Method for detecting assembly quality of mechanical connector

Abstract

The application provides a mechanical connector assembly quality detection method, and relates to the technical field of assembly detection. The application collects acoustic emission signals and tactile array signals in the assembly period of the mechanical connector, constructs a three-dimensional sound field model, a three-dimensional deformation model and a curvature mutation area. The two types of models are aligned through a reference coordinate system, the effective stress direction is determined, the target acoustic data are filtered, the stress feature points are projected and positioned, the spatial stress field is constructed through interpolation, the dynamic stress model is formed, the plastic deformation area and the defect propagation path are extracted, and the assembly quality state is determined according to the connected domain area and the spatial aggregation form. The application can realize multi-modal in-situ identification and quality determination of assembly defects in the blind area.

Description

Technical Field

[0001] This application relates to the field of assembly inspection technology, and in particular to a method for inspecting the assembly quality of mechanical connectors. Background Technology

[0002] Mechanical connector assembly quality inspection methods can be used to determine whether connectors have bias, loosening, deformation or local damage after assembly. As humanoid robot flexible joint modules develop towards miniaturization and high integration, connectors are often placed in narrow and difficult-to-observe locations, and their assembly quality directly affects the joint force transmission stability and operational reliability.

[0003] In the prior art, the patent with publication number CN119618998A discloses a solution for acquiring images of the outer rivets by shooting components and detecting the assembly quality of the internal rivet expansion parts by pressure sensors. This solution can improve the problem that it is difficult to observe the internal connection status by traditional manual or single visual inspection.

[0004] However, these methods are mainly aimed at inspecting the structural condition after rivet installation. The inspection information is mostly concentrated on appearance images or local pressure feedback, which is difficult to continuously reflect the stress changes, local irreversible deformation, and defect propagation trends of the connector during the insertion, tightening, and locking processes in confined spaces. Therefore, existing technologies have the technical problem of making it difficult to accurately determine the assembly damage status of mechanical connectors in confined blind zones in situ. Summary of the Invention

[0005] The purpose of this application is to provide a method for inspecting the assembly quality of mechanical connectors, so as to solve the problem in the prior art that it is difficult to accurately determine the damage status of mechanical connectors in restricted blind areas.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a method for inspecting the assembly quality of a mechanical connector, the method comprising:

[0007] Acquire acoustic emission signals and tactile array signals from the mechanical connector during the assembly cycle;

[0008] The spatial attenuation characteristics of the acoustic emission signal are mapped to a three-dimensional sound field model, and a three-dimensional deformation model is constructed based on the tactile array signal. The curvature change region on the three-dimensional deformation model is divided according to the three-dimensional curvature distribution of the three-dimensional deformation model.

[0009] Align the spatial positions of the three-dimensional sound field model and the three-dimensional deformation model in the reference coordinate system to determine the effective force direction corresponding to the curvature change region, and perform spatial direction filtering on the three-dimensional sound field model based on the effective force direction to obtain the target acoustic data;

[0010] The target acoustic data is projected onto the aligned curvature abrupt change region to locate stress feature points, and spatial interpolation compensation is performed on the stress feature points to construct a spatial stress field.

[0011] A dynamic stress model is constructed based on the time evolution sequence of the spatial stress field during the assembly cycle, and the plastic deformation region and the extended defect propagation path are extracted based on the stress gradient characteristics of the dynamic stress model.

[0012] The assembly quality status of the mechanical connector is determined based on the area of ​​the connected domain and the spatial aggregation pattern of the plastic deformation region and the defect propagation path on the three-dimensional deformation model.

[0013] Optionally, aligning the spatial positions of the three-dimensional sound field model and the three-dimensional deformation model in the reference coordinate system to determine the effective force direction corresponding to the curvature abrupt change region includes:

[0014] A reference coordinate system is established with the joint rotation center of the flexible joint module as the reference origin. The assembly axis of the mechanical connector in the flexible joint module is used as the direction reference. The sound field node coordinates of the three-dimensional sound field model and the deformation node coordinates of the three-dimensional deformation model are uniformly mapped to the reference coordinate system.

[0015] In the reference coordinate system, the local tangent planes of each deformation node in the curvature change region are solved, and the local normal vectors perpendicular to the local tangent planes are extracted;

[0016] Calculate the three-dimensional curvature gradient of each of the deformation nodes, and determine the direction weighting coefficient of the corresponding local normal vector based on the magnitude of the three-dimensional curvature gradient.

[0017] The directional weighting coefficient is used to perform vector weighting fusion on all the local normal vectors to generate a fused directional vector, and the fused directional vector is used as the effective force direction corresponding to the curvature change region. The effective force direction represents the main local force direction of the mechanical connector during the assembly process in the restricted blind zone.

[0018] Optionally, the step of calculating the three-dimensional curvature gradient of each of the deformed nodes and determining the direction weighting coefficient of the corresponding local normal vector based on the magnitude of the three-dimensional curvature gradient includes:

[0019] Extract the maximum and minimum principal curvatures of each deformation node on different cross sections;

[0020] The absolute difference between the maximum principal curvature and the minimum principal curvature is used as the three-dimensional curvature gradient.

[0021] Determine the geometric center of the curvature abrupt change region in three-dimensional space, and calculate the spatial distance from each deformation node to the geometric center of the region;

[0022] A spatial penalty coefficient is constructed based on the spatial distance, and the spatial penalty coefficient is used to perform a product calculation on the three-dimensional curvature gradient to obtain the joint gradient feature;

[0023] The joint gradient features are numerically transformed using a preset exponential decay function to calculate the basic weight values ​​that characterize the degree of local deformation.

[0024] Calculate the weight ratio of the basic weight value of each deformation node in the curvature change region, and use the weight ratio as the direction weighting coefficient of the corresponding local normal vector.

[0025] Optionally, the step of performing spatial directional filtering on the three-dimensional sound field model based on the effective force direction to obtain target acoustic data includes:

[0026] Extract the propagation direction vector and acoustic energy gradient of each spatial data point in the three-dimensional sound field model;

[0027] Calculate the directional deviation angle between the propagation direction vector and the effective force direction;

[0028] Based on the magnitude of the acoustic energy gradient, an adaptive dynamic angle threshold is matched for the spatial data points;

[0029] Spatial data points whose directional deviation angle is greater than the dynamic angle threshold are identified as noise interference points and removed. The remaining spatial data points are used as target acoustic data to characterize the acoustic response features of the mechanical connector in the effective force direction.

[0030] The mechanical connector assembly quality inspection method provided in this application has the following beneficial effects: By collecting acoustic emission signals and tactile array signals of the mechanical connector during the assembly cycle, this application can obtain acoustic changes and contact pressure information during the blind zone assembly process; by constructing a three-dimensional sound field model and a three-dimensional deformation model and dividing the curvature change region, it can convert the dispersed detection signals into spatialized data and find the potential stress concentration location; by aligning the three-dimensional sound field model and the three-dimensional deformation model and performing spatial direction filtering, it can reduce irrelevant acoustic interference and retain the target acoustic data related to the actual stress direction; by projecting the target acoustic data, locating stress feature points and performing spatial interpolation compensation, it can form a continuous spatial stress field; by constructing a dynamic stress model and extracting the plastic deformation region and defect propagation path, it can identify the location of irreversible deformation and its possible extension direction; by determining the assembly quality state through the area of ​​the connected domain and the spatial aggregation pattern, it can convert the damage distribution in the blind zone into an outputtable quality judgment result.

[0031] Furthermore, this application establishes a reference coordinate system with the joint rotation center and the assembly axis, uniformly mapping the coordinates of the sound field nodes and the coordinates of the deformation nodes. It also extracts local tangent planes, local normal vectors, three-dimensional curvature gradients, and direction weighting coefficients in the curvature change region, and then performs vector weighted fusion on all local normal vectors. This can obtain the effective force direction that represents the main force direction of the local force, improve the correspondence between the three-dimensional sound field model and the three-dimensional deformation model, and enhance the reliability of subsequent stress positioning and assembly quality judgment. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart illustrating a mechanical connector assembly quality inspection method disclosed in this application;

[0034] Figure 2 This is a schematic diagram of the restricted blind zone detection data acquisition structure disclosed in this application;

[0035] Figure 3 This is a schematic diagram showing the spatial alignment of the three-dimensional sound field model and the three-dimensional deformation model disclosed in this application.

[0036] Figure 4 This is a schematic diagram showing the relationship between the regional boundary and the plastic deformation region disclosed in this application.

[0037] Figure label:

[0038] 20-Flexible joint module; 21-Joint rotation center; 22-Mechanical connector; 23-Assembly axis; 24-Acoustic emission acquisition channel; 25-Tactile array; 26-Restricted blind zone; 30-Three-dimensional sound field model; 31-Sound field node; 32-Three-dimensional deformation model; 33-Deformation node; 34-Curvature change region; 40-Region boundary; 41-Plastic deformation region; 42-Surrounding region without irreversible deformation; 43-Edge node; 44-Defect propagation direction. Detailed Implementation

[0039] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0040] Figure 1 This is a flowchart illustrating a method for inspecting the assembly quality of a mechanical connector disclosed in this application. The method includes:

[0041] S101, Acquire acoustic emission signals and tactile array signals of the mechanical connector during the assembly cycle.

[0042] In one specific implementation, the detection controller first synchronously reads the output data from the acoustic emission acquisition channel and the tactile array during the assembly cycle, and adds event timestamps to each set of data. Next, it performs bandpass filtering and event truncation on the acoustic emission signals to retain effective acoustic segments related to contact, friction, or micro-impact. Then, it performs zero-point calibration and pressure conversion on the tactile array signals to obtain tactile pressure data that can be used for subsequent deformation analysis. Through event timestamps, the detection controller can correlate acoustic and tactile responses within the same assembly stage.

