Air tightness detection method adaptive to multi-angle bent pipe

Abstract

The invention discloses an adaptive multi-angle elbow airtightness detection method, which comprises the following steps: arranging a flexible pressure sensor array and a micro inertial measurement unit on the inner wall of an elbow, collecting multi-point sealing pressure and attitude signals in real time, fusing geometric parameters and a spatial mapping relation, and completing denoising, normalization and multi-curvature region mapping of pressure data; a distributed pressure-attitude database is combined, a function model associated with sealing pressure and pipeline geometric parameters is established, inflation, exhaust and deformation of each node are dynamically adjusted through an intelligent closed-loop regulation and control algorithm, collaborative optimization of multi-node pressure and a fitting state is achieved, and the sealing performance of the pipeline is improved. According to the method, the accuracy and the self-adaptive adjusting capacity of sealing pressure distribution under a complex bent pipe structure are effectively improved, the tiny leakage real-time positioning and self-learning optimization capacity is achieved, and the method is suitable for sealing quality guarantee of various types of curved surface pipelines.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pipeline airtightness detection and intelligent sealing control, and particularly relates to a method for detecting airtightness of a multi-angle elbow pipe. BACKGROUND

[0002] Currently, pipeline airtightness detection and sealing control technology is widely used in the petroleum chemical industry, energy transportation, municipal pipe network, food and drug industries, etc., for ensuring the safety of the pipeline system and the non-leakage transmission of the medium. The existing mainstream airtightness detection method mainly uses traditional rigid sealing elements, annular inflation sealing rings or mechanical compression sealing components as the main means, realizes sealing and airtightness evaluation of straight pipe sections or regular curve small elbow pipes through pressure maintenance and leak detection instrument fixed point detection. In the aspect of sealing quality regulation, the current main technical route is concentrated in the integrated annular structure, relies on single-point adjustment of the overall inflation pressure, supplemented by a small number of center / edge pressure feedback units, and combines conventional pressure gauges, manual pressure injection and fixed program structure adjustment method to evaluate the overall sealing performance.

[0003] At present, the existing technology mainly has the following outstanding technical bottlenecks and unmet needs in the scene of complex curvature and variable pipe diameter of elbow pipes:

[0004] (1) The traditional rigid or semi-flexible annular sealing group structure is difficult to adapt to the space change of the elbow pipe with large, continuous and gradually changing shape, and the local area (such as high-curvature turning and diameter transition) often forms a sealing dead angle and an uncontrollable leakage blind area due to uneven pressure distribution or structural deformation out of position;

[0005] (2) The existing multi-point pressure acquisition device is mainly coarse resolution point distribution, which can only reflect the overall trend, and is difficult to real-time feedback of the small deformation and local pressure mutation in the complex space, and does not have the ability of closed-loop pressure / strain dynamic adjustment at the node level;

[0006] (3) In the face of different batches and different geometric characteristics of pipes, the existing scheme has high adjustment cost and low efficiency, lacks an intelligent mapping model based on the actual geometric parameters (such as curvature radius, pipe diameter, bending angle, etc.) of the pipe, and thus the universal detection and batch high-precision sealing of complex special-shaped pipes are difficult to achieve;

[0007] (4) For high requirement scenes (such as food, pharmaceutical, pressure energy pipeline, etc.), the traditional scheme is difficult to realize effective and rapid positioning and accurate compensation for “sub-millimeter level” leakage, special-shaped surface blind area in space bending area and other high-risk points; the isolated analysis of acoustic / pressure data is easily affected by environmental noise and system interference, and the system closed-loop self-learning and multi-modal fusion ability are generally lacking. SUMMARY

[0008] The present application provides a method for detecting airtightness of a multi-angle elbow pipe to solve the above technical problems.

[0009] The technical scheme of the present application is implemented as follows: a method for detecting the air tightness of a multi-angle elbow pipe, comprising:

[0010] S1: Collecting sealing pressure signals and attitude signals distributed along the full length of the multi-curvature elbow pipe, the pressure signals being obtained by a flexible pressure sensor array, and the attitude signals being obtained by a micro inertial measurement unit embedded therein, to obtain the real-time contact state of the sealing unit and the pipe wall of each region;

[0011] S2: Denoising and normalizing the sealing pressure signals, and calibrating the spatial mapping relationship of each sensor node according to the position of each sensing point and the curvature characteristics of the contact region, to realize the standardized input of multi-position and multi-curvature data;

[0012] S3: Based on the normalized pressure signals and attitude signals, extracting the contact pressure distribution characteristics and normal deformation characteristics between each sealing unit and the curved surface of the elbow pipe, to form a multi-dimensional sealing pressure distribution data set;

[0013] S4: Constructing a distributed sealing pressure and spatial geometry parameter mapping model according to the curvature radius, pipe diameter and bending angle of the multiple regions, to realize the function modeling of the sealing pressure distribution and geometric curvature of different regions;

[0014] S5: Comparing the multi-point sealing pressure distribution data with the mapping model, calculating the pressure deviation and non-fitting index of each region, to determine whether each contact node meets the uniformity and fitting threshold conditions of the sealing pressure;

[0015] S6: Based on the pressure deviation and non-fitting index, generating pressure adjustment target values and deformation adjustment instructions for each sealing node through a distributed intelligent control algorithm, for implementing closed-loop control;

[0016] S7: According to the pressure adjustment target values and deformation adjustment instructions, automatically performing inflation, deflation or flexible material form adjustment operations of each sealing node, to realize dynamic balance compensation of the sealing pressure of the multi-curvature complex surface pipe;

[0017] S8: Continuously collecting the adjusted sealing pressure signals and attitude signals, updating the multi-dimensional pressure distribution data set, and determining whether the new distribution state eliminates the local pressure abnormality and non-fitting region, if not, feeding back the correction to the distributed intelligent control algorithm, and optimizing the parameters in a loop;

[0018] S9: After the above dynamic control closed loop is ended, combining the sealing pressure distribution data and acoustic emission signals, implementing multi-point air tightness detection and micro leakage identification, locating and quantifying the smallest leakage point, and inputting the detection results into an adaptive model to realize iterative self-learning of the sealing parameters.

