Intelligent docking and adaptive tightening control method and device for flange of loading arm based on multi-sensor fusion

Abstract

The application provides a method and device for intelligent docking and adaptive tightening of a flange based on multi-sensor fusion. Flange hole position information is extracted from a flange image collected by a two-dimensional vision sensor to control a robot arm to perform initial coarse positioning of the flange. Contact force feedback information collected by a six-dimensional moment sensor mounted at the end of the robot arm is obtained in the contact stage, and a mobility control model is established and the docking pose of the robot arm is adjusted to complete intelligent docking. In the bolt tightening process, the tightening speed is dynamically adjusted using torque-angle correlation data feature points of the tightening shaft. The application improves the flexibility and safety of flange docking through multi-sensor fusion of vision and force sensation, and realizes adaptive regulation and control of the tightening process, effectively avoiding problems such as docking jamming and uneven pre-tightening force.

Description

Technical Field

[0001] This application relates to the field of automated loading and unloading control technology, and more specifically, to a method and device for intelligent docking and adaptive tightening control of loading arm flanges based on multi-sensor fusion. Background Technology

[0002] In the transportation of liquid media such as petroleum and chemicals, loading arms are key equipment connecting storage tanks and transport vehicles. The level of automation in their flange connections directly affects loading and unloading efficiency and operational safety. With the development of automation technology, using robotic arms to replace manual flange connections and bolt tightening has become an industry trend.

[0003] Existing flange docking solutions typically employ machine vision-based positioning and control methods. A camera mounted at the end of a robotic arm identifies the flange's hole positions and guides the arm's movement. This approach first uses visual algorithms to identify the center coordinates and hole distribution of the target flange; then, it calculates the robotic arm's trajectory based on the identification results; finally, it drives the robotic arm to move the mating joint to the target position to complete the docking.

[0004] However, this vision-based positioning solution has significant technical drawbacks in practical applications. Due to the inherent accuracy error in visual recognition and the rigid positioning of the robotic arm during docking, even slight deviations in the flange hole position or unevenness of the sealing surface can easily generate enormous instantaneous contact forces during the contact phase, leading to flange jamming or robotic arm overload alarms. Furthermore, existing bolt tightening processes often employ a constant speed mode, failing to dynamically adjust according to actual load changes during tightening (such as thread wear or foreign object resistance). This can easily result in low tightening efficiency or uneven preload distribution, making it difficult to meet the stringent sealing and safety requirements of loading arm flanges. Summary of the Invention

[0005] This application provides a method and device for intelligent docking and adaptive tightening control of loading arm flanges based on multi-sensor fusion, so as to at least alleviate the above-mentioned technical problems.

[0006] A method for intelligent docking and adaptive tightening control of loading arm flanges based on multi-sensor fusion includes: The system acquires flange images from a 2D vision sensor, extracts flange hole position information from the flange images, and controls a robotic arm to perform initial coarse positioning of the flange based on the flange hole position information. During the contact phase after the initial coarse positioning is completed, contact force feedback information is acquired by a six-dimensional torque sensor installed at the end of the robotic arm. An admittance control model is established based on the contact force feedback information, and the docking posture of the robotic arm is adjusted based on the admittance control model to complete the intelligent docking of the flange. During the bolt tightening process after intelligent docking, the torque and angle correlation data of the tightening shaft are collected in real time. The tightening speed is dynamically adjusted according to the feature points of the torque and angle correlation data to achieve adaptive tightening control.

[0007] Optionally, the steps of extracting flange hole position information from the flange image and controlling the robotic arm to perform initial coarse positioning of the flange based on the flange hole position information include: Feature contours are extracted from the flange image to obtain the initial contour data of the target flange; Identify the target flange hole on the target flange based on the initial contour data, and extract the two-dimensional coordinate data of the target flange hole as flange hole position information; Convert two-dimensional coordinate data into three-dimensional spatial coordinate data in the base coordinate system of the robotic arm; An initial movement trajectory is generated based on the three-dimensional coordinate data in space. The robotic arm is then controlled to move along the initial movement trajectory to the preset docking preparation position to complete the initial coarse positioning of the flange.

[0008] Optionally, the steps for obtaining contact force feedback information collected by a six-dimensional torque sensor installed at the end of the robotic arm include: During the docking operation, the raw force signals output by the six-dimensional torque sensor are periodically read. The original force sensor signal is smoothed to obtain a smoothed force sensor signal. The smoothed force sensor signal is then transformed into a coordinate system according to a preset coordinate system mapping rule to generate three-dimensional force data and three-dimensional torque data in the coordinate system of the robotic arm end effector. The three-dimensional force data and three-dimensional torque data are combined into contact force feedback information.

[0009] Optionally, the steps for establishing an admittance control model based on contact force feedback information include: Obtain data on the ideal stress state of the flange during the docking process; Calculate the force deviation between the contact force feedback information and the ideal force state data; Obtain the preset mass parameters, preset damping parameters, and preset stiffness parameters of the robotic arm; A multi-dimensional compliance mapping relationship is constructed based on preset mass parameters, preset damping parameters, preset stiffness parameters, and force deviation values, and this multi-dimensional compliance mapping relationship is used as the admittance control model.

[0010] Optionally, the steps for adjusting the docking pose of the robotic arm based on the admittance control model to complete the intelligent docking of the flange include: The force deviation value is input into the admittance control model so that the robotic arm exhibits the compliant physical characteristics of spring damping, and the pose compensation amount of the robotic arm is output. Obtain the current docking pose of the robotic arm, correct the current docking pose according to the pose compensation amount, and obtain the target docking pose; Control the robotic arm to move to the target docking position.

[0011] Optionally, after the step of controlling the robotic arm to move to the target docking pose, the method further includes: Real-time monitoring of new contact force feedback information after moving to the target docking position; If the new contact force feedback information is greater than the preset collision threshold, it is determined that there is a risk of flange jamming. The robotic arm is controlled to retreat according to the preset retraction trajectory and the docking posture is readjusted. If the new contact force feedback information is less than the preset fitting threshold, the intelligent docking of the flange is determined to be complete.

[0012] Optionally, during the bolt tightening process after intelligent docking, the steps for real-time acquisition of torque and angle correlation data of the tightening shaft include: Control the tightening shaft to perform the tightening action on the target bolt; During the tightening action, the real-time output torque of the tightening shaft is collected by a torque sensor; The real-time rotation angle of the tightening shaft is acquired by an angle encoder; Match and bind the real-time output torque and real-time rotation angle at the same time point to generate torque-angle correlation data.

[0013] Optionally, the step of dynamically adjusting the tightening speed of the tightening shaft based on the feature points of the torque-angle correlation data includes: A real-time monitoring curve of torque versus rotation angle is generated based on torque-angle correlation data; The slope changes of the real-time monitoring curve are analyzed to identify the contact point and yield point on the real-time monitoring curve, and the contact point and yield point are used as feature points. The entire bolt tightening process is divided into multiple different tightening stages based on the contact point and yield point, and a corresponding target tightening speed is matched for each tightening stage.

[0014] Optionally, the process of dividing the entire bolt tightening process into multiple different tightening stages based on the contact point and yield point, and matching a corresponding target tightening speed for each tightening stage, includes the following steps: In the first tightening stage before reaching the contact point, the tightening speed of the tightening shaft is set to a preset high-speed value as the target tightening speed; In the second tightening stage, as the tightening point passes the contact point and approaches the yield point, the tightening speed of the tightening shaft is gradually reduced so that the target tightening speed in the second tightening stage is less than the preset high speed value. In the third tightening stage when the yield point is reached, the tightening shaft is controlled to stop rotating in order to achieve a uniform distribution of preload on each target bolt.

[0015] A smart docking and adaptive tightening control device for loading arm flanges based on multi-sensor fusion includes: The visual positioning and initial coarse adjustment module is used to acquire flange images collected by a two-dimensional vision sensor, extract flange hole position information based on the flange images, and control the robotic arm to perform initial coarse positioning of the flange based on the flange hole position information. The force-sensing docking and admittance control module is used to acquire contact force feedback information collected by the six-dimensional torque sensor installed at the end of the robotic arm during the contact phase after the initial coarse positioning is completed. The bolt adaptive tightening control module is used to establish an admittance control model based on contact force feedback information, and adjust the docking posture of the robotic arm based on the admittance control model to complete the intelligent docking of the flange. During the bolt tightening process after the intelligent docking is completed, the torque-angle correlation data of the tightening shaft is collected in real time, and the tightening speed is dynamically adjusted according to the feature points of the torque-angle correlation data to achieve adaptive tightening control.

[0016] Optionally, the visual positioning and initial coarse adjustment module is also used for: Feature contour extraction is performed on the flange image to obtain the initial contour data of the target flange; the target flange hole on the target flange is identified based on the initial contour data, and the two-dimensional coordinate data of the target flange hole is extracted as the flange hole position information; the two-dimensional coordinate data is converted into spatial three-dimensional coordinate data in the base coordinate system of the robotic arm; an initial movement trajectory is generated based on the spatial three-dimensional coordinate data, and the robotic arm is controlled to move to the preset docking preparation position according to the initial movement trajectory to complete the initial coarse positioning of the flange.

[0017] Optionally, the force-sensing docking and admittance control module is also used for: During the docking action of the robotic arm, the raw force signals output by the six-dimensional torque sensor are periodically read; the raw force signals are smoothed to obtain smoothed force signals, and the smoothed force signals are transformed into coordinate systems according to the preset coordinate system mapping rules to generate three-dimensional force data and three-dimensional torque data in the coordinate system of the robotic arm end effector; the three-dimensional force data and three-dimensional torque data are combined into contact force feedback information.

[0018] Optionally, the force-sensing docking and admittance control module is also used for: Obtain ideal stress state data of the flange during the docking process; calculate the stress deviation value between the contact force feedback information and the ideal stress state data; obtain the preset mass parameters, preset damping parameters and preset stiffness parameters of the robotic arm; construct a multi-dimensional compliance mapping relationship as an admittance control model based on the preset mass parameters, preset damping parameters, preset stiffness parameters and stress deviation value.

[0019] Optionally, the force-sensing docking and admittance control module is also used for: The force deviation value is input into the admittance control model to make the robotic arm exhibit the compliant physical characteristics of spring damping, and the pose compensation amount of the robotic arm is output; the current docking pose of the robotic arm is obtained, and the current docking pose is corrected according to the pose compensation amount to obtain the target docking pose; the robotic arm is controlled to move to the target docking pose.

[0020] Optionally, the force-sensing docking and admittance control module is also used for: Real-time monitoring of new contact force feedback information after moving to the target docking position; if the new contact force feedback information is greater than the preset collision threshold, it is determined that there is a risk of flange jamming, and the robotic arm is controlled to retreat according to the preset retraction trajectory and the docking position is readjusted; if the new contact force feedback information is less than the preset fitting threshold, it is determined that the intelligent docking of the flange is completed.

[0021] Optionally, the bolt adaptive tightening control module is also used for: The tightening shaft is controlled to perform the tightening action on the target bolt. During the tightening action, the real-time output torque of the tightening shaft is collected by a torque sensor. The real-time rotation angle of the tightening shaft is collected by an angle encoder. The real-time output torque and real-time rotation angle at the same time point are matched and bound to generate torque-angle correlation data.

[0022] Optionally, the bolt adaptive tightening control module is also used for: Based on torque-angle correlation data, a real-time monitoring curve of torque changing with rotation angle is generated; the slope change of the real-time monitoring curve is analyzed, and the contact point and yield point on the real-time monitoring curve are identified as feature points; the entire bolt tightening process is divided into multiple different tightening stages according to the contact point and yield point, and a corresponding target tightening speed is matched for each tightening stage.

[0023] Optionally, the bolt adaptive tightening control module is also used for: In the first tightening stage before reaching the contact point, the tightening speed of the tightening shaft is set to a preset high-speed value as the target tightening speed. In the second tightening stage, after passing the contact point and approaching the yield point, the tightening speed of the tightening shaft is gradually reduced so that the target tightening speed of the second tightening stage is less than the preset high-speed value. In the third tightening stage when the yield point is reached, the tightening shaft is controlled to stop rotating so as to achieve a uniform distribution of preload on each target bolt.

[0024] The technical advantages of the technical solution provided in this application are: This application provides a multi-sensor fusion-based intelligent docking and adaptive tightening control method for loading arm flanges. It addresses the technical shortcomings of traditional vision-based flange docking schemes, such as rigid positioning leading to jamming, high collision risk, and low adaptability to fixed speed tightening. By introducing a multi-sensor fusion control mechanism, it effectively improves the smoothness and intelligence of the operation.

[0025] First, this application solves the problem of hard collisions and jamming caused by accuracy deviations during the contact phase in traditional vision-based positioning schemes by establishing an admittance control model and adjusting the pose based on contact force feedback information from a six-dimensional torque sensor. Compared to the "rigid" motion of traditional schemes that rely solely on visual guidance, the admittance control model endows the robotic arm with spring-like damping physical characteristics, enabling it to adaptively correct its pose based on force deviations. This compliant docking method allows the robotic arm to dynamically compensate for hole position errors during the docking process, significantly reducing instantaneous contact forces and improving the success rate of flange docking and equipment safety.

[0026] Secondly, this application solves the problem that traditional fixed-speed-ratio tightening schemes cannot cope with complex load changes by dynamically adjusting the tightening speed through real-time acquisition of torque and angle correlation data of the tightening shaft and identification of feature points. Traditional schemes, due to their constant speed, are prone to significant impacts during the high-load phase after the bolt contact point, and it is difficult to accurately control the preload. However, the contact point and yield point identified in this application can accurately divide the tightening stage, achieving rapid screwing in before contact and precise tightening by slowing down after contact. This not only improves overall work efficiency but also ensures the uniformity and reliability of the bolt preload, better meeting the stringent sealing quality requirements of loading arm flanges. Attached Figure Description

[0027] Figure 1 This is a schematic diagram illustrating an application scenario of an intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion, as described in an embodiment of this application.

[0028] Figure 2 This application provides an embodiment of an intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion.

[0029] Figure 3 This application provides an embodiment of an intelligent docking and adaptive tightening control device for loading arm flanges based on multi-sensor fusion.

[0030] Figure 4 This is an electronic device according to an embodiment of the present application. Detailed Implementation

[0031] like Figure 1 The image shows an application scenario of the intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion, as described in this application embodiment. Figure 2As shown in the figure, an intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion is provided in an embodiment of this application, which includes the following steps: The flange image is acquired by a two-dimensional vision sensor, the flange hole position information is extracted from the flange image, and the robotic arm is controlled to perform initial coarse positioning of the flange based on the flange hole position information. During the contact phase after the initial coarse positioning is completed, contact force feedback information is acquired by a six-dimensional torque sensor installed at the end of the robotic arm. An admittance control model is established based on the contact force feedback information, and the docking posture of the robotic arm is adjusted based on the admittance control model to complete the intelligent docking of the flange. During the bolt tightening process after the intelligent docking is completed, the torque-angle correlation data of the tightening shaft is collected in real time. The tightening speed of the tightening shaft is dynamically adjusted according to the feature points of the torque-angle correlation data to achieve adaptive tightening control.

[0032] Optionally, the step of extracting flange hole position information from the flange image and controlling the robotic arm to perform initial coarse positioning of the flange based on the flange hole position information includes: Feature contour extraction is performed on the flange image to obtain the initial contour data of the target flange; The target flange hole on the target flange is identified based on the initial contour data, and the two-dimensional coordinate data of the target flange hole is extracted as the flange hole position information; The two-dimensional coordinate data is converted into three-dimensional spatial coordinate data in the base coordinate system of the robotic arm; An initial movement trajectory is generated based on the spatial three-dimensional coordinate data, and the robotic arm is controlled to move to the preset docking preparation position according to the initial movement trajectory to complete the initial coarse positioning of the flange.

[0033] Preferably, in the specific technical implementation of extracting flange hole position information based on the flange image, the flange image comes from a two-dimensional vision sensor configured at the end of the robotic arm. Before the robotic arm sends the flange connection part into the flange mating position, the two-dimensional vision sensor performs a frontal imaging of the target flange in the clamping state of the customized tooling platform to form a flange image including the outer circle boundary of the flange, the center through hole boundary, the bolt mounting hole boundary, and the tooling obstruction boundary. The flange connection part includes the flange connection part of an elbow flange and a straight pipe assembly flange. The target flange includes the flange end face that needs to be mated with the flange connection part from the flange face of the rotary joint, the mating face of the elbow flange, or the mating face of the straight pipe assembly flange connection part. The flange image is not simply used as a planar recognition image, but as the geometric source for the subsequent generation of the initial movement trajectory by the robotic arm. Therefore, when acquiring the flange image, the effective image area of ​​the flange in the flange image is first determined based on the clamping position of the customized tooling platform, the assembly plane where the target flange is located, and the imaging direction of the two-dimensional vision sensor. Then, the effective image area of ​​the flange is used as the input content for feature contour extraction, so that the initial contour data obtained later can correspond to the actual bolt mounting hole distribution of the target flange, rather than to the non-interlocking interference boundary formed by the material rack, customized tooling platform, elbow edge, or rotary joint shell.