[0043] The acoustic emission acquisition channel is deployed on the joint housing, support, or adjacent mounting surface around the restricted blind zone to receive transient acoustic emission signals generated by the mechanical connector during the processes of entering contact, generating friction, experiencing micro-impact, and completing locking. The tactile array is deployed on the assembly contact path of the mechanical connector or adjacent support surface to collect the spatial position coordinates, tactile pressure values, normal direction, and zero-point calibration values ​​corresponding to each tactile sensing unit. The assembly cycle refers to the continuous time period from when the mechanical connector enters the assembly contact state until it completes clamping, locking, or positioning and enters the holding state. The acoustic emission signal includes transient acoustic data generated by the mechanical connector during contact, friction, micro-impact, or local damage. The tactile array signal includes the contact pressure, pressure position, acquisition time, and sensing unit number collected by the tactile sensing unit, which reflects the local contact and deformation of the mechanical connector within the restricted blind zone.

[0044] like Figure 2 As shown, the flexible joint module 20 has a joint rotation center 21, and the mechanical connector 22 enters the restricted blind zone 26 along the assembly axis 23. An acoustic emission acquisition channel 24 is arranged around the restricted blind zone 26 to receive acoustic emission signals generated by contact, friction, or micro-impact. A tactile array 25 is arranged along the assembly contact path of the mechanical connector 22 or near the support surface to collect contact pressure and local deformation response. Through this arrangement, the detection controller can obtain both acoustic and tactile responses within the same assembly cycle.

[0045] S102. Map the spatial attenuation characteristics of the acoustic emission signal into a three-dimensional sound field model, and construct a three-dimensional deformation model based on the tactile array signal. Divide the curvature change region on the three-dimensional deformation model according to the three-dimensional curvature distribution of the three-dimensional deformation model.

[0046] Among them, the spatial attenuation characteristics include the reception time difference and energy attenuation value when the acoustic emission signal arrives at different acoustic emission acquisition channels, which are used to reflect the approximate location and propagation intensity of the acoustic emission event in the confined blind zone; the three-dimensional sound field model is a spatial data model composed of sound source location coordinates, sound source energy estimation value and acoustic emission event time, which is used to represent the acoustic response distribution generated by the mechanical connector during the assembly process; the three-dimensional deformation model is a spatial deformation model obtained by converting tactile array signals, which is used to represent the pressure deformation of the mechanical connector on the assembly contact path; the three-dimensional curvature distribution is used to describe the change in the surface curvature of the three-dimensional deformation model; the curvature abrupt change region is used to represent the location where the surface deformation change reaches the preset curvature change threshold and is continuously distributed.

[0047] In this embodiment of the application, S102 specifically includes:

[0048] S1021. Obtain the reception time difference of the acoustic emission signal arriving at different acquisition channels.

[0049] The reception time difference refers to the difference in reception time between multiple acoustic emission acquisition channels when the same acoustic emission event is received by multiple acoustic emission acquisition channels.

[0050] In this embodiment, the detection controller first extracts the effective band corresponding to the same acoustic emission event from the acoustic emission signal, then selects an acoustic emission acquisition channel as a reference channel, and compares the reception time of other acoustic emission acquisition channels with that reference channel to obtain the reception time difference. Based on the positive and negative relationship of the reception time difference and the absolute value of the time difference, the approach direction and distance difference of the acoustic emission event relative to each acquisition channel are determined. In order to obtain the sound source positioning coordinates in three-dimensional space, the acoustic emission acquisition channel includes at least four non-coplanar effective acquisition channels. If there are fewer than four effective acquisition channels that can participate in the positioning, or if the calibration coordinates of the four effective acquisition channels are in the same plane, the detection controller performs time stamping and energy filtering on the acoustic emission event, and temporarily does not use the event as the three-dimensional sound source positioning coordinates for modeling.

[0051] S1022. Based on the receiving time difference and the energy attenuation value of the acoustic emission signal, calculate the sound source location coordinates in three-dimensional space.

[0052] Among them, the energy attenuation value refers to the degree of energy reduction of the acoustic emission signal during its propagation to the acoustic emission acquisition channel; the sound source location coordinates refer to the estimated location of the acoustic emission event in the three-dimensional space of the restricted blind zone.

[0053] In this embodiment, the detection controller first uses a time-difference-of-arrival (TDOA) positioning method to process the receiving time difference, reading the calibration coordinates of each acoustic emission acquisition channel in the reference coordinate system and the propagation speed of the acoustic emission signal in the connector material. Next, it establishes propagation time difference constraints from candidate sound source locations to each acoustic emission acquisition channel and calculates the propagation time difference residual for each candidate location. If the propagation time difference residual does not exceed the time difference residual threshold, the detection controller retains the candidate location as a time-constrained candidate location. Then, it compares the distance attenuation relationship from the candidate location to each acoustic emission acquisition channel using energy attenuation values ​​and calculates the energy attenuation residual for that candidate location. If the energy attenuation residual does not exceed the energy attenuation residual threshold, and the propagation time difference residual does not exceed the time difference residual threshold, the detection controller determines the candidate location as the sound source positioning coordinates. The time difference residual threshold can be determined based on the sampling period of the acoustic emission acquisition channel and the acoustic velocity calibration error of the connector material, while the energy attenuation residual threshold can be determined based on the acoustic energy attenuation calibration results of connector materials of the same model. This processing method enables the sound source location coordinates to reflect both the temporal relationship of the acoustic emission event arriving at different acquisition channels and the attenuation relationship of acoustic energy after propagation within the restricted blind zone.

[0054] S1023. Construct a three-dimensional sound field model reflecting the energy attenuation distribution characteristics according to the sound source location coordinates and the corresponding energy attenuation value.

[0055] In this embodiment, the detection controller treats each sound source location coordinate as a sound field node and writes the corresponding energy attenuation value, sound source energy estimate, and event time into the sound field node. After multiple sound field nodes are arranged according to their positional relationship in the reference space, a three-dimensional sound field model is formed. Therefore, the three-dimensional sound field model can reflect the spatial distribution of acoustic emission response in the restricted blind zone, rather than just retaining the acoustic intensity result of a single channel.

[0056] S1024. Based on the spatial position coordinates of each tactile sensing unit in the tactile array signal and the tactile pressure value, construct a three-dimensional deformation model.

[0057] Among them, the tactile sensing unit is a single detection point in the tactile array used to sense local pressure; the tactile pressure value is used to represent the pressure magnitude when the corresponding position is squeezed or contacted by the mechanical connector.

[0058] In this embodiment, the detection controller first reads the spatial coordinates and tactile pressure values ​​of each tactile sensing unit; then, it reads a pre-calibrated correspondence between pressure and deformation. This correspondence can be obtained by applying a known normal load to a tactile array of the same model and recording the corresponding displacement, or it can be represented by a calibration table or monotonic fitting curve of pressure and deformation. Then, the detection controller converts the tactile pressure values ​​into normal deformation values ​​at the corresponding positions according to this correspondence, and displaces the original spatial position of the tactile sensing unit along its normal direction, obtaining multiple deformation nodes. These multiple deformation nodes together constitute a three-dimensional deformation model, used to represent the spatial deformation state of the mechanical connector in the assembly contact area.

[0059] S1025. Calculate the local surface curvature of each tactile sensing unit on the three-dimensional deformation model to obtain the three-dimensional curvature distribution.

[0060] Among them, local surface curvature is used to represent the degree of bending of the three-dimensional deformation model within a certain small range; three-dimensional curvature distribution is used to represent the bending changes of multiple deformation nodes on the spatial surface.

[0061] In this embodiment, the detection controller selects the surrounding adjacent deformation nodes as the center of each deformation node and performs local surface fitting on these deformation nodes. Through local surface fitting, it can be determined whether the position is a gentle deformation or a sharp bend, thereby obtaining the local surface curvature of the corresponding position of each tactile sensing unit. After arranging all the local surface curvatures according to their spatial positions, a three-dimensional curvature distribution is obtained.

[0062] S1026. Based on the three-dimensional curvature distribution, extract the stress concentration areas in the three-dimensional deformation model where the surface curvature value is higher than the preset curvature change threshold.

[0063] The preset curvature change threshold is a criterion for distinguishing between normal deformation and abnormal concentrated deformation. It can be set according to the qualified assembly samples and structural tolerances of the same type of mechanical connector. In this embodiment, the value of the threshold is not limited. The stress concentration part refers to the local area where the surface curvature value is higher than the preset curvature change threshold. In the assembly scenario, it corresponds to the stress concentration position of the pressure transition part, locking contact part or thin-walled connection part.

[0064] In this embodiment of the application, the detection controller compares the surface curvature values ​​of each surface in the three-dimensional curvature distribution with a preset curvature change threshold. When the surface curvature value at a certain position is higher than the threshold, it indicates that the position has reached the abnormal concentrated deformation judgment condition, and the detection controller extracts the position as a stress concentration part.

[0065] S1027. The extracted stress concentration areas are used as curvature change regions on the three-dimensional deformation model.

[0066] In this embodiment, the detection controller performs spatial connectivity judgment on the extracted stress concentration locations. If the multiple stress concentration locations are adjacent or continuously distributed on the three-dimensional deformation model, the detection controller merges these locations into a curvature change region. The curvature change region is used as the key analysis object for subsequent determination of effective force direction and screening of target acoustic data.

[0067] As an example, when the mechanical connector enters the restricted blind zone of the flexible joint module along the assembly axis, the acquisition channel first acquires acoustic emission signals generated by contact, friction, or micro-impact. The detection controller determines the sound source location coordinates of the acoustic emission event in the restricted blind zone by comparing the reception time difference of the acoustic emission signals arriving at different acquisition channels and combining the energy attenuation value. Then, it combines the location coordinates of multiple sound sources and the corresponding energy information to form a three-dimensional sound field model. At the same time, the tactile array acquires the spatial position coordinates and tactile pressure values ​​of each tactile sensing unit. The detection controller converts the tactile pressure into normal deformation to form a three-dimensional deformation model. Then, the detection controller calculates the local surface curvature of the three-dimensional deformation model and identifies continuous areas with surface curvature values ​​higher than a preset curvature change threshold as curvature abrupt change regions. The above example is only one example of this application. In practical applications, it can also be set according to the structural dimensions of the mechanical connector, the layout of the tactile array, and the assembly process requirements. This application does not limit this.

[0068] This application converts acoustic emission signals into spatial acoustic distributions and tactile array signals into spatial deformation distributions, enabling the simultaneous acquisition of acoustic response and deformation concentration locations within a confined blind zone, thus providing a basis for subsequent stress analysis and assembly quality assessment.

