segment, identifying upward trend segments, downward trend segments, and stable segments within the segment using a sliding window trend identification method, calculating the absolute value of the slope, the number of continuous sampling points, and the fluctuation frequency for each trend segment, and generating a pressure segment feature vector containing multi-dimensional trend parameters. This includes: initializing sliding window parameters, with the window length determined according to the proportion of the total number of sampling points in the pressure data segment and the window step size being a single sampling point; traversing the pressure sampling value sequence within the pressure data segment; performing linear fitting on the pressure sampling values ​​within each window, determining the trend type by the sign of the slope of the fitted line: a positive slope indicates an upward trend window, a negative slope indicates a downward trend window, and an absolute slope value less than a trend threshold indicates a stable trend window; merging consecutive windows with the same trend type into trend segments, merging upward trend windows into upward trend segments, and merging downward trend windows into... The downward trend segment and the stable trend window are merged into a stable trend segment, and the starting and ending sampling point indices of each trend segment are recorded. The absolute value of the representative slope of each trend segment is calculated by taking the arithmetic mean of the absolute values ​​of the slopes of all fitted lines within the merged window. The number of sampling points contained in the trend segment is calculated based on the starting and ending sampling point indices as the number of continuous sampling points. The fluctuation frequency is calculated by dividing the number of extreme points of the pressure sampling values ​​within the trend segment by the number of continuous sampling points. The absolute value of the slope of the upward trend segment, the number of continuous sampling points, the fluctuation frequency, the corresponding parameters of the downward trend segment, and the corresponding parameters of the stable segment are arranged in a fixed order to obtain the pressure segment feature vector containing multiple dimensions of parameters, where each dimension parameter is determined by the cross-index of the trend segment type and the parameter type. The parameters of each dimension in the pressure segment feature vector are normalized, and the parameter values ​​are mapped to a preset numerical range.

4. The method according to claim 3, characterized in that, The process of analyzing the morphological features of the instantaneous flow fluctuation waveform, identifying the start and end points of the rising edge, peak plateau segment, and the start and end points of the falling edge in the waveform through multi-scale edge detection, recording the timestamp coordinates of each feature point and the corresponding waveform curvature value, and generating a flow feature event sequence arranged in chronological order includes: performing Gaussian filtering on the instantaneous flow fluctuation waveform, setting multiple filtering parameters with different smoothness levels for multi-scale filtering, generating a set of flow waveforms containing different smoothness levels, the filtering parameters being determined through the correlation between the waveform sampling frequency and the gas compression frequency of the air pump; calculating the second derivative of the smoothed flow waveform at each scale to obtain the curvature change curve of the waveform, identifying potential edge points of the waveform through zero-crossing point detection of the curvature change curve, fusing edge points at different scales through non-maximum suppression to eliminate false edge points; and selecting the start point of the rising edge of the flow waveform from the fused edge points, marking the corresponding sampling point as the rising edge when the curvature change curve changes from negative to positive and the waveform slope begins to exceed the rising threshold. Record the timestamp coordinates of the starting point; identify the rising edge endpoint: when the waveform slope after the rising edge starting point reaches its maximum upward angle and then begins to decrease, the corresponding sampling point is marked as the rising edge endpoint, and its timestamp coordinates and waveform curvature value are recorded. The waveform segment between the rising edge starting point and the rising edge endpoint is the rising edge; identify the peak plateau segment: after the rising edge endpoint, if the absolute value of the waveform slope is continuously less than the plateau threshold within a preset duration, the waveform segment is confirmed as the peak plateau segment, and the start timestamp, end timestamp, and average waveform curvature value of the plateau segment are recorded; identify the falling edge starting point and falling edge endpoint: after the peak plateau segment ends, when the waveform slope changes from zero to a negative value and the absolute value is greater than the falling threshold, it is marked as the falling edge starting point; when the absolute value of the waveform slope after the falling edge is less than the falling threshold, it is marked as the falling edge endpoint, and the timestamp coordinates and waveform curvature values ​​of both are recorded; define the feature combination containing the rising edge starting point, rising edge endpoint, peak plateau segment, falling edge starting point, and falling edge endpoint as a flow characteristic event, and generate a flow characteristic event sequence by arranging them in timestamp order.