[0019] The application provides a multi-angle elbow gas tightness detection method.

[0020] (1) Through the accurate array fitting algorithm and space node mapping, the accurate physical positioning of the sealing pressure sensor in the multi-curvature and multi-variable pipe diameter area is realized, the problems of local non-fitting and pressure dead angle formation caused by poor structural compliance of the traditional rigid or non-self-adaptive sealing ring are broken through, the full coverage ability of the special-shaped pipeline surface pressure field is fundamentally improved, the edge / angle / transition area which is usually missed in the past is effectively covered, and the integrity of the gas tightness detection is improved;

[0021] (2) Based on the real-time acquisition and fusion of multi-point pressure and posture signals, a sealing pressure-geometric parameter distribution model under complex curvature is constructed, combined with a distributed intelligent regulation algorithm and a multi-arm decision strategy, node-level pressure and shape adaptive adjustment are realized. Compared with the traditional scheme of manual adjustment or one-time fixed pressure, dynamic response and accurate compensation are realized;

[0022] (3) Through the fusion of multi-modal data such as pressure distribution, posture deformation and acoustic emission signals, high-sensitivity intelligent identification of micro leakage points in complex special-shaped elbow pipes is realized. BRIEF DESCRIPTION OF DRAWINGS

[0023] Fig. 1 A flowchart of the multi-angle elbow gas tightness detection method of the application;

[0024] Fig. 2 A sub-flowchart of the multi-angle elbow gas tightness detection method of the application;

[0025] Fig. 3 Another sub-flowchart of the multi-angle elbow gas tightness detection method of the application. DETAILED DESCRIPTION

[0026] In order to make the purpose and advantages of the application more clear and obvious, the application will be further described below in combination with examples; it should be understood that the specific examples described herein are only used to explain the application, and do not limit the application.

[0027] The preferred implementation methods of the application will be described below with reference to the drawings. Those skilled in the art should understand that these implementation methods are only used to explain the technical principles of the application, and do not limit the protection scope of the application.

[0028] When used herein, the singular forms of “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising / including” or “having,” etc., specify the presence of the stated features, wholes, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0029] Please see Figs. 1-3 As shown, a method for testing the airtightness of multi-angle bends includes:

[0030] S1: Collect sealing pressure and attitude signals distributed along the entire length of the multi-curvature bend. The pressure signal is acquired through a flexible pressure sensor array, and the attitude signal is acquired through an embedded micro inertial measurement unit to obtain the real-time contact status between the sealing unit and the pipe wall in each area.

[0031] S2: The sealing pressure signal is denoised and normalized, and the spatial mapping relationship of each sensor node is calibrated according to the position of each sensing point and the curvature characteristics of the contact area, so as to realize the standardized input of multi-position and multi-curvature data.

[0032] S3: Based on the normalized pressure signal and attitude signal, extract the contact pressure distribution characteristics and normal deformation characteristics between each sealing unit and the curved surface of the bend to form a multi-dimensional sealing pressure distribution dataset.

[0033] S4: Based on the curvature radius, pipe diameter and bending angle of multiple regions, construct a distributed sealing pressure and spatial geometric parameter mapping model to realize the functional modeling of sealing pressure distribution and geometric curvature in different regions;

[0034] S5: Compare the multi-point sealing pressure distribution data with the mapping model, calculate the pressure deviation and non-adhesion index of each area, and determine whether each contact node meets the sealing pressure uniformity and adhesion threshold conditions.

[0035] S6: Based on the pressure deviation and non-fit index, a distributed intelligent control algorithm is used to generate pressure adjustment target values ​​and deformation adjustment commands for each sealing node, which are then used to implement closed-loop control.

[0036] S7: Based on the pressure regulation target value and deformation adjustment command, automatically execute the inflation, deflation or flexible material shape adjustment operations of each sealing node to achieve dynamic balance compensation of sealing pressure for pipelines with complex curvature surfaces.

[0037] S8: Continuously collect the adjusted sealing pressure signal and attitude signal, update the multi-dimensional pressure distribution dataset, determine whether the new distribution state eliminates local pressure anomalies and non-fitting areas, and if it does not meet expectations, feed back to the distributed intelligent control algorithm for correction and iteratively optimize parameters.

[0038] S9: After the above dynamic control closed loop is completed, multi-point airtightness detection and micro-leakage identification are carried out by combining the sealing pressure distribution data and acoustic emission signal, the minute leakage point is located and quantified, and the detection results are input into the adaptive model to realize the iterative self-learning of sealing parameters.

[0039] Step S1: Acquire sealing pressure and attitude signals distributed along the entire length of the multi-curvature bend. The pressure signals are acquired through a flexible pressure sensor array, and the attitude signals are acquired through an embedded micro inertial measurement unit to obtain the real-time contact status between the sealing unit and the pipe wall in each area. Specifically, this includes:

[0040] S1.1: Physically divide the preset detection range of the multi-curvature bend, and determine the spatial arrangement scheme of the flexible pressure sensor array and the embedded micro-inertial measurement unit (micro-IMU) based on the geometric parameter database to obtain the detection area distribution strategy of the multi-curvature bend;

[0041] S1.2: The flexible pressure sensor array is precisely attached to each region of the multi-curvature bend pipe wall. Using the automatic structure recognition algorithm, each sensing node and the specific detection area are physically located, and a node spatial index is established to obtain the node region mapping structure.

[0042] The physical division of the preset detection zone for the multi-curvature bend has been completed, and the geometric parameters of the bending area of ​​the target bend, the spatial coordinates of the detection area, and the preliminary arrangement of the sensor array have been obtained.

[0043] A flexible pressure sensor array bonding algorithm (parameters: sensor unit spacing, array compliance, target curvature distribution, bonding force threshold) is adopted to precisely physically attach the flexible pressure sensor array along the multi-curvature pipe wall of the detection section. Local fine-tuning is achieved through a controllable actuator to ensure that each sensing node is tightly bonded to the surface of the curved pipe, achieving a cavity-free bonding state.