[0069] S103. Align the spatial positions of the three-dimensional sound field model and the three-dimensional deformation model in the reference coordinate system to determine the effective force direction corresponding to the curvature change region, and perform spatial direction filtering on the three-dimensional sound field model based on the effective force direction to obtain the target acoustic data.

[0070] The reference coordinate system is a three-dimensional spatial coordinate system established with the joint rotation center of the flexible joint module as the three-dimensional coordinate origin and the assembly axis of the mechanical connector as the direction reference. It is used to compare the three-dimensional sound field model and the three-dimensional deformation model on the same spatial scale. The effective force direction is used to represent the local compression trend in the curvature change region. The spatial direction filtering is used to remove acoustic data from the three-dimensional sound field model whose propagation direction deviates from the dynamic angle threshold. The target acoustic data is the acoustic data retained after spatial direction filtering and is used for subsequent location of stress feature points.

[0071] like Figure 3 As shown, the sound field node 31 in the three-dimensional sound field model 30 and the deformation node 33 in the three-dimensional deformation model 32 are first uniformly mapped to the reference coordinate system so that the acoustic response position and the tactile deformation position can correspond in the same three-dimensional space; the curvature change region 34 is located on the three-dimensional deformation model 32 and is used to provide spatial constraints for determining the effective force direction and spatial direction filtering.

[0072] In this embodiment of the application, S103 specifically includes:

[0073] S1031. Establish a reference coordinate system with the joint rotation center of the flexible joint module as the reference origin, and use the assembly axis of the mechanical connector in the flexible joint module as the direction reference to uniformly map the sound field node coordinates of the three-dimensional sound field model and the deformation node coordinates of the three-dimensional deformation model to the reference coordinate system.

[0074] Among them, the sound field node coordinates are used to represent the spatial location of the sound emission event in the three-dimensional sound field model; the deformation node coordinates are used to represent the spatial location of the tactile deformation location in the three-dimensional deformation model; and the unified mapping refers to the detection controller transforming the two types of node coordinates into the same reference coordinate system according to the pre-calibrated installation position relationship.

[0075] In this embodiment, the detection controller first reads the calibration positions of the acoustic emission acquisition channel and the tactile array within the flexible joint module, using the joint rotation center as the origin of the three-dimensional coordinate system and the assembly axis as the directional reference. Then, based on the common positioning point obtained before assembly calibration, it calculates the rigid body transformation relationship from the acoustic emission acquisition channel installation coordinate system, the tactile array installation coordinate system to the reference coordinate system. This rigid body transformation relationship includes rotation and translation relationships. Finally, the detection controller uses this rigid body transformation relationship to perform a three-dimensional coordinate transformation on the sound field node coordinates and the deformation node coordinates, so that the acoustic positions in the three-dimensional sound field model and the deformation positions in the three-dimensional deformation model can correspond in the same space.

[0076] S1032. Under the reference coordinate system, solve for the local tangent plane of each deformation node in the curvature change region, and extract the local normal vector perpendicular to the local tangent plane.

[0077] Among them, the local tangent plane is a planar approximation of the surface morphology near a deformation node in the curvature abrupt change region, used to represent the local surface direction near that location; the local normal vector is a direction vector perpendicular to the local tangent plane, used to represent that the location is prone to deformation along that direction when subjected to compression or pushing.

[0078] In this embodiment, the detection controller first determines whether the curvature change region is located on the blind contact surface, locking edge, or pressure transition part of the mechanical connector. If the curvature change region falls into the above-mentioned structural parts, the curvature change region is taken as the effective force direction analysis object. Then, each deformation node in the curvature change region is taken as the analysis object, and the adjacent deformation nodes around each deformation node are selected. The local tangent plane is solved by the least squares plane fitting method, as shown in formula (1). The local tangent plane can be expressed as:

[0079] (1)

[0080] in, This represents the coordinate value of the neighboring deformation node along the first coordinate axis in the reference coordinate system; This represents the coordinate value of the neighboring deformation node along the second coordinate axis in the reference coordinate system; This represents the coordinate value of the neighboring deformation node along the third coordinate axis in the reference coordinate system; This represents the plane coefficient of the local tangent plane along the first coordinate axis. This represents the plane coefficient of the local tangent plane along the second coordinate axis. This represents the plane coefficient of the local tangent plane along the third coordinate axis. The constant term represents the local tangent plane.

[0081] As an example, in a reference coordinate system with millimeters as the coordinate unit, if the local tangent plane obtained by fitting the neighborhood of a certain deformation node after normalization is... ,but , , , The above values ​​are only used to illustrate the meaning of the plane coefficient and constant term. The actual values ​​are obtained by fitting the coordinates of the neighborhood deformation nodes, and this application does not limit them.

[0082] The detection controller obtains the planar parameters of the local tangent plane by minimizing the distance deviation from the neighboring deformed nodes to the local tangent plane; after obtaining the planar parameters, the local normal vector can be expressed as... , and The controller also adjusts the local normal vectors to the pressure side corresponding to the assembly axis to avoid directional cancellation during subsequent directional fusion.

[0083] S1033. Calculate the three-dimensional curvature gradient of each of the deformation nodes, and determine the direction weighting coefficient of the corresponding local normal vector based on the magnitude of the three-dimensional curvature gradient.

[0084] S1033 may specifically include the following steps:

[0085] A1. Extract the maximum and minimum principal curvatures of each deformation node on different cross sections; A2. Use the absolute difference between the maximum and minimum principal curvatures as the three-dimensional curvature gradient; A3. Determine the geometric center of the curvature abrupt change region in three-dimensional space, and calculate the spatial distance from each deformation node to the geometric center of the region; A4. Construct a spatial penalty coefficient based on the spatial distance, and use the spatial penalty coefficient to perform a product calculation on the three-dimensional curvature gradient to obtain a joint gradient feature; A5. Perform numerical transformation on the joint gradient feature using a preset exponential decay function to calculate the basic weight value representing the degree of local deformation; A6. Calculate the weight ratio of the basic weight value of each deformation node in the curvature abrupt change region, and use the weight ratio as the direction weighting coefficient of the corresponding local normal vector.

[0086] Among them, the three-dimensional curvature gradient is used to represent the strength of the surface curvature change at the deformation node; the direction weighting coefficient is used to represent the proportion of a certain local normal vector in the subsequent direction fusion. Deformation nodes whose curvature change reaches the preset curvature change threshold and are close to the stress concentration position are given a higher direction weighting coefficient for their local normal vectors in the direction fusion than that of the surrounding nodes.

[0087] In this embodiment, the detection controller first extracts the maximum and minimum principal curvatures of each deformed node at different cross-sections. These different cross-sections include an axial cross-section along the assembly axis, a radial cross-section passing through the contact edge of the mechanical connector, and a contact cross-section passing through the locking contact. The detection controller can first fit a local surface based on the deformed node and its neighboring deformed nodes, and then read the principal curvature changes of this local surface within the aforementioned different cross-sections to obtain the maximum and minimum principal curvatures. Then, the absolute difference between the maximum and minimum principal curvatures is calculated, and this difference is used as the three-dimensional curvature gradient of the deformed node. Since this process compares the differences under the same curvature dimension, it can directly reflect the bending differences of the deformed node at different cross-sections.

[0088] Next, the geometric center of the curvature abrupt change region in three-dimensional space is determined. The geometric center can be obtained by taking the geometric mean of the three-dimensional coordinates of all deformation nodes within the curvature abrupt change region, and is used to represent the spatial concentration location of the curvature abrupt change region. In this application, the geometric center can be used as the spatial representative point of the force concentration location in the blind zone assembly of the mechanical connector. Further, the spatial distance from each deformation node to the geometric center is calculated, and a spatial penalty coefficient is constructed based on this spatial distance, as shown in formula (2). The spatial penalty coefficient can be expressed as:

[0089] (2)

[0090] in, Indicates the first The spatial penalty coefficient corresponding to each deformation node; Indicates the first Spatial distance from each deformation node to the geometric center of the region; The reference distance can be determined based on the average distance from each deformation node to the geometric center of the region within the curvature abrupt change area, the spacing between adjacent tactile sensing units in the tactile array, or the average side length of the mesh in the three-dimensional deformation model. As shown in the formula above, as the distance from the deformation node to the geometric center of the region increases, the spatial penalty coefficient decreases accordingly, and its influence on the effective force direction weakens.

[0091] The joint gradient feature is obtained by multiplying the three-dimensional curvature gradient using the spatial penalty coefficient. The joint gradient feature considers both the strength of curvature change and the spatial position relationship of the deformation nodes, so that the direction weighting coefficient is biased towards the deformation nodes where the curvature change reaches the preset curvature change threshold and is close to the force concentration position. As shown in formula (3), the joint gradient feature can be expressed as:

[0092] (3)

[0093] in, Indicates the first Joint gradient characteristics of each deformable node; Indicates the first Three-dimensional curvature gradient of each deformable node; The reference curvature gradient can be determined based on the average three-dimensional curvature gradient of each deformation node in the curvature abrupt change region, the curvature gradient reference value of a qualified assembly sample, or a preset curvature change threshold. Indicates the first The spatial penalty coefficient corresponding to each deformation node. and All of them are dimensionless quantities, therefore It is a dimensionless quantity.

[0094] The joint gradient features are numerically transformed using a preset exponential decay function to calculate the basic weight value. In one specific implementation, the preset exponential decay function can be implemented using a transformation function that includes an exponential decay term, so that the larger the joint gradient features are, the closer the basic weight value is to the stable upper limit, avoiding excessive influence of a single abnormal deformation node on subsequent directional fusion. As shown in formula (4), the basic weight value can be expressed as:

[0095] (4)

[0096] in, Indicates the first The basic weight values ​​of each deformable node. It is a dimensionless quantity; Indicates the first Joint gradient characteristics of each deformation node.

