Automatic alignment method and device for internal welding machine and internal welding machine

Abstract

The invention provides an automatic alignment method and device for an internal welding machine and the internal welding machine, and relates to the technical field of intelligent pipeline welding. When the internal welding machine advances in a pipeline and makes contact with a bent pipe, driving force of a plurality of branch chains is calculated through an impedance controller according to the actual position of a posture adjusting mechanism; and the conical head mechanism is driven to deflect and turn. The current actual thrust is predicted through a prediction model, the pose adjustment amount is calculated through the current actual thrust in combination with a real-time dynamic positive solution model and an admittance controller, and therefore the length of each branch chain is adjusted, and the alignment guide block reaches the groove limit; after alignment is completed, the rear expansion shoe is lifted to fix the relative position of the conical head mechanism and the groove, and the alignment guide block is withdrawn; when another pipeline groove is in butt joint, the front expansion shoe is lifted to fix the conical head and the groove so as to facilitate welding. The contact state of the internal welding machine and the pipeline is adjusted in real time, adaptability and stability are improved, high-precision automatic alignment is achieved, and welding quality is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent pipeline welding, and particularly to an automatic pipe alignment method, device and internal welding machine for an internal welding machine. Background Art

[0002] Traditional internal welding machines face great challenges in the automatic welding of bent pipes, especially the problems of pipe alignment positioning and attitude adjustment for bent pipes. Due to the bending radius and deformation of the pipes, it is difficult for the mechanical structure of the internal welding machine to adapt, resulting in the inability to accurately dock the pipe orifices, and usually manual intervention is required for positioning and correction. In addition, errors in the pipes (such as the perpendicularity error of groove shaping) also make the internal welding machine lack flexibility and autonomy in attitude adjustment, which has a negative impact on welding accuracy and efficiency.

[0003] The prior art usually relies on sensors such as lasers, vision, and rangefinders to assist the internal welding machine in precise pipe alignment positioning and attitude adjustment. However, the working performance of these sensors is greatly affected by the external environment (such as strong light, dust, high temperature, etc.), resulting in frequent measurement errors and equipment failures. In addition, this solution relying on a complex sensor system not only increases the equipment cost, but also increases the difficulty of maintenance and operation, and its adaptability is poor, and it cannot work stably in complex terrains and extreme environments.

[0004] Therefore, solving the problem of precise pipe alignment of the internal welding machine in a complex environment is still a technical problem to be urgently solved. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an automatic pipe alignment method, device and internal welding machine for an internal welding machine to solve the problem that it is difficult for the internal welding machine to accurately align the pipes in a complex environment.

[0006] To achieve the above object, embodiments of the present invention provide the following technical solutions:

[0007] The first aspect of the present invention discloses an automatic pipe alignment method for an internal welding machine, and the method includes:

[0008] When the internal welding machine travels inside the pipe to a bent pipe through the body movement system, an impedance controller calculates the driving forces of multiple branched chains in the attitude adjustment mechanism according to the actual position of the attitude adjustment mechanism; the control strategy of the impedance controller is pre-created according to the target inertia, target stiffness value and target damping;

[0009] Drive the cone head mechanism in the internal welding machine to deflect according to the driving forces so that the internal welding machine passes through the bent pipe;

[0010] When the internal welding machine travels to the pipe orifice end, raise the pipe alignment guide block to perform a pipe alignment operation with the groove of the pipe orifice end, and the pipe alignment operation includes:

[0011] Using a prediction model, based on the motion data and current data of the branch chains in the attitude adjustment mechanism, predict the current actual thrust of each branch chain; the prediction model is pre-trained by using a six-axis dynamics model on a neural network model; the six-axis dynamics model is pre-constructed according to a six-axis platform;

[0012] Through a real-time dynamics forward solution model, according to the current actual thrust of each branch chain, calculate the load force and load torque in the moving platform coordinate system; the six-axis platform includes a moving platform and a static platform; the moving platform coordinate system is a coordinate system constructed according to the moving platform;

[0013] Using an admittance controller, according to the load force and the load torque, calculate the pose adjustment amount of the attitude adjustment mechanism, and adjust the lengths of the branch chains during the butt joint process, so that the butt joint guiding block reaches the groove limit position; the admittance controller is pre-established based on a second-order differential model;

[0014] After completing the butt joint operation, raise the rear expansion shoe to fix the relative position of the cone head mechanism and the groove, and retract the butt joint guiding block;

[0015] When the groove of another pipeline coincides with the groove, raise the front expansion shoe to fix the relative position of the cone head mechanism and the groove of the other pipeline, so that the welding torch in the cone head mechanism welds the two pipelines.

[0016] Preferably, the process of pre-constructing a six-axis dynamics model according to a six-axis platform includes:

[0017] Define the structure of the six-axis platform, and the structure at least includes a static platform, a moving platform, the cylinder barrel of the branch chain, and the piston rod of the branch chain;

[0018] Based on the static platform coordinate system, define the motion physical quantities, and combine the force data of the six-axis platform to create the Jacobian matrix of the moving platform, the Jacobian matrix of the cylinder barrel of the branch chain, and the Jacobian matrix of the piston rod of the branch chain; the static platform coordinate system is a coordinate system constructed according to the static platform;

[0019] Construct the inertia matrix of the cone head mechanism, the inertia matrix of the cylinder barrel of the branch chain, and the inertia matrix of the piston rod of the branch chain;

[0020] Combine all the Jacobian matrices and all the inertia matrices to construct a six-axis dynamics model.