[0034] Preferably, in the specific technical implementation of feature contour extraction of the flange image, grayscale equalization, edge continuity filtering, and tooling background removal are first performed on the effective image area of ​​the flange to form a flange contour enhancement image. Specifically, grayscale equalization is used to reduce the influence of stainless steel flange surface reflection on the hole wall edge; edge continuity filtering is used to retain the flange outer circle boundary and bolt mounting hole boundary continuously distributed along the circumferential direction; and tooling background removal is used to exclude tooling occlusion boundaries and non-docking interference boundaries formed by the clamping cylinder, jaws, and positioning surfaces of the customized tooling platform in the flange image. Subsequently, the brightness abrupt change boundaries in the flange contour enhancement image are judged for closure. Brightness abrupt change boundaries that can enclose the flange outer circle, central through hole, and bolt mounting hole are selected as candidate flange contours. These candidate flange contours are then organized according to their positional relationship in the flange contour enhancement image to obtain the initial contour data of the target flange. Thus, the initial contour data originates from the flange outer circle boundary, central through hole boundary, and bolt mounting hole boundary in the flange image and continues to serve as the basis for identifying the target flange hole.

[0035] Preferably, when identifying the target flange hole on the target flange based on the initial contour data, not all approximately circular boundaries in the initial contour data are directly identified as target flange holes. Instead, the flange center reference position of the target flange is first determined based on the flange outer circle boundary in the initial contour data, and then the annular search area where the bolt mounting holes should be distributed is determined based on the flange center reference position. Subsequently, candidate flange contours located within the annular search area, with closed hole wall boundaries and adjacent hole spacing conforming to the target flange mounting hole layout, are identified as target flange holes. The boundary of the center through hole located at the flange center, the tooling obstruction boundary located in the contact area of ​​the customized tooling platform, and the discontinuous boundary formed by weld reinforcement reflection are excluded from the target flange holes. For structural schemes where the target flange mounting hole layout is adjusted from more holes to fewer holes in the automatic loading arm assembly scenario (e.g., from 12 mounting holes to 8 mounting holes), the annular search area is also verified for the number of holes and circumferential spacing according to the adjusted target flange mounting hole layout to ensure that the target flange hole is consistent with the bolt mounting hole alignment requirements in the subsequent automatic bolt assembly process.

[0036] Preferably, when extracting the two-dimensional coordinate data of the target flange holes, edge point aggregation is performed on the hole wall boundary of each target flange hole, and the image center position of the target flange hole in the flange image is obtained based on the aggregated hole wall boundary. The image center positions of multiple target flange holes are arranged in circumferential order around the flange center reference position to form the two-dimensional coordinate data. The two-dimensional coordinate data includes not only the image center position of each target flange hole in the flange image, but also the circumferential arrangement relationship between multiple target flange holes. The circumferential arrangement relationship continues to participate in the generation of subsequent three-dimensional spatial coordinate data, so that when the robotic arm performs the initial coarse positioning of the flange, it can not only align with the spatial position of the flange center, but also with the circumferential posture of the flange, thereby reducing the probability of bolt installation hole deviation when the bolts enter the bolt installation holes.

[0037] Preferably, the flange hole position information is formed based on the two-dimensional coordinate data, and the flange hole position information is also combined with the flange outer circle boundary and the center through hole boundary in the initial contour data for consistency verification. Specifically, the flange center reference position obtained from the flange outer circle boundary is compared with the target flange hole surrounding center obtained from multiple target flange holes. If the deviation between the flange center reference position and the target flange hole surrounding center in the flange image is within the allowable image recognition range, then the two-dimensional coordinate data is used as the flange hole position information; if the deviation between the flange center reference position and the target flange hole surrounding center in the flange image exceeds the allowable image recognition range, then the tooling background removal and edge continuity filtering are re-performed on the effective image area of ​​the flange to obtain the initial contour data again, and the target flange hole is identified again based on the re-obtained initial contour data. Through the above processing, the flange hole position information can complete image-level self-verification before entering the base coordinate system transformation of the robotic arm, reducing misreading of target flange holes caused by local reflections, occlusion by the customized tooling platform, or interference from weld edges.

[0038] Preferably, when converting the two-dimensional coordinate data into spatial three-dimensional coordinate data in the base coordinate system of the robotic arm, the hand-eye calibration relationship between the two-dimensional vision sensor and the robotic arm end effector, the current posture record of the robotic arm, and the assembly plane constraint corresponding to the assembly plane where the target flange is located are first read. The hand-eye calibration relationship expresses the positional correspondence between the imaging plane of the two-dimensional vision sensor and the robotic arm end effector; the current posture record of the robotic arm expresses the position and orientation of the robotic arm end effector in the base coordinate system of the robotic arm; and the assembly plane constraint expresses the spatial position of the target flange end face in the customized tooling platform, rotary joint flange face, elbow flange mating face, or straight pipe assembly flange connection mating face. Subsequently, the image center position of each target flange hole in the two-dimensional coordinate data is projected onto the assembly plane constraint along the imaging direction of the two-dimensional vision sensor to form the projected center position of each target flange hole. Then, combining the hand-eye calibration relationship and the current posture record of the robotic arm, the projected center position of each target flange hole is converted to the base coordinate system of the robotic arm to form spatial three-dimensional coordinate data. Since the spatial three-dimensional coordinate data is jointly determined by the two-dimensional coordinate data, the hand-eye calibration relationship, the current posture record of the robotic arm, and the assembly plane constraints, it is possible to explain how the hole positions in the two-dimensional image become the basis for the spatial positioning of the robotic arm.

[0039] Preferably, the spatial three-dimensional coordinate data includes the spatial positions of the center holes of multiple target flange holes in the base coordinate system of the robotic arm, the spatial position of the flange center determined by the multiple center hole positions, and the circumferential posture of the flange determined by the circumferential arrangement of the multiple center hole positions. The center hole positions are used to express the positions of the target flange holes in the actual assembly space, the flange center position is used to express the translation reference when the robotic arm carries the flange connection part close to the flange mating point, and the flange circumferential posture is used to express the angular distribution relationship of the bolt mounting holes around the flange center position. Subsequently, the spatial proximity area to which the robotic arm end needs to move is determined based on the flange center position, and the spatial proximity area is used as the termination area of ​​the subsequent translational guide trajectory segment; at the same time, the rotation direction to which the robotic arm end needs to be adjusted is determined based on the flange circumferential posture, and the rotation direction is used as the posture adjustment basis for the subsequent angle correction trajectory segment, so that the spatial three-dimensional coordinate data directly participates in the generation of the initial movement trajectory, rather than being saved only as coordinate records.

[0040] Preferably, when generating the initial movement trajectory based on the spatial three-dimensional coordinate data, firstly, a translational guide trajectory segment is generated based on the spatial position of the flange center, moving the robotic arm end from the current gripping position to near the flange docking point. Then, an angle correction trajectory segment is generated based on the flange's circumferential attitude, aligning the bolt mounting holes on the flange connection with the target flange hole direction. Finally, an axial approach trajectory segment is generated based on the normal direction of the assembly plane where the target flange is located, approaching the docking preparation position. The translational guide trajectory segment, the angle correction trajectory segment, and the axial approach trajectory segment form the initial movement trajectory in the order of translation, attitude adjustment, and approach. Specifically, the translational guide trajectory segment reduces the handling deviation of the robotic arm from the material rack area to the flange docking point; the angle correction trajectory segment reduces the circumferential deviation between the bolt mounting holes on the flange connection and the target flange hole; and the axial approach trajectory segment keeps the robotic arm at the docking preparation position before entering the contact stage. Therefore, the initial movement trajectory and the spatial three-dimensional coordinate data have a direct technical connection.

[0041] Preferably, during the generation of the initial movement trajectory, interference zone avoidance processing is performed on the initial movement trajectory based on the elbow shape, rotary joint housing, and the entry direction of the automatic tightening tool in the automatic assembly scenario of the loading arm. Specifically, the area directly below the elbow that is prone to interference with the automatic tightening tool, the area occupied by the clamping cylinder of the customized tooling platform, and the clamping areas on both sides of the rotary joint are designated as trajectory avoidance areas. The spatial relationship between the trajectory avoidance areas and the translational guide trajectory segment, the angle correction trajectory segment, and the axial approach trajectory segment is compared. When any trajectory segment enters the trajectory avoidance area, the intermediate transition position of the trajectory segment is adjusted to obtain the adjusted trajectory segment, and the adjusted trajectory segment is recombined into the initial movement trajectory. Through this processing, the initial movement trajectory is not only positioned based on the flange hole information, but also takes into account the spatial occupancy relationship of the loading arm elbow, target flange, rotary joint housing, and customized tooling platform in the automatic assembly station, thus better meeting the actual assembly requirements of automatic loading arm flange docking.

[0042] Preferably, when controlling the robotic arm to move to the preset docking preparation position according to the initial movement trajectory, the robotic arm first moves the clamped flange connection part to the vicinity of the flange docking point according to the translational guide trajectory segment, then adjusts the circumferential posture of the clamped flange connection part according to the angle correction trajectory segment, so that the direction of the bolt mounting holes on the clamped flange connection part corresponds to the direction of the target flange hole, and then moves to the docking preparation position according to the axial approach trajectory segment. The docking preparation position is the docking position before the robotic arm enters the contact stage, and a contact adjustment margin is reserved between it and the target flange end face for subsequent adjustment of the robotic arm's docking posture based on contact force feedback information; the contact adjustment margin does not participate in the compliant control of the initial coarse positioning in this segment, but provides the initial space conditions for subsequent adjustment of the robotic arm's docking posture based on contact force feedback information. Thus, the initial coarse positioning of the flange and the subsequent adjustment of the robotic arm's docking posture based on contact force feedback information are technically interconnected, but their processing objects and control bases remain distinct.

[0043] Preferably, in one implementation scenario, when the rotary joint has been clamped by a customized tooling platform and the robotic arm needs to move the elbow flange to the flange mating point on the rotary joint flange face, a two-dimensional vision sensor first acquires a flange image of the rotary joint flange face and obtains initial contour data from the flange image. The initial contour data is then used to identify the target flange hole, forming two-dimensional coordinate data. This two-dimensional coordinate data is converted into three-dimensional spatial coordinate data through hand-eye calibration, the robotic arm's current posture recording, and assembly plane constraints. The three-dimensional spatial coordinate data further generates an initial movement trajectory, and the robotic arm moves the elbow flange to the mating preparation position on the rotary joint flange face according to this initial movement trajectory. At this point, the subsequent automatic bolt assembly robot can move the bolt above the target flange hole and perform the pre-tightening hole alignment action based on the completed initial coarse positioning of the flange. Therefore, the flange hole position information not only serves the initial coarse positioning of the robotic arm but also provides a low-deviation assembly starting point for subsequent adaptive tightening control.

[0044] Optionally, the step of acquiring contact force feedback information collected by a six-dimensional torque sensor installed at the end of the robotic arm includes: During the docking operation, the raw force signals output by the six-dimensional torque sensor are periodically read. The original force signal is smoothed to obtain a smoothed force signal. The smoothed force signal is then transformed according to a preset coordinate system mapping rule to generate three-dimensional force data and three-dimensional torque data in the coordinate system of the robotic arm end effector. The three-dimensional force data and the three-dimensional torque data are combined to form the contact force feedback information.

[0045] Preferably, during the docking operation, after the robotic arm completes the initial coarse positioning of the flange, it moves the clamped flange connection to the docking preparation position and brings the flange connection into the contact stage along the normal approach direction of the target flange end face. The flange connection includes elbow flanges and straight pipe assembly flanges, and the target flange end face includes the flange end face that needs to dock with the flange connection, such as the swivel joint flange face, the elbow flange mating face, or the straight pipe assembly flange mating face. The normal approach direction of the target flange end face indicates the main direction of movement when the flange connection approaches the target flange end face. A six-dimensional torque sensor is installed between the end of the robotic arm and the end-gripping mechanism that holds the flange connection. This ensures that the contact force generated when the flange connection contacts the target flange end face is first transmitted to the end-gripping mechanism, and then from the end-gripping mechanism to the six-dimensional torque sensor, which outputs the raw force signal. Therefore, the original force sensor signal is not an independently acquired external detection quantity, but directly originates from the contact action between the flange connection part and the target flange end face when the robotic arm performs the docking action. Subsequent reading, correction, filtering, smoothing and coordinate system transformation of the original force sensor signal are all carried out around the contact action.

[0046] Preferably, the raw force sensor signal includes a force component and a torque component formed along the measurement coordinate direction of the six-dimensional torque sensor itself. The force component reflects the linear contact action experienced by the flange connection during approach, contact, and slight offset. The torque component reflects the rotational contact action formed when the flange connection relative to the target flange end face undergoes circumferential deflection, end face tilting, or edge-first contact. During the docking operation performed by the robotic arm, the six-dimensional torque sensor is periodically read according to the motion control cycle of the robotic arm to form a force sensory sampling sequence based on the sampling time sequence of the continuously read raw force sensor signals. The force sensory sampling sequence includes the raw force sensor signals corresponding to multiple sampling times, and each raw force sensor signal corresponding to a sampling time includes both the force component and the torque component. The force sensory sampling sequence continues to serve as input for subsequent benchmark correction and smoothing processing, enabling subsequent benchmark correction and smoothing processing to retain the temporal relationship of the contact action changing from small to large during the contact phase, rather than only retaining the instantaneous force sensor reading at a certain moment.

[0047] Preferably, before forming the force sampling sequence, while the robotic arm is stationary at the docking preparation position and the flange connection has not yet contacted the target flange end face, the non-contact reference force signal output by the six-dimensional torque sensor is read. This non-contact reference force signal characterizes the initial influence of the end-effector, the weight of the flange connection itself, and the installation offset on the six-dimensional torque sensor. The non-contact reference force signal also includes force and torque components formed along the measurement coordinate direction of the six-dimensional torque sensor itself. Subsequently, during the docking operation, the original force signal corresponding to each sampling moment in the force sampling sequence is corrected in relation to the non-contact reference force signal to form a corrected reference force signal. This corrected reference force signal retains the sampling time sequence from the force sampling sequence and continues to serve as input for subsequent validity screening and smoothing processes. This ensures that subsequent validity screening and smoothing processes are geared towards the contact action between the flange connection and the target flange end face, rather than towards the weight of the end-effector or the base reading formed by the installation offset.

[0048] Preferably, before smoothing the original force signal, the reference correction force signal is first filtered for effectiveness based on the speed change of the robotic arm during docking, the clamping state of the end effector, and the range of the six-dimensional torque sensor, to form a filtered force signal. The effectiveness filtering includes removing short clamping impact readings caused by the clamping moment of the end effector, removing non-contact fluctuation readings caused by the start-stop jitter of the robotic arm, and removing saturation readings exceeding the usable range of the six-dimensional torque sensor. The retained reference correction force signal forms the filtered force signal, which retains the corresponding sampling time sequence and continues to participate in subsequent smoothing. Through effectiveness filtering, the filtered force signal can correspond more closely to the contact action between the flange connection and the target flange end face during the contact phase, thereby reducing the impact of the clamping moment of the end effector, the start-stop jitter of the robotic arm, and the range saturation of the six-dimensional torque sensor on the contact force feedback information.

[0049] Preferably, when smoothing the filtered force signals, multiple adjacent filtered force signals are selected according to the sampling time sequence of the filtered force signals, and time-continuous smoothing is performed on the adjacent multiple filtered force signals to generate a smoothed force signal. The time-continuous smoothing suppresses jumps in readings between adjacent sampling times, but retains the continuous force change formed when the flange connection gradually approaches the target flange end face. The smoothed force signal continues to carry the force and torque components corresponding to each sampling time, so that subsequent coordinate system transformation can synchronously map the force and torque components at the same sampling time, avoiding the incorrect combination of force changes at different sampling times into the same contact state. Thus, the smoothed force signal is formed by time-continuous smoothing of the filtered force signals and continues to serve as the direct input for the coordinate system transformation.