[0097] Calculate the weight ratio of the basic weight value of each deformation node in the curvature change region, and use this weight ratio as the direction weighting coefficient of the corresponding local normal vector, as shown in formula (5). The direction weighting coefficient can be expressed as:

[0098] (5)

[0099] in, Indicates the first The directional weighting coefficients of the local normal vectors corresponding to each deformation node; Indicates the first The basic weight values ​​of each deformable node; This represents the index of the deformation node participating in the weight normalization calculation, with a value ranging from 1 to... ; This indicates the number of deformation nodes involved in the calculation within the curvature abrupt change region; Indicates the first The basic weight values ​​of each deformable node.

[0100] Through the above processing method, the detection controller can convert the curvature change, spatial distance and local deformation degree of each deformation node into comparable directional weighting coefficients, which are then used for vector weighting fusion of the local normal vectors.

[0101] S1034. Using the direction weighting coefficient, perform vector weighting fusion on all the local normal vectors to generate a fused direction vector, and use the fused direction vector as the effective force direction corresponding to the curvature change region. The effective force direction represents the main local force direction of the mechanical connector during the assembly process in the restricted blind zone.

[0102] Among them, the fused direction vector is the direction result obtained by merging multiple local normal vectors according to the direction weighting coefficient, and is used to represent the dominant force direction in the curvature change region as a whole.

[0103] In this embodiment, the detection controller first multiplies each local normal vector by its corresponding direction weighting coefficient, so that the local normal vector whose curvature change reaches the preset curvature change threshold and whose position is close to the geometric center of the region occupies the dominant weight in the fusion; then, all the weighted local normal vectors are vector merged, and the fusion result is normalized in direction to obtain the fusion direction vector, which is the effective force direction corresponding to the curvature change region, as shown in formula (6):

[0104] (6)

[0105] in, Indicates the effective direction of force; Indicates the first The direction weighting coefficients of each local normal vector; Indicates the first Local normal vectors; This represents the number of local normal vectors involved in the fusion. After the detection controller normalizes the weighted fusion result, it obtains the effective force direction that only represents the direction and does not introduce the length dimension.

[0106] S1035. Extract the propagation direction vector and acoustic energy gradient of each spatial data point in the three-dimensional sound field model.

[0107] Among them, the spatial data points correspond to the acoustic field node positions in the three-dimensional acoustic field model formed by the acoustic response of mechanical connectors in the confined blind zone through assembly contact, friction, or micro-impact; the propagation direction vector is used to represent the propagation direction of the acoustic emission signal in space; and the acoustic energy gradient is used to represent the trend of acoustic energy change in spatial position.

[0108] In this embodiment, the detection controller first reads the coordinates of each spatial data point and the estimated energy value of the sound source in the three-dimensional sound field model, and calculates the direction of change of acoustic energy from strong to weak according to the positional relationship between adjacent spatial data points. This direction of change is used as the propagation direction vector or used to correct the propagation direction vector. At the same time, the acoustic energy gradient is obtained according to the amplitude of energy change.

[0109] S1036. Calculate the directional deviation angle between the propagation direction vector and the effective force direction.

[0110] The directional deviation angle is used to represent the degree of deviation between the propagation direction vector and the effective force direction. The smaller the directional deviation angle, the closer the acoustic propagation direction of the spatial data point is to the local principal force direction of the curvature change region.

[0111] In this embodiment, the detection controller compares the angle between the propagation direction vector and the effective force direction of each spatial data point. First, it unifies the two into a unit direction vector, and then obtains the direction deviation angle based on the vector angle relationship. This calculation only compares the directional relationship between the propagation direction vector and the effective force direction, without introducing new physical dimensions, and is used for subsequent comparison with the dynamic angle threshold.

[0112] S1037. Based on the magnitude of the acoustic energy gradient, adaptively match a dynamic angle threshold for the spatial data points.

[0113] Among them, the dynamic angle threshold is a direction screening benchmark that is adjusted according to the acoustic energy gradient, in order to avoid using the same fixed angle standard for all spatial data points.

[0114] In this embodiment of the application, when the dynamic angle threshold is expressed as a continuous function, as shown in formula (7), the dynamic angle threshold can be expressed as:

[0115] (7)

[0116] in, Indicates the first Dynamic angle thresholds corresponding to each spatial data point; Indicates the lower limit of the angle threshold; Indicates the upper limit of the angle threshold; Indicates the first Acoustic energy gradient of each spatial data point; This represents the reference acoustic energy gradient. It should be noted that... , and Calibration can be performed based on the acoustic direction statistics of qualified and abnormal assembly samples.

[0117] S1038. Spatial data points with directional deviation angles greater than the dynamic angle threshold are treated as noise interference points and removed. The remaining spatial data points are used as target acoustic data to characterize the acoustic response features of the mechanical connector in the effective force direction.

[0118] In this embodiment, the detection controller first compares the directional deviation angle and the dynamic angle threshold of each spatial data point: if the directional deviation angle is greater than the dynamic angle threshold, the spatial data point is marked as a noise interference point and removed from the subsequent analysis data of the three-dimensional sound field model; if the directional deviation angle is not greater than the dynamic angle threshold, the spatial data point is retained and included in the target acoustic data. After this filtering, the target acoustic data is aggregated according to the effective force direction, which can represent the acoustic response characteristics of the mechanical connector in the effective force direction.

[0119] As an example, when the mechanical connector enters the restricted blind zone of the flexible joint module along the assembly axis, the detection controller first transforms the coordinates of the sound field nodes in the three-dimensional sound field model and the coordinates of the deformation nodes in the three-dimensional deformation model into a reference coordinate system based on the joint rotation center and the assembly axis. Secondly, it determines whether the curvature abrupt change region is located on the blind zone contact surface, locking edge, or pressure transition area, and solves for the local tangent plane and local normal vector of the deformation nodes within this curvature abrupt change region. Then, it calculates the direction weighting coefficient by combining the three-dimensional curvature gradient, the region's geometric center, and the spatial penalty coefficient, and merges all local normal vectors into an effective force direction.

[0120] Next, spatial data points corresponding to the acoustic response positions of assembly contact, friction, or micro-impact are extracted from the three-dimensional sound field model. The propagation direction vector, acoustic energy gradient, and directional deviation angle are calculated, and a dynamic angle threshold is matched for each spatial data point. Finally, noise interference points with directional deviation angles greater than the dynamic angle threshold are removed, and the remaining spatial data points are used as the target acoustic data. The above example is only one example of this application. In practical applications, it can also be set according to the mechanical connector structure, confined blind zone space, and acoustic emission acquisition channel layout. This application does not limit this.

[0121] This application improves the correspondence between acoustic response and assembly stress state by aligning the reference coordinate system, determining the structural position of the curvature change region, calculating the effective force direction, and filtering the spatial direction. This enables the target acoustic data to be filtered around the actual force concentration position of the mechanical connector in the restricted blind zone.

[0122] S104. Project the target acoustic data onto the aligned curvature abrupt change region to locate stress feature points, and perform spatial interpolation compensation on the stress feature points to construct a spatial stress field.

[0123] Among them, the stress feature point includes its position coordinates in the reference coordinate system and the corresponding stress response value. The stress response value is used to represent the local stress intensity at that position, which can be determined by the tactile pressure value corresponding to that position and corrected by combining the acoustic energy attenuation value of the target acoustic data. Spatial interpolation compensation is used to supplement the area where there is no directly collected data between adjacent positions based on the existing discrete stress feature points. The spatial stress field is the spatial distribution result formed by the stress feature points and the compensated interpolation points, which is used to represent the stress change state in the curvature change region.

[0124] In this embodiment of the application, S104 specifically includes:

[0125] S1041. Extract the coordinates of the neighborhood points of each of the discretely distributed stress feature points, and determine the corresponding covariance features based on the neighborhood point coordinates.

[0126] Among them, the neighborhood point coordinates refer to the three-dimensional coordinates of several stress feature points around a target interpolation point in the reference coordinate system; the covariance feature is used to represent the distribution direction and dispersion of these neighborhood points in space, that is, to determine which direction these points are mainly arranged along and in which direction the change range reaches the neighborhood comparison benchmark.

[0127] In this embodiment, the detection controller first projects the target acoustic data along the effective force direction onto the aligned curvature abrupt change region. If the projected position falls within the curvature abrupt change region, and the deformation nodes near that position have corresponding tactile pressure responses, then the projected position is determined as a stress feature point. After determining the stress feature point, the detection controller determines the deformation nodes or the center of the curved mesh within the curvature abrupt change region that are not directly covered by the target acoustic data projection as target interpolation points. Then, taking the target interpolation point to be compensated as the center, stress feature points within a preset neighborhood range are selected as neighborhood points. Finally, the coordinates of these neighborhood points are read, and the covariance characteristics are determined based on the relative positional relationship between the neighborhood points. Through the covariance characteristics, the detection controller can determine whether the neighborhood stress feature points are concentrated along the assembly axis direction or expand along the contact edge direction, thereby providing a spatial directional basis for subsequent interpolation coefficient calculation.

[0128] S1042. Calculate the interpolation coefficients of the target interpolation points corresponding to each stress feature point based on the covariance characteristics.

[0129] The target interpolation point refers to the location in the curvature abrupt change region where stress data needs to be compensated; the interpolation coefficient is used to represent the degree of influence of a certain stress feature point on the target interpolation point.

[0130] In this embodiment, the interpolation coefficient of each stress feature point corresponding to the target interpolation point refers to the spatial weight of each surrounding stress feature point to the target interpolation point calculated by the detection controller. The detection controller uses the covariance feature to determine the distance influence in different directions: if the spatial relationship between a stress feature point and the target interpolation point conforms to the main distribution direction of the neighboring points, then the stress feature point obtains the interpolation coefficient with a higher ranking; if the distance between a stress feature point and the target interpolation point exceeds the neighborhood screening range, or deviates from the main distribution direction of the neighboring points, then the stress feature point does not participate in the interpolation calculation or obtains the interpolation coefficient with a lower ranking. The interpolation coefficient can be calculated using the anisotropic distance weighting method, as shown in formulas (8) and (9):

[0131] (8)

[0132] (9)

[0133] in, Indicates the target interpolation point and the first Anisotropic distance between stress characteristic points; This represents the three-dimensional coordinates of the target interpolation point in the reference coordinate system; Indicates the first The three-dimensional coordinates of each stress feature point in the reference coordinate system; It represents the covariance characteristic determined by the coordinates of neighboring points, and is used to reflect the spatial distribution direction of neighboring points; This represents the regularization coefficient, used to avoid instability in the calculation of covariance features; Represents the identity matrix; Indicates the first The interpolation coefficients of each stress feature point corresponding to the target interpolation point; This indicates the number of neighborhood stress feature points involved in the interpolation calculation; This represents the index of the neighborhood stress feature point participating in the interpolation calculation, with a value ranging from 1 to... ; Indicates the target interpolation point and the first Anisotropic distance between stress characteristic points.