[0021] Preferably, the process of pre-training a neural network model by using a six-axis dynamics model to obtain the prediction model includes:

[0022] Sample a plurality of loads from a preset load space, and for each load, obtain the corresponding motion sequence data from a preset position space and a preset dynamic space;

[0023] Input all loads and all motion sequence data into a six-axis dynamics model to obtain initial training data including branch chain motion data, current data, and branch chain thrust data, and preprocess the initial training data to obtain training data;

[0024] Input the training data into a long short-term memory deep neural network model to output the predicted value of the branch chain thrust;

[0025] Calculate the loss function of the long short-term memory deep neural network model according to the predicted value;

[0026] If the loss function converges, determine the long short-term memory deep neural network model as the trained prediction model;

[0027] If the loss function does not converge, adjust the parameters of the long short-term memory deep neural network model, and return to execute the step of inputting the training data into the long short-term memory deep neural network model to output the predicted value of the branch chain thrust.

[0028] Preferably, the method further includes:

[0029] Obtain the branch chain driving force data of the attitude adjustment mechanism in multiple different postures;

[0030] According to multiple groups of branch chain driving force data, calculate the load force and load torque in the moving platform coordinate system through a real-time dynamics forward solution model; the moving platform coordinate system is a coordinate system established according to the moving platform in the attitude adjustment mechanism;

[0031] Establish a linear equation of the load force, load torque, cone head weight, and center of gravity coordinates received by the moving platform of the attitude adjustment mechanism;

[0032] Calculate the cone head weight and center of gravity coordinates according to the least squares method and the linear equation;

[0033] Correspondingly, the use of the impedance controller to calculate the driving forces of multiple branch chains in the attitude adjustment mechanism according to the actual position of the attitude adjustment mechanism includes:

[0034] Obtain the actual position of the attitude adjustment mechanism, where the actual position includes the angle between each branch chain of the attitude adjustment mechanism and gravity, the deflection data, and the offset data of the branch chain;

[0035] Use the real-time dynamics forward solution model to calculate the gravity compensation amount of the moving platform to the cone head in the moving platform lower coordinate system according to the angle, the deflection data, the offset data, the cone head weight, and the center of gravity coordinates;

[0036] Through the real-time kinematic inverse solution model, convert the gravity compensation amount of the moving platform on the conical head into the gravity compensation amount of the support chain on the conical head in the current posture; the gravity compensation amount of the support chain on the conical head is the force exerted by the gravity of the conical head on each support chain.

[0037] Input the included angle, the deflection data, the offset data, and the gravity compensation amount of the support chain on the conical head into the impedance controller, and calculate the driving forces of multiple support chains in the posture adjustment mechanism.

[0038] Preferably, the using the admittance controller to calculate the pose adjustment amount of the pose adjustment mechanism according to the load force and the load torque, and adjusting the lengths of the respective support chains during the butt joint process so that the butt joint guiding block reaches the groove limit position includes:

[0039] Using the admittance controller to calculate the pose adjustment amount of the pose adjustment mechanism according to the load force, the load torque, and the gravity compensation amount of the moving platform on the conical head;

[0040] Using the position controller to adjust the lengths of the respective support chains during the butt joint process according to the pose adjustment amount so that the butt joint guiding block reaches the groove limit position; the position controller is pre-established based on a proportional-derivative controller.

[0041] Preferably, the using the admittance controller to calculate the pose adjustment amount of the pose adjustment mechanism according to the load force, the load torque, and the gravity compensation amount of the moving platform on the conical head includes:

[0042] Calculate the difference between the load force and the load torque minus the gravity compensation amount of the moving platform on the conical head to obtain the actual contact resultant force of several butt joint guiding blocks in the conical head mechanism in contact with the groove;

[0043] Obtain the actual positions and actual speeds of the respective support chains during the butt joint process;

[0044] Input the preset desired contact resultant force, the actual contact resultant force, the actual positions, and the actual speeds into the admittance controller;

[0045] Combined with the ordinary differential equation set in the admittance controller, perform iterative solution according to the time step of the control period to obtain the pose adjustment amount of the pose adjustment mechanism.

[0046] Preferably, the using the position controller to adjust the lengths of the respective support chains during the butt joint process according to the pose adjustment amount includes:

[0047] Using the position controller to calculate the butt joint output force of the pose adjustment mechanism according to the pose adjustment amount;

[0048] Adjust the lengths of the respective support chains during the butt joint process through the butt joint output force.

[0049] The second aspect of the present invention discloses an automatic pipe alignment device for an internal welding machine, and the device includes:

[0050] A first calculation unit, configured to, when the internal welding machine travels to a bend in the pipeline through the fuselage movement system, calculate the driving forces of multiple branch chains in the attitude adjustment mechanism by using an impedance controller according to the actual position of the attitude adjustment mechanism; the control strategy of the impedance controller is pre-created according to the target inertia, target stiffness value, and target damping;

[0051] A driving unit, configured to drive the cone head mechanism in the internal welding machine to deflect according to the driving force, so that the internal welding machine passes through the bend;

[0052] A pipe alignment unit, configured to, when the internal welding machine travels to the pipe orifice end, raise the pipe alignment guiding block and the bevel of the pipe orifice end, and perform pipe alignment operations based on a prediction unit, a second calculation unit, and an adjustment unit;

[0053] A prediction unit, configured to predict the current actual thrust of each branch chain by using a prediction model based on the branch chain movement data and current data in the attitude adjustment mechanism; the prediction model is pre-trained by using a six-axis dynamics model for a neural network model; the six-axis dynamics model is pre-constructed according to a six-axis platform;

[0054] A second calculation unit, configured to calculate the load force and load torque in the moving platform coordinate system according to the current actual thrust of each branch chain through a real-time dynamics forward solution model; the six-axis platform includes a moving platform and a static platform;

[0055] An adjustment unit, configured to calculate the pose adjustment amount of the attitude adjustment mechanism according to the load force and the load torque by using an admittance controller, and adjust the lengths of the branch chains during the pipe alignment process, so that the pipe alignment guiding block reaches the bevel limit position; the admittance controller is pre-established based on a second-order differential model;

[0056] A first fixing unit, configured to, after completing the pipe alignment operation, raise the rear expansion shoe to fix the relative position of the cone head mechanism and the bevel, and retract the pipe alignment guiding block;

[0057] A second fixing unit, configured to, when the bevel of another pipeline coincides with the bevel, raise the front expansion shoe to fix the relative position of the cone head mechanism and the bevel of the other pipeline, so that the welding torch in the cone head mechanism welds the two pipelines.