[0050] Preferably, when transforming the smoothed force signal according to a preset coordinate system mapping rule, the following steps are taken: first, the installation posture relationship of the six-dimensional torque sensor relative to the end effector of the robotic arm is read; second, the end effector posture record of the robotic arm when performing the docking action is read; and third, the tool center offset relationship of the end effector clamping mechanism relative to the end effector of the robotic arm is read. The installation posture relationship expresses the directional correspondence between the measurement coordinate direction of the six-dimensional torque sensor and the coordinate system of the end effector of the robotic arm. The end effector posture record expresses the posture change of the end effector during the contact phase. The tool center offset relationship expresses the positional offset of the actual contact position of the flange connection relative to the end effector of the robotic arm. Subsequently, based on the installation posture relationship, the end effector posture record, and the tool center offset relationship, the force and torque components in the smoothed force signal are unified in direction and their point of application is converted to generate three-dimensional force data and three-dimensional torque data in the coordinate system of the end effector of the robotic arm. Both the three-dimensional force data and the three-dimensional torque data originate from the smoothed force signal and correspond to the same sampling time in the coordinate system of the end effector of the robotic arm.

[0051] Preferably, the three-dimensional force data includes axial force data related to the movement of the flange connection part along the normal approach direction of the target flange end face in the coordinate system of the robotic arm end, and transverse and longitudinal force data related to the offset of the flange connection part within the target flange end face; the axial force data reflects the degree of contact when the flange connection part approaches the target flange end face, and the transverse and longitudinal force data reflect the in-face offset trend of the flange connection part relative to the target flange end face. The three-dimensional torque data includes circumferential torque data formed around the normal approach direction of the target flange end face in the coordinate system of the robotic arm end, and end face tilting torque data formed in two directions within the target flange end face; the circumferential torque data reflects the torsional trend generated when there is a circumferential deviation between the bolt mounting holes on the flange connection part and the target flange holes on the target flange end face, and the end face tilting torque data reflects the tilting trend generated when the flange connection part first contacts the target flange end face locally. Therefore, the three-dimensional force data and the three-dimensional torque data can respectively express the influence of translational deviation and attitude deviation on the contact stage.

[0052] Preferably, after generating the three-dimensional force data and three-dimensional torque data in the coordinate system of the robotic arm's end effector, the three-dimensional force data and three-dimensional torque data corresponding to the same sampling time are synchronously combined to form the contact force feedback information. The contact force feedback information includes the axial force data, the transverse force data of the end face, the longitudinal force data of the end face, the circumferential torque data, and the tilting torque data of the end face, and retains the sampling time sequence corresponding to the above data. The contact force feedback information continues to participate in the subsequent process of establishing an admittance control model based on the contact force feedback information, so that the admittance control model can adjust the docking posture of the robotic arm based on the force and torque changes under the same contact state, rather than adjusting only based on the magnitude of the contact force in a single direction. Since the contact force feedback information is formed by synchronously combining the three-dimensional force data and three-dimensional torque data corresponding to the same sampling time, the contact force feedback information can maintain the temporal consistency between the force components and the torque components.

[0053] Preferably, during the formation of the contact force feedback information, the three-dimensional force data and the three-dimensional torque data are also recorded in correspondence with the docking action stages when the robotic arm performs the docking action. The docking action stages include the approach stage, which begins from the docking preparation position and moves closer to the target flange end face; the initial contact stage, where the flange connection part initially contacts the target flange end face; and the fitting stage, where the flange connection part makes slight contact after contact. The docking action stages and the contact force feedback information together characterize the force evolution relationship of the contact stage, so that when the contact force feedback information is used subsequently, it is possible to distinguish the normal force increase caused by approaching along the normal approach direction of the target flange end face, the increase in transverse and longitudinal force data of the end face caused by the offset of the flange connection part within the target flange end face, and the increase in end face tilt torque data caused by the non-parallelism of the flange connection part and the target flange end face. This makes the contact force feedback information more suitable for subsequent docking posture adjustment.

[0054] Preferably, in one implementation scenario, the rotary joint flange face has been clamped by a customized tooling platform. After the robotic arm clamps the elbow flange and moves it to the docking preparation position according to the initial movement trajectory, the robotic arm performs a docking action along the normal approach direction of the rotary joint flange face. During the process of the elbow flange approaching the rotary joint flange face, the six-dimensional torque sensor periodically outputs the original force sensor signal. The original force sensor signal is corrected by the non-contact reference force sensor signal to form a reference corrected force sensor signal. The reference corrected force sensor signal is filtered for validity to form a filtered force sensor signal. The filtered force sensor signal is smoothed to form a smoothed force sensor signal. The smoothed force sensor signal is then converted into three-dimensional force data and three-dimensional torque data in the coordinate system of the robotic arm end effector according to the coordinate system mapping rules. If there is a circumferential deviation between the bolt mounting holes of the elbow flange and the target flange hole on the rotary joint flange surface, the circumferential torque data in the three-dimensional torque data will increase as the contact stage progresses. If there is partial initial contact between the elbow flange and the rotary joint flange surface, the transverse end-face force data, longitudinal end-face force data, and tilting end-face torque data in the three-dimensional force data will change as the contact stage progresses. After the aforementioned three-dimensional force data and three-dimensional torque data are synchronously combined into the contact force feedback information, it can provide a force feedback basis for adjusting the docking posture of the robotic arm during the contact stage.

[0055] Optionally, the step of establishing an admittance control model based on the contact force feedback information includes: Obtain the ideal stress state data of the flange during the docking process; Calculate the force deviation between the contact force feedback information and the ideal force state data; Obtain the preset mass parameters, preset damping parameters, and preset stiffness parameters of the robotic arm; A multi-dimensional compliance mapping relationship is constructed based on the preset mass parameters, the preset damping parameters, the preset stiffness parameters, and the force deviation value, and the multi-dimensional compliance mapping relationship is used as the admittance control model.

[0056] Preferably, when acquiring the ideal stress state data of the flange during the docking process, the docking condition configuration data, which is pre-configured and associated with the flange connection part, the target flange, the docking preparation position, and the initial movement trajectory, is first read according to the flange connection part, the target flange, the docking preparation position, and the initial movement trajectory. The docking condition configuration data includes the workpiece category of the flange connection part, the flange end face category of the target flange, the axial approach direction of the flange connection part when it enters the contact stage along the flange end face of the target flange, the circumferential correspondence between the target flange hole and the bolt mounting hole of the flange connection part, and the clamping posture relationship of the end clamping mechanism on the flange connection part. The workpiece category of the flange connection is used to distinguish between elbow flanges and straight pipe assembly flange connections. The flange end face category of the target flange is used to distinguish between the flange face of a rotary joint, the mating face of an elbow flange, or the mating face of a straight pipe assembly flange connection. The axial approach direction continues to participate in the determination of the axial ideal force data in the ideal force state data. The circumferential correspondence continues to participate in the determination of the circumferential ideal torque data in the ideal force state data. The clamping posture relationship continues to participate in the determination of the end face tilt ideal torque data in the ideal force state data. Thus, the ideal force state data is not fixed general data, but corresponds to the workpiece category of the flange connection, the flange end face category of the target flange, and the clamping posture relationship of the end clamping mechanism on the flange connection.

[0057] Preferably, the ideal stress state data includes axial ideal stress data, end face transverse ideal stress data, end face longitudinal ideal stress data, circumferential ideal torque data, and end face tilting ideal torque data; the axial ideal stress data is determined based on the allowable pressing tendency when the flange connection part approaches the target flange along the axial approach direction, and the axial ideal stress data is used to represent the target contact state where the flange connection part can enter the fitting stage but does not form excessive pressing against the target flange; the end face transverse ideal stress data and the end face longitudinal ideal stress data are determined based on the end face offset constraint data in two directions within the flange end face of the target flange, and the end face offset constraint data comes from the flange end face type of the target flange and the circumferential correspondence between the target flange hole and the bolt mounting hole of the flange connection part. The end-face offset constraint data continues to participate in the determination of the ideal transverse force data and the ideal longitudinal force data of the end face. These ideal transverse and longitudinal force data represent a target contact state where the flange connection portion does not experience continuous lateral thrust relative to the target flange within the flange end face of the target flange. The ideal circumferential torque data is determined based on the circumferential correspondence, representing a target contact state where the bolt mounting holes of the flange connection portion and the target flange holes are in a tightenable and aligned position in the circumferential direction. The ideal end-face tilt torque data is determined based on the clamping posture relationship and the flange end-face type of the target flange, representing a target contact state where the end face of the flange connection portion and the flange end face of the target flange are in a relatively parallel position. The ideal axial force data, the ideal transverse force data, the ideal longitudinal force data, the ideal circumferential torque data, and the ideal tilt torque data together constitute the ideal force state data and continue to serve as the comparison benchmark for the force deviation value.

[0058] Preferably, the ideal force state data is further segmented according to the docking action stages; in the approaching stage, starting from the docking preparation position and moving closer to the target flange, the ideal force state data is configured as the approaching stage ideal force state data, which corresponds to the low contact state when the flange connection part has not yet formed stable contact, so that the axial ideal force data, the end face transverse ideal force data, the end face longitudinal ideal force data, the circumferential ideal torque data, and the end face inclined ideal torque data in the approaching stage ideal force state data are all used to identify undesired contact; in the initial contact stage, when the flange connection part initially contacts the target flange, the ideal force state data is configured as the initial contact stage ideal force state data, which corresponds to the over-contact stage when the flange connection part begins to receive the reaction force of the target flange but has not yet fully adhered. During the initial contact phase, the axial ideal force data in the ideal force data is allowed to gradually increase as contact progresses. The end-face transverse ideal force data, end-face longitudinal ideal force data, circumferential ideal torque data, and end-face tilting ideal torque data in the initial contact phase are still used to limit offset and deflection. During the bonding phase at the flange connection point after contact, when a small amount of contact is made, the ideal force data is configured as the bonding phase ideal force data. The bonding phase ideal force data corresponds to the stable contact state between the flange connection point and the target flange before it can be tightened. The axial ideal force data, end-face transverse ideal force data, end-face longitudinal ideal force data, circumferential ideal torque data, and end-face tilting ideal torque data in the bonding phase ideal force data jointly constrain the pose correction direction after bonding. By segmenting the ideal stress state data during the docking operation phase, the stress deviation value can be calculated in a way that distinguishes between the increased stress caused by normal axial approach and the abnormal stress changes caused by end face offset, circumferential misalignment, or end face tilt.

[0059] Preferably, when calculating the force deviation between the contact force feedback information and the ideal force state data, the axial force data, end face transverse force data, end face longitudinal force data, circumferential torque data, and end face tilting torque data in the contact force feedback information are read in the order of sampling time. Then, according to the docking action stage to which the sampling time belongs, the axial ideal force data, end face transverse ideal force data, end face longitudinal ideal force data, circumferential ideal torque data, and end face tilting ideal torque data corresponding to the sampling time are read from the ideal force state data. Subsequently, the axial force data and the ideal axial force data at the same sampling time are compared with each other to form axial force deviation data; the transverse force data and the ideal transverse force data at the same sampling time are compared with each other to form transverse force deviation data; the longitudinal force data and the ideal longitudinal force data at the same sampling time are compared with each other to form longitudinal force deviation data; the circumferential torque data and the ideal circumferential torque data at the same sampling time are compared with each other to form circumferential torque deviation data; and the tilting torque data and the ideal tilting torque data at the same sampling time are compared with each other to form tilting torque deviation data. The axial force deviation data, the transverse force deviation data, the longitudinal force deviation data, the circumferential torque deviation data, and the tilting torque deviation data together constitute the force deviation value, which further participates in the construction of the multi-dimensional compliance mapping relationship.

[0060] Preferably, in the process of forming the force deviation value, force data and torque data of different dimensions are not directly superimposed. Instead, the directional attributes and physical dimensions of the axial force deviation data, the end face lateral force deviation data, the end face longitudinal force deviation data, the circumferential torque deviation data, and the end face tilt torque deviation data are retained separately. When it is necessary to compare the priority of different deviations in the docking posture adjustment of the robotic arm, the axial force deviation data, the end face lateral force deviation data, the end face longitudinal force deviation data, the circumferential torque deviation data, and the end face tilt torque deviation data are matched with corresponding deviation evaluation benchmarks according to the docking condition configuration data, so as to obtain the axial deviation evaluation result, the end face lateral deviation evaluation result, the end face longitudinal deviation evaluation result, the circumferential deviation evaluation result, and the end face tilt deviation evaluation result. The axial deviation evaluation results, the end face lateral deviation evaluation results, the end face longitudinal deviation evaluation results, the circumferential deviation evaluation results, and the end face tilt deviation evaluation results are only used to determine the priority correction direction of the docking posture of the robotic arm. They do not replace the axial force deviation data, the end face lateral force deviation data, the end face longitudinal force deviation data, the circumferential torque deviation data, and the end face tilt torque deviation data in the force deviation values. This ensures that the force deviation values ​​retain the physical meaning of the actual contact action and can participate in multi-directional coordinated adjustment in the subsequent multi-dimensional compliance mapping relationship.

[0061] Preferably, when acquiring the preset mass parameters, preset damping parameters, and preset stiffness parameters of the robotic arm, virtual mechanical parameter configuration data is first formed based on the workpiece type of the flange connection, the clamping posture relationship of the end-effector, the change in motion speed of the robotic arm carrying the flange connection when executing the initial moving trajectory, and the flange end face type of the target flange. This virtual mechanical parameter configuration data is used to express the virtual inertia, virtual energy dissipation, and virtual self-alignment characteristics that the robotic arm should exhibit during admittance control, rather than directly representing the physical mass, physical damping, and physical stiffness of the robotic arm or the flange connection. The virtual inertia corresponds to the preset mass parameter, the virtual energy dissipation corresponds to the preset damping parameter, and the virtual self-alignment characteristic corresponds to the preset stiffness parameter. The virtual inertia, virtual energy dissipation, and virtual self-alignment characteristic continue to participate in the determination of the preset mass parameter, the preset damping parameter, and the preset stiffness parameter through the virtual mechanical parameter configuration data. Subsequently, parameters corresponding to the axial approach direction, the transverse direction within the flange end face of the target flange, the longitudinal direction within the flange end face of the target flange, the circumferential direction, and the end face tilt direction are read from the virtual mechanical parameter configuration data to form the preset mass parameter, the preset damping parameter, and the preset stiffness parameter. The preset mass parameter is used to constrain the rate of change of the pose compensation amount caused by the force deviation value, the preset damping parameter is used to constrain the change range of the pose compensation amount between continuous sampling moments, and the preset stiffness parameter is used to constrain the degree of deviation of the pose compensation amount from the current docking pose. The preset mass parameter, the preset damping parameter, and the preset stiffness parameter continue to jointly participate in the construction of the multi-dimensional compliance mapping relationship.

[0062] Preferably, the preset mass parameters are preset according to the load state of the flange connection part on the end clamping mechanism and the motion state of the robotic arm when performing docking action; when the flange connection part is the elbow flange, the elbow flange has a curved shape and offset clamping characteristics relative to the end of the robotic arm, and the preset mass parameters are configured in the end face tilt direction and the circumferential direction to make the change of the pose compensation amount relatively gradual, so as to reduce the amplification of the swing caused by the offset clamping of the elbow flange during the contact stage; when the flange connection part is the straight pipe assembly flange connection part, the axial length and support state of the straight pipe assembly flange connection part will affect the contact stability during the axial approach process, and the preset mass parameters are configured in the axial approach direction to make the pose compensation amount gradually change in the axial approach direction, so as to reduce the instantaneous pressure of the straight pipe assembly flange connection part on the target flange. After the preset quality parameters are configured according to the workpiece type of the flange connection and the motion state of the robotic arm when performing the docking action, they continue to work together with the force deviation value to determine the change rhythm of the pose compensation amount. This ensures that the robotic arm does not immediately perform a rigid jump correction after detecting the force deviation value, but rather forms a continuously executable pose correction process according to the load state of the flange connection.

[0063] Preferably, the preset damping parameter is preset according to the motion control cycle of the robotic arm, the sampling time sequence of the contact force feedback information, and the docking action stage; in the approaching stage, the preset damping parameter is configured to allow the robotic arm to maintain a relatively continuous approaching action along the axial approach direction, so that the pose compensation amount is mainly used to suppress undesirable contact; in the initial contact stage, the preset damping parameter is configured to improve the suppressive effect of the pose compensation amount on the jump force, so that the axial force deviation data, the end face lateral force deviation data, the end face longitudinal force deviation data, the circumferential torque deviation data, and the end face tilt torque deviation data will not directly cause excessive single pose correction; in the bonding stage, the preset damping parameter is configured to make the pose compensation amount between continuous sampling times tend to change smoothly, so that the robotic arm can complete the micro-bonding while maintaining contact between the flange connection part and the target flange. After being configured under the guidance of the docking action phase, the preset damping parameter continues to influence the multi-dimensional compliance mapping relationship together with the preset mass parameter and the preset stiffness parameter, so that the multi-dimensional compliance mapping relationship can simultaneously take into account the temporal changes of the contact phase and the continuous motion constraints of the robotic arm.