[0134] Through the above processing method, the detection controller first retains the stress feature points whose spatial distance from the target interpolation point is no greater than the preset neighborhood range as the neighborhood stress feature points to participate in the interpolation calculation; then, it judges the spatial direction relationship between the neighborhood stress feature points and the target interpolation point based on the covariance characteristics. If a certain neighborhood stress feature point is close to the target interpolation point along the main distribution direction of the neighborhood points, then the neighborhood stress feature point will obtain a relatively high influence weight on the target interpolation point.

[0135] S1043. Using the interpolation coefficients, perform spatial interpolation calculations on the target interpolation points based on the stress feature points to fill data gaps and construct a spatial stress field.

[0136] Among them, data voids refer to locations within curvature abrupt change areas that are not directly covered by the target acoustic data projection and do not directly form stress feature points. These locations do not have directly measured stress feature points, but they are spatially located in continuously compressed areas and are to be compensated.

[0137] In this embodiment, the detection controller first reads the stress response values ​​of each stress feature point around the target interpolation point. These stress response values ​​are calculated from the tactile pressure values ​​at the corresponding locations and corrected using the energy attenuation values ​​of the target acoustic data at the same projection location. Then, the stress response values ​​are weighted according to the interpolation coefficients to obtain the stress estimation result of the target interpolation point. It should be noted that the covariance feature is used to calculate the interpolation coefficients, the interpolation coefficients are used to determine the contribution ratio of each stress feature point to the target interpolation point, and the stress response values ​​are the weighted objects. All three are used together to complete the spatial interpolation calculation. Finally, the original stress feature points and the compensated target interpolation points are arranged together in the curvature abrupt change region to form a spatial stress field. This spatial stress field is spatial distribution data composed of multiple spatial locations and their corresponding stress responses, which is used to subsequently construct a dynamic stress model and identify plastic deformation regions.

[0138] As an example, when the mechanical connector enters the restricted blind zone along the assembly axis, the detection controller first projects the target acoustic data along the effective force direction to the curvature abrupt change region, and determines the stress feature point when the projection position corresponds to the tactile pressure response. Next, it selects neighborhood stress feature points around the target interpolation point that needs compensation, and determines the covariance feature based on the coordinates of the neighborhood points. Then, it uses the covariance feature to calculate the interpolation coefficients of each stress feature point to the target interpolation point. Finally, using the interpolation coefficients, it weights the stress response values ​​of each stress feature point in the neighborhood of the target interpolation point to obtain the stress estimation result of the target interpolation point, and arranges the original stress feature points and the compensated target interpolation point together in the curvature abrupt change region to construct a spatial stress field. The above example is only one example of this application. In practical applications, it can also be set according to the tactile array deployment density, the mechanical connector structure, and the spatial range of the restricted blind zone; this application does not limit this.

[0139] This application projects the target acoustic data onto the curvature abrupt region and performs spatial interpolation compensation on discrete stress feature points, which enables a more continuous stress distribution within the confined blind zone, providing a basis for subsequent extraction of plastic deformation regions and defect propagation paths.

[0140] S105. Construct a dynamic stress model based on the time evolution sequence of the spatial stress field during the assembly cycle, and extract the plastic deformation region and the extended defect propagation path based on the stress gradient characteristics of the dynamic stress model.

[0141] Among them, the time evolution sequence refers to the data sequence formed by the detection controller arranging multiple spatial stress fields in chronological order within the assembly cycle, which is used to represent the stress changes of the mechanical connector during insertion, clamping, locking, and locking retention processes; the dynamic stress model is a time-series spatial data model composed of spatial stress fields at multiple time points, which is used to record the changes of stress in spatial location and time process; the plastic deformation region refers to the local area of ​​the mechanical connector that undergoes irreversible deformation within the confined blind zone; the defect propagation path refers to the continuous trajectory extending outward from the edge of the plastic deformation region, which is used to represent the propagation trend of cracks, crushing, or permanent bending.

[0142] In this embodiment of the application, S105 specifically includes:

[0143] S1051. Calculate the stress difference of the dynamic stress model at continuous time nodes based on the time difference algorithm, and perform spatial partial derivative calculation on the stress difference to extract the boundary of the region where irreversible deformation occurs.

[0144] Among them, the time difference algorithm is used to compare the stress changes at the same spatial location at adjacent time nodes; consecutive time nodes refer to adjacent time points obtained in the order of sampling time within the above assembly process; stress difference is used to represent the stress increase or decrease at this location between consecutive assembly stages; spatial partial derivative is used to represent the degree of change of stress difference along different spatial directions; region boundary is used to define the outer contour of the subsequent plastic deformation region.

[0145] In this embodiment, the detection controller first arranges multiple spatial stress fields according to the sequence of insertion, clamping, locking, and locking retention of the mechanical connector to form a dynamic stress model. Then, it performs differential comparison on the stress values ​​at adjacent time points at the same spatial location to obtain the stress difference. If the residual stress in a certain area exceeds a preset residual stress threshold after clamping or locking, and the corresponding deformation node does not return to the elastic recovery range, then the area is judged as a suspected irreversible deformation area. The preset residual stress threshold and elastic recovery range can be calibrated based on qualified assembly samples of the same type of mechanical connector, material elasticity limits, and the allowable deformation range of the assembly process. Then, it compares the stress difference changes of adjacent nodes along the surface direction of the three-dimensional deformation model, extracts the boundary position where the change is prominent, and uses this boundary as the boundary of the area where irreversible deformation occurs. The area boundary is used to describe the division between the irreversible deformation area and the surrounding areas where no irreversible deformation has occurred.

[0146] S1052. Delineate the plastic deformation region representing the plastic damage state in the dynamic stress model according to the region boundary. The plastic deformation region is located in the pressure transition part, locking contact part, or thin-walled connection part of the mechanical connector.

[0147] Among them, the pressure-bearing transition section is the part of the mechanical connector where the force-bearing surface transitions to the supporting structure; the locking contact section is the part of the mechanical connector that contacts the mating structure when it completes locking; and the thin-walled connection section is the connection part of the mechanical connector whose wall thickness is lower than that of the main support section and which is easily bent under pressure.

[0148] In this embodiment, the detection controller first delineates a continuous region within the boundary in the dynamic stress model based on the region boundary. Then, the position of this continuous region is compared with the structural partitions of the mechanical connector. If the residual stress in this continuous region exceeds a preset residual stress threshold after compression or locking, the deformation node does not return to the elastic recovery range, and this continuous region is located in a pressure transition section, a locking contact section, or a thin-walled connection section, then the detection controller identifies it as a plastic deformation region. The relationship between the plastic deformation region and the region boundary is as follows: the region boundary forms the outer contour of the plastic deformation region, and the plastic deformation region is the internal continuous region formed by the outer contour surrounding or connecting these regions.

[0149] like Figure 4 As shown, the region boundary 40 separates the plastic deformation region 41 from the surrounding non-irreversible deformation region 42. The region boundary 40 is used to represent the outer contour of the irreversible deformation region in the dynamic stress model. The plastic deformation region 41 is an internal continuous region formed by the region boundary 40 surrounding or connecting the regions. The edge nodes 43 are distributed on the region boundary 40 and can be used as the starting tracking point for subsequent defect propagation path tracing. The defect propagation direction 44 extends outward from the region boundary 40 and is used to represent the direction of continued plastic damage development. Through this correspondence, the region boundary can be connected with the plastic deformation region and the starting tracking point.

[0150] S1053. Using the edge nodes of the plastic deformation region as the starting tracking points, extract the principal stress directions of the starting tracking points in the dynamic stress model.

[0151] Among them, edge nodes are deformation nodes located on the outer contour of the plastic deformation region; starting tracking point is the position where the defect propagation path begins to be tracked; principal stress direction is used to indicate the direction in which stress mainly continues to be transmitted or propagated near the starting tracking point.

[0152] In this embodiment, the detection controller first selects edge nodes from the outer contour of the plastic deformation region and uses edge nodes with stress gradients exceeding a preset starting threshold as starting tracking points. The preset starting threshold can be calibrated based on the stress gradient distribution of edge nodes in qualified assembly samples, the stress gradient distribution of defect initiation positions in abnormal assembly samples, and the material strength requirements of mechanical connectors. Then, it reads the principal stress components in the dynamic stress model or the stress direction information obtained from assembly load calibration around the starting tracking points and uses the corresponding direction as the principal stress direction. It should be noted that the stress gradient direction is used to represent the direction of the fastest change in stress response value, and the principal stress direction is used to represent the main direction of stress action. When the dynamic stress model does not store the principal stress direction separately, the detection controller uses the stress gradient direction as an approximate direction for path tracking and maintains the same processing rules in the same batch of detection samples.

[0153] S1054. Solve the spatial displacement vector from the starting tracking point to the surrounding adjacent nodes on the three-dimensional surface of the dynamic stress model, and calculate the angular deviation between the direction of the spatial displacement vector and the direction of the principal stress.

[0154] Among them, the spatial displacement vector is used to represent the spatial direction and distance from the starting tracking point or the current tracking point to the adjacent node; the included angle deviation is used to represent the degree of deviation between the spatial displacement vector and the principal stress direction; the surrounding adjacent nodes refer to the nodes that are directly adjacent to the current tracking point on the three-dimensional curved surface mesh of the dynamic stress model.

[0155] When a curved mesh has been established, neighboring nodes can be deformable nodes that share a mesh edge with the current tracking point. When no curved mesh edge connection relationship has been established, neighboring nodes can be deformable nodes whose spatial distance from the current tracking point is no greater than a preset neighborhood radius. The preset neighborhood radius can be determined based on the spacing between adjacent tactile sensing units in the tactile array, the average edge length of the mesh in the 3D deformable model, or the sampling resolution of the restricted blind zone contact surface.