[0058] Preferably, the device further includes:

[0059] A definition unit, configured to define the structure of the six-axis platform, and the structure at least includes a static platform, a moving platform, the cylinder barrel of the branch chain, and the piston rod of the branch chain;

[0060] A first creation unit, configured to define kinematic physical quantities based on a static platform coordinate system, and create a Jacobian matrix of a moving platform, a Jacobian matrix of a cylinder of a link, and a Jacobian matrix of a piston rod of the link in combination with the force data of the six-axis platform; the static platform coordinate system is a coordinate system constructed according to the static platform;

[0061] A second creation unit, configured to construct an inertia matrix of a cone head mechanism, an inertia matrix of a cylinder of a link, and an inertia matrix of a piston rod of the link;

[0062] A construction unit, configured to construct a six-axis dynamics model by combining all the Jacobian matrices and all the inertia matrices.

[0063] A third aspect of the present invention discloses an internal welding machine, including a cone head mechanism, a fuselage motion system, an attitude adjustment mechanism, a memory, and a processor;

[0064] The cone head mechanism includes a butt joint guiding block;

[0065] The attitude adjustment mechanism is located between the cone head mechanism and the fuselage motion system; the attitude adjustment mechanism includes at least a plurality of links and attitude guiding wheels;

[0066] The memory stores a computer program, and when the processor calls the computer program in the memory, the steps of the internal welding machine automatic butt joint method disclosed in the first aspect of the present invention are implemented.

[0067] Based on the internal welding machine automatic butt joint method, device and internal welding machine provided in the above embodiments of the present invention, when the internal welding machine travels in a pipeline and contacts a bent pipe, an impedance controller is used to calculate the driving forces of a plurality of links according to the actual position of the attitude adjustment mechanism; the cone head mechanism is driven to deflect and turn; when reaching the pipe orifice end, the butt joint guiding block is lifted to perform a groove butt joint operation: a prediction model obtained by training a neural network using a six-axis dynamics model is used to predict the current actual thrust, and a real-time dynamics forward solution model is used to calculate the load force and load torque in the moving platform coordinate system according to the current actual thrust of each link; a admittance controller is used to calculate the pose adjustment amount of the attitude adjustment mechanism according to the load force and load torque in the moving platform coordinate system, so as to adjust the lengths of the links, so that the butt joint guiding block reaches the groove limit; after the butt joint is completed, the rear expansion shoe is lifted to fix the relative position of the cone head mechanism and the groove and the butt joint guiding block is retracted; when the grooves of another pipeline are butted, the front expansion shoe is lifted to fix the cone head and the groove for welding. The contact state between the internal welding machine and the pipeline is adjusted in real time, the adaptability and stability are improved, high-precision automatic butt joint is realized, and the welding quality is improved. Description of the Drawings

[0068] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0069] Figure 1 Structural schematic diagram of an internal welding machine provided by an embodiment of the present invention;

[0070] Figure 2 Schematic diagram of the butt joint guiding block and the bevel provided by an embodiment of the present invention;

[0071] Figure 3 Flowchart of an automatic butt joint method for an internal welding machine provided by an embodiment of the present invention;

[0072] Figure 4 Flowchart of the butt joint operation provided by an embodiment of the present invention;

[0073] Figure 5 Control flowchart provided by an embodiment of the present invention;

[0074] Figure 6 Schematic diagram of the data curve of the control process of the automatic butt joint method for an internal welding machine provided by an embodiment of the present invention;

[0075] Figure 7 Schematic diagram of the change of the pose of the six-axis platform over time in the moving platform coordinate system (b system) provided by an embodiment of the present invention;

[0076] Figure 8 Schematic diagram of the change of the contact force of the six-axis platform over time in the static platform coordinate system (w system) provided by an embodiment of the present invention;

[0077] Figure 9 Schematic diagram of the relationship between the distance between the butt joint device and the bevel and time provided by an embodiment of the present invention;

[0078] Figure 10 Schematic diagram of the relationship between the magnitude of the contact force between the butt joint device and the bevel and time provided by an embodiment of the present invention;

[0079] Figure 11 Structural block diagram of an automatic butt joint device for an internal welding machine provided by an embodiment of the present invention;

[0080] Among them, 1 is the cone head mechanism, 2 is the fuselage motion system, 3 is the force sensor, 4 is the support chain, 5 is the attitude guiding wheel, 6 is the traveling mechanism, 7 is the cornering attitude guiding wheel, 8 is the butt joint guiding block, 9 is the butt joint device base, 10 is the pipe bevel, 1101 is the first calculation unit, 1102 is the driving unit, 1103 is the butt joint unit, 1104 is the first fixing unit, and 1105 is the second fixing unit. Specific embodiments

[0081] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0082] In this application, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "including one..." does not exclude the presence of additional identical elements in the process, method, article or device including the element.

[0083] As can be seen from the background art, traditional internal welders rely on lasers, vision and ranging sensors for precise positioning and adjustment, but are greatly affected by the external environment, resulting in frequent errors and failures, increasing equipment costs, maintenance difficulties, and poor adaptability.