[0044] Furthermore, by using a structure automatic recognition algorithm based on three-dimensional point clouds (parameters: initial spatial coordinates of the array, three-dimensional scanning data of the pipeline, tolerance threshold), the actual landing point coordinates of the flexible sensing nodes are collected in real time. By comparing with the three-dimensional geometric model of the curved pipe, the physical detection area of ​​each sensing node on the pipe wall of the curved pipe is located, and the spatial positioning error of the node does not exceed 0.15mm, preferably 0.03mm.

[0045] Furthermore, a node physical positioning calibration method (parameters: IMU-assisted positioning, signal strength mapping, real-time correction compensation) is adopted to correct the actual spatial position of the sensing nodes by using attitude data output by the micro inertial measurement unit to address the array micro-displacement and deformation caused by flexible bonding, thereby improving the node positioning accuracy.

[0046] Furthermore, a unique spatial index label is generated for each sensing node through a hierarchical spatial index modeling algorithm (parameters: region number, node number, spatial index rule), mapping the spatial location of the node to the specific detection area and local curvature parameters one by one.

[0047] By using the node-region spatial mapping template, the spatial index, physical location and curvature of all sensing nodes are hierarchically archived, and the node-region mapping structure is output to realize the precise management of multiple nodes in the detection area and the subsequent spatial association of multiple signals.

[0048] Through the above chain processing, the precise bonding of sensors, physical positioning of nodes, and spatial mapping are seamlessly connected, establishing a high-precision spatial mapping relationship between the nodes of the sealing sensor array on complex curved surfaces with multiple curvatures and the detection area. This enables the physical alignment of the pressure signals of multiple nodes, which plays a key role in improving the integrity and spatial resolution of the detection of sealing areas in complex bends.

[0049] For example, in a municipal water supply system, the outer diameter of the bend is 150mm, the bending radius is 200mm, and the maximum spatial curvature is 0.005mm. -1 The overall pipe section length is 800mm. The flexible pressure sensor array uses a single-node sensing area of ​​10×10mm. 2 The array spacing is 12mm, and the overall compliance is controlled within the adhesion force range of 200mN·m. Three-dimensional coordinate data of the pipe section was obtained through 3D point cloud scanning. Using a structure recognition algorithm, 40 sensor nodes were spatially distributed across three curvature zones: 16 nodes in the high curvature zone, 14 nodes in the medium curvature zone, and 10 nodes in the low curvature zone. With the help of the embedded IMU, the node positioning error was corrected from 0.15mm to 0.03mm, completing the spatial index mapping. The final output node region mapping structure ensures that subsequent pressure signals can be uniquely locked in terms of spatial location and local curvature, improving the system's multi-point data processing accuracy and the complete coverage of complex surface detection.

[0050] S1.3: Real-time output acquisition of flexible pressure sensors at each node, using a high-precision analog-to-digital converter to acquire raw sealing pressure signals, generating distributed pressure sensing raw data streams, providing a signal basis for precise acquisition of multi-point pressure fields;

[0051] S1.4: Perform synchronous acquisition of vector attitude data for micro-inertial measurement units (micro-IMUs) adjacent to each node, acquire local normal changes and spatial attitude information through six-axis / nine-axis attitude sensing, generate synchronous attitude signal stream, and realize the collaborative perception of pressure and spatial state.

[0052] S1.5: Based on the synchronously acquired distributed pressure sensing raw data stream and synchronous attitude signal stream, data fusion is performed using timestamps and node spatial indexes to form a multi-node high-dimensional sealing pressure-attitude raw dataset, laying the data foundation for subsequent signal processing and dynamic modeling of sealing status.

[0053] Step S2: The sealing pressure signal is denoised and normalized, and the spatial mapping relationship of each sensor node is calibrated according to the position of each sensing point and the curvature characteristics of the contact area, so as to achieve standardized input of multi-position, multi-curvature data. Specifically, this includes:

[0054] S2.1: The raw sealing pressure signal output by the flexible pressure sensor array is processed by a bandpass filtering algorithm to remove external noise introduced by environmental interference, physical jitter and high / low frequency noise sources during the signal acquisition process, so as to achieve preliminary noise suppression of the sealing pressure signal and obtain a noise-suppressed sealing pressure signal, providing a basic input for subsequent signal normalization and spatial mapping.

[0055] S2.2: Based on the sealing pressure signal with noise suppression, a normalization processing algorithm (such as z-score normalization or min-max normalization) is implemented to standardize and adjust the amplitude fluctuations of different sensing nodes originating from the pressure sensor array, eliminate the amplitude imbalance caused by differences in sensor sensitivity or structural layout, and output a standardized sealing pressure signal, laying an unbiased foundation for the next step of spatial mapping.

[0056] S2.3: Obtain the preset position parameters of each node in the flexible pressure sensor array and the real-time attitude signal obtained by the micro inertial measurement unit. Using the node space index table, perform node pairing based on physical space distribution for the standardized sealing pressure signal, establish a one-to-one correspondence between the sealing pressure signal and the actual curved pipe spatial position and curvature region, and generate node space mapping data.

[0057] S2.4: Integrate node spatial mapping data, standardized sealing pressure signals and local curvature characteristic parameters corresponding to each node, and adopt a multivariate spatial mapping model (such as three-dimensional interpolation, spatial curvature partition calibration algorithm) to calibrate the actual working area of ​​each pressure sensor node on the complex curved surface, and output a multi-curvature pressure signal set with spatial curvature weights and node distribution labels to prepare for the realization of multi-position, multi-curvature standardized data input;

[0058] S2.5: Input the multi-curvature pressure signal set with spatial curvature weights and node distribution labels into the unified data structure archiving module. Based on the standardized input template, perform multi-position and multi-curvature feature indexing and sorting to achieve standardized alignment and spatial multi-point indexing of pressure signals in different regions. Output a standardized multi-dimensional pressure signal dataset that can be directly used for the extraction of lower-level sealing pressure distribution features and function modeling.