[0156] In this embodiment, the detection controller first takes the starting tracking point as the current node and selects multiple candidate nodes adjacent to the current node on the three-dimensional surface of the dynamic stress model. Each candidate node corresponds to a candidate defect extension direction, which is determined by the direction of the spatial displacement vector from the current tracking point to the candidate node. Then, the spatial displacement vector from the current node to each candidate node is calculated, and the angle between the direction of the spatial displacement vector and the principal stress direction is compared. The smaller the angle deviation, the closer the candidate defect extension direction is to the stress transmission direction. The angle relationship is shown in formula (10):

[0157] (10)

[0158] in, Indicates the current tracking point; Indicates candidate nodes around the current tracking point; Indicates starting from the current node Pointing to candidate nodes The angle between the spatial displacement vector and the direction of the principal stress; Indicates the current node Pointing to candidate nodes The spatial displacement vector; Indicates the current node The direction of the principal stress at the location; This represents the dot product of the spatial displacement vector and the principal stress direction, used to calculate the cosine of the angle between them.

[0159] S1055. Apply a nonlinear penalty to the magnitude of the spatial displacement vector using the included angle deviation to generate a path cost.

[0160] Among them, the path cost value is used to represent the selection cost of extending from the current node to a certain candidate node. The path cost value is not the defect extension direction itself, but a value used to evaluate the priority of different candidate defect extension directions. The lower the path cost value, the more suitable the candidate node is as the next node of the defect extension path. The nonlinear penalty is used to reduce the priority of candidate nodes whose direction deviation exceeds the preset allowable deviation in path selection.

[0161] In this embodiment, the detection controller generates a path cost value based on the magnitude of the spatial displacement vector from the current tracking point to the candidate node and the included angle deviation. If the candidate node shares a grid edge with the current tracking point, or the spatial distance between them is not greater than a preset neighborhood radius, the detection controller uses the candidate node as a candidate node to participate in the path cost value calculation. For candidate nodes whose spatial displacement vector magnitude is not greater than the reference length and whose included angle deviation is not greater than a preset allowable deviation, the detection controller assigns a path cost value to the node with the highest ranking. For candidate nodes whose included angle deviation is greater than a preset allowable deviation, even if they share a grid edge with the current tracking point or the spatial distance is not greater than a preset neighborhood radius, the detection controller increases their path cost value through nonlinear penalty. Thus, the direction of the spatial displacement vector is used to form the candidate defect extension direction, the principal stress direction is used to provide a judgment benchmark, and the path cost value is used to select the extension direction with the lowest path cost value among multiple candidate defect extension directions. The path cost value is shown in formula (11).

[0162] (11)

[0163] in, Indicates starting from the current node Extending to candidate nodes The path value; Indicates the current node To candidate nodes The magnitude of the spatial displacement vector between them; Indicates the reference length, used to convert the magnitude of a spatial displacement vector into a relative distance for easier comparison; This represents the nonlinear penalty coefficient, used to control the intensity of the impact of directional deviation on the path cost; It represents the cosine value of the angle between the direction of the spatial displacement vector and the direction of the principal stress. and Both are units of length; dividing them results in a dimensionless quantity. and All of these are dimensionless quantities, therefore the path cost value is a dimensionless comparison quantity.

[0164] S1056. Determine the defect extension direction according to the path cost from small to large, connect a continuous extension trajectory in the dynamic stress model, and use the continuous extension trajectory as the defect extension path to characterize the crack, crushing or permanent bending extension trend of the mechanical connector when the assembly quality is abnormal in the restricted blind zone.

[0165] Among them, the defect extension direction is the direction determined after evaluating the spatial displacement vector direction of the current tracking point to the candidate node; the extension trajectory is a path formed by connecting multiple tracking nodes in sequence.

[0166] In this embodiment, the detection controller first selects the node with the lowest path cost among the candidate nodes as the next tracking point and connects this node to the current node. Then, the next tracking point is used as the new current tracking node, and the surrounding adjacent candidate nodes are selected and the path cost is calculated. Nodes that have already been used as tracking nodes will not participate in the selection of candidate nodes for the same path again to avoid forming a loop path. The above tracking process is repeated until the stress gradient of the candidate node is lower than the preset termination threshold, or the tracking point reaches the boundary of the three-dimensional deformation model. The preset termination threshold can be set according to the distribution range of edge stress gradient in the qualified assembly sample, the mesh resolution of the three-dimensional deformation model, and the assembly quality acceptance requirements. Finally, the continuous trajectory obtained by connecting them in sequence is used as the defect expansion path.

[0167] As an example, after the mechanical connector enters the restricted blind zone and completes locking, the detection controller first arranges multiple spatial stress fields according to the assembly stages of insertion, clamping, locking, and locking retention to form a dynamic stress model. Secondly, it compares the stress changes at adjacent time nodes using a time-difference algorithm and searches for boundaries with prominent stress difference changes along the three-dimensional surface direction to extract the region boundary. Next, it delineates the continuous internal region based on the region boundary and confirms whether this continuous internal region is located in a pressure transition section, locking contact section, or thin-walled connection section to determine the plastic deformation region. Then, it selects a starting tracking point from the edge of the plastic deformation region, extracts the principal stress direction, and calculates the spatial displacement vector, included angle deviation, and path cost for the surrounding adjacent nodes. Finally, it connects the tracking nodes in ascending order of path cost to form a defect propagation path. The above example is only one example of this application. In practical applications, it can be set according to the mechanical connector structure, the accuracy of the dynamic stress model, and the assembly process requirements. This application does not limit this.

[0168] This application identifies irreversible deformation regions through a dynamic stress model and tracks the defect propagation path along the principal stress direction, thus linking local plastic damage during the assembly process with subsequent quality assessment.

[0169] S106. Determine the assembly quality status of the mechanical connector based on the area of ​​the connected domain and the spatial aggregation pattern of the plastic deformation region and the defect propagation path on the three-dimensional deformation model.

[0170] Among them, the connected domain area refers to the area of ​​continuous damaged range formed by the plastic deformation region on the surface of the three-dimensional deformation model, which is used to represent the coverage of irreversible deformation on the surface of the mechanical connector or contact structure; the spatial aggregation morphology refers to the relative distribution state of the plastic deformation region and the defect propagation path on the three-dimensional deformation model, which is used to indicate whether the defect is concentrated at the local contact position or extends outward along the pressure transition part, locking contact part or thin-walled connection part; the assembly quality status is used to represent the quality judgment result of the mechanical connector after the current assembly cycle, such as qualified status, pending re-inspection status or unqualified status.

[0171] In this embodiment of the application, S106 specifically includes:

[0172] S1061. Calculate the area of ​​the connected domain of the plastic deformation region on the three-dimensional deformation model, and obtain the geometric spatial shape of the defect propagation path.

[0173] The geometric spatial morphology includes the starting position, ending position, extension direction, bending trend of the defect propagation path, and its positional relationship relative to the mechanical connector assembly axis and structural edge.

[0174] In this embodiment, the detection controller first marks the plastic deformation region and the defect propagation path onto the three-dimensional deformation model according to the node correspondence in the reference coordinate system. Since the three-dimensional deformation model is composed of deformation nodes corresponding to the tactile sensing units and surface meshes, the detection controller can determine which deformation nodes and which surface mesh units are covered by the plastic deformation region.

[0175] Then, using whether adjacent deformation nodes share the edge of the surface mesh as the connectivity condition, the deformation nodes in the plastic deformation region are marked with connected domains: for surface mesh elements belonging to the same connected domain, their surface areas are accumulated to obtain the connected domain area of ​​the plastic deformation region on the three-dimensional deformation model; if the three-dimensional deformation model is not represented by a surface mesh, the adjacency relationship can also be determined according to the spacing of the tactile sensing elements, and the areas of the pressure elements corresponding to the adjacent tactile sensing elements are accumulated.

[0176] After obtaining the area of ​​the connected region, the spatial position of each tracking node on the 3D deformation model is read sequentially along the defect propagation path, and the structural parts traversed as the defect propagation path extends outward from the edge of the plastic deformation region are recorded. Through the above processing method, the detection controller can obtain the geometric spatial shape of the defect propagation path.

[0177] S1062. Based on the numerical range of the area of ​​the connected domain and the geometric spatial shape, determine whether the defect propagation path exhibits an outwardly divergent spatial distribution characteristic.

[0178] The numerical interval is an area range divided according to a preset area threshold, used to distinguish the coverage degree of the plastic deformation area. The numerical interval can include a first numerical interval, a second numerical interval, and a third numerical interval: the first numerical interval indicates that the area of ​​the connected domain does not exceed the first preset area threshold, and the corresponding plastic deformation area is still limited to the local contact trace range; the second numerical interval indicates that the area of ​​the connected domain exceeds the first preset area threshold but does not exceed the second preset area threshold, and the corresponding plastic deformation area has exceeded the local contact trace range but has not yet reached the unqualified judgment range; the third numerical interval indicates that the area of ​​the connected domain exceeds the second preset area threshold, and the corresponding plastic deformation area has reached the unqualified judgment range.

[0179] The outward-spreading spatial distribution characteristic refers to the distribution state of the defect propagation path starting from the edge of the plastic deformation region, extending in a direction away from the geometric center of the plastic deformation region, and gradually approaching the edge of the blind contact surface, locking edge, or thin-walled connection boundary of the mechanical connector.

[0180] In this embodiment, the detection controller first compares the area of ​​the connected region with a first preset area threshold and a second preset area threshold to determine the numerical range to which the area of ​​the connected region belongs. The first preset area threshold is less than the second preset area threshold. The first preset area threshold is used to distinguish between local contact marks and plastic deformation ranges that need to be re-inspected. The second preset area threshold is used to distinguish between re-inspectable states and unqualified states. The first preset area threshold and the second preset area threshold can be pre-calibrated based on qualified assembly samples, abnormal assembly samples, mechanical connector structure dimensions, and assembly process requirements.