5. The method according to claim 1, characterized in that, The method, based on the correspondence and the three-dimensional structural model of the internal sealed cavity of the air pump, reconstructs the spatiotemporal distribution of pressure gradient and flow peak during gas flow, traces the flow trajectory of gas in the pipeline, and identifies the spatial mapping relationship between abnormal flow fluctuation regions and pressure decay change regions. This includes: obtaining a three-dimensional solid model of the internal sealed cavity of the air pump; uniformly dividing the model space into cubic mesh units through three-dimensional meshing; combining the continuous change law of pressure gradient in the correspondence with a pre-established spatial decay model of pressure gradient, assigning a pressure gradient magnitude estimate to each three-dimensional spatial mesh unit at each time point, and constructing a pressure gradient spatiotemporal distribution data structure; and combining the occurrence time of flow peak in the correspondence with a pre-established flow propagation model, assigning a pressure gradient magnitude estimate to each three-dimensional spatial mesh unit at each time point. Flow amplitude estimation is performed, and a spatiotemporal distribution data structure for peak flow is constructed. Spatiotemporal correlation analysis is conducted between the pressure gradient spatiotemporal distribution data structure and the peak flow spatiotemporal distribution data structure to calculate the co-variation intensity index at each spatiotemporal location, generating co-variation intensity spatiotemporal distribution data. Based on the co-variation intensity spatiotemporal distribution data and a preset threshold, active trajectories representing gas flow are identified, and optimized and matched according to a preset pipeline topology diagram inside the air pump to generate gas flow paths. On these gas flow paths, segments with co-variation intensity indices significantly lower than the overall level are identified as candidate regions for abnormal flow fluctuations. Similarly, segments with pressure gradient magnitudes significantly lower than the overall level in the pressure gradient spatiotemporal distribution data are identified as candidate regions for pressure attenuation. The spatial intersection of the two candidate regions is calculated to generate the spatial mapping relationship.

6. The method according to claim 5, characterized in that, The method combines the continuous change law of pressure gradient in the correspondence with the pre-established spatial attenuation model of pressure gradient, assigns a pressure gradient magnitude estimate to each three-dimensional spatial grid cell under each time stamp, and constructs a spatiotemporal distribution data structure of pressure gradient. This includes: obtaining a three-dimensional solid model of the sealed cavity inside the air pump; uniformly dividing the model space into cubic grid cells through three-dimensional meshing; determining the side length of each grid cell through the correlation between the minimum feature size of the model and the preset number of grid cells; assigning a unique three-dimensional coordinate index to each grid cell; extracting all time stamps and their corresponding pressure gradient change values ​​from the correspondence to generate a pressure gradient time series; supplementing the pressure gradient values ​​of missing time stamps through time interpolation to ensure that the time stamps are uniformly distributed at fixed time intervals; and obtaining the pre-established spatial attenuation model of pressure gradient. An inter-time decay model is used to define the attenuation characteristics of pressure gradients propagating along different pipe sections, joints, and cavity structures. For each time stamp, the pressure gradient estimate of each three-dimensional spatial grid cell within its influence range is calculated based on the current pressure gradient value and the decay model. For each grid cell, all time stamps and all relevant pressure gradient sources at that location are integrated, and the estimated pressure gradients are fused using the rules provided by the decay model or based on the confidence level of the estimated values ​​to obtain the final pressure gradient magnitude estimate of that grid cell at each time stamp. The pressure gradient magnitude estimates of all three-dimensional spatial grid cells corresponding to each time stamp are arranged in time stamp order as the first dimension of the data structure, and the three-dimensional spatial grid coordinates are used as the second to fourth dimensions of the data structure, thus completing the spatiotemporal distribution data structure of the pressure gradient.

7. The method according to claim 5, characterized in that, The step of performing spatiotemporal correlation analysis on the pressure gradient spatiotemporal distribution data structure and the flow peak spatiotemporal distribution data structure, calculating the cooperative change intensity index for each spatiotemporal location, and generating spatiotemporal distribution data of cooperative change intensity includes: determining a processing window for spatiotemporal correlation analysis, which includes the current timestamp and a preset number of timestamps before and after it in the time dimension, and the current three-dimensional spatial grid cell and its coordinate adjacent grid cells in the spatial dimension; assigning weight distributions to the processing window, wherein the weight values ​​are determined according to the spatiotemporal distance between each spatiotemporal location within the window and the center location, with the weight value being larger the closer the spatiotemporal distance, and all weight values ​​being normalized; and traversing the pressure gradient spatiotemporal distribution. For each spatiotemporal location of the distributed data structure and the spatiotemporal distribution data structure of the peak flow rate, extract the pressure gradient magnitude estimation data subset and the flow amplitude estimation data subset within the corresponding processing window. Based on the pressure gradient magnitude estimation data subset and the flow amplitude estimation data subset, calculate a scalar value characterizing the spatiotemporal correlation strength between the two, as the co-change intensity index of the current spatiotemporal location. Organize the co-change intensity indices of all spatiotemporal locations according to the dimensional order of the pressure gradient spatiotemporal distribution data structure to generate spatiotemporal distribution data of co-change intensity. Post-process the generated spatiotemporal distribution data of co-change intensity to reduce the impact of estimation noise and random fluctuations.