[0059] Step S3: Based on the normalized pressure and attitude signals, extract the contact pressure distribution characteristics and normal deformation characteristics between each sealing unit and the curved surface of the bend, forming a multi-dimensional sealing pressure distribution dataset. For example... Fig. 2 As shown, it specifically includes:

[0060] S3.1: Perform block division processing on the normalized pressure signal data group to obtain the pressure signal subsets of each independent sealing unit in different spatial regions (such as high curvature region, low curvature region, multi-axis bending point) of the curved pipe surface, so as to provide input conditions for the subsequent directional extraction of regional features;

[0061] S3.2: For the segmented pressure signal subset, a pressure distribution identification algorithm (such as partition statistics and spatial interpolation) is used to extract the contact pressure distribution characteristic parameters between each sealing unit and the local bend surface, including the maximum pressure, minimum pressure, average pressure and pressure gradient, in order to obtain the core data parameters reflecting the sealing effect.

[0062] S3.3: For the normalized attitude signal data set, use attitude change analysis algorithms (such as IMU-based normal recovery and spatial tracking) to calculate the normal change and local deformation of each sealing unit at different spatial positions, so as to achieve high-precision quantitative output of the normal deformation characteristic parameters of the sealing unit.

[0063] The normalized attitude signal data set consists of spatial attitude information continuously output by an embedded micro-inertial measurement unit (micro-IMU). The data distribution covers the triaxial or multiaxial acceleration and angular velocity components of various regions of the multi-curvature bend pipe wall. A normal vector recovery algorithm based on IMU attitude data (input parameters: triaxial acceleration, triaxial gyroscope angular velocity, sampling period T) is used to realize real-time inversion of the normal vector of each sealing unit in the local area of ​​the pipe wall.

[0064] Furthermore, by using Euler angle-quaternion conversion or direction cosine matrix (DCM) splicing method, the original attitude data stream of IMU acquired at discrete time moments is transformed into a local normal vector time series matrix to obtain the normal change of the sealed node in the detection interval;

[0065] A vector difference and cumulative integration algorithm (parameters: initial normal value N0, sampling step size Δt) is used to perform cumulative dynamic analysis of the attitude of each sealing unit, which is specifically described by the following mathematical expression:

[0066] ΔN i =N i -N i-1

[0067] Where, ΔN i Let N be the normal change at the i-th sampling time. i The normal vector measured at the i-th sampling time;

[0068] Furthermore, to obtain the local deformation of the sealing unit, based on the integral of the acceleration and angular velocity output from the IMU, a surface tracking algorithm (such as Kalman filter attitude tracking) is used, combined with the nodal spatial coordinates, to calculate the displacement vector of the sealing unit on the pipe wall surface. The following integration model can be used:

[0069]

[0070] Where, δs i Let v(t) be the local deformation (displacement) of the sealing element in the i-th segment, and v(t) be the instantaneous velocity function calculated from the IMU acceleration.

[0071] After the above multi-step parameter processing, the attitude signal data stream is parsed into the normal variation and local deformation sequence of the sealing unit in multi-curvature space, realizing the high-precision quantification of the normal deformation characteristic parameters of each node;

[0072] Through the collaborative processing of attitude change analysis algorithm and surface tracking algorithm, the attitude signal normalized in the previous step is formally transformed into structured normal change and deformation output data, so as to achieve accurate quantification of the three-dimensional fitting state of complex curved pipe sealing unit.

[0073] For example, a food-grade stainless steel bent pipe with a radius of curvature of 120 mm, a diameter of 40 mm, and a maximum bending angle of 75° is tested. Sealing units are spaced 20 mm apart, and the IMU sampling frequency is set to 200 Hz. For a specific node, a segment of continuously acquired triaxial acceleration and angular velocity signals is bandpass filtered and normalized. Quaternion calculations are used to obtain the normal vector sequence at each sampling time. For 20 frames of data, the maximum normal change angle θ is calculated. max =7.5°, average angle of change Kalman filtering attitude tracking was used to estimate the local deformation range δs = 0.85 mm. The results were archived and output to a multi-dimensional sealing pressure distribution dataset, directly demonstrating the node's three-dimensional spatial following capability in that region. In actual testing, the normal variation range of the node in different curvature sections was 5°–12°, and the maximum local deformation range was 0.7–1.5 mm. Test verification results show that the normal deformation variation error is less than 3%, effectively supporting the optimization of the sealing contact area distribution and providing high-precision structured input for subsequent blind zone identification and pressure adaptive compensation algorithms.

[0074] S3.4: Synchronize the pressure distribution characteristic parameters and normal deformation characteristic parameters of each sealing unit in time and spatial correspondence, and use multimodal feature fusion methods (such as weighted aggregation and multidimensional feature splicing) to generate a unified multidimensional sealing pressure distribution dataset, so as to realize the correlation mapping of pressure and deformation data.

[0075] The input is a normalized time-series dataset of pressure distribution characteristic parameters and normal deformation characteristic parameters of each sealing unit on the wall of a multi-curvature bend.

[0076] A multimodal temporal synchronization method (parameters: node number, timestamp, pressure / normal data) is adopted to achieve accurate correspondence of multi-source feature data at sampling time;

[0077] Furthermore, through a spatial location mapping algorithm (dependent on: node spatial index table, block region label), the pressure distribution characteristics and normal deformation characteristics of the same physical region are spatially aligned and archived to obtain a one-to-one pairing structure of node space and feature parameters;

[0078] A multimodal feature fusion method (such as weighted aggregation, principal component analysis (PCA), and multidimensional feature stitching) is adopted. First, the fusion weight parameters α and β for pressure features and normal deformation features are set, and the node multidimensional feature fusion vector is obtained using the following weighted fusion formula:

[0079]

[0080] in, Let be the pressure distribution vector of the i-th sealing unit. Its normal deformation vector;

[0081] Furthermore, through multi-dimensional feature concatenation, the weighted fused features are concatenated with the original multi-component features to form the following extended feature vector:

[0082]

[0083] Where [~] denotes the vector concatenation operation;

[0084] A unified feature archiving mechanism is adopted, using node number, spatial coordinates, and region label as primary keys, to store the extended feature vectors of all nodes in a unified multi-dimensional sealing pressure distribution dataset;

[0085] By using the above-mentioned multimodal feature fusion and archiving methods, the pressure and normal deformation data are mapped in time and space to output a structured high-dimensional feature dataset, providing basic data for subsequent core links such as anomaly analysis, pressure control model, blind spot identification and compensation;

[0086] For example, in the testing scenario of stainless steel bends for food use, the pressure characteristic parameter group includes 10 dimensions such as maximum pressure, average pressure, and pressure gradient at each node, and the normal deformation characteristic parameter includes 6 dimensions such as maximum normal change angle, average normal change angle, and local cumulative deformation.