[0084] Therefore, an automatic butt-joint method, device and internal welding machine are provided in an embodiment of the present invention. When the internal welding machine travels inside a pipeline and contacts a bent pipe, an impedance controller calculates the driving forces of multiple branched chains according to the actual position of the attitude adjustment mechanism; the conical head mechanism is driven to deflect and turn; when reaching the pipe orifice end, the butt-joint guiding block is lifted to perform a groove butt-joint operation: a prediction model obtained by training a neural network using a six-axis dynamics model predicts the current actual thrust, and a real-time dynamics forward solution model calculates the load force and load torque in the moving platform coordinate system according to the current actual thrust of each branched chain; a admittance controller calculates the pose adjustment amount of the attitude adjustment mechanism according to the load force and load torque in the moving platform coordinate system, thereby adjusting the lengths of the branched chains to make the butt-joint guiding block reach the groove limit; after the butt-joint is completed, the rear expansion shoe is lifted to fix the relative position of the conical head mechanism and the groove and the butt-joint guiding block is retracted; when the grooves of another pipeline are butted, the front expansion shoe is lifted to fix the conical head and the groove for welding. By adjusting the contact state between the internal welding machine and the pipeline in real time, the adaptability and stability of the internal welding machine are significantly improved, the influence of harsh environments on the construction process is effectively reduced, the high quality of welding operations is ensured, and the overall work efficiency is enhanced.

[0085] See Figure 1 , which shows a structural schematic diagram of an internal welding machine provided in an embodiment of the present invention.

[0086] It should be noted that this internal welding machine is applicable to the internal welding scenarios of butt-joints of any attitude plane groove or inclined groove of various bent pipelines. It has high adaptability, can walk freely inside the bend and complete high-precision automatic groove butt-joint tasks, ensuring reliable welding quality; at the same time, it is difficult for field environmental factors to affect the stability of the internal welding machine.

[0087] As Figure 1 shown, the internal welding machine at least includes a conical head mechanism 1, a fuselage motion system 2, an attitude adjustment mechanism, a memory and a processor.

[0088] In some preferred embodiments, the internal welding machine further includes a force sensor (such as the component "3" shown as Figure 1 ). The force sensor 3 can be arranged on the attitude adjustment mechanism or at the position of the butt-joint guiding block. The thrust data of the branched chain 4 can be collected through the force sensor 3 arranged on the attitude adjustment mechanism. This helps to improve the control effect during the turning and butt-joint process of the internal welding machine.

[0089] Specifically, in an embodiment of the present invention, the conical head mechanism 1 is composed of several welding units, a tightening mechanism (including a front expansion shoe and a rear expansion shoe) and a butt-joint device. Among them, a welding torch is installed on the welding unit, and the welding unit moves on a track. Secondly, the butt-joint device includes a hydraulic cylinder, a butt-joint guiding block 8 and a butt-joint device base 9.

[0090] In specific applications, the butt-joint device includes multiple butt-joint guiding blocks 8.

[0091] It can be understood that for the structural schematic diagram of the mouthpiece, refer to the embodiments of the present invention. Figure 2 Shown is the schematic diagram of the butt joint guiding block and the bevel groove. Among them, the butt joint guiding block 8 is installed on the mouthpiece base 9. By pushing the mouthpiece base 9 through a hydraulic cylinder, the butt joint guiding block 8 moves closer to the pipe bevel groove 10 and reaches a preset positioning position (i.e., the bevel groove limiting position).

[0092] Specifically, as Figure 1 shown, the attitude adjustment mechanism is located between the cone head mechanism 1 and the fuselage movement system 2; the attitude adjustment mechanism includes at least a plurality of linkages 4 and bend attitude guiding wheels 5.

[0093] It should be noted that the attitude adjustment mechanism is constructed based on a six-axis platform. Among them, the attitude adjustment mechanism includes six linkages 4, and the linkages 4 are electric cylinders. The attitude adjustment mechanism includes, but is not limited to, a moving platform and a static platform.

[0094] In practical applications, the processor controls the lengths of the respective linkages 4 (i.e., electric cylinders), so that the cone head mechanism 1 can be flexibly offset and deflected.

[0095] In addition, the fuselage movement system 2 is composed of mechanisms such as a traveling mechanism 6, a braking mechanism, and bottom carrying wheels. The function of the fuselage movement system 2 is to meet the requirements of the internal welder for forward movement, backward movement, and stop movement inside the pipe.

[0096] In a specific embodiment, as Figure 1 shown, the cone head mechanism 1 further includes a plurality of bend attitude guiding wheels 7.

[0097] It should be noted that a computer program is stored in the memory. When the processor calls the computer program in the memory, the automatic butt joint method of the internal welder as shown in the embodiments of the present invention is implemented. Figure 3 Shown is the automatic butt joint method of the internal welder.

[0098] Next, in combination with Figure 3 , an explanation is given for the automatic butt joint method of the internal welder provided by the embodiments of the present invention. The method includes:

[0099] Step S301: When the internal welder travels inside the pipe through the fuselage movement system to a bend pipe, use an impedance controller to calculate the driving forces of multiple linkages in the attitude adjustment mechanism according to the actual position of the attitude adjustment mechanism.

[0100] In the process of specifically implementing step S301, when the internal welder travels inside the pipe through the fuselage movement system to a bend pipe, obtain the actual position of the attitude adjustment mechanism, input the actual position into the impedance controller, and calculate the driving forces of multiple linkages in the attitude adjustment mechanism.

[0101] It should be noted that the fuselage motion system drives the internal welder to move back and forth inside the pipeline through mechanisms such as the traveling mechanism, braking mechanism, and bottom carrier wheels.