[0064] Preferably, the preset stiffness parameter is preset based on the flange end face type of the target flange, the alignment requirements of the target flange hole and the bolt mounting holes of the flange connection part, and the sealing and fitting requirements of the target flange. In the scenario where the rotary joint flange face and the elbow flange are mated, the preset stiffness parameter is configured in the circumferential direction to allow the circumferential torque deviation data to guide the robotic arm to make a small-amplitude circumferential posture correction, thereby gradually bringing the bolt mounting holes of the flange connection part closer to the target flange hole. In the scenario where the elbow flange mating surface or the straight pipe assembly flange connection part mating surface participates in the fitting, the preset stiffness parameter is configured in the end face tilt direction to allow the end face tilt torque deviation data to guide the robotic arm to make end face posture correction, thereby reducing the impact of localized initial contact on the subsequent bolt tightening process. In the axial approach direction, the preset stiffness parameter is configured to limit excessive axial pressure, so that the axial force deviation data does not guide the robotic arm to continuously apply pressure to the target flange. The preset stiffness parameters, after being configured by the flange end face type of the target flange, the alignment requirements of the bolt mounting holes of the target flange hole and the flange connection part, and the sealing fit requirements of the target flange, continue to participate in the multi-dimensional compliance mapping relationship, so that the multi-dimensional compliance mapping relationship can correspond the force deviation value to different docking posture adjustment directions.

[0065] Preferably, when constructing a multi-dimensional compliance mapping relationship based on the preset mass parameters, the preset damping parameters, the preset stiffness parameters, and the force deviation values, the axial force deviation data, the end face lateral force deviation data, the end face longitudinal force deviation data, the circumferential torque deviation data, and the end face tilt torque deviation data in the force deviation values ​​are first mapped to the axial compensation direction, the end face lateral compensation direction, the end face longitudinal compensation direction, the circumferential attitude compensation direction, and the end face tilt attitude compensation direction of the robotic arm, respectively, to form a deviation direction correspondence relationship; the deviation direction correspondence relationship is further combined with the preset mass parameters to determine the rate of change of the pose compensation amount in each compensation direction; the deviation direction correspondence relationship is further combined with the preset damping parameters to determine the smooth constraint of the pose compensation amount between adjacent sampling times; the deviation direction correspondence relationship is further combined with the preset stiffness parameters to determine the allowable deviation of the pose compensation amount relative to the current docking pose. The resulting multi-dimensional compliance mapping relationship represents how the force deviation value is converted into the pose compensation amount of the robotic arm in different directions, rather than simply converting the contact force feedback information into a single-direction displacement.

[0066] Preferably, the multi-dimensional compliance mapping relationship includes an axial compliance mapping relationship, a transverse compliance mapping relationship, a longitudinal compliance mapping relationship, a circumferential compliance mapping relationship, and a tilt compliance mapping relationship. The axial compliance mapping relationship determines the feed reduction or retraction compensation amount of the robotic arm along the axial approach direction based on the axial force deviation data. The transverse compliance mapping relationship determines the translation compensation amount of the robotic arm in the transverse direction within the flange end face of the target flange based on the transverse force deviation data. The longitudinal compliance mapping relationship determines the translation compensation amount of the robotic arm in the longitudinal direction within the flange end face of the target flange based on the longitudinal force deviation data. The circumferential compliance mapping relationship determines the circumferential attitude compensation amount of the robotic arm around the axial approach direction based on the circumferential torque deviation data. The tilt compliance mapping relationship determines the end face tilt compensation amount of the robotic arm relative to the target flange based on the end face tilt torque deviation data. The axial compliance mapping relationship, the transverse compliance mapping relationship, the longitudinal compliance mapping relationship, the circumferential compliance mapping relationship, and the inclined compliance mapping relationship together constitute the multi-dimensional compliance mapping relationship, and participate in the compensation calculation in the corresponding direction when the admittance control model outputs the pose compensation amount.

[0067] Preferably, the multi-dimensional compliant mapping relationship also constrains the sequential action relationship between different compensation directions; when the axial force deviation data exceeds the deviation evaluation benchmark corresponding to the axial ideal force data, the multi-dimensional compliant mapping relationship preferentially restricts the robotic arm from continuing to feed along the axial approach direction, and causes the axial compliant mapping relationship to output the retraction compensation amount or the feed reduction amount; when the end face lateral force deviation data and the end face longitudinal force deviation data are large and the axial deviation evaluation result does not indicate a flange jamming risk, the multi-dimensional compliant mapping relationship causes the end face lateral compliant mapping relationship and the end face longitudinal compliant mapping relationship to... The relationship prioritizes outputting the translation compensation amount to cause the flange connection part to undergo a slight translation within the flange end face of the target flange. As the circumferential torque deviation data continuously increases with the progress of the contact stage, the multi-dimensional compliance mapping relationship outputs the circumferential attitude compensation amount to bring the bolt mounting holes of the flange connection part closer to the circumferential position of the target flange hole. When the end face tilt torque data corresponds to the end face tilt torque deviation data, the multi-dimensional compliance mapping relationship outputs the end face tilt compensation amount to reduce the tilting tendency formed by the initial local contact of the flange connection end face. This sequential action relationship allows the multi-dimensional compliance mapping relationship to select the appropriate docking posture adjustment method based on the source of the force deviation value.

[0068] Preferably, when the multi-dimensional compliance mapping relationship is used as the admittance control model, the admittance control model includes at least a force reference part formed by the ideal force state data, a deviation input part formed by the force deviation value, a virtual mechanical parameter part formed by the preset mass parameter, the preset damping parameter and the preset stiffness parameter, and a pose compensation output part formed by the axial compliance mapping relationship, the end face lateral compliance mapping relationship, the end face longitudinal compliance mapping relationship, the circumferential compliance mapping relationship and the end face tilt compliance mapping relationship. The force reference section provides a comparison benchmark to the deviation input section at the same sampling time and during the same docking action phase. The deviation input section provides the axial force deviation data, end-face lateral force deviation data, end-face longitudinal force deviation data, circumferential torque deviation data, and end-face tilt torque deviation data to the virtual mechanical parameter section. The virtual mechanical parameter section constrains the influence of the axial force deviation data, end-face lateral force deviation data, end-face longitudinal force deviation data, circumferential torque deviation data, and end-face tilt torque deviation data on the pose compensation amount. The pose compensation output section outputs the pose compensation amount that matches the current docking pose of the robotic arm. Thus, the admittance control model can convert the multi-directional force and torque changes in the contact force feedback information into a basis for the robotic arm to perform docking pose adjustments.

[0069] Preferably, in one implementation scenario, the robotic arm clamps the elbow flange and moves it above the rotary joint flange face, which is clamped and positioned by a customized tooling platform. After the elbow flange enters the contact stage along the normal direction of the rotary joint flange face, the axial force data in the contact force feedback information is used to reflect the degree of pressure of the elbow flange against the rotary joint flange face, the transverse force data and the longitudinal force data are used to reflect the in-face offset trend of the elbow flange relative to the rotary joint flange face, the circumferential torque data is used to reflect the circumferential misalignment trend between the bolt mounting holes of the elbow flange and the target flange holes, and the end face tilt torque data is used to reflect the local pre-contact trend between the elbow flange and the rotary joint flange face. The admittance control model compares the contact force feedback information with the ideal force state data to form the force deviation value. Then, based on the preset mass parameters, preset damping parameters, and preset stiffness parameters, it generates the pose compensation amount, enabling the robotic arm to first reduce axial pressure, and then perform minor corrections in the corresponding directions based on the end-face offset trend, the circumferential misalignment trend, and the local pre-contact trend. In the above implementation scenario, the pose compensation amount continues to participate in the generation of the target docking pose, ensuring that the elbow flange and the rotary joint flange face enter a contact state that satisfies the subsequent bolt tightening conditions.

[0070] Optionally, the step of adjusting the docking posture of the robotic arm based on the admittance control model to complete the intelligent docking of the flange includes: The force deviation value is input into the admittance control model so that the robotic arm exhibits the compliant physical characteristics of spring damping, and the pose compensation amount of the robotic arm is output. The current docking pose of the robotic arm is obtained, and the current docking pose is corrected according to the pose compensation amount to obtain the target docking pose; Control the robotic arm to move to the target docking position.

[0071] Preferably, when inputting the force deviation value into the admittance control model, the axial force deviation data, the transverse force deviation data, the longitudinal force deviation data, the circumferential torque deviation data, and the tilting torque deviation data from the force deviation value are first sent to the corresponding axial compliance mapping relationship, the transverse compliance mapping relationship, the longitudinal compliance mapping relationship, the circumferential compliance mapping relationship, and the tilting compliance mapping relationship in the admittance control model, respectively; the admittance control model does not input the axial force deviation data, The transverse force deviation data, longitudinal force deviation data, circumferential torque deviation data, and tilting torque deviation data of the end face are directly converted into various compensation components in the pose compensation amount. Instead, the preset mass parameter is first used to limit the rate of change of the pose compensation amount in each compensation direction, the preset damping parameter is used to limit the smoothness of the change of the pose compensation amount between consecutive sampling moments, and the preset stiffness parameter is used to limit the deviation of the pose compensation amount from the current docking pose. This makes the robotic arm exhibit the compliant physical characteristics of spring damping. The compliant physical characteristics of spring damping indicate that after the robotic arm comes into contact with the target flange at the flange connection point, it no longer rigidly presses against it along a fixed trajectory, but instead forms a continuously executable pose compensation amount based on the direction, magnitude, and continuous changing trend of the force deviation value. The pose compensation amount continues to participate in the correction of the current docking pose.

[0072] Preferably, when the admittance control model outputs the pose compensation amount of the robotic arm, the pose compensation amount is divided into an axial compensation component, an end-face lateral compensation component, an end-face longitudinal compensation component, a circumferential attitude compensation component, and an end-face tilt compensation component; the axial compensation component originates from the processing result of the axial compliance mapping relationship on the axial force deviation data, and the axial compensation component is used to adjust the feed reduction or retraction compensation amount formed by the robotic arm along the axial approach direction; the end-face lateral compensation component originates from the processing result of the end-face lateral compliance mapping relationship on the end-face lateral force deviation data, and the end-face lateral compensation component is used to adjust the end-face lateral compensation component formed by the robotic arm in the lateral direction within the flange end face of the target flange; The longitudinal compensation component of the end face originates from the processing result of the longitudinal compliance mapping relationship on the longitudinal force deviation data of the end face. This longitudinal compensation component is used to adjust the longitudinal compensation component of the end face formed by the robotic arm in the longitudinal direction within the flange end face of the target flange. The circumferential attitude compensation component originates from the processing result of the circumferential compliance mapping relationship on the circumferential torque deviation data. This circumferential attitude compensation component is used to adjust the circumferential attitude compensation component formed by the robotic arm around the axial approach direction. The end face tilt compensation component originates from the processing result of the end face tilt torque deviation data on the end face tilt compliance mapping relationship. This end face tilt compensation component is used to adjust the end face tilt compensation component formed by the robotic arm relative to the target flange. The axial compensation component, the transverse compensation component, the longitudinal compensation component, the circumferential attitude compensation component, and the end face tilt compensation component together constitute the pose compensation amount and continue to participate in the correction of the current docking pose.

[0073] Preferably, when the axial compliance mapping relationship outputs the axial compensation component, the axial force deviation data and the axial deviation evaluation result are first read, and it is determined whether the axial force deviation data corresponds to normal contact advancement or excessive pressing trend based on the docking action stage. When the axial force deviation data corresponds to normal contact advancement, the axial compliance mapping relationship, combined with the preset damping parameter, outputs the feed reduction amount, causing the robotic arm to continue approaching the target flange along the axial approach direction but reducing the single feed amplitude. When the axial force deviation data corresponds to excessive pressing trend, the axial compliance mapping relationship, combined with the preset mass parameter and the preset stiffness parameter, outputs the retraction compensation amount, causing the robotic arm to reduce the excessive pressing trend in the direction opposite to the axial approach direction. The feed reduction amount or the retraction compensation amount is used as the axial compensation component in the pose compensation amount, so that the pose compensation amount can preferentially constrain the pressing state of the flange connection part against the target flange.

[0074] Preferably, when the transverse compliance mapping relationship and the longitudinal compliance mapping relationship output the transverse compensation component and the longitudinal compensation component of the end face, the transverse force deviation data, the longitudinal force deviation data, the transverse deviation evaluation result, and the longitudinal deviation evaluation result of the end face are read first, respectively. When the transverse force deviation data indicates that the flange connection part has a transverse lateral thrust tendency in the flange end face of the target flange, the transverse compliance mapping relationship outputs the end face compensation component opposite to the transverse lateral thrust tendency according to the preset stiffness parameter. Lateral compensation component; when the longitudinal force deviation data of the end face indicates that the flange connection part has a longitudinal lateral thrust tendency in the flange end face of the target flange, the longitudinal compliance mapping relationship of the end face outputs the longitudinal compensation component of the end face that is opposite to the longitudinal lateral thrust tendency according to the preset stiffness parameter; the lateral compensation component and the longitudinal compensation component of the end face continue to enter the pose compensation amount, so that the robotic arm can perform a small translation in the end face while the flange connection part is in contact with the target flange, instead of completely disengaging from the contact first and then finding a new docking position.

[0075] Preferably, when the circumferential compliance mapping relationship outputs the circumferential attitude compensation component, the circumferential torque deviation data, the circumferential deviation evaluation result, and the circumferential correspondence are first read. The circumferential torque deviation data reflects the misalignment trend between the bolt mounting holes of the flange connection and the target flange hole in the circumferential direction. The circumferential correspondence defines the circumferential correction direction for the bolt mounting holes of the flange connection to approach the target flange hole. Based on the circumferential correction direction and in conjunction with the preset mass parameter and the preset damping parameter, the circumferential compliance mapping relationship outputs the circumferential attitude compensation component, ensuring that the circumferential attitude compensation component does not generate excessive single rotation due to instantaneous changes in the circumferential torque deviation data. The circumferential attitude compensation component then enters the pose compensation amount, enabling the pose compensation amount to convert the circumferential torque deviation data caused by the circumferential misalignment trend into a basis for the attitude correction of the robotic arm around the axial approach direction.

[0076] Preferably, when the end face tilt compliance mapping relationship outputs the end face tilt compensation component, the end face tilt torque deviation data, the end face tilt deviation evaluation result, the clamping posture relationship, and the flange end face category of the target flange are read first. The end face tilt torque deviation data is used to reflect the local pre-contact trend between the end face of the flange connection and the flange end face of the target flange. The clamping posture relationship is used to distinguish whether the flange connection forms an offset clamping or approximately centered clamping on the end clamping mechanism. The flange end face category of the target flange is used to distinguish whether the flange end face of the target flange belongs to a rotary joint flange face, an elbow flange mating face, or a straight pipe assembly flange connection mating face. The end face tilt compliance mapping relationship outputs the end face tilt compensation component based on the local pre-contact trend and in combination with the preset stiffness parameter, so that the end face tilt compensation component can guide the robotic arm to adjust the end face orientation of the flange connection. The end face tilt compensation component continues to enter the pose compensation amount, so that the pose compensation amount can convert the end face tilt torque deviation data caused by the local pre-contact trend into the end face posture correction basis.

[0077] Preferably, before outputting the pose compensation amount, the admittance control model also forms pose compensation constraint data based on the docking condition configuration data. The pose compensation constraint data originates from the workpiece category of the flange connection part, the flange end face category of the target flange, the axial approach direction, the circumferential correspondence, and the clamping posture relationship. The pose compensation constraint data is used to limit the range and sequence of action of the axial compensation component, the end face lateral compensation component, the end face longitudinal compensation component, the circumferential posture compensation component, and the end face tilt compensation component. The current loading arm assembly scenario is jointly characterized by the workpiece category of the flange connection part, the flange end face category of the target flange, the axial approach direction, the circumferential correspondence, and the clamping posture relationship. The pose compensation constraint data corresponds to the current loading arm assembly scenario. Specifically, when the axial force deviation data corresponds to an excessive pressing trend, the pose compensation constraint data causes the axial compensation component to preferentially act on the pose compensation amount; when the end face lateral force deviation data and the end face longitudinal force deviation data correspond to an in-end face offset trend, the pose compensation constraint data causes the end face lateral compensation component and the end face longitudinal compensation component to participate in the pose compensation amount; when the circumferential torque deviation data and the end face tilt torque deviation data occur simultaneously, the pose compensation constraint data determines the order of action of the circumferential pose compensation component and the end face tilt compensation component according to the circumferential correspondence and the clamping posture relationship. The pose compensation constraint data continues to act on the pose compensation amount, making the pose compensation amount correspond to the current loading arm assembly scenario.