[0181] During calibration, the detection controller statistically analyzes the area distribution of connected regions in qualified and abnormal assembly samples. A first preset area threshold is determined based on the upper limit of the area of ​​qualified assembly samples or a preset statistical quantile. A second preset area threshold is determined based on the lower limit of the area of ​​abnormal assembly samples, the allowable material damage range, or the upper limit of process acceptance. When the structural dimensions of the mechanical connector, material batch, or calibration parameters of the tactile array change, the detection controller re-executes the threshold calibration. When an abnormality occurs in the tactile array or acoustic emission acquisition channel, the detection controller stops outputting a qualified status and outputs a status requiring re-inspection or a prompt for re-acquisition.

[0182] Next, the geometric spatial morphology of the defect propagation path is analyzed: if multiple tracking nodes of the defect propagation path are mainly distributed near the plastic deformation area, and the path termination position is still near the pressure concentration area, then it is determined that the defect propagation path does not show an outward diverging spatial distribution characteristic; if the defect propagation path continues to extend from the edge of the plastic deformation area to the edge of the blind contact surface, the locking edge, or the boundary of the thin-walled connection, then it is determined that the defect propagation path shows an outward diverging spatial distribution characteristic.

[0183] S1063. Based on the determination results of the numerical range and the spatial distribution characteristics, output the assembly quality status of the mechanical connector according to the preset determination rules.

[0184] Among them, the determination results of spatial distribution characteristics include spatial distribution characteristics that show outward divergence and spatial distribution characteristics that do not show outward divergence; the detection controller obtains the assembly quality status of the mechanical connector by combining the numerical range of the connected domain area and the determination results of spatial distribution characteristics.

[0185] The preset judgment rules include: outputting a qualified state when there is no outward divergence and the area of ​​the connected domain belongs to the first numerical range; outputting a state pending re-inspection when there is no outward divergence and the area of ​​the connected domain belongs to the second numerical range; and outputting a unqualified state when there is outward divergence or the area of ​​the connected domain belongs to the third numerical range.

[0186] In this embodiment, the detection controller outputs the assembly quality status according to the following preset judgment rules: if the defect propagation path exhibits an outward diverging spatial distribution characteristic, the assembly quality status of the mechanical connector is judged as unqualified; if the defect propagation path does not exhibit an outward diverging spatial distribution characteristic, and the area of ​​the connected domain belongs to the first numerical range, the assembly quality status of the mechanical connector is judged as qualified; if the defect propagation path does not exhibit an outward diverging spatial distribution characteristic, and the area of ​​the connected domain belongs to the second numerical range, the assembly quality status of the mechanical connector is judged as pending re-inspection; if the area of ​​the connected domain belongs to the third numerical range, the assembly quality status of the mechanical connector is judged as unqualified.

[0187] As an example, in a detection scenario where the area is measured in square millimeters, the first preset area threshold can be set to 2 square millimeters, and the second preset area threshold can be set to 6 square millimeters. In this case, the first numerical range is for connected regions with an area not exceeding 2 square millimeters, the second numerical range is for connected regions with an area greater than 2 square millimeters but not exceeding 6 square millimeters, and the third numerical range is for connected regions with an area greater than 6 square millimeters. The above values ​​are merely examples of range divisions; in actual applications, they can be set according to the size, material, and acceptance standards of the mechanical connector. This application does not impose any limitations on this.

[0188] The aforementioned preset judgment rules indicate that whether the defect propagation path diverges outward has a priority in judgment; if the defect propagation path does not diverge outward, the qualified state, the state awaiting re-inspection, and the unqualified state are further distinguished based on the numerical range of the connected region area. The above rules are merely an example of this application, and in practical applications, they can also be set according to the material, structural dimensions, assembly process, and quality acceptance requirements of the mechanical connector. This application does not limit this.

[0189] As an example, after the mechanical connector completes locking and retention, the plastic deformation area is first marked on the three-dimensional deformation model, and the continuous damaged range is identified according to the adjacency relationship of the curved mesh to obtain the area of ​​the connected domain; then, the tracking nodes of the defect propagation path on the three-dimensional deformation model are read to determine whether the path stops near the locking contact or continues to extend to the thin-walled connection or the edge of the blind contact surface.

[0190] Then, the numerical interval of the connected region area is combined with whether the defect propagation path exhibits an outward diverging spatial distribution characteristic for judgment: if the coverage of the plastic deformation region is limited and the defect propagation path does not diverge outward, the output is a qualified state; if the plastic deformation region enters the re-inspection range but the defect propagation path does not diverge outward, the output is a state pending re-inspection; if the coverage of the plastic deformation region exceeds the unqualified judgment requirement, or the defect propagation path has diverged towards the edge of the structure, the output is an unqualified state. The above example is only one example of this application. In practical applications, it can also be adjusted according to the detection accuracy and assembly quality standards, and this application does not limit it in this regard.

[0191] This application determines the assembly quality status by combining the area of ​​the connected domain and the spatial aggregation pattern, so that the quality judgment not only considers the range of the plastic deformation area, but also the extension trend of the defect propagation path, thereby improving the correspondence between the assembly quality judgment and the actual damage state.

[0192] After determining the assembly quality state of the mechanical connector based on the area of ​​the connected domain and the spatial aggregation pattern of the defect propagation path on the three-dimensional deformation model, the method further includes: B1, when the assembly quality state is determined to be unqualified, extracting the tangent extension direction of the defect propagation path; B2, calculating the pose compensation amount of the flexible joint module in the current assembly pose based on the tangent extension direction; B3, generating a reverse motion command based on the pose compensation amount, and calling the reverse motion command to drive the flexible joint module to perform a pose adjustment action to suppress the defect propagation trend.

[0193] Among them, the tangential extension direction refers to the local extension direction of the defect propagation path near the current extension end, which is used to indicate the main direction of the defect's continued development. The tangential extension direction can be determined by multiple tracking nodes near the end of the defect propagation path. Multiple tracking nodes near the end refer to a group of consecutive nodes in the defect propagation path that are close to the end position of the path, which are used to reflect the local direction in which the defect continues to develop outward. The current assembly pose refers to the spatial position and attitude of the flexible joint module when the mechanical connector is judged to be in a non-conforming state. The pose compensation amount includes the translation and attitude adjustment amount that the flexible joint module needs to adjust under the current assembly pose, which is used to indicate how the flexible joint module should yield or deflect to reduce the risk of the mechanical connector continuing to be subjected to force along the defect propagation path.

[0194] The reverse motion command is a control command generated based on the pose compensation amount. This command includes four types of control data: motion direction, adjustment range, motion speed, and execution sequence. It is used to drive the flexible joint module to adjust its pose according to the pose compensation amount in the direction that weakens the defect propagation trend. The pose adjustment action is the yielding, deflection, or unloading action executed by the flexible joint module according to the reverse motion command. The defect propagation trend refers to the trend of continued development of cracks, crushing, or permanent bending as represented by the defect propagation path.

[0195] In this embodiment, the detection controller first reads the assembly quality status: if the assembly quality status is qualified or awaiting re-inspection, only the detection result is recorded, and no reverse motion command is generated; if the assembly quality status is unqualified, the tracking nodes of the defect propagation path on the three-dimensional deformation model are read; then, continuous tracking nodes are selected from the end of the defect propagation path, and local linear fitting is performed on these tracking nodes to obtain the extension trend of the defect propagation path near the end; the extension trend is adjusted to the direction from the plastic deformation region to the path termination position, and it is used as the tangent extension direction of the defect propagation path; if the number of tracking nodes contained in the defect propagation path is less than the preset number of fitting nodes, the direction of the line connecting the last two tracking nodes is directly used as the tangent extension direction.

[0196] Next, the tangent extension direction is transformed from the reference coordinate system to the motion control coordinate system of the flexible joint module, so that the defect expansion direction corresponds to the movable direction of the flexible joint module. The opposite direction of the tangent extension direction is taken as the unloading direction, which indicates that the flexible joint module should preferentially move away from the direction in which the defect continues to expand, thereby reducing the abnormal compression of the mechanical connector in the restricted blind zone. Then, combined with the current assembly posture, assembly axis direction, adjustable degrees of freedom of the flexible joint module and preset posture adjustment limit, the translation compensation amount and attitude compensation amount are calculated. The translation compensation amount is used to control the flexible joint module to yield along the unloading direction, and the attitude compensation amount is used to control the flexible joint module to deflect around the adjustable joint axis.

[0197] Then, the inverse kinematics calculation method commonly used in robot motion control is used to convert the translation compensation amount and posture compensation amount into the pose compensation amount that the flexible joint module can execute. If the calculated pose compensation amount exceeds the preset pose adjustment limit, the pose compensation amount is limited to the preset pose adjustment limit and the pressing action is stopped to avoid the adjustment action introducing new assembly damage.

[0198] In another specific implementation, if the flexible joint module is only allowed to retract along the assembly axis, the detection controller can use the reverse projection of the tangent extension direction on the assembly axis as the main source of pose compensation; if the flexible joint module allows multi-degree-of-freedom adjustment, the detection controller can simultaneously calculate the axial retraction amount and the attitude deflection amount, so that the pose compensation amount matches the actual assembly space within the restricted blind zone.

[0199] Then, the pose compensation amount is converted into control commands that the flexible joint module can execute: the detection controller writes the translation compensation amount into the corresponding linear motion control item and the pose compensation amount into the corresponding rotation control item according to the motion interface format of the flexible joint module, and configures speed limits and stroke limits for the reverse motion command to avoid the pose adjustment action causing new impact on the mechanical connector; then the detection controller calls the reverse motion command to drive the flexible joint module to perform the pose adjustment action, which may include retraction along the unloading direction, deflection around the joint axis, or stopping the current clamping action.

[0200] After the pose adjustment action is performed, the acoustic emission signal and tactile array signal can be re-acquired to confirm whether the plastic deformation area and defect propagation path continue to expand. If the re-detection result shows that the defect propagation trend has weakened or stopped, the execution result of the pose adjustment action is recorded. If the defect propagation trend still exists, the unqualified state can be maintained and assembly can be stopped.