[0102] Specifically, the control strategy of the impedance controller is pre-created according to the target inertia, target stiffness value, and target damping.

[0103] In some preferred embodiments, during the entire cornering control process, in order to avoid the influence of the cone head gravity (the mass of the cone head is close to 1 ton) on cornering, it is necessary to pre-correct the influence of the cone head gravity on the driving force of the attitude adjustment mechanism and adjust the magnitude of the driving force.

[0104] It can be understood that the cone head mechanism is a load object, and under the action of gravity, a force is generated. Through the known included angle, deflection data, and offset data, it is possible to calculate how the gravity of the cone head mechanism acts on each branch chain of the attitude adjustment mechanism. This force is the load force that the branch chain needs to bear.

[0105] The process of the attitude adjustment mechanism compensating for the cone head gravity is first to identify the cone head gravity parameters. Specifically, it is realized based on the linear equations of the load force, load torque, cone head weight, and center of gravity coordinates under the moving platform coordinate system of the attitude adjustment mechanism. Specifically, the establishment process of the linear equation is as follows:

[0106] First, obtain the branch chain driving force data of the attitude adjustment mechanism in multiple different postures. This process requires using a prediction model (see the relevant content of step S401 shown below Figure 4 to convert the branch chain drive current into the actual driving force).

[0107] For example: Obtain no less than 3 groups of branch chain driving force data of the attitude adjustment mechanism in different postures.

[0108] Secondly, convert multiple groups of branch chain driving force data into the load force and load torque in the moving platform coordinate system through the real-time dynamics forward solution model.

[0109] It should be noted that since the real-time feedback control mode is adopted, the accuracy requirement for the load force feedback is relatively low. Therefore, the model can be simplified to reduce the calculation amount. Based on this, the following real-time dynamics forward solution model (as shown in formula (1)) is established:

[0110] (1)

[0111] In formula (1), r i is the coordinate of the upper hinge point of the electric cylinder on the attitude adjustment mechanism (i.e., the six-axis platform, generally having 6 hinge points) in the reference coordinate system, represents the coordinates of the upper hinge points of the six electric cylinders in the reference coordinate system, is the unit direction vector of the electric cylinder, Represents the unit direction vectors of the six electric cylinders; Indicates the thrust of the electric cylinder, that is, the branch chain driving force data, It is the sum of the contact forces generated by the gravity of the cone head and the contact between several matching guide blocks and the groove, so gravity compensation is required to obtain the resultant contact force of several guide blocks.

[0112] Then, according to the least square method and the load force and load moment in the moving platform coordinate system, a linear equation of the load force, load moment on the moving platform of the attitude adjustment mechanism and the cone head weight and center of gravity coordinates is established.

[0113] Finally, the cone head weight and the coordinates of the cone head center of gravity are calculated based on the least squares method.

[0114] It can be understood that with the cone head weight and center of gravity coordinate parameters, the force and torque applied by the cone head to the posture adjustment mechanism in any posture can be calculated.

[0115] The process of the attitude adjustment mechanism performing cone head gravity compensation is to compensate for the influence of the force and torque generated by the cone head mechanism on the attitude adjustment mechanism. Specifically, the force and torque applied by the cone head to the attitude adjustment mechanism are converted into the load force of each branch chain according to the real-time dynamic inverse solution model, and the load force is used as the cone head gravity compensation amount.

[0116] Based on the above description of cone head gravity compensation, it can be known that gravity compensation is performed during impedance control, which is to add the cone head gravity compensation amount to the driving force calculated by the impedance controller to obtain the sum value, and the sum value is used as the target output force of each branch of the attitude adjustment mechanism. Gravity compensation is performed during admittance control, which is to subtract the cone head gravity compensation amount from the branch load force to obtain the branch load force and pass it into the real-time dynamics forward solution model to obtain the force and torque, that is, the contact force and contact torque generated by the contact between the matching guide block and the groove.

[0117] It should be noted that the specific process of impedance control and admittance control is detailed in the following content.

[0118] It is understandable that the contact between the attitude guide wheel and the inner wall of the pipe changes rapidly due to the high speed of the internal welding machine passing through the curved pipe and the lack of prior modeling of the curved pipe. In order to cope with this change, the six-axis motor operates in thrust control mode when passing through the curve.

[0119] When the cone head is subjected to external forces such as squeezing from the inner wall of the pipe, it will move, but it cannot completely follow the external force. For example, when the internal welding machine is moving forward, if the cone head mechanism encounters backward resistance, if it is not controlled, the six branches of the posture adjustment mechanism may shrink backwards to the shortest possible length, and then lose the ability to move.

[0120] To solve this problem, the embodiment of the present invention adopts an impedance controller. By adjusting the driving forces of multiple branches, the six-axis motor can move flexibly with the contact force only in the four required degrees of freedom (up and down translation, left and right translation, up and down rotation, left and right rotation), while the other two degrees of freedom (front and back translation, self-rotation) maintain a relatively high stiffness. That is, in these two directions, even if the external force is large, the displacement is very small. This stiffness is adjusted by the target stiffness coefficient in the impedance controller. At the same time, the target damping and target inertia coefficient are used to control the stability of the six-axis attitude change during the contact with the pipeline, ensuring that even when under a large pressure, the compression amount in the axial direction can be maintained within a very limited range.

[0121] Based on the above, the specific process of calculating the driving forces of multiple branches in the attitude adjustment mechanism by using the impedance controller according to the actual position of the attitude adjustment mechanism is as follows (Process A1 to Process A4):

[0122] Process A1: Obtain the actual position of the attitude adjustment mechanism.

[0123] It can be understood that the actual position includes the angle between each branch of the attitude adjustment mechanism and gravity, the deflection data and offset data of the branch.