[0078] Preferably, when acquiring the current docking pose of the robotic arm, the current position of the robotic arm end effector, the current posture of the robotic arm end effector, and the clamping posture relationship of the flange connection relative to the robotic arm end effector are read, and the current position of the robotic arm end effector, the current posture of the robotic arm end effector, and the clamping posture relationship are used together as the pose source data of the current docking pose. The current position of the robotic arm end effector is used to represent the spatial position of the robotic arm end effector in the base coordinate system of the robotic arm; the current posture of the robotic arm end effector is used to represent the orientation of the robotic arm end effector relative to the flange end face of the target flange; and the clamping posture relationship is used to represent the installation direction and offset state of the flange connection relative to the robotic arm end effector. After the current docking pose is formed by the pose source data, the current docking pose is further combined with the pose compensation amount, so that the correction of the current docking pose considers not only the current position and current posture of the robotic arm end effector, but also the clamping posture relationship of the flange connection after being clamped by the end effector clamping mechanism.

[0079] Preferably, when correcting the current docking pose based on the pose compensation amount, the axial compensation component is first applied to the axial proximity position in the current docking pose to form the axially corrected docking pose; then the end face lateral compensation component and the end face longitudinal compensation component are applied to the end face position in the axially corrected docking pose to form the end face in-plane corrected docking pose; subsequently, the circumferential attitude compensation component is applied to the circumferential attitude in the end face in-plane corrected docking pose to form the circumferentially corrected docking pose; and finally, the end face tilt compensation component is applied to the end face orientation in the circumferentially corrected docking pose to form a candidate docking pose. The candidate docking pose inherits the original position and original attitude of the current docking pose, and at the same time superimposes the axial compensation component, the end face lateral compensation component, the end face longitudinal compensation component, the circumferential attitude compensation component, and the end face tilt compensation component in the pose compensation amount. The candidate docking pose continues to enter the pose executability verification to avoid the situation where the pose compensation amount meets the contact force adjustment requirements but is not suitable for the continuous execution of the robotic arm.

[0080] Preferably, when verifying the executability of the candidate docking pose, pose correction boundary data is formed based on the initial movement trajectory, the docking preparation position, the circumferential correspondence between the target flange hole and the bolt mounting holes of the flange connection part, and the sealing fit requirements of the target flange. The pose correction boundary data is used to limit the deviation range of the candidate docking pose from the initial movement trajectory, the circumferential alignment range of the candidate docking pose relative to the target flange hole, and the end face fit range of the candidate docking pose relative to the flange end face of the target flange. The deviation range, the circumferential alignment range, and the end face fit range all participate in the pose executability verification. After comparing the candidate docking pose with the pose correction boundary data, if the candidate docking pose is within the allowable range of the pose correction boundary data, then the candidate docking pose is taken as the target docking pose; if the candidate docking pose exceeds the allowable range of the pose correction boundary data, then the excess portion of the candidate docking pose is limited and corrected according to the pose correction boundary data to form the target docking pose. The target docking pose thus simultaneously retains the correction result of the pose compensation amount on the candidate docking pose and the spatial constraint result of the pose correction boundary data on the candidate docking pose.

[0081] Preferably, after the target docking pose is formed, the difference between the target docking pose and the current docking pose is analyzed to generate robotic arm pose adjustment data. The robotic arm pose adjustment data includes axial movement adjustment data, end-face lateral movement adjustment data, end-face longitudinal movement adjustment data, circumferential posture adjustment data, and end-face tilt posture adjustment data. The axial movement adjustment data originates from the axial position difference between the target docking pose and the current docking pose. The end-face lateral movement adjustment data and the end-face longitudinal movement adjustment data originate from the in-end-face position difference between the target docking pose and the current docking pose. The circumferential posture adjustment data originates from the circumferential posture difference between the target docking pose and the current docking pose. The end-face tilt posture adjustment data originates from the end-face orientation difference between the target docking pose and the current docking pose. The robotic arm pose adjustment data continues to participate in the control process of the robotic arm moving to the target docking pose, ensuring that the robotic arm does not jump to the target docking pose all at once, but rather performs continuous pose adjustments according to the robotic arm pose adjustment data.

[0082] Preferably, when controlling the robotic arm to move to the target docking posture, a segmented execution trajectory is first generated based on the robotic arm posture adjustment data. The segmented execution trajectory includes an axial release segment, an in-face translation segment, a posture fine-tuning segment, and a fitting confirmation segment. The axial release segment is generated based on the axial movement adjustment data; the in-face translation segment is generated based on the end-face lateral movement adjustment data and the end-face longitudinal movement adjustment data; the posture fine-tuning segment is generated based on the circumferential posture adjustment data and the end-face tilt posture adjustment data; and the fitting confirmation segment is generated based on the ideal force state data and the posture correction boundary data. The segmented execution trajectory continues to drive the robotic arm to move from the current docking posture to the target docking posture. During the movement, the robotic arm first reduces the excessive pressing tendency through the axial release segment, then corrects its in-face position through the in-face translation segment, then corrects its circumferential posture and end-face orientation through the posture fine-tuning segment, and finally enters a fitting state suitable for bolt tightening through the fitting confirmation segment.

[0083] Preferably, during the movement of the robotic arm according to the segmented execution trajectory, the contact force feedback information is continuously read at the sampling time, and the new contact force feedback information is re-inputted into the admittance control model. The admittance control model recalculates the force deviation value based on the new contact force feedback information and updates the pose compensation amount based on the recalculated force deviation value. The updated pose compensation amount continues to correct the segmented execution trajectory that has not yet been completed, so that the segmented execution trajectory can be adjusted according to changes in the contact state. In this way, the process of the robotic arm moving to the target docking pose is not based solely on a fixed movement process formed by a single sampling, but is dynamically corrected based on the continuous technical interaction relationship between the contact force feedback information, the force deviation value, the pose compensation amount, the segmented execution trajectory, and the target docking pose.

[0084] Preferably, in the implementation scenario where the elbow flange and the rotary joint flange face are mated, the robotic arm clamps the elbow flange and approaches the rotary joint flange face along the normal direction of the rotary joint flange face; when the contact force feedback information reflects that there is a local pre-contact tendency between the elbow flange and the rotary joint flange face, the admittance control model outputs the end face tilt compensation component according to the end face tilt torque deviation data, and outputs the feed reduction amount or the retraction compensation amount according to the axial force deviation data; the end face tilt compensation component and the feed reduction amount or the retraction compensation amount together form the pose compensation amount, and the pose compensation amount further corrects the current docking pose to form the target docking pose. After the robotic arm moves according to the target docking posture, the end face of the elbow flange tends to be in a relatively parallel target contact state relative to the flange face of the rotary joint. The bolt mounting hole of the elbow flange and the target flange hole also tend to approach the target contact state that can be tightened and aligned with the circumferential posture compensation component, thereby providing a target contact state basis for the subsequent bolt tightening action of the tightening shaft.

[0085] Preferably, in the implementation scenario where the flange connection of the straight pipe assembly is docked with the target flange, the robotic arm clamps the flange connection of the straight pipe assembly and displaces it towards the target flange. When the contact force feedback information reflects a continuous lateral pushing trend of the flange connection of the straight pipe assembly within the flange end face of the target flange, the admittance control model outputs the lateral compensation component of the end face based on the lateral force deviation data of the end face, and outputs the longitudinal compensation component of the end face based on the longitudinal force deviation data of the end face. The lateral compensation component and the longitudinal compensation component of the end face jointly participate in the pose compensation amount, which corrects the position within the end face of the current docking pose to form the target docking pose. After the robotic arm moves to the target docking pose, the continuous lateral pushing trend of the flange connection of the straight pipe assembly relative to the flange end face of the target flange is reduced, and the target docking pose continues to serve as the pose basis for subsequent judgments to complete the intelligent docking of the flange and enter the bolt tightening process.

[0086] Optionally, after the step of controlling the robotic arm to move to the target docking pose, the method further includes: Real-time monitoring of new contact force feedback information after moving to the target docking pose; If the new contact force feedback information is greater than the preset collision threshold, it is determined that there is a risk of flange jamming. The robotic arm is then controlled to retreat according to the preset retraction trajectory and the docking posture is readjusted. If the new contact force feedback information is less than the preset fitting threshold, then the intelligent docking of the flange is determined to be complete.

[0087] Preferably, after controlling the robotic arm to move to the target docking posture, when monitoring the new contact force feedback information after moving to the target docking posture in real time, the new contact force feedback information is not treated as a single value for overall judgment. Instead, according to the componentization method of the contact force feedback information in the admittance control model, the new axial force data, new end-face transverse force data, new end-face longitudinal force data, new circumferential torque data, and new end-face tilt torque data in the new contact force feedback information are read. The new axial force data, new end-face transverse force data, new end-face longitudinal force data, new circumferential torque data, and new end-face tilt torque data are respectively correlated with the axial force data corresponding to the same docking action stage in the ideal force state data. Ideal force data, ideal end-face lateral force data, ideal end-face longitudinal force data, ideal circumferential torque data, and ideal end-face tilt torque data are compared with each other using the same dimensions to form new axial force deviation data, new end-face lateral force deviation data, new end-face longitudinal force deviation data, new circumferential torque deviation data, and new end-face tilt torque deviation data. The new axial force deviation data, the new end-face lateral force deviation data, the new end-face longitudinal force deviation data, the new circumferential torque deviation data, and the new end-face tilt torque deviation data together constitute a new force deviation value. The new force deviation value continues to participate in the judgment of the preset collision threshold and the preset contact threshold, thereby avoiding the direct superposition of force data and torque data with different dimensions for judgment.

[0088] Preferably, the preset collision threshold is pre-configured based on the workpiece type of the flange connection, the flange end face type of the target flange, the axial approach direction, the circumferential correspondence, the clamping posture relationship, and the docking action stage; the preset collision threshold includes axial collision judgment criteria, end face transverse collision judgment criteria, end face longitudinal collision judgment criteria, circumferential collision judgment criteria, and end face tilting collision judgment criteria. The axial collision judgment criteria are used to determine whether the new axial force deviation data corresponds to an excessive pressing trend; the end face transverse collision judgment criteria are used to determine whether the new end face transverse force deviation data corresponds to a transverse lateral pushing trend; the end face longitudinal collision judgment criteria are used to determine whether the new end face longitudinal force deviation data corresponds to a longitudinal lateral pushing trend; and the circumferential collision threshold is... The judgment criterion is used to determine whether the new circumferential torque deviation data corresponds to a circumferential misalignment trend. The end face tilting collision judgment criterion is used to determine whether the new end face tilting torque deviation data corresponds to a local first contact trend. The lateral thrust trend and the longitudinal thrust trend together characterize the continuous thrust trend within the end face. The continuous thrust trend within the end face indicates that the flange connection part continuously generates lateral contact offset relative to the flange end face of the target flange in the direction within the end face. The axial collision judgment criterion, the end face lateral collision judgment criterion, the end face longitudinal collision judgment criterion, the circumferential collision judgment criterion, and the end face tilting collision judgment criterion together constitute the preset collision threshold, so that the preset collision threshold corresponds to the force direction, torque direction, and contact stage in the current loading arm assembly scenario.

[0089] Preferably, when the new contact force feedback information is greater than the preset collision threshold, technically, this means that at least one deviation data in the new force deviation value exceeds the collision judgment benchmark corresponding to its physical dimension, and the deviation data exceeding the collision judgment benchmark maintains the same trend of change in the same direction during continuous sampling time; when the new axial force deviation data exceeds the axial collision judgment benchmark, it is determined that the flange connection part has an excessive pressing tendency against the target flange along the axial approach direction; when the new end face lateral force deviation data exceeds the end face lateral collision judgment benchmark, it is determined that the flange connection part has a lateral pushing tendency relative to the flange end face of the target flange; when the new end face longitudinal force deviation data exceeds the end face longitudinal collision judgment benchmark, it is determined that the flange connection part has a longitudinal pushing tendency relative to the flange end face of the target flange; when the lateral... When at least one of the lateral thrust trend and the longitudinal lateral thrust trend maintains the same trend during continuous sampling time, it is determined that there is a continuous in-face lateral thrust trend of the flange connection part relative to the flange end face of the target flange; when the new circumferential torque deviation data exceeds the circumferential collision judgment benchmark, it is determined that there is a circumferential misalignment trend between the bolt mounting hole of the flange connection part and the target flange hole; when the new end face tilt torque deviation data exceeds the end face tilt collision judgment benchmark, it is determined that there is a local first contact trend between the end face of the flange connection part and the flange end face of the target flange; the excessive pressing trend, the continuous in-face lateral thrust trend, the circumferential misalignment trend, and the local first contact trend jointly participate in forming the flange jamming risk judgment result, and the flange jamming risk judgment result is further used to select the retraction trajectory segment and retraction direction data in the preset retraction trajectory.

[0090] Preferably, the preset retraction trajectory is not a fixed reverse straight line trajectory, but rather, after generating the initial movement trajectory, it is pre-configured based on the initial movement trajectory, the docking preparation position, the target docking pose, the axial approach direction, the pose correction boundary data, the circumferential correspondence, and the clamping posture relationship to form retraction trajectory configuration data. The retraction trajectory configuration data includes an axial unloading segment, an in-end-face unloading segment, a circumferential unloading segment, an in-end-face tilt unloading segment, the docking preparation pose after retraction, and retraction direction data. The axial unloading segment is formed based on the axial approach direction and the docking preparation position, and the in-end-face unloading segment is formed based on the flange end of the target flange. The in-plane lateral direction and the in-plane longitudinal direction of the flange end face of the target flange are formed. The circumferential unloading segment is formed according to the circumferential correspondence. The end face inclined unloading segment is formed according to the clamping posture relationship and the flange end face category of the target flange. The retraction and docking preparation posture is formed according to the docking preparation position and the posture correction boundary data. The retraction direction data is formed according to the unloading direction corresponding to the axial unloading segment, the in-plane unloading segment, the circumferential unloading segment and the end face inclined unloading segment. The retraction trajectory configuration data continues to serve as the source data of the preset retraction trajectory, so that the preset retraction trajectory can correspond to different flange jamming risk judgment results.

[0091] Preferably, when controlling the robotic arm to retract according to the preset retraction trajectory, the corresponding retraction trajectory segment and retraction direction data are first selected from the retraction trajectory configuration data based on the flange jamming risk determination result; when the flange jamming risk determination result indicates an excessive pressing trend, the axial unloading segment and the retraction direction data corresponding to the axial unloading segment are preferentially selected, so that the robotic arm reduces the pressing state of the flange connection part against the target flange in a direction opposite to the axial approach direction; when the flange jamming risk determination result indicates a continuous lateral pushing trend within the end face, the in-end face unloading segment and the retraction direction data corresponding to the in-end face unloading segment are selected after the axial unloading segment, so that the robotic arm continuously pushes the target flange within the flange end face along the direction opposite to the in-end face. A slight translation is performed in the opposite direction of the trend; when the flange jamming risk determination result indicates the circumferential misalignment trend, after reducing the axial pressure, the circumferential unloading segment and the corresponding retraction direction data are selected, so that the robotic arm performs a circumferential attitude release around the axial approach direction; when the flange jamming risk determination result indicates the local first contact trend, after reducing the axial pressure, the end face tilting unloading segment and the corresponding retraction direction data are selected, so that the robotic arm adjusts the end face orientation of the flange connection part; the selected retraction trajectory segment and the selected retraction direction data together constitute the current retraction execution trajectory, and the current retraction execution trajectory is used to control the robotic arm to retract from the target docking posture to the retraction-prepared docking posture.

[0092] Preferably, during the retraction of the robotic arm along the current retraction execution trajectory, it continues to periodically read the new contact force feedback information and converts the new contact force feedback information during the retraction process back into the new force deviation value; the new force deviation value is compared with the preset collision threshold in the same dimension to form a collision change result during the retraction process. The collision change result during the retraction process is used to determine whether the current retraction execution trajectory is reducing the abnormal contact state corresponding to the flange jamming risk determination result; if the collision change result during the retraction process indicates that the new force deviation value still maintains an increasing trend, then according to the retraction trajectory... The execution ratio of the axial unloading segment in the current retraction execution trajectory is increased by setting the data, so that the current retraction execution trajectory first releases the axial pressure state of the flange connection part on the target flange; if the collision change result of the retraction process indicates that the new force deviation value tends to decrease, the current retraction execution trajectory continues to be executed until the retraction docking preparation posture is reached; the retraction docking preparation posture continues to participate in the re-adjustment of the docking posture, so that the robot arm does not return to the state before the initial coarse positioning after retraction, but re-enters the docking posture adjustment process based on retaining the flange hole position information and the circumferential correspondence.