[0087] The actual pressure / normal weight parameters were set as α = 0.7 and β = 0.3. The nodes were synchronously archived at a sampling frequency of 200Hz. Using the weighted fusion formula and feature concatenation algorithm, the fused feature vector of each node was calculated to have a length of up to 27 dimensions.

[0088] The multidimensional feature vectors of all nodes are indexed by spatial location and regional curvature labels, and batched into a unified multidimensional dataset. Test results show that after archiving through multimodal feature fusion, the consistency of the node structure of the dataset is improved by 15%, the temporal synchronization error is less than 2ms, and the specificity index of pressure-deformation coupling in adjacent node regions is improved by 24%, providing high-precision multidimensional input for subsequent distributed blind spot identification and control algorithms.

[0089] S3.5: For the fused multi-dimensional sealing pressure distribution dataset, use data consistency verification algorithms (such as normalized residual analysis) to identify abnormal pressure points and abnormal normal deformation points, and output correction prompts or abnormal area indexes, providing a structured feature basis for subsequent distributed sealing pressure control and blind zone compensation.

[0090] Step S4: Based on the curvature radius, pipe diameter, and bending angle of multiple regions, a distributed sealing pressure and spatial geometric parameter mapping model is constructed to achieve functional modeling of the sealing pressure distribution and geometric curvature in different regions. Fig. 3 As shown, it specifically includes:

[0091] S4.1: Correlation analysis is performed on the normalized multi-point sealing pressure distribution dataset, attitude data and spatial coordinate data to extract the basic feature mapping relationship between the sealing pressure value and the local curvature radius, pipe diameter and bending angle of each pipe body, so as to provide a data foundation for the next step of constructing the mapping function;

[0092] S4.2: Based on the local radius of curvature, pipe diameter and normalized sealing pressure distribution data obtained by feature extraction, a high-order nonlinear fitting algorithm (such as polynomial regression, piecewise spline fitting or neural network regression model) is used to establish the functional relationship between the sealing pressure distribution of each region and the corresponding spatial geometric parameters, and obtain a preliminary mapping function model;

[0093] S4.3: Input the initially established functional relationship model into the parameter optimization module, and use the minimum mean square error criterion or other adaptive function approximation algorithms to iteratively optimize the multi-point historical pressure distribution data and real-time collected data, so that the mapping model can learn to converge to the best fit and improve the model's generalization ability in diverse and complex curved pipe scenarios.

[0094] S4.4: Spatial partitioning discretizes the optimized mapping model. Based on the curvature change and geometric discontinuities of the bend structure, the sealing pressure mapping function is divided into regions in physical space to generate a distributed sealing pressure mapping model, thereby enabling the output of local pressure control parameters for each spatial partition.

[0095] S4.5: The output parameters of the distributed sealing pressure mapping model are dynamically compared with the pressure control target database to generate regional pressure balance deviation standards and local control weights for intelligent control. These serve as inputs for subsequent pressure compensation and real-time closed-loop feedback, achieving optimal dynamic adaptation of sealing pressure under all complex spatial structures.

[0096] Step S5: Compare the multi-point sealing pressure distribution data with the mapping model, calculate the pressure deviation and non-adhesion index of each region, and determine whether each contact node meets the sealing pressure uniformity and adhesion threshold conditions. Specifically, this includes:

[0097] S5.1: Based on the optimal theoretical sealing pressure distribution corresponding to each region in the sealing pressure and spatial geometric parameter mapping model obtained in the previous step S4, and combined with the sealing pressure distribution dataset formed in step S3, data pairing is performed based on the node spatial coordinates to obtain the actual sealing pressure and theoretical optimal pressure pairing data for each measurement point.

[0098] S5.2: The point-by-point residual analysis algorithm is used to calculate the pressure deviation between the actual sealing pressure and the theoretical optimal pressure of each sealing unit, forming a sealing pressure deviation distribution map, and based on this, the local over-limit area, under-pressure area and possible non-fitting area are disclosed.

[0099] For the actual sealing pressure and theoretical optimal pressure data of the mating system, a point-by-point residual analysis algorithm (parameters: actual pressure value, theoretical optimal pressure value) is used to calculate the deviation of the sealing pressure at the node level.

[0100] Furthermore, the pressure deviation of each sealing unit is generated through differential calculation. The specific calculation method is as follows:

[0101] ΔP i =P i,act -P i,opt

[0102] Wherein, ΔP i Let P be the pressure deviation at node i. i,act P represents the actual measured sealing pressure. i,opt This represents the theoretically optimal sealing pressure for the same node.

[0103] Furthermore, through the residual vector set {ΔP1,ΔP2,...,ΔP} N The spatial distribution mapping method is used to map the pressure deviation of each node back to the spatial location of the sealing node, thereby realizing the spatial visualization of the pressure deviation.

[0104] Furthermore, a threshold determination algorithm is used (parameter: upper limit threshold T). high Lower limit threshold T low All node pressure deviations are classified according to the following rules:

[0105] When ΔP i >T high This area was determined to be a localized area exceeding the standard.

[0106] When ΔP i <T low This area was identified as a localized undervoltage zone.

[0107] When T low ≤ΔP i ≤T high It was determined to be in the normal zone.

[0108] Furthermore, the excessive and under-voltage areas are merged using a spatial clustering algorithm, and any non-aligned areas are marked, providing a structured regional index for subsequent quantification of composite indicators.

[0109] Through the residual analysis and spatial visualization algorithm described above, the node pairing data obtained in the previous step is transformed into a sealing pressure deviation distribution map with spatial distribution labels, enabling rapid identification and accurate positioning of abnormal sealing conditions in multiple areas of complex bends.