[0124] Process A2: Use the real-time dynamic forward solution model to calculate the gravity compensation amount of the cone head under the moving platform coordinate system in the moving platform according to the angle, deflection data, offset data, cone head weight and center of gravity coordinates.

[0125] Process A3: Through the real-time dynamic inverse solution model, convert the gravity compensation amount of the cone head under the moving platform into the gravity compensation amount of the cone head by the branches in the current attitude.

[0126] It should be noted that the gravity compensation amount of the cone head by the branches is the force exerted by the cone head gravity on each branch.

[0127] It can be understood that the real-time dynamic inverse solution model is the inverse operation of formula (1), as shown in formula (2):

[0128] (2)

[0129] Among them, the meanings of each parameter in formula (2) are as shown in formula (1), which will not be elaborated here. In addition, in formula (2) represents the sum of the cone head gravity and the contact forces generated by the contact of several butt guiding blocks with the groove.

[0130] Process A4: Input the angle, deflection data, offset data and the gravity compensation amount of the cone head by the branches into the impedance controller to calculate the driving forces of multiple branches in the attitude adjustment mechanism.

[0131] The specific calculation method of step S301 is shown in formula (3).

[0132] (3)

[0133] In formula (3), x is the actual position. is the derivative of the attitude adjustment mechanism. is the expected position of the attitude adjustment mechanism, which can be set to the initial middle position of the mechanism. is the contact force (in practical applications, this value can be 0 when no force sensor is installed). F is the driving force of the branch chain of the attitude adjustment mechanism. The target stiffness K d and the target damping B d and the target inertia M d are parameters for controlling the dynamic response of the attitude adjustment mechanism during the cornering process. m is the actual mass of the cone head (i.e., the load weight).

[0134] Step S302: Drive the cone head mechanism in the internal welder to deflect according to the driving force, so that the internal welder passes through the elbow pipe.

[0135] In the process of specifically implementing step S302, drive the cone head mechanism in the internal welder to deflect according to the driving force, change the shape of the internal welder, so that the position of the attitude adjustment mechanism of the internal welder in the pipeline changes with the curvature of the pipeline.

[0136] Step S303: When the internal welder travels to the pipe orifice end, raise the butt joint guiding block to perform a butt joint operation with the groove of the pipe orifice end.

[0137] It should be noted that the pipe orifice end is the area near the pipe groove (such as Figure 2 the pipe groove 10 shown).

[0138] In the process of specifically implementing step S303, when the internal welder travels to the pipe orifice end in the pipeline, raise the butt joint guiding block in the cone head mechanism, and perform a butt joint operation with the groove of the pipe orifice end through the butt joint guiding block.

[0139] It can be understood that when the butt joint guiding block is just raised, the position between the cone head mechanism and the groove does not meet the positional relationship required for welding. Therefore, a butt joint operation is required to make the butt joint guiding block reach the groove limiting position. At this time, the position between the cone head mechanism and the groove does not meet the positional relationship required for welding.

[0140] Next, in combination with the content shown in the embodiments of the present invention Figure 4 the detailed process of the butt joint operation will be described. Figure 4 It includes:

[0141] Step S401: Based on the motion data and current data of the branch chains in the posture adjustment mechanism, use the prediction model to predict the current actual thrust of each branch chain.

[0142] It should be noted that the prediction model is pre-trained using a six-axis dynamics model for the neural network model. And the six-axis dynamics model is pre-constructed according to the six-axis platform.

[0143] It can be understood that the six-axis dynamics model describes the output forces required by each branch chain under the given displacement, velocity, and acceleration of the posture adjustment mechanism. The six-axis dynamics model is used to train the prediction model. The content of constructing the six-axis dynamics model will be described in detail below through processes B1 to B4.

[0144] Process B1: Define the structure of the six-axis platform.

[0145] Specifically, the structure at least includes a static platform, a moving platform, the cylinder of the branch chain, and the piston rod of the branch chain.

[0146] Process B2: Define the motion physical quantities based on the static platform coordinate system, and combine the force data of the six-axis platform to create the Jacobian matrix of the moving platform, the Jacobian matrix of the cylinder of the branch chain, and the Jacobian matrix of the piston rod of the branch chain.

[0147] It can be understood that the static platform coordinate system is a coordinate system constructed based on the static platform.

[0148] It should be noted that since the six-axis platform holds the cone head mechanism in a horizontal state and the six branch chains of the six-axis platform are placed horizontally, the cone head mechanism part is used as a load, which is the sum of the gravity, inertia force, and external force acting at the centroid of the cone head body, as shown in formula (4):

[0149] (4)

[0150] In formula (4), represents the total force acting at the centroid of the cone head composite body. represents the mass of the cone head mechanism. represents the acceleration vector of the cone head mechanism. g represents the gravitational acceleration vector, usually pointing towards the center of the earth. represents the external force applied to the centroid of the cone head composite body.

[0151] More specifically, the motion physical quantities defined with reference to the six-axis static platform coordinate system include:

[0152] The translational acceleration of the moving platform is expressed as: a p ;

[0153] The rotational angular velocity of the moving platform is expressed as: ω p ;

[0154] The rotational angular acceleration of the moving platform is expressed as: α p ;

[0155] The translational acceleration of the centroid at the cylinder end of the branch chain i is expressed as: a 1i ;

[0156] The translational acceleration of the centroid at the piston rod end of the branch chain i is expressed as: a 2i ;

[0157] The rotational angular velocity of the connecting rod i of the branch chain is expressed as: ω i ;

[0158] The rotational angular acceleration of the connecting rod i of the branch chain is expressed as: α i ;

[0159] The acceleration at the centroid of the conical head composite mass is expressed as: a C ;

[0160] The external force acting on the centroid of the conical head composite mass e .