[0201] As an example, after the mechanical connector completes locking and holding, the detection controller determines the assembly quality status as unqualified: First, it reads the tracking nodes near the end of the defect propagation path and performs local straight-line fitting on these tracking nodes to obtain the tangent extension direction of the defect propagation path; second, it transforms the tangent extension direction into the motion control coordinate system of the flexible joint module, and uses the reverse of the tangent extension direction as the unloading direction; then, it calculates the pose compensation amount by combining the current assembly pose and the adjustable degrees of freedom of the flexible joint module; then, it generates a reverse motion command based on the pose compensation amount and drives the flexible joint module to perform a yielding or deflection action to reduce the risk of the mechanical connector continuing to be subjected to force along the defect propagation path. The above example is only one example of this application. In practical applications, it can also be set according to the structure of the flexible joint module, the control interface, and the assembly safety requirements. This application does not limit this.

[0202] This application converts the defect propagation path into a pose compensation amount and further generates a reverse motion command, enabling the detection results to be directly used for unloading and avoidance control during the assembly process, thereby suppressing the defect propagation trend.

[0203] This application acquires and fuses acoustic emission signals and tactile array signals through S101-S106 to construct a three-dimensional sound field model, a three-dimensional deformation model, a spatial stress field, and a dynamic stress model. This enables continuous identification of the stress, plastic deformation area, and defect propagation path of mechanical connector assembly within a confined blind zone. Furthermore, it outputs the assembly quality status based on the area of ​​the connected domain and the spatial aggregation pattern, and generates a reverse motion command to perform pose adjustment when the assembly is unqualified, thereby improving the accuracy of blind zone assembly quality detection and the ability to suppress defect propagation.

[0204] The above provides a detailed description of a mechanical connector assembly quality inspection method provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for inspecting the assembly quality of mechanical connectors, applied to the blind zone of flexible joint modules in humanoid robots, characterized in that, include: Acquire acoustic emission signals and tactile array signals from the mechanical connector during the assembly cycle; The spatial attenuation characteristics of the acoustic emission signal are mapped to a three-dimensional sound field model, and a three-dimensional deformation model is constructed based on the tactile array signal. The curvature change region on the three-dimensional deformation model is divided according to the three-dimensional curvature distribution of the three-dimensional deformation model. Align the spatial positions of the three-dimensional sound field model and the three-dimensional deformation model in the reference coordinate system to determine the effective force direction corresponding to the curvature change region, and perform spatial direction filtering on the three-dimensional sound field model based on the effective force direction to obtain the target acoustic data; The target acoustic data is projected onto the aligned curvature abrupt change region to locate stress feature points, and spatial interpolation compensation is performed on the stress feature points to construct a spatial stress field. A dynamic stress model is constructed based on the time evolution sequence of the spatial stress field during the assembly cycle, and the plastic deformation region and the extended defect propagation path are extracted based on the stress gradient characteristics of the dynamic stress model. The assembly quality status of the mechanical connector is determined based on the area of ​​the connected domain and the spatial aggregation pattern of the plastic deformation region and the defect propagation path on the three-dimensional deformation model.

2. The method according to claim 1, characterized in that, Aligning the spatial positions of the three-dimensional sound field model and the three-dimensional deformation model in the reference coordinate system to determine the effective force direction corresponding to the curvature abrupt change region includes: A reference coordinate system is established with the joint rotation center of the flexible joint module as the reference origin. The assembly axis of the mechanical connector in the flexible joint module is used as the direction reference. The sound field node coordinates of the three-dimensional sound field model and the deformation node coordinates of the three-dimensional deformation model are uniformly mapped to the reference coordinate system. In the reference coordinate system, the local tangent planes of each deformation node in the curvature change region are solved, and the local normal vectors perpendicular to the local tangent planes are extracted; Calculate the three-dimensional curvature gradient of each of the deformation nodes, and determine the direction weighting coefficient of the corresponding local normal vector based on the magnitude of the three-dimensional curvature gradient. The directional weighting coefficient is used to perform vector weighting fusion on all the local normal vectors to generate a fused directional vector, and the fused directional vector is used as the effective force direction corresponding to the curvature change region. The effective force direction represents the main local force direction of the mechanical connector during the assembly process in the restricted blind zone.

3. The method according to claim 2, characterized in that, The calculation of the three-dimensional curvature gradient of each of the deformed nodes, and the determination of the direction weighting coefficient of the corresponding local normal vector based on the magnitude of the three-dimensional curvature gradient, includes: Extract the maximum and minimum principal curvatures of each deformation node on different cross sections; The absolute difference between the maximum principal curvature and the minimum principal curvature is used as the three-dimensional curvature gradient. Determine the geometric center of the curvature abrupt change region in three-dimensional space, and calculate the spatial distance from each deformation node to the geometric center of the region; A spatial penalty coefficient is constructed based on the spatial distance, and the spatial penalty coefficient is used to perform a product calculation on the three-dimensional curvature gradient to obtain the joint gradient feature; The joint gradient features are numerically transformed using a preset exponential decay function to calculate the basic weight values ​​that characterize the degree of local deformation. Calculate the weight ratio of the basic weight value of each deformation node in the curvature change region, and use the weight ratio as the direction weighting coefficient of the corresponding local normal vector.

4. The method according to claim 1, characterized in that, The process of performing spatial directional filtering on the three-dimensional sound field model based on the effective force direction to obtain target acoustic data includes: Extract the propagation direction vector and acoustic energy gradient of each spatial data point in the three-dimensional sound field model; Calculate the directional deviation angle between the propagation direction vector and the effective force direction; Based on the magnitude of the acoustic energy gradient, an adaptive dynamic angle threshold is matched for the spatial data points; Spatial data points whose directional deviation angle is greater than the dynamic angle threshold are identified as noise interference points and removed. The remaining spatial data points are used as target acoustic data to characterize the acoustic response features of the mechanical connector in the effective force direction.

5. The method according to claim 1, characterized in that, The step of extracting the plastic deformation region and the extended defect propagation path based on the stress gradient characteristics of the dynamic stress model includes: The stress difference of the dynamic stress model at continuous time nodes is calculated based on the time difference algorithm, and the spatial partial derivative is solved on the stress difference to extract the boundary of the region where irreversible deformation occurs. According to the boundary of the region, a plastic deformation region representing the plastic damage state is delineated in the dynamic stress model. The plastic deformation region is located in the pressure transition part, locking contact part or thin-walled connection part of the mechanical connector. Using the edge nodes of the plastic deformation region as the starting tracking points, the principal stress directions of the starting tracking points in the dynamic stress model are extracted; Solve the spatial displacement vector from the starting tracking point to the surrounding adjacent nodes on the three-dimensional surface of the dynamic stress model, and calculate the angular deviation between the direction of the spatial displacement vector and the direction of the principal stress. The magnitude of the spatial displacement vector is nonlinearly penalized using the included angle deviation to generate the path cost. The defect extension direction is determined according to the path cost from small to large. A continuous extension trajectory is connected in the dynamic stress model. The continuous extension trajectory is used as the defect extension path to characterize the crack, crushing or permanent bending propagation trend of the mechanical connector when the assembly quality is abnormal in the confined blind zone.

6. The method according to claim 1, characterized in that, The step of performing spatial interpolation compensation on the stress feature points to construct a spatial stress field includes: Extract the coordinates of the neighborhood points of each of the discretely distributed stress feature points, and determine the corresponding covariance features based on the neighborhood point coordinates; Calculate the interpolation coefficients of the target interpolation points corresponding to each stress feature point based on the covariance characteristics; Using the interpolation coefficients, spatial interpolation calculations are performed on the target interpolation points based on the stress feature points to fill data gaps and construct a spatial stress field.

7. The method according to claim 1, characterized in that, The step of mapping the spatial attenuation characteristics of the acoustic emission signal to a three-dimensional sound field model includes: Obtain the time difference between the arrival of the acoustic emission signal at different acquisition channels; By combining the receiving time difference and the energy attenuation value of the acoustic emission signal, the sound source energy is located in three-dimensional space, and the sound source coordinates are calculated. A three-dimensional sound field model reflecting the energy attenuation distribution characteristics is constructed based on the sound source location coordinates and the corresponding energy attenuation values.

8. The method according to claim 1, characterized in that, The construction of a three-dimensional deformation model based on the tactile array signals, and the division of curvature abrupt change regions on the three-dimensional deformation model according to the three-dimensional curvature distribution of the three-dimensional deformation model, includes: A three-dimensional deformation model is constructed based on the spatial coordinates of each tactile sensing unit in the tactile array signal and the tactile pressure value. Calculate the local surface curvature of each tactile sensing unit on the three-dimensional deformation model to obtain the three-dimensional curvature distribution; Based on the three-dimensional curvature distribution, stress concentration areas in the three-dimensional deformation model where the surface curvature value is higher than a preset curvature change threshold are extracted; The extracted stress concentration areas are used as curvature abrupt change regions on the three-dimensional deformation model.

9. The method according to claim 1, characterized in that, The process of determining the assembly quality state of the mechanical connector based on the area of ​​the connected domains and the spatial aggregation pattern on the three-dimensional deformation model according to the plastic deformation region and the defect propagation path includes: Calculate the area of ​​the connected domain of the plastic deformation region on the three-dimensional deformation model, and obtain the geometric spatial morphology of the defect propagation path; Based on the numerical range of the area of ​​the connected domain and the geometric spatial shape, it is determined whether the defect propagation path exhibits an outwardly divergent spatial distribution characteristic. Based on the determination results of the numerical range and the spatial distribution characteristics, the assembly quality status of the mechanical connector is output according to the preset determination rules.

10. The method according to claim 1, characterized in that, After determining the assembly quality state of the mechanical connector based on the area of ​​the connected domains and the spatial aggregation pattern on the three-dimensional deformation model according to the plastic deformation region and the defect propagation path, the method further includes: When the assembly quality is determined to be unqualified, the tangent extension direction of the defect propagation path is extracted; Calculate the pose compensation amount of the flexible joint module in the current assembly pose based on the tangent extension direction. Based on the pose compensation amount, a reverse motion command is generated, and the reverse motion command is invoked to drive the flexible joint module to perform pose adjustment actions in order to suppress the trend of defect expansion.

Images