[0093] Preferably, when readjusting the docking posture, the retracted docking preparation posture is used as the new current docking posture, and the new contact force feedback information reacquired after the retraction is input into the admittance control model. The admittance control model recalculates the new force deviation value based on the new contact force feedback information reacquired after the retraction, and updates the posture compensation amount based on the new force deviation value. The updated posture compensation amount is used to correct the retracted docking preparation posture to form the retracted target docking posture. The retracted target docking posture is further compared with the posture correction boundary data. If the retracted target docking posture is within the range allowed by the posture correction boundary data, the robotic arm is controlled to move towards the retracted target docking posture. If the retracted target docking posture exceeds the range allowed by the posture correction boundary data, the retracted target docking posture is limited and corrected according to the posture correction boundary data, and the docking posture adjustment continues with the limited and corrected retracted target docking posture.

[0094] Preferably, the preset fitting threshold is pre-configured based on the flange end face type of the target flange, the circumferential correspondence between the target flange hole and the bolt mounting holes of the flange connection part, the sealing fitting requirements of the target flange, and the docking action stage; the preset fitting threshold includes axial fitting judgment criteria, transverse end face fitting judgment criteria, longitudinal end face fitting judgment criteria, circumferential fitting judgment criteria, and inclined end face fitting judgment criteria. The axial fitting judgment criteria are used to determine whether the new axial force deviation data is within the allowable fitting pressure range. The transverse end face fitting judgment criteria are used to determine whether the flange connection part is within the flange end face of the target flange without a transverse pushing tendency. The longitudinal end face fitting judgment criteria are used to determine whether the... The criteria for determining whether the flange connection is within the flange end face of the target flange without longitudinal lateral pushing tendency are as follows: the circumferential fit judgment criterion is used to determine whether the bolt mounting holes of the flange connection are in a tightenable alignment state relative to the target flange holes; the end face tilt fit judgment criterion is used to determine whether the end face of the flange connection is in a relatively parallel target contact state relative to the flange end face of the target flange; the axial fit judgment criterion, the end face transverse fit judgment criterion, the end face longitudinal fit judgment criterion, the circumferential fit judgment criterion, and the end face tilt fit judgment criterion together constitute the preset fit threshold, so that the preset fit threshold can simultaneously constrain the pressing state, transverse lateral pushing tendency, longitudinal lateral pushing tendency, circumferential misalignment tendency, and local first contact tendency.

[0095] Preferably, when the new contact force feedback information is less than the preset fitting threshold, technically speaking, this means that the new axial force deviation data, the new end-face lateral force deviation data, the new end-face longitudinal force deviation data, the new circumferential torque deviation data, and the new end-face tilt torque deviation data in the new force deviation value are respectively within the allowable range of the corresponding axial fitting judgment benchmark, the end-face lateral fitting judgment benchmark, the end-face longitudinal fitting judgment benchmark, the circumferential fitting judgment benchmark, and the end-face tilt fitting judgment benchmark, and the variation amplitude of the new force deviation value during continuous sampling time is in a stable fitting state; the stable fitting state is used to represent the new axial force deviation. The data, including the new end-face lateral force deviation data, the new end-face longitudinal force deviation data, the new circumferential torque deviation data, and the new end-face tilt torque deviation data, do not form new excessive pressing trends, lateral pushing trends, longitudinal pushing trends, circumferential misalignment trends, or local pre-contact trends during continuous sampling. The stable fitting state continues to be correlated with the target docking posture. If the target docking posture simultaneously satisfies the circumferential alignment range and end-face fitting range in the posture correction boundary data, a smart docking completion determination result is formed. The smart docking completion determination result is used to determine the completion of the smart docking of the flange and continues to serve as the trigger basis for entering the bolt tightening process.

[0096] Preferably, before determining the completion of the intelligent docking of the flange, the intelligent docking completion determination result is checked for consistency with the flange hole position information. The consistency check first reads the spatial three-dimensional coordinate data converted from the two-dimensional coordinate data of the target flange hole, then reads the spatial position of the bolt mounting holes of the flange connection part corresponding to the target docking posture, and determines whether the spatial position of the bolt mounting holes of the flange connection part is in a tightenable alignment state with the target flange hole according to the circumferential correspondence. If the spatial position of the bolt mounting holes of the flange connection part is in a tightenable alignment state with the target flange hole, the intelligent docking completion determination result is maintained. If the spatial position of the bolt mounting holes of the flange connection part is not in a tightenable alignment state with the target flange hole, the new circumferential torque deviation data is re-input into the circumferential compliance mapping relationship to regenerate the circumferential attitude compensation component. The circumferential attitude compensation component continues to participate in the correction of the target docking posture, so that the determination of the completion of the intelligent docking of the flange is based not only on the force state, but also on the flange hole position information and the circumferential correspondence.

[0097] Preferably, in the scenario where the elbow flange and the swivel joint flange face are in contact, the preset retraction trajectory is configured based on the bending shape of the elbow flange, the offset clamping state of the elbow flange on the end clamping mechanism, the normal direction of the swivel joint flange face, and the circumferential correspondence between the bolt mounting holes of the elbow flange and the target flange hole; the normal direction of the swivel joint flange face serves as the axial approach direction when the elbow flange and the swivel joint flange face are in contact, and the axial approach direction continues to participate in the formation of the axial unloading segment and the retraction direction data; when the new end face tilt torque deviation data exceeds the When determining the end face tilt collision benchmark, the flange jamming risk determination result indicates that there is a local pre-contact tendency between the elbow flange and the rotary joint flange face. The current retraction execution trajectory first selects the axial unloading segment to reduce the pressure state of the elbow flange on the rotary joint flange face, and then selects the end face tilt unloading segment to adjust the end face orientation of the elbow flange. After the current retraction execution trajectory is completed, the retraction and docking preparation posture continues to retain the circumferential correspondence between the bolt mounting holes of the elbow flange and the target flange hole, so that when the robotic arm readjusts the docking posture, it can still correct around the tightenable alignment state.

[0098] Preferably, in the scenario where the flange connection of the straight pipe assembly is connected to the target flange, the preset retraction trajectory is configured based on the axial length of the flange connection of the straight pipe assembly, the support state of the flange connection of the straight pipe assembly, the transverse direction within the flange end face of the target flange, and the longitudinal direction within the flange end face of the target flange. The transverse direction within the flange end face of the target flange continues to participate in the formation of the transverse unloading direction in the unloading section within the end face, and the longitudinal direction within the flange end face of the target flange continues to participate in the formation of the longitudinal unloading direction in the unloading section within the end face. The transverse unloading direction and the longitudinal unloading direction jointly participate in forming the retraction direction data corresponding to the unloading section within the end face. When the new transverse force deviation data of the end face exceeds the transverse collision judgment benchmark of the end face, or the new end face... When the longitudinal force deviation data exceeds the longitudinal collision judgment benchmark of the end face, the flange jamming risk judgment result indicates that the flange connection part of the straight pipe assembly has a continuous lateral pushing trend within the flange end face of the target flange. The current retraction execution trajectory first reduces the axial pressing state between the flange connection part of the straight pipe assembly and the target flange through the axial unloading section, and then causes the flange connection part of the straight pipe assembly to make a slight translation in the opposite direction to the continuous lateral pushing trend within the end face through the end face unloading section. The retraction and docking preparation posture after the end face unloading section is executed continues as the new current docking posture, and the updated posture compensation amount continues to be generated through the admittance control model so that the flange connection part of the straight pipe assembly approaches the fitting position of the target flange again.

[0099] Preferably, after the intelligent docking completion determination result is formed, the target docking posture, the new contact force feedback information, the new force deviation value, the preset fitting threshold, and the flange hole position information are associated and recorded to form docking completion status data. The docking completion status data is used to characterize that the robotic arm has moved the flange connection part to the target docking posture that meets the requirements of force fitting, hole alignment, and end face posture. The force fitting is jointly characterized by the new force deviation value and the preset fitting threshold. The hole alignment is jointly characterized by the flange hole position information and the circumferential correspondence. The end face posture requirement is jointly characterized by the end face tilt fitting determination benchmark and the target docking posture. The docking completion status data continues to participate in the subsequent tightening action of the tightening shaft on the target bolt. Through the docking completion status data, the bolt tightening process can be started when the flange connection part and the target flange are in a contact state with a low risk of jamming, thereby reducing the influence of excessive pressure tendency, circumferential misalignment tendency, or local pre-contact tendency caused by direct tightening after fixed trajectory docking on the tightening process. Optionally, the step of collecting torque and angle correlation data of the tightening shaft in real time during the bolt tightening process after the intelligent docking is completed includes: Control the tightening shaft to perform the tightening action on the target bolt; During the tightening action, the real-time output torque of the tightening shaft is collected by a torque sensor; The real-time rotation angle of the tightening shaft is acquired by an angle encoder; The real-time output torque and the real-time rotation angle at the same time point are matched and bound to generate the torque-angle correlation data.

[0100] Preferably, during the bolt tightening process after the intelligent docking is completed, before controlling the tightening shaft to perform the tightening action on the target bolt, the intelligent docking completion judgment result, the docking completion status data, the flange hole position information, and the circumferential correspondence are read first. The docking completion status data is used to indicate that the flange connection part has moved to the target docking position, the flange hole position information is used to indicate the spatial position of the target flange hole, and the circumferential correspondence is used to indicate the correspondence order between the bolt mounting holes of the flange connection part and the target flange hole. Based on the docking completion status data, the flange hole position information, and the circumferential correspondence, the bolt hole position data to be tightened is determined, and the bolt already placed at the bolt hole position corresponding to the bolt hole position data to be tightened is taken as the target bolt. The bolt hole position data to be tightened continues to participate in the alignment control of the tightening shaft, so that the tightening shaft establishes a correspondence with the bolt hole position corresponding to the bolt hole position data to be tightened before performing the tightening action, instead of simply performing tightening according to a preset fixed point.

[0101] Preferably, when controlling the tightening shaft to perform the tightening action on the target bolt, a tightening shaft alignment posture is first generated based on the bolt hole position data, the target mating posture, and the end face posture requirements of the flange connection. The tightening shaft alignment posture is used to indicate the approach direction of the tightening shaft axis relative to the target bolt, the engagement direction of the tightening shaft sleeve, and the avoidance direction of the tightening shaft end face. After the tightening shaft alignment posture is formed, the tightening shaft is controlled to carry a sleeve tool that matches the head structure of the target bolt close to the target bolt, and the sleeve tool is made to form a fitting contact with the target bolt according to the tightening shaft alignment posture. The fitting contact continues to generate tightening shaft engagement state data, which is used to determine that the tightening shaft has entered an engagement state capable of transmitting rotational driving force. The tightening shaft engagement state data continues to serve as a prerequisite for starting the tightening action.

[0102] Preferably, after the tightening shaft engagement state data is generated, the tightening shaft is not immediately driven to rotate continuously at a constant speed. Instead, the bolt assembly sequence data corresponding to the target bolt, the bolt hole position data of the target bolt, and the sealing fit requirements of the target flange are read first. The bolt assembly sequence data is derived from the circumferential correspondence, and the sealing fit requirements are used to characterize the end face pressure state that the target flange and the flange connection part need to maintain during bolt tightening. Tightening action execution data is generated based on the bolt assembly sequence data, the bolt hole position data, and the sealing fit requirements. The tightening action execution data includes the tightening start state of the target bolt, the rotation direction of the tightening shaft, and the initial tightening speed of the tightening shaft. The tightening action execution data is then used to drive the tightening shaft to perform the tightening action, so that the tightening action corresponds to the bolt hole sequence, end face fit state, and sleeve engagement state.

[0103] Preferably, during the tightening action, when the torque sensor acquires the real-time output torque of the tightening shaft, the torque sensor reads the tightening reaction torque generated by the tightening shaft during the rotation of the target bolt, and converts the tightening reaction torque into torque sampling data. The torque sampling data, after zero-point offset correction and sampling jitter suppression, forms the real-time output torque. The zero-point offset correction is used to eliminate the initial torque offset when the tightening shaft is engaged under no-load conditions, and the sampling jitter suppression is used to reduce the influence of the contact gap between the sleeve tool and the head structure of the target bolt on the torque reading. The real-time output torque is further correlated with sampling clock data, which is used to identify the sampling time of the real-time output torque during the tightening action. The real-time output torque and the sampling clock data together form torque timing sampling data.

[0104] Preferably, during the tightening action, when the real-time rotation angle of the tightening shaft is acquired by the angle encoder, the angle encoder reads the rotational displacement of the tightening shaft relative to the tightening start state and converts the rotational displacement into rotation angle sampling data. The rotation angle sampling data forms the real-time rotation angle after the start angle is zeroed and the backlash is eliminated. The start angle zeroing is used to take the angle position when the sleeve tool and the target bolt form the tightening shaft engagement state data as the starting position of the rotation angle. The backlash elimination is used to exclude the invalid rotation amount generated between the sleeve tool and the target bolt in the initial engagement stage. The real-time rotation angle is further associated with the sampling clock data, and the real-time rotation angle and the sampling clock data together form the rotation angle timing sampling data.

[0105] Preferably, when matching and binding the real-time output torque and the real-time rotation angle at the same sampling moment, the sampling clock data in the torque timing sampling data is read first, and then the sampling clock data in the rotation angle timing sampling data is read. The real-time output torque and the real-time rotation angle are matched according to the principle that the sampling moments are consistent or fall within the same sampling window. The sampling window is formed based on the sampling period corresponding to the sampling clock data and is used to limit the allowed matching time range when there is a sampling moment offset between the torque timing sampling data and the rotation angle timing sampling data. When the real-time output torque and the real-time rotation angle correspond to the same sampling moment, they are directly bound as single-point torque-rotation angle data. When the real-time output torque and the real-time rotation angle correspond to different sampling moments within the same sampling window, the real-time output torque and the real-time rotation angle are time-aligned according to the sampling order within the sampling window to form the single-point torque-rotation angle data. The single-point torque-rotation angle data continues to be arranged according to the sampling order during the tightening action to generate the torque-rotation angle associated data.

[0106] Preferably, the torque-angle correlation data does not record the real-time output torque or the real-time rotation angle separately, but rather associates and saves the target bolt, the bolt hole position data to be tightened, the tightening shaft engagement state data, the single-point torque-angle data, and the sampling clock data. The target bolt is used to define the tightening object corresponding to the torque-angle correlation data; the bolt hole position data to be tightened is used to define the source of the target bolt's hole position in the target flange hole; the tightening shaft engagement state data is used to define the engagement state between the tightening shaft and the target bolt when the real-time output torque and the real-time rotation angle are generated; the single-point torque-angle data is used to express the correspondence between torque change and angle change at the same sampling moment; and the sampling clock data is used to maintain the chronological order of the single-point torque-angle data. The torque-angle correlation data formed in this way continues to participate in the feature point recognition of subsequent torque-angle correlation data, so that the subsequently identified contact point and yield point can be traced back to the corresponding target bolt and bolt hole position data to be tightened.

[0107] Preferably, in a scenario where multiple target bolts sequentially perform the tightening action, the tightening sequence among the multiple target bolts is first determined according to the circumferential correspondence, and the tightening action execution data corresponding to each target bolt is generated sequentially according to the tightening sequence. During the tightening action of each target bolt, torque timing sampling data, angle timing sampling data, and torque-angle correlation data corresponding to that target bolt are respectively generated. The torque-angle correlation data corresponding to each target bolt is further correlated according to the tightening sequence to form multi-bolt torque-angle correlation data. The multi-bolt torque-angle correlation data is used to characterize the difference in tightening state of multiple target bolts under the same target docking posture. The multi-bolt torque-angle correlation data continues to participate in the subsequent tightening stage matching between different target bolts, so that the tightening speed adjustment of the tightening shaft can be judged by combining the torque change of a single target bolt and the tightening sequence among multiple target bolts.

[0108] Preferably, in the assembly scenario where the target bolt is an external hexagonal bolt, the target bolt head structure includes an external hexagonal bolt head contour. The tightening shaft first aligns the sleeve tool with the external hexagonal bolt head contour according to the tightening shaft alignment posture, and then initiates the tightening action based on the tightening shaft engagement state data. The engagement contact between the external hexagonal bolt head contour and the sleeve tool will produce a change in engagement gap in the early stage of the tightening action. The change in engagement gap is represented by an angle change within a low torque range through the torque sampling data and the angle sampling data. After the initial angle is zeroed and the reverse gap is eliminated, the angle change within the low torque range is not used as the basis for judging whether the target bolt has entered the target bolt force tightening stage, but is recorded as part of the tightening shaft engagement state data in the torque-angle correlation data. By associating the tightening shaft engagement state data with the torque-angle correlation data, the sleeve tool engagement stage and the target bolt force tightening stage can be distinguished, reducing the possibility of misjudging the change in engagement gap as a change in bolt force.