[0161] It should be noted that the data of these motion physical quantities are known during the data acquisition stage.

[0162] Based on this, when implementing process B2, combining the force data of the six-axis platform, create the Jacobian matrix J of the pose velocity of the moving platform and the telescopic velocities of the 6 branch chains; and create the Jacobian matrix J 1i of the velocity of the centroid of the cylinder of the branch chain i and the velocities of each branch chain, and create the Jacobian matrix J 2i of the velocity of the centroid of the piston rod of the branch chain i and the velocities of each branch chain.

[0163] Process B3: Construct the inertia matrix of the conical head mechanism, the inertia matrix of the cylinder of the branch chain, and the inertia matrix of the piston rod of the branch chain.

[0164] Specifically, construct the inertia matrix I C of the conical head mechanism about its centroid; construct the inertia matrix I 1i of the cylinder of the branch chain i about its centroid; and construct the inertia matrix I 2i of the piston rod of the branch chain i about its centroid.

[0165] Process B4: Combine all the Jacobian matrices and all the inertia matrices to construct a six-axis dynamic model.

[0166] It can be understood that by combining all the Jacobian matrices and all the inertia matrices, the established six-axis dynamic model is shown in formula (5).

[0167] (5)

[0168] In Equation (5), represents the thrust of six branched chains (i.e., electric cylinders); J C represents the Jacobian matrix of the velocity of the centroid of the conical head's overall physical properties and the velocity of the moving platform; represents the load force and load torque acting on the conical head mechanism; c has no specific meaning.

[0169] Specifically, as shown in Equation (6).

[0170] (6)

[0171] It should be noted that the above Equation (4) only describes the total force (load force) acting on the centroid of the conical head's overall physical properties, which belongs to a simplified model and is applicable to simplified analysis. While Equation (6) describes the load force and load torque acting on the conical head mechanism, which belongs to a complete dynamic model.

[0172] Meanwhile, in Equation (4), is the classical representation of inertial force (opposite to the direction of acceleration), while in Equation (6), is due to differences in coordinate system definition or derivation of dynamic equations (such as moving the inertial force to the other side of the equation).

[0173] As shown in Equation (7).

[0174] (7)

[0175] where, m 1 represents the mass of the cylinder end.

[0176] As shown in Equation (8).

[0177] (8)

[0178] where, m 2 represents the mass of the piston rod end.

[0179] Based on this, the six-axis dynamic model is established. Using the six-axis dynamic model to design training data helps to provide more accurate and comprehensive input data during the training process of the prediction model. Through the six-axis dynamic model, the mechanical behavior of the system under different states can be simulated and predicted, thus generating a set of training data covering various working conditions. These data can not only effectively supplement the deficiencies of the model but also improve the generalization ability of the prediction model, enabling it to adapt to more changing actual scenarios. By combining the physical characteristics of the six-axis dynamic model with real-time feedback control, it can be ensured that the prediction model still maintains a high prediction accuracy and stability when facing a complex dynamic environment.

[0180] The six-axis dynamic model of the attitude adjustment mechanism is established by the above method, aiming to provide guidance for the training of the deep neural network so as to correct the friction force and other unknown non-linear factors inside the branch chain. Finally, the thrust of the branch chain can be accurately predicted through the driving current and motion data.

[0181] The content of training the prediction model is described in detail below through processes C1 to C6.

[0182] Process C1: Sample multiple loads from the preset load space. For each load, obtain the corresponding motion sequence data from the preset position space and preset dynamic space.

[0183] It should be noted that during the data collection process, a scheme for the motion trajectory of the cone head mechanism and the load change of the cone head mechanism is designed, and the thrust of the branch chain is calculated through the six-axis dynamic model. The data collection space is divided into three parts: position space, dynamic space, and load space.

[0184] Position space: Use the telescopic stroke range of the branch chain as the data collection space.

[0185] Dynamic space: Based on the process of the branch chain in the butt contact, record the extreme values of its speed and acceleration to define the dynamic space.

[0186] Load space: Define the load space by adding or reducing loads with known masses at the centroid of the cone head mechanism.

[0187] In the specific implementation of Process C1, in these preset spaces, data collection is carried out by random sampling. After sampling a load in the load space each time, according to the sampled load in the position space and dynamic space, the motion trajectory of the branch chain is smoothly generated to determine the target position, speed, and acceleration. The branch chain moves according to the generated trajectory, and the motion sequence data of the branch chain is obtained through the feedback of the motor encoder as the input data for training the model.

[0188] It should be noted that the motion sequence data of the branch chain As shown in formula (9) for example:

[0189] (9)

[0190] In formula (9), - represents the acceleration sequence data, - represents the speed sequence data, - represents the telescopic displacement sequence data.

[0191] It is understandable that this method ensures extensive data collection under different working conditions, which helps to optimize the model training process.

[0192] In some embodiments, it is also possible to obtain the time-series data of the current of the branch chain.

[0193] Process C2: Input all loads and all motion sequence data into a six-axis dynamics model to obtain initial training data including branch chain motion data, current data, and branch chain thrust data, and preprocess the initial training data to obtain training data.

[0194] In the specific implementation of Process C2, the motion sequence data of all branch chains and the motion sequence data of the branch chain corresponding load magnitude F C are input into the six-axis dynamics model to obtain initial training data including branch chain motion data, current data, and branch chain thrust data (as shown in Equation (10)).

[0195] (10)

[0196] In Equation (10), - represents the branch chain thrust sequence data, - represents the current sequence data.

[0197] Specifically, when preprocessing the initial training data, data standardization can be performed, for example, scaling the initial training data to a range with the same scale.