[0110] For example, in a scenario involving the detection of a composite stainless steel bend with a total length of 1200mm and a maximum bending angle of 90°, a flexible sealing array is deployed with N=24 sensing nodes along the entire length of the pipe. The measured sealing pressure values ​​at each node range from [79, 102] kPa, and the theoretical optimal pressure values, based on the spatial curvature model, are 88 kPa in the bend region and 95 kPa in the straight pipe region. (The last sentence appears to be incomplete and possibly refers to a different scenario.)high Set to +6 kPa, T low Set to -7 kPa. According to the above formula, the measured P at node 16... 16,act =97kPa, theoretical P 16,opt =88kPa, therefore ΔP 16 =9kPa, judged as exceeding the standard area; measured P at node 21 21,act =87kPa, theoretical P 21,opt =95kPa, ΔP 21 = -8kPa, identified as an undervoltage zone. By visualizing the deviation space of all N=24 nodes, the deviation is automatically calculated using pipe coordinates L. i ,θ i Three areas exceeding the standard and two areas under-pressure were marked, and a sealing pressure deviation distribution map was output. In multiple rounds of testing experiments, this step achieved accurate location of abnormal nodes, quantified the respective pressure deviation ranges, and the consistency confidence level between the pressure distribution map and manual verification reached 99.4%.

[0111] S5.3: Based on the attitude signal of each measurement unit (originating from the micro inertial measurement unit IMU) and combined with the pressure deviation result, the surface adhesion index of each node is quantified. The adhesion index function (such as a correction function based on the normal angle and deformation) is used to quantitatively characterize the degree of non-adhesion, so as to obtain the non-adhesion index distribution of each region.

[0112] S5.4: The sealing pressure deviation generated in step S5.2 and the non-adhesion index obtained in step S5.3 are combined and archived. The threshold judgment algorithm is used to judge the pressure uniformity and adhesion respectively, identify the key nodes and spatial areas that do not meet the sealing threshold standard, and accurately locate them in the sealing unit distribution structure.

[0113] S5.5: For the identified unqualified nodes and areas, the sealing pressure deviation and non-adhesion index are used as closed-loop feedback inputs to provide the distributed intelligent control algorithm module with node-level pressure adjustment targets and adhesion status adjustment references, thereby enabling efficient and accurate command generation for subsequent S6 intelligent control operations.

[0114] Step S6: Based on the pressure deviation and non-fitting index, a distributed intelligent control algorithm is used to generate pressure adjustment target values ​​and deformation adjustment commands for each sealing node, for implementing closed-loop control. Specifically, this includes:

[0115] S6.1: Perform data processing and vectorization on each pressure deviation and non-fit index output by the multi-point sealing pressure distribution data and spatial mapping model to obtain a high-dimensional set of state variables for input to the distributed intelligent control algorithm, ensuring that the multi-node decision factors fully express the pressure unevenness characteristics between actual sealing nodes;

[0116] S6.2: Based on a high-dimensional set of state variables, the state estimation module in the distributed intelligent control algorithm is applied to perform Bayesian filter prediction and dynamic state update on the historical pressure adjustment response sequence of each sealing node, and output the optimal state probability distribution of each node to improve the time consistency and robustness of the current pressure distribution.

[0117] S6.3: Using the optimal state probability distribution output by the state estimation module, as well as the pressure deviation and non-fit index of each node, a multi-arm decision-making mechanism is adopted to generate the pressure adjustment target value of each sealing node, ensuring that the adjustment target takes into account both global pressure uniformity and local adaptability, forming the first type of execution instruction set;

[0118] S6.4: Based on the generation of pressure regulation target, combined with attitude signal and spatial curvature mapping data, the deformation adaptive modeling algorithm is applied to calculate the amount of flexible material deformation adjustment required for each node, thereby generating a second type of deformation adjustment instruction set to realize the node's ability to dynamically fit a large range of complex curved surfaces.

[0119] S6.5: Perform consistency detection on the pressure regulation target value and deformation adjustment command, package the two with a distributed command orchestration mechanism, and output a joint sealing control command set for all sealing nodes. This provides standardized execution parameters for the next step of automatic inflation and deflation and flexible response, realizing coordinated adaptive closed-loop control of pressure and deformation of multiple nodes.

[0120] Step S7: Based on the pressure adjustment target value and deformation adjustment command, automatically execute the inflation, deflation, or flexible material shape adjustment operations of each sealing node to achieve dynamic balance compensation of the sealing pressure of pipelines with complex curvatures. Specifically, this includes:

[0121] S7.1: Based on the pressure regulation target value output by the distributed intelligent control algorithm, the pressure status of the current sealing node is judged to determine the inflation, deflation or maintenance status of each sealing node;

[0122] S7.2: For sealing nodes that are determined to require inflation, perform a micro-inflation operation, and adjust the internal pressure of the flexible sealing material through a high-precision gas injection module so that the pressure of the node gradually approaches the pressure adjustment target value;

[0123] S7.3: For sealing nodes that are determined to require venting, perform a micro-venting operation to release excess pressure inside the flexible sealing material through a precision venting control unit, thereby adjusting the local pressure downward to the target pressure regulation value;

[0124] S7.4: For the deformation adjustment command of flexible sealing material, drive the shape adjustment unit (such as flexible actuator array) to perform posture or contour reconstruction operation on the contact surface of the sealing node according to the deformation adjustment command, so as to further optimize its fit with the tube surface.

[0125] S7.5: After performing the above inflation, deflation and shape adjustment operations, the pressure status and attitude data of each sealing node are collected in real time, and the latest pressure distribution and deformation status are used as feedback input to update the multidimensional sealing pressure distribution dataset.

[0126] Step S8: Continuously collect the adjusted sealing pressure signal and attitude signal, update the multi-dimensional pressure distribution dataset, and determine whether the new distribution state eliminates local pressure anomalies and non-fitting areas. If it does not meet expectations, feedback is sent to the distributed intelligent control algorithm for correction, and the parameters are iteratively optimized. Specifically, this includes:

[0127] S8.1: The sealing pressure signal and attitude signal output after the previous round of dynamic equalization compensation are continuously and in real time acquired based on a flexible pressure sensor array and an embedded micro inertial measurement unit to obtain the latest sealing pressure state data, which serves as the direct input for subsequent judgment and feedback optimization algorithms.