[0109] Preferably, during the bolt tightening process after the elbow flange and the rotary joint flange face are aligned, the target flange holes are circumferentially distributed around the rotary joint flange face. The curved shape of the elbow flange will create a clearance requirement for the approach direction of the sleeve tool for some of the target bolts. After generating the bolt hole position data to be tightened based on the flange hole position information, the circumferential correspondence, and the target mating posture, the clearance direction of the tightening shaft end face in the tightening shaft alignment posture is further adjusted according to the curved shape of the elbow flange, so that the tightening shaft avoids the curved shape of the elbow flange when approaching the target bolt. The adjusted tightening shaft alignment posture continues to participate in the formation of the tightening shaft engagement state data. The tightening shaft engagement state data continues to serve as a prerequisite for collecting the real-time output torque and the real-time rotation angle, so that the torque-angle correlation data corresponds to the tightening action that has been aligned and engaged.

[0110] Preferably, during component assembly, after the target bolt is inserted into the bolt hole corresponding to the bolt hole position data to be tightened, the tightening shaft does not simply perform the tightening action based on the bolt hole position data to be tightened. Instead, it continues to read the target docking posture and the docking completion status data to determine whether the flange connection part and the target flange are still in a suitable end-face contact state for tightening. If the docking completion status data indicates that the flange connection part and the target flange are in a contact state with a low risk of flange jamming, the tightening shaft is controlled to perform the tightening action according to the tightening action execution data. If the docking completion status data indicates that a contact deviation affecting the tightening action reappears between the flange connection part and the target flange, the tightening action is paused and the process of adjusting the docking posture is returned. The docking completion status data thus continues to participate in the pre-determination of the torque angle correlation data, so that the tightening process corresponding to the torque angle correlation data is based on the actual contact state after the intelligent docking.

[0111] Optionally, the step of dynamically adjusting the tightening speed of the tightening shaft based on the feature points of the torque angle correlation data includes: A real-time monitoring curve of torque versus rotation angle is generated based on the torque-angle correlation data; The slope change of the real-time monitoring curve is analyzed to identify the contact point and yield point on the real-time monitoring curve, and the contact point and yield point are used as the feature points. The entire bolt tightening process is divided into multiple different tightening stages based on the contact point and the yield point, and a corresponding target tightening speed is matched for each tightening stage.

[0112] Preferably, when generating a real-time monitoring curve of torque versus rotation angle based on the torque-angle correlation data, the single-point torque-angle data arranged in the sampling order in the torque-angle correlation data are first read, and the real-time output torque, the real-time rotation angle, the sampling clock data, the target bolt, the bolt hole position data to be tightened, and the tightening shaft engagement state data are extracted from each single-point torque-angle data; the real-time rotation angle is used to form the angle change benchmark of the real-time monitoring curve, the real-time output torque is used to form the torque change benchmark of the real-time monitoring curve, and the sampling clock data is used to maintain adjacent... The sequential relationship between single-point torque and rotation angle data, the target bolt and the bolt to be tightened hole position data are used to define the tightening object and hole position source corresponding to the real-time monitoring curve, and the tightening shaft engagement state data are used to identify the sleeve tool engagement stage data formed when the sleeve tool has not yet stably transmitted the rotational driving force; the real-time monitoring curve generated thereby is not a curve that reflects the torque magnitude alone, but reflects the real-time output torque change process generated by the target bolt at the corresponding bolt to be tightened hole position data as the real-time rotation angle advances, and the real-time monitoring curve continues to serve as the curve source for identifying the contact point and the yield point.

[0113] Preferably, before generating the real-time monitoring curve, the torque angle correlation data is preprocessed based on the tightening shaft engagement state data. The preprocessing includes engagement stage data identification, low torque range data marking, and effective tightening data filtering. The engagement stage data identification is used to identify the engagement gap change between the sleeve tool and the target bolt head structure in the initial stage of contact. The low torque range data marking is used to mark the single-point torque angle data corresponding to the engagement gap change as the sleeve tool engagement stage data. The effective tightening data filtering is used to extract the single-point torque angle data after the target bolt has entered the target bolt force tightening stage from the torque angle correlation data. The sleeve tool engagement stage data continues to be retained as a component of the tightening shaft engagement state data. The single-point torque angle data obtained from the effective tightening data filtering continues to participate in the generation of the real-time monitoring curve, so that the real-time monitoring curve mainly reflects the correspondence between the real-time output torque and the real-time rotation angle in the target bolt force tightening stage.

[0114] Preferably, in the assembly scenario where the target bolt is an external hexagonal bolt, there is an initial freewheeling interval caused by the fitting gap between the socket tool and the head contour of the external hexagonal bolt. This initial freewheeling interval is represented in the torque angle correlation data as a data segment where the real-time rotation angle has changed but the real-time output torque has changed relatively little. After identifying the initial freewheeling interval data segment, it is associated with the tightening shaft engagement state data, rather than being used as a candidate data segment for the contact point. After the initial freewheeling interval data segment ends, the single-point torque angle data obtained from the effective tightening data filtering continues to be read, and the single-point torque angle data is sorted according to the increasing relationship of the real-time rotation angle to form the real-time monitoring curve corresponding to the target bolt's tightening stage. By incorporating the initial freewheeling interval data segment into the socket tool engagement stage data, the socket tool engagement stage and the target bolt tightening stage can be distinguished, reducing the possibility of misjudging the initial freewheeling interval data segment as the bolt contact change corresponding to the contact point.

[0115] Preferably, when analyzing the slope change of the real-time monitoring curve, the real-time monitoring curve is first divided into multiple adjacent angle analysis segments, each of which consists of multiple adjacent single-point torque angle data. The degree of change of the real-time output torque relative to the real-time rotation angle in each adjacent angle analysis segment is calculated to obtain the local torque growth state data corresponding to that adjacent angle analysis segment. The local torque growth state data is further arranged according to the sequential relationship between the sampling clock data and the real-time rotation angle to form a slope change sequence. The slope change sequence is used to express the change process of the target bolt from rapid screwing in, end face contact, preload growth to near yielding, so that the identification of the contact point and the yield point does not depend on the instantaneous magnitude of a single real-time output torque, but on the change of the local torque growth state data of the real-time monitoring curve in the continuous adjacent angle analysis segments.

[0116] Preferably, when identifying the contact point on the real-time monitoring curve, the local torque growth state data in the slope change sequence is read first, and it is determined whether the local torque growth state data has changed from a low growth state to a continuous growth state. The low growth state is used to characterize that the target bolt is still in the screw-in and pushing state in the corresponding bolt hole position data to be tightened, and the continuous growth state is used to characterize that the target bolt head structure has begun to press the flange connection part and the target flange towards the sealing and contact direction. When the continuous local torque growth state data all show a trend of transitioning from the low growth state to the continuous growth state, and the real-time rotation angle corresponding to the transition position is within the force tightening stage of the target bolt, the single-point torque angle data corresponding to the transition position is marked as the contact point candidate data. The contact point candidate data continues to be consistent with the end face contact state in the docking completion state data. If the end face contact state does not show a reappearance of contact deviation that affects the tightening action, the position corresponding to the contact point candidate data is identified as the contact point, and the contact point continues to participate in the division of the tightening stage.

[0117] Preferably, when identifying the yield point on the real-time monitoring curve, the position where the real-time output torque reaches a certain fixed torque value is not directly taken as the yield point. Instead, the slope change sequence after the contact point is read, and the position where the real-time output torque changes from a stable growth state to a slowing growth state as the real-time rotation angle continues to increase is identified based on the slope change sequence. The stable growth state is used to characterize the stage where the target bolt forms a preload between the target flange hole and the bolt mounting hole at the flange connection. The slowing growth state is used to characterize the decrease in the growth rate of the real-time output torque as the real-time rotation angle continues to increase. When the slowing growth state is maintained in multiple consecutive adjacent angle analysis segments, and the real-time output torque does not show a sudden drop caused by the slippage of the sleeve tool, the single-point torque angle data corresponding to the change position is identified as the yield point. The yield point continues to participate in the division of the tightening stage as the feature point to avoid judging the tightening end position of different target bolts based solely on a fixed torque threshold.

[0118] Preferably, during the process of identifying the contact point and the yield point, the data on the bolt hole position and the bolt assembly sequence corresponding to the target bolt are read. The bolt assembly sequence is determined by the circumferential correspondence between the target flange hole and the bolt mounting hole at the flange connection. The data on the bolt hole position and the bolt assembly sequence are then associated with the real-time monitoring curve. The data on the bolt hole position is used to distinguish the pressure differences on the flange end face at different hole positions in the target flange hole. The bolt assembly sequence is used to distinguish the sequential positions of the target bolts during the automatic tightening process. During the bolt tightening process after the elbow flange and the rotary joint flange face are connected, some of the target bolts are affected by the bending shape of the elbow flange and have different sleeve tool access conditions. The data on the bolt hole position and the bolt assembly sequence are then used to verify whether the contact point and the yield point match the force change of the corresponding hole position, so that different target bolts under the same circumferential distribution of the target flange hole can form the corresponding feature points respectively.

[0119] Preferably, when the entire bolt tightening process is divided into multiple different tightening stages based on the contact point and the yield point, the curve segment from the tightening start state to the contact point is taken as the first tightening stage, the curve segment from the contact point to the yield point is taken as the second tightening stage, and the stop control interval at the yield point and thereafter is taken as the third tightening stage. The first tightening stage corresponds to the rapid screwing process when the target bolt has not yet formed an effective end face contact; the second tightening stage corresponds to the pre-tightening growth process when the target bolt pushes the flange connection part to gradually form a sealed contact with the target flange; and the third tightening stage corresponds to the stopping rotation process after the target bolt reaches the yield point. The first, second, and third tightening stages are all defined by the single-point torque angle data on the real-time monitoring curve, so that the division of the tightening stages comes from the actual torque angle change of the target bolt, rather than from a preset fixed time period. The tightening stages continue to serve as the stage basis for matching the target tightening speed.

[0120] Preferably, when matching the target tightening speed for each tightening stage, the real-time monitoring curve, the slope change sequence, and the bolt hole position data of the target bolt are read according to the first tightening stage, the second tightening stage, and the third tightening stage, respectively; in the first tightening stage, the target tightening speed is matched to a rapid screw-in speed according to the low growth state in the real-time monitoring curve, so that the tightening shaft can complete the screw-in and advance before the target bolt applies significant pressure to the flange connection; in the third tightening stage... In the second tightening stage, the target tightening speed is matched to a deceleration preload speed based on the continuously increasing state in the slope change sequence, so that the tightening shaft reduces rotational impact after the target bolt begins to form end face pressure; in the third tightening stage, the target tightening speed is matched to a stop control speed based on the single-point torque angle data corresponding to the yield point, so that the tightening shaft stops rotating and the tightening action of the target bolt ends; the rapid screw-in speed, the deceleration preload speed, and the stop control speed together constitute the target tightening speed corresponding to the tightening stage.

[0121] Preferably, in a scenario where multiple target bolts sequentially perform the tightening action, each target bolt generates a corresponding real-time monitoring curve, slope change sequence, contact point, yield point, and tightening stage. The real-time monitoring curves corresponding to each target bolt are further correlated according to the bolt assembly sequence data to form multi-bolt curve correlation data. The multi-bolt torque angle correlation data is data formed by correlating the torque angle correlation data corresponding to each target bolt according to the bolt assembly sequence data. The multi-bolt curve correlation data is further correlated with the multi-bolt torque angle correlation data to compare the contact point appearance order, yield point appearance position, and local torque growth state data of different target bolts. When the contact point of a target bolt lags behind the contact points of other target bolts in the same assembly position, the hole position data of the bolt to be tightened and the docking completion state data are used to determine whether there is insufficient end face contact at the hole position of the target bolt, and the determination result is used to adjust the target tightening speed and bolt assembly sequence data of subsequent target bolts.

[0122] Preferably, in the assembly scenario after the elbow flange and the rotary joint flange face are mated, the target flange holes are circumferentially distributed around the rotary joint flange face, and the curved shape of the elbow flange creates a clearance requirement for the approach direction of the sleeve tool for some of the target bolts; therefore, the generation of the real-time monitoring curve and the identification of the feature points are also verified in conjunction with the alignment posture of the tightening shaft and the engagement status data of the tightening shaft. When the alignment posture of the tightening shaft shows that the sleeve tool has completed the approach according to the clearance direction of the tightening shaft end face, and the engagement status data of the tightening shaft shows... Only when the sleeve tool has formed an engagement state with the target bolt capable of transmitting rotational driving force is the corresponding torque angle correlation data incorporated into the real-time monitoring curve; thus, the contact point and the yield point can correspond to the target bolt under tension tightening stage after the avoidance approach is completed, reducing curve misjudgment caused by local interference of the elbow flange or insufficient engagement of the sleeve tool; the real-time monitoring curve, the contact point, and the yield point continue to be correlated with the hole position data of the bolt to be tightened, so as to maintain the source of hole positions of each target bolt in the elbow flange assembly scenario.

[0123] Preferably, during the flange assembly process of different components, each component corresponds to a different flange connection position in the loading arm assembly. The target bolts establish a hole position relationship with the corresponding target flange holes through the bolt hole position data to be tightened. The real-time monitoring curve, the contact point, the yield point, the tightening stage, and the target tightening speed are all associated and saved with the bolt hole position data to be tightened. When it is necessary to determine the tightening status of different flange connection positions in the same loading arm assembly, the real-time monitoring curve corresponding to the bolt hole position data to be tightened is read first, then the contact point and the yield point corresponding to the real-time monitoring curve are read, and then the tightening stage and the target tightening speed defined by the contact point and the yield point are read. Through the above association and saving method, the feature point recognition result of the torque angle association data can be traced back to the corresponding target bolt, target flange hole, and tightening action, so that the tightening speed adjustment of the tightening shaft corresponds to the hole alignment, end face contact, and partial automatic tightening process in the loading arm flange assembly scenario.

[0124] Optionally, the step of dividing the entire bolt tightening process into multiple different tightening stages based on the contact point and the yield point, and matching a corresponding target tightening speed for each tightening stage, includes: In the first tightening stage before reaching the contact point, the tightening speed of the tightening shaft is set to a preset high-speed value as the target tightening speed; In the second tightening stage, as the screw passes the contact point and approaches the yield point, the tightening speed of the tightening shaft is gradually reduced so that the target tightening speed in the second tightening stage is less than the preset high speed value. In the third tightening stage when the yield point is reached, the tightening shaft is controlled to stop rotating in order to achieve a uniform distribution of preload on each of the target bolts.

[0125] Preferably, when dividing the entire bolt tightening process into multiple different tightening stages based on the contact point and the yield point, the real-time monitoring curve, the slope change sequence, the contact point, the yield point, the tightening start state, the target bolt, the bolt hole position data to be tightened, the bolt assembly sequence data, and the docking completion state data are read first, and the contact point and the yield point are used as stage boundary data on the real-time monitoring curve; the stage boundary data is used to define the start and end positions of the first tightening stage, the second tightening stage, and the third tightening stage on the real-time monitoring curve, the tightening start state is used to define the start position of the first tightening stage, and the bolt hole position data to be tightened are used to define the starting position of the first tightening stage. The position data is used to define the source of the target bolt's hole position in the target flange hole; the bolt assembly sequence data is used to define the sequential position of the target bolt during the automatic tightening process; and the docking completion status data is used to define the end face contact state between the flange connection part and the target flange when the tightening stage is divided. Therefore, the tightening stage is not divided according to a preset fixed time, but according to the torque angle change process formed by the target bolt at the corresponding bolt hole position data. The torque angle change process is further expressed by the real-time monitoring curve, the slope change sequence, the contact point, and the yield point. The tightening stage continues to serve as the stage basis for matching the target tightening speed.

[0126] Preferably, when dividing the first tightening stage, the tightening start state is used as the starting boundary of the first tightening stage, and the contact point is used as the ending boundary of the first tightening stage. The single-point torque angle data between the tightening start state and the contact point is read to form the first tightening stage data. The first tightening stage data corresponds to the process of the target bolt being screwed into the bolt hole corresponding to the bolt hole position data to be tightened. During this process, the real-time rotation angle continuously increases, while the real-time output torque remains in a low-growth state. This low-growth state is used to characterize the target bolt being screwed into the bolt hole. Before the target bolt has formed an effective end-face pressing state with the flange connection and the target flange, the effective end-face pressing state is used to distinguish between the screwing-in process of the target bolt and the pre-tightening process of the target bolt. After the data of the first tightening stage is confirmed to belong to the screwing-in process of the target bolt, the tightening speed of the tightening shaft is set to the preset high-speed value as the target tightening speed, so that the tightening shaft completes the screwing-in process before the target bolt has formed the effective end-face pressing state. The preset high-speed value continues to participate in the speed reduction process of the second tightening stage.