8. The method according to claim 1, characterized in that, The process of comparing the gas flow path with that under standard sealing conditions, analyzing the path differences segment by segment, and locating key nodes of seal failure based on the spatial mapping relationship includes: acquiring a gas flow path dataset under standard sealing conditions, which contains a sequence of standard pipe segments arranged in the topological order of the air pump pipeline system. Each standard pipe segment includes a standard spatial coordinate range, a standard inner wall roughness distribution, a standard cooperative variation intensity curve, and a standard pressure gradient threshold range; decomposing the gas flow path of the current air pump into a current pipe segment sequence corresponding to the standard pipe segment sequence, establishing a one-to-one correspondence through pipe segment identifiers, matching the number and spatial location of the decomposed current pipe segments with the standard pipe segment sequence, and generating pipe segment comparison index data; for each corresponding pipe segment in the pipe segment comparison index data, calculating the spatial deviation between the trajectory point set of the current pipe segment in the current gas flow path and the preset spatial coordinate range of the standard pipe segment, and the inner wall roughness distribution of the current pipe segment and the standard inner wall roughness distribution... The deviation degree of roughness distribution, the similarity between the current pipeline segment's coordinated change intensity curve and the standard coordinated change intensity curve, and the proportion of time during which the current pipeline segment's pressure gradient value exceeds the standard pressure gradient threshold range are considered. Spatial deviation, deviation degree, similarity degree, and time proportion are combined into a pipeline segment difference feature combination. Feature selection processing is used to filter out key difference features that significantly affect sealing failure, generating a key difference feature vector. Based on the key difference feature vector, a pipeline segment failure probability assessment data model is constructed. This model is trained using historical failed pipeline segment samples, outputting the failure probability value for each current pipeline segment. Pipeline segments with failure probability values ​​higher than the probability threshold are marked as suspected sealing failure pipeline segments. The spatial coordinate set of the abnormal flow fluctuation area and pressure attenuation change area in the spatial mapping relationship is queried. The spatial overlap volume between the spatial coordinate range of the suspected sealing failure pipeline segment and the spatial coordinate set of the abnormal area is calculated. The two connecting nodes of the suspected sealing failure pipeline segment with the largest overlap volume are located as key nodes of sealing failure.

9. The method according to claim 8, characterized in that, For each corresponding pipeline segment pair in the pipeline segment comparison index data, the spatial deviation between the trajectory point set of the pipeline segment in the current gas flow path and the preset spatial coordinate range of the standard pipeline segment is calculated; the deviation between the inner wall roughness distribution of the current pipeline segment and the standard inner wall roughness distribution is calculated; the similarity between the cooperative change intensity curve of the current pipeline segment and the standard cooperative change intensity curve is calculated; and the proportion of time when the pressure gradient value of the current pipeline segment exceeds the standard pressure gradient threshold range is calculated. This includes: obtaining the three-dimensional spatial boundary defined by the preset spatial coordinate range of the standard pipeline segment; obtaining the three-dimensional spatial coordinate set of all trajectory points corresponding to the current pipeline segment in the gas flow path; calculating the shortest distance from each point in the trajectory point set to the three-dimensional spatial boundary of the standard pipeline segment; and calculating a statistic based on the shortest distance of all trajectory points as the spatial deviation, wherein the statistic includes the average distance, the maximum distance, or the proportion of points whose distance exceeds the threshold; obtaining the current pipeline... The roughness distribution data of the inner wall of the pipeline segment is obtained by laser scanning to acquire three-dimensional point cloud data of the inner wall of the pipeline segment. Roughness values ​​are sampled at equal intervals along the axis of the pipeline segment to generate the current roughness sequence. The standard roughness sequence of the standard pipeline segment is read, and the sum of the absolute differences between corresponding positions of the two sequences is calculated and divided by the sequence length to obtain the degree of deviation. The coordinated change intensity curve of the current pipeline segment and the standard coordinated change intensity curve of the standard pipeline segment are extracted. The time axis of the two curves is adjusted by dynamic time alignment, and the sum of the differences between corresponding points of the aligned curves is calculated. The reciprocal of the sum of the differences is used as the similarity index. The pressure gradient magnitude of the current pipeline segment at all timestamps is extracted from the pressure gradient spatiotemporal distribution data structure to form a pressure gradient time series. The standard pressure gradient threshold range of the standard pipeline segment is read, and the number of timestamps with values ​​greater than the upper limit of the threshold or less than the lower limit of the threshold in the pressure gradient time series is counted and divided by the total number of timestamps to obtain the time proportion.

10. A computer system, characterized in that, The computer device includes a processor and a memory, the memory storing a computer program, which is loaded and executed by the processor to implement the airtightness detection method for an air pump based on TPMS feedback and flow modeling as described in any one of claims 1 to 9.

Images