[0128] S8.2: For the continuously acquired sealing pressure signal and attitude signal, a digital filtering and normalization preprocessing algorithm is used to eliminate high-frequency noise and system interference during the acquisition process, realize high-fidelity update of the multi-dimensional pressure distribution dataset, and provide standardized input for multi-point pressure state discrimination and further analysis;

[0129] S8.3: Based on the updated multidimensional pressure distribution dataset, and using the multi-curvature space mapping model, extract the local pressure anomaly parameters and non-fitting index of each sealing node, systematically identify the local low-pressure points and non-fitting areas in the current sealing pressure distribution, so as to form a basis for judging sealing quality.

[0130] S8.4: Local pressure anomaly parameters and non-fit index are used as feedback variables and input into the distributed intelligent control algorithm module. Combined with the multi-dimensional pressure distribution dataset, the pressure regulation target value and deformation adjustment command are dynamically corrected to realize the adaptive adjustment of distributed intelligent control parameters and support closed-loop optimization.

[0131] S8.5: Based on the latest round of distributed intelligent control parameters, automatically control each sealing unit to perform inflation, deflation or flexible shape readjustment actions again, continuously and dynamically compensate for pressure and fit, and finally form a new multi-dimensional pressure distribution dataset. If the distribution state does not meet the optimal threshold of uniformity and fit, repeat the aforementioned closed-loop acquisition, judgment and optimization steps to achieve cyclic dynamic self-optimization.

[0132] Step S9: After the above dynamic control closed loop is completed, multi-point airtightness detection and micro-leakage identification are performed by combining the sealing pressure distribution data and acoustic emission signals. The smallest leak point is located and quantified, and the detection results are input into the adaptive model to achieve iterative self-learning of the sealing parameters. Specifically, this includes:

[0133] S9.1: Collect multi-point sealing pressure distribution data after closed-loop sealing pressure balance regulation to fully obtain the pressure status and abnormal distribution information of each node, and provide high-precision basic data for subsequent airtightness testing and micro-leakage identification.

[0134] S9.2: Simultaneously acquire multi-point acoustic emission signals. By using an acoustic emission sensor array arranged at the sealing node, monitor and collect the original acoustic emission waveform signals during pipeline operation, providing signal basis for the extraction of acoustic mode features and anomaly identification of potential leakage points.

[0135] S9.3: Based on abnormal pressure distribution or abrupt changes in acoustic emission characteristics, a data fusion discrimination model is adopted to jointly analyze the collected multi-point sealing pressure distribution data and acoustic emission signals to achieve preliminary screening of airtight anomaly areas and spatial location of suspected leakage areas.

[0136] S9.4: Conduct in-depth analysis of the sealing pressure change curve and acoustic emission characteristic parameters of the suspected leak point area, and use statistical feature extraction and pattern recognition algorithms to quantify the leak point characteristics (including leak intensity, duration and occurrence node) to obtain a set of parameters for minute leaks under precise location.

[0137] S9.5: Using the quantified set of minute leakage parameters as input, the current distributed sealing pressure control parameters are corrected through an adaptive model parameter correction mechanism, thereby achieving adaptive parameter optimization for local weak sealing areas and continuous elimination of leakage blind spots.

[0138] S9.6: Based on the compensated and optimized sealing pressure distribution results, continuously collect and verify pressure and acoustic emission signals, dynamically correct adaptive model parameters, so as to realize the iterative update of sealing parameters self-learning ability and ensure that the system can adapt to complex scenarios of multiple batches and multiple types of bends for a long time.

[0139] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0140] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and rules of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for testing the airtightness of multi-angle bends, characterized in that, Includes the following steps: S1: Collect sealing pressure and attitude signals distributed along the entire length of the multi-curvature bend to obtain the real-time contact status between the sealing unit and the pipe wall in each area. S2: Normalize the sealing pressure signal and the attitude signal, and calibrate the spatial mapping relationship of each sensor node according to the position of each sensing point and the curvature characteristics of the contact area. S3: Based on the normalized pressure signal and attitude signal, extract the contact pressure distribution characteristics and normal deformation characteristics between each sealing unit and the curved surface of the bend to form a multi-dimensional sealing pressure distribution dataset. S4: Based on the radius of curvature, pipe diameter and bending angle of multiple regions, construct a distributed sealing pressure and spatial geometric parameter mapping model; S5: Compare the multi-point sealing pressure distribution data with the mapping model, calculate the pressure deviation and non-adhesion index of each region, and determine whether each contact node meets the sealing pressure uniformity and adhesion threshold conditions. S6: Based on the pressure deviation and the non-adhesion index, a distributed intelligent control algorithm is used to generate pressure adjustment target values ​​and deformation adjustment instructions for each sealing node. S7: Based on the pressure adjustment target value and deformation adjustment command, automatically execute the inflation, deflation or flexible material shape adjustment operations of each sealing node.

2. The method for testing the airtightness of multi-angle bends according to claim 1, characterized in that, The process following step S7 also includes: S8: Continuously collect the adjusted sealing pressure signal and attitude signal, update the multi-dimensional pressure distribution dataset, determine whether the new distribution state eliminates local pressure anomalies and non-fitting areas, and if it does not meet expectations, feed back to the distributed intelligent control algorithm for correction. S9: After the dynamic control closed loop ends, multi-point airtightness detection and micro-leakage identification are carried out by combining sealing pressure distribution data and acoustic emission signals, locating and quantifying the smallest leak point, and inputting the detection results into the adaptive model.

3. The method for testing the airtightness of multi-angle bends according to claim 1, characterized in that, Step S1 specifically includes: The pre-defined detection range of the multi-curvature bend is physically divided. Based on the geometric parameter database, the spatial arrangement scheme of the flexible pressure sensor array and the embedded micro inertial measurement unit is determined, and the distribution strategy of the detection area of ​​the multi-curvature bend is obtained. The flexible pressure sensor array is precisely attached to each region of the wall of the multi-curvature bend pipe. Using the automatic structure recognition algorithm, each sensing node and the specific detection area are physically located, and a node spatial index is established to obtain the node region mapping structure. Real-time output acquisition of flexible pressure sensors at each node is performed, and a high-precision analog-to-digital conversion module is used to acquire the original sealing pressure signal and generate a distributed pressure sensing raw data stream. Vector attitude data is synchronously acquired for the miniature inertial measurement units adjacent to each node. Local normal changes and spatial attitude information are obtained through six-axis / nine-axis attitude sensing to generate a synchronous attitude signal stream. Based on the distributed pressure sensing raw data stream and the synchronous attitude signal stream, data fusion is performed using timestamps and node spatial indexes to form a multi-node high-dimensional sealing pressure-attitude raw dataset.