[0127] Preferably, the preset high-speed value in the first tightening stage is not a preset fixed speed directly invoked from the hole position of the target bolt, but rather is read and confirmed based on the bolt hole position data, the bolt assembly sequence data, the tightening shaft engagement state data, and the data of the first tightening stage, to form the speed confirmation data for the first tightening stage; when the tightening shaft engagement state data indicates that the sleeve tool has formed an engagement state with the target bolt capable of transmitting rotational driving force, and the real-time output torque in the first tightening stage data maintains the low-growth state, the first tightening stage... The speed confirmation data confirms the preset high-speed value as the target tightening speed of the first tightening stage, and makes the target tightening speed of the first tightening stage participate in the tightening speed control of the tightening shaft; when torque fluctuations caused by insufficient engagement of the sleeve tool appear in the data of the first tightening stage, the single-point torque angle data corresponding to the torque fluctuation is included in the data of the sleeve tool engagement stage, instead of using the torque fluctuation as a basis for reducing the target tightening speed of the first tightening stage, so that the target tightening speed of the first tightening stage corresponds to the actual screwing and pushing process of the target bolt.

[0128] Preferably, when dividing the second tightening stage, the contact point is used as the starting boundary of the second tightening stage, and the yield point is used as the ending boundary of the second tightening stage. The single-point torque angle data between the contact point and the yield point is read to form the second tightening stage data. The second tightening stage data corresponds to the pre-tightening growth process after the target bolt head structure starts pressing the flange connection and the target flange towards the sealing contact direction. During the pre-tightening growth process, the real-time output torque enters a continuous growth state from the low growth state as the real-time rotation angle advances. The continuous growth state is further expressed by the slope change sequence. Based on the second tightening stage data and the slope change sequence, the tightening speed of the tightening shaft is gradually reduced from the preset high-speed value so that the target tightening speed of the second tightening stage is less than the preset high-speed value. The target tightening speed of the second tightening stage continues to participate in the pre-tightening growth process of the target bolt.

[0129] Preferably, the gradual reduction in the second tightening stage is not performed according to a preset fixed deceleration time, but rather by segmenting and matching the target tightening speed based on the local torque growth state data in the slope change sequence to form second tightening stage speed matching data. When the local torque growth state data indicates that the real-time output torque has just entered the continuous growth state from the low growth state, the second tightening stage speed matching data reduces the target tightening speed from the preset high-speed value to a first deceleration pre-tightening speed. The first deceleration pre-tightening speed is used to reduce the rotational impact when the target bolt just enters the end face for pressing. When the local torque growth state data indicates that the real-time output torque continues to grow with the real-time rotation angle, the second tightening stage speed matching data further reduces the first deceleration preload speed to a second deceleration preload speed. The second deceleration preload speed is used to ensure that the target bolt maintains a relatively stable torque growth during the process of forming preload between the target flange hole and the bolt mounting hole at the flange connection. Both the first deceleration preload speed and the second deceleration preload speed belong to the target tightening speed of the second tightening stage and continue to be constrained by the yield point identification result.

[0130] Preferably, in the second tightening stage, the target tightening speed is further corrected based on the bolt hole position data and the bolt assembly sequence data to form second tightening stage hole position correction data. For the assembly scenario after the elbow flange and rotary joint flange face are mated, the target flange holes are circumferentially distributed around the rotary joint flange face. The curved shape of the elbow flange creates an avoidance requirement for the approach direction of the sleeve tool for some of the target bolts. The bolt hole position data is used to identify whether the target bolt is located in a hole position requiring end-face avoidance direction adjustment. The bolt assembly sequence data is used to identify the tightening sequence of the target bolt among multiple target bolts. When the hole position corresponding to the bolt hole position data is significantly affected by the curved shape of the elbow flange, the second tightening stage hole position correction data, combined with the tightening shaft alignment posture and the tightening shaft engagement state data, corrects the target tightening speed in the second tightening stage, ensuring that the target tightening speed in the second tightening stage corresponds to the sleeve tool approach conditions and end-face contact state of the hole where the target bolt is located.

[0131] Preferably, when dividing the third tightening stage, the position corresponding to the yield point is used as the starting boundary of the third tightening stage, and the single-point torque angle data, the real-time output torque, the real-time rotation angle, the sampling clock data, and the slope change sequence corresponding to the yield point are read to form the third tightening stage data; the third tightening stage data corresponds to the process of the real-time output torque changing from a stable growth state to a slowing growth state as the real-time rotation angle continues to increase. The stable growth state is used to characterize that the target bolt is already in the pre-tightening growth process, and the slowing growth state is used to characterize that the growth rate of the real-time output torque begins to decrease when the real-time rotation angle continues to increase; when the third tightening stage data is consistent with the yield point identification result, and the real-time output torque does not show a sudden drop caused by the slippage of the sleeve tool, the tightening shaft is controlled to stop rotating according to the third tightening stage data, and the target tightening speed corresponding to the stop of rotation is used as the target tightening speed of the third tightening stage, so that the tightening action of the target bolt ends at the position corresponding to the yield point.

[0132] Preferably, when controlling the tightening shaft to stop rotating, a stop command is not only sent to the tightening shaft, but the consistency judgment is first made between the single-point torque angle data corresponding to the yield point and the data of the third tightening stage to form a consistency judgment result, and then stop control data is generated based on the consistency judgment result. The stop control data includes the target bolt, the hole position data of the bolt to be tightened, the yield point, the third tightening stage, the target tightening speed, and the tightening end state data. The tightening end state data is used to indicate that the target bolt has completed the tightening action. The stop control data continues to be associated with and saved with the torque angle association data, so that when the torque angle association data is read later, the third tightening stage and the tightening end state data of the target bolt can be read at the same time, thereby tracing the position where the tightening shaft stops rotating to the corresponding target bolt and bolt hole position data.

[0133] Preferably, in a scenario where multiple target bolts sequentially perform the tightening action, each target bolt forms a corresponding first tightening stage, a second tightening stage, a third tightening stage, and a corresponding target tightening speed. The first tightening stage, the second tightening stage, and the third tightening stage corresponding to each target bolt are further associated according to the bolt assembly sequence data to form multi-bolt stage association data. The multi-bolt stage association data is further correlated with the multi-bolt torque angle association data to compare the order of appearance of the contact point, the position of the yield point, the local torque growth state data of the second tightening stage, and the tightening end state data of the third tightening stage for different target bolts. When the contact point of a target bolt lags behind the contact points of other target bolts in the same flange connection position, the hole position data of the bolt to be tightened and the docking completion state data are used to determine whether there is insufficient end face contact at the hole position of the target bolt, and the insufficient end face contact is used to adjust the target tightening speed of the subsequent target bolts.

[0134] Preferably, during the assembly of elbow flanges and rotary joints, the assembly of loading arm flange connectors with straight pipe assemblies that have completed previous assembly, the assembly of loading arm flange connectors with rotary joints that have completed previous assembly, and the assembly between multiple loading arm flange connectors that have completed previous assembly, the target bolts establish a hole position relationship with the corresponding target flange holes through the bolt hole position data to be tightened. The first tightening stage, the second tightening stage, the third tightening stage, and the target tightening speed are all associated and saved with the bolt hole position data to be tightened. During the above flange assembly process, the automatic bolt assembly robot will assemble multiple... The target bolts are automatically tightened in sections. The multi-bolt stage correlation data generated during the automatic tightening process is used to determine the preload distribution state of each target bolt. After each target bolt in the same flange connection position has formed the corresponding third tightening stage and tightening end state data, the yield point and the local torque growth state data of the second tightening stage corresponding to each target bolt are read to determine whether the preload of each target bolt is in a similar preload distribution state, and the similar preload distribution state is used as the basis for judging the uniform distribution of the preload of the target bolts.

[0135] Preferably, in the assembly scenario where the target bolt is an external hexagonal bolt, the engagement contact state between the socket tool and the head structure of the target bolt will affect the matching of the target tightening speed in the first tightening stage and the second tightening stage. Therefore, before matching the target tightening speed, the tightening shaft engagement state data and the socket tool engagement stage data are read again. The socket tool engagement stage data is used to exclude the initial idle interval data segment when the target bolt has not yet entered the target bolt force tightening stage. The tightening shaft engagement state data is used to confirm that the socket tool can transmit rotational driving force to the target bolt. When the initial idle interval data segment has been included in the socket tool engagement stage data, and the tightening shaft engagement state data indicates that the engagement state meets the tightening action start condition, the tightening stage is divided according to the contact point and the yield point, and the target tightening speed is matched for the tightening stage to reduce the situation where the socket tool engagement process is mistakenly regarded as the first tightening stage or the second tightening stage.

[0136] like Figure 3 As shown, this embodiment of the present application discloses an intelligent docking and adaptive tightening control device for loading arm flanges based on multi-sensor fusion, which includes: The visual positioning and initial coarse adjustment module is used to acquire flange images collected by a two-dimensional vision sensor, extract flange hole position information based on the flange images, and control the robotic arm to perform initial coarse positioning of the flange based on the flange hole position information. The force-sensing docking and admittance control module is used to acquire contact force feedback information collected by a six-dimensional torque sensor installed at the end of the robotic arm during the contact phase after the initial coarse positioning is completed. The bolt adaptive tightening control module is used to establish an admittance control model based on the contact force feedback information, and adjust the docking posture of the robotic arm based on the admittance control model to complete the intelligent docking of the flange; during the bolt tightening process after the intelligent docking is completed, the torque-angle correlation data of the tightening shaft is collected in real time, and the tightening speed of the tightening shaft is dynamically adjusted according to the feature points of the torque-angle correlation data to achieve adaptive tightening control.

[0137] like Figure 4 As shown, an electronic device according to an embodiment of this application includes: a processor and a memory; the memory is used to store computer programs; the processor is used to execute the program stored in the memory to implement the functions of each module of the device described in the embodiment of this application, or to implement the steps of the method described.

[0138] Figures 2-4 For an exemplary description, please refer to the above. Figure 1 This will not be elaborated upon here.

Claims

1. A method for intelligent docking and adaptive tightening control of loading arm flanges based on multi-sensor fusion, characterized in that, Includes the following steps: The flange image is acquired by a two-dimensional vision sensor, the flange hole position information is extracted from the flange image, and the robotic arm is controlled to perform initial coarse positioning of the flange based on the flange hole position information. During the contact phase after the initial coarse positioning is completed, contact force feedback information is acquired by a six-dimensional torque sensor installed at the end of the robotic arm. An admittance control model is established based on the contact force feedback information, and the docking posture of the robotic arm is adjusted based on the admittance control model to complete the intelligent docking of the flange. During the bolt tightening process after the intelligent docking is completed, the torque-angle correlation data of the tightening shaft is collected in real time. The tightening speed of the tightening shaft is dynamically adjusted according to the feature points of the torque-angle correlation data to achieve adaptive tightening control.

2. The intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion as described in claim 1, characterized in that, The steps of extracting flange hole position information from the flange image and controlling the robotic arm to perform initial coarse positioning of the flange based on the flange hole position information include: Feature contour extraction is performed on the flange image to obtain the initial contour data of the target flange; The target flange hole on the target flange is identified based on the initial contour data, and the two-dimensional coordinate data of the target flange hole is extracted as the flange hole position information; The two-dimensional coordinate data is converted into three-dimensional spatial coordinate data in the base coordinate system of the robotic arm; An initial movement trajectory is generated based on the spatial three-dimensional coordinate data, and the robotic arm is controlled to move to the preset docking preparation position according to the initial movement trajectory to complete the initial coarse positioning of the flange.

3. The intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion as described in claim 1, characterized in that, The step of acquiring contact force feedback information collected by the six-dimensional torque sensor installed at the end of the robotic arm includes: During the docking operation, the raw force signals output by the six-dimensional torque sensor are periodically read. The original force signal is smoothed to obtain a smoothed force signal. The smoothed force signal is then transformed according to a preset coordinate system mapping rule to generate three-dimensional force data and three-dimensional torque data in the coordinate system of the robotic arm end effector. The three-dimensional force data and the three-dimensional torque data are combined to form the contact force feedback information.

4. The intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion as described in claim 3, characterized in that, The step of establishing the admittance control model based on the contact force feedback information includes: Obtain the ideal stress state data of the flange during the docking process; Calculate the force deviation between the contact force feedback information and the ideal force state data; Obtain the preset mass parameters, preset damping parameters, and preset stiffness parameters of the robotic arm; A multi-dimensional compliance mapping relationship is constructed based on the preset mass parameters, the preset damping parameters, the preset stiffness parameters, and the force deviation value, and the multi-dimensional compliance mapping relationship is used as the admittance control model.

5. The intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion as described in claim 4, characterized in that, The step of adjusting the docking posture of the robotic arm based on the admittance control model to complete the intelligent docking of the flange includes: The force deviation value is input into the admittance control model so that the robotic arm exhibits the compliant physical characteristics of spring damping, and the pose compensation amount of the robotic arm is output. The current docking pose of the robotic arm is obtained, and the current docking pose is corrected according to the pose compensation amount to obtain the target docking pose; Control the robotic arm to move to the target docking position.

6. The intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion as described in claim 5, characterized in that, After the step of controlling the robotic arm to move to the target docking pose, the method further includes: Real-time monitoring of new contact force feedback information after moving to the target docking pose; If the new contact force feedback information is greater than the preset collision threshold, it is determined that there is a risk of flange jamming. The robotic arm is then controlled to retreat according to the preset retraction trajectory and the docking posture is readjusted. If the new contact force feedback information is less than the preset fitting threshold, then the intelligent docking of the flange is determined to be complete.

7. The intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion as described in claim 1, characterized in that, The step of collecting real-time torque-angle correlation data of the tightening shaft during the bolt tightening process after the intelligent docking is completed includes: Control the tightening shaft to perform the tightening action on the target bolt; During the tightening action, the real-time output torque of the tightening shaft is collected by a torque sensor; The real-time rotation angle of the tightening shaft is acquired by an angle encoder; The real-time output torque and the real-time rotation angle at the same time point are matched and bound to generate the torque-angle correlation data.

8. The intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion as described in claim 7, characterized in that, The step of dynamically adjusting the tightening speed of the tightening shaft based on the feature points of the torque-angle correlation data includes: A real-time monitoring curve of torque versus rotation angle is generated based on the torque-angle correlation data; The slope change of the real-time monitoring curve is analyzed to identify the contact point and yield point on the real-time monitoring curve, and the contact point and yield point are used as the feature points. The entire bolt tightening process is divided into multiple different tightening stages based on the contact point and the yield point, and a corresponding target tightening speed is matched for each tightening stage.

9. The intelligent docking and adaptive tightening control method for loading arm flanges based on multi-sensor fusion as described in claim 8, characterized in that, The step of dividing the entire bolt tightening process into multiple different tightening stages based on the contact point and the yield point, and matching a corresponding target tightening speed for each tightening stage, includes: In the first tightening stage before reaching the contact point, the tightening speed of the tightening shaft is set to a preset high-speed value as the target tightening speed; In the second tightening stage, as the screw passes the contact point and approaches the yield point, the tightening speed of the tightening shaft is gradually reduced so that the target tightening speed in the second tightening stage is less than the preset high speed value. In the third tightening stage when the yield point is reached, the tightening shaft is controlled to stop rotating in order to achieve a uniform distribution of preload on each of the target bolts.

10. A smart docking and adaptive tightening control device for loading arm flanges based on multi-sensor fusion, characterized in that, include: The visual positioning and initial coarse adjustment module is used to acquire flange images collected by a two-dimensional vision sensor, extract flange hole position information based on the flange images, and control the robotic arm to perform initial coarse positioning of the flange based on the flange hole position information. The force-sensing docking and admittance control module is used to acquire contact force feedback information collected by a six-dimensional torque sensor installed at the end of the robotic arm during the contact phase after the initial coarse positioning is completed. The bolt adaptive tightening control module is used to establish an admittance control model based on the contact force feedback information, and adjust the docking posture of the robotic arm based on the admittance control model to complete the intelligent docking of the flange; during the bolt tightening process after the intelligent docking is completed, the torque-angle correlation data of the tightening shaft is collected in real time, and the tightening speed of the tightening shaft is dynamically adjusted according to the feature points of the torque-angle correlation data to achieve adaptive tightening control.

Images