4. The method for testing the airtightness of multi-angle bends according to claim 3, characterized in that, The flexible pressure sensor array uses an automatic structural recognition algorithm and three-dimensional point cloud spatial modeling to perform unique mapping and high-precision positioning of each pressure sensing node and each physical detection area on the curved surface of the pipe, with a node spatial positioning error of no more than 0.15mm.

5. The method for testing the airtightness of multi-angle bends according to claim 1, characterized in that, Step S2 specifically includes: The raw sealing pressure signal output by the flexible pressure sensor array is processed by a bandpass filtering algorithm to obtain a sealing pressure signal with noise suppression. Based on the sealing pressure signal with noise suppression, a normalization processing algorithm is implemented to standardize and adjust the amplitude fluctuations of different sensing nodes originating from the pressure sensor array, and output a standardized sealing pressure signal. The system acquires the preset position parameters of each node in the flexible pressure sensor array and the real-time attitude signals acquired by the micro inertial measurement unit. Using the node space index table, it achieves node pairing based on physical space distribution for standardized sealing pressure signals, establishes a one-to-one correspondence between sealing pressure signals and actual curved pipe spatial positions and curvature regions, and generates node space mapping data. By integrating node spatial mapping data, the standardized sealing pressure signal and the local curvature characteristic parameters corresponding to each node, and using a multivariable spatial mapping model, the actual working area of ​​each pressure sensor node on the complex curved surface is calibrated, and a multi-curvature pressure signal set with spatial curvature weights and node distribution labels is output. The multi-curvature pressure signal set with spatial curvature weights and node distribution labels is input into a unified data structure archiving module. Based on the standardized input template, multi-position and multi-curvature feature indexing is performed to output a standardized multi-dimensional pressure signal dataset.

6. The method for testing the airtightness of multi-angle bends according to claim 1, characterized in that, Step S3 specifically includes: The normalized pressure signal data group is divided into blocks to obtain the pressure signal subsets of each independent sealing unit in different spatial regions of the curved surface of the bend. For the segmented pressure signal subset, a pressure distribution identification algorithm is used to extract the contact pressure distribution characteristic parameters between each sealing unit and the curved surface of the local bend, so as to obtain the core data parameters reflecting the sealing effect. For the normalized attitude signal data set, the attitude change analysis algorithm is used to calculate the normal change and local deformation of each sealing unit at different spatial positions; The pressure distribution characteristic parameters and normal deformation characteristic parameters of each sealing unit are synchronized in time and spatially, and a multi-modal feature fusion method is used to generate a multi-dimensional sealing pressure distribution dataset. For the multi-dimensional sealing pressure distribution dataset, an abnormal pressure point and an abnormal normal deformation point are identified using a data consistency verification algorithm, and a correction prompt or an abnormal area index is output.

7. The method for testing the airtightness of multi-angle bends according to claim 6, characterized in that, The contact pressure distribution characteristic parameters include maximum pressure, minimum pressure, average pressure, and pressure gradient.

8. The method for testing the airtightness of multi-angle bends according to claim 1, characterized in that, Step S4 specifically includes: Correlation analysis was performed on the normalized multi-point sealing pressure distribution dataset, attitude data and spatial coordinate data to extract the basic feature mapping relationship between the sealing pressure value and the local curvature radius, pipe diameter and bending angle of each pipe body. Based on the local radius of curvature, pipe diameter and normalized sealing pressure distribution data obtained by feature extraction, a high-order nonlinear fitting algorithm is used to establish the functional relationship between the sealing pressure distribution of each region and the corresponding spatial geometric parameters, and a preliminary mapping function model is obtained. The preliminary mapping function model is input into the parameter optimization module, and the adaptive function approximation algorithm is used to iteratively optimize the multi-point historical pressure distribution data and real-time collected data. The optimized mapping model is spatially partitioned and discretized. Based on the curvature change and geometric discontinuity of the pipe structure, the sealing pressure mapping function is divided into regions in physical space to generate a distributed sealing pressure mapping model. The output parameters of the distributed sealing pressure mapping model are dynamically compared with the pressure regulation target database to generate regional pressure balance deviation standards and local control weights for intelligent regulation.

9. The method for testing the airtightness of multi-angle bends according to claim 1, characterized in that, Step S5 specifically includes: For the optimal theoretical sealing pressure distribution corresponding to each region in the obtained sealing pressure and spatial geometric parameter mapping model, combined with the sealing pressure distribution dataset, data pairing is performed based on the node spatial coordinates to obtain the actual sealing pressure and theoretical optimal pressure pairing data of each measurement point. A point-by-point residual analysis algorithm is used to calculate the pressure deviation between the actual sealing pressure and the theoretical optimal pressure of each sealing unit, and a sealing pressure deviation distribution map is formed. Based on the attitude signal of each measurement unit and combined with the pressure deviation result, the surface adhesion index of each node is quantified. The adhesion index function is used to quantitatively characterize the degree of non-adhesion and obtain the non-adhesion index distribution of each region. The sealing pressure deviation and non-adhesion index are comprehensively archived, and the pressure uniformity and adhesion are judged by a threshold judgment algorithm. The key nodes and spatial areas that do not meet the sealing threshold standard are identified and precisely located in the sealing unit distribution structure. For the identified non-conforming nodes and areas, the sealing pressure deviation and non-adhesion index are used as closed-loop feedback inputs.

10. The method for testing the airtightness of a multi-angle bend according to claim 2, characterized in that, Step S9 includes using a data fusion discrimination algorithm to combine pressure distribution data and acoustic modal characteristics to perform precise spatial localization and quantitative characterization of local micro-leakage points, and inputting the leakage information into an adaptive model to cyclically correct the sealing control parameters.

Images