[0198] It is understandable that for the training data set , is the i-th training data sequence.

[0199] The data after standardizing each training data sequence (i.e., the training data) is , .

[0200] Among them, Equation (11) is used to represent the average value of all samples, and Equation (12) is used to represent the standard deviation of the original data.

[0201] It should be noted that i represents the training data sequence, and j represents the j-th data in the i-th sequence.

[0202] Process C3: Input the training data into a long short-term memory deep neural network model to output the predicted value of the branch chain thrust.

[0203] It should be noted that in the embodiments of the present invention, a Long Short-Term Memory (LSTM) deep neural network is used to process the current, speed, and position data of the motor, so as to predict the thrust magnitude of the branch chain in the attitude adjustment mechanism.

[0204] The LSTM network structure used in the embodiments of the present invention is composed of several stacked LSTM layers. Each LSTM layer contains several memory units and can effectively capture the dependencies hidden in the sequence data of the branch chain motion state. Compared with the traditional Recurrent Neural Network (RNN), LSTM can maintain the flow of important information in a long time series by introducing a gating mechanism, thus effectively solving the problem of gradient vanishing or explosion, being able to capture dependencies over a long time span, and having faster training convergence. In the embodiments of the present invention, the three gates of LSTM are specifically defined as follows:

[0205] The specific form of the forget gate is shown in formula (13); the specific form of the input gate is shown in formula (14); the specific form of the candidate memory unit is shown in formula (15); the specific form of the updated memory unit is shown in formula (16); the specific form of the output gate is shown in formula (17); the specific form of the hidden state is shown in formula (18).

[0206] (13)

[0207] (14)

[0208] (15)

[0209] (16)

[0210] (17)

[0211] (18)

[0212] In formulas (13) to (18), represents the hidden state at the current time, represents the hidden state at the previous time of the current time, carrying the memory information of the current time and the previous times. represents the input data at the current time. σ represents the sigmoid activation function. tanh represents the hyperbolic tangent activation function.

[0213] represents the forget gate weight matrix; represents the input gate weight matrix; represents the candidate memory weight matrix; represents the output gate weight matrix.

[0214] represents the forget gate bias term; represents the input gate bias term; represents the candidate memory bias term; represents the output gate bias term.

[0215] represents the output of the forget gate; represents the output of the input gate; represents the output of the output gate.

[0216] represents the long-term memory inside the LSTM cell. represents the potential memory at the current moment, which participates in updating the hidden state at the current moment. represents the potential memory at the previous moment of the current moment.

[0217] It should be particularly noted that, in order to further improve the performance of the model, the embodiments of the present invention optimize the model performance by carefully tuning hyperparameters. Specifically, the tuning content includes the number of hidden layers, the number of neurons, the activation function, the regularization method, the learning rate, the batch size, the time step, and the number of training epochs for each LSTM layer, aiming to ensure that the network can exhibit the best prediction ability under different data conditions.

[0218] Process C4: Calculate the loss function of the long short-term memory deep neural network model according to the predicted value.

[0219] It should be noted that in the embodiments of the present invention, the loss function is the mean squared error, as shown in formula (19).

[0220] (19)

[0221] In formula (19), represents the predicted value; represents the true value; N represents the number of samples.

[0222] It can be understood that based on the predicted value, the mean squared error method is used to calculate the loss function of the long short-term memory deep neural network model. It can effectively measure the gap between the predicted value and the true value and provide a basis for adjusting the model during the training process.

[0223] Process C5: If the loss function converges, determine the long short-term memory deep neural network model as the trained prediction model.

[0224] It can be understood that when the loss function converges, it indicates that the prediction model can make relatively accurate predictions, has high generalization ability, and can provide reliable prediction results in practical applications.

[0225] Process C6: If the loss function does not converge, adjust the parameters of the long short-term memory deep neural network model, and return to execute Process C3.

[0226] It can be understood that if the loss function does not converge, the error backpropagation algorithm is used to iteratively update the weights and bias values in the long short-term memory deep neural network model. Continue training and evaluation until the loss function converges. Through continuous optimization and adjustment, the prediction performance of the model can be effectively improved, the risk of overfitting or underfitting can be reduced, thereby ensuring the stability and accuracy of the model in various data environments.

[0227] Based on the above content, the specific implementation method of step S402 is: during the alignment process, use the trained prediction model to continuously predict the current actual thrust of each branch chain based on the motion data and current data of the branch chains in the pose adjustment mechanism.

[0228] Step S402: Calculate the load force and load torque in the moving platform coordinate system according to the current actual thrust of each branch chain through the real-time dynamic forward solution model.

[0229] In the process of specifically implementing step S402, use the real-time dynamic forward solution model to calculate the load force and load torque in the moving platform coordinate system in real time according to the current actual thrust of each branch chain during the alignment process.

[0230] It can be understood that the six-axis platform includes a moving platform and a static platform; the moving platform coordinate system is a coordinate system constructed based on the moving platform.

[0231] It should be noted that the coordinates of the electric cylinder and the unit direction vector of the electric cylinder are determined by inverse kinematics, which will not be elaborated in the embodiments of the present invention.

[0232] Step S403: Use the admittance controller to calculate the pose adjustment amount of the pose adjustment mechanism according to the load force and load torque, and adjust the lengths of the respective branch chains during the alignment process so that the alignment guide block reaches the groove limit position.

[0233] In the process of specifically implementing step S403, first use the admittance controller to calculate the pose adjustment amount according to the load force received by the moving platform (i.e., the load force and load torque in the moving platform coordinate system). Secondly, use the position controller to adjust the lengths of the respective branch chains during the alignment process according to the pose adjustment amount so that the alignment guide block reaches the groove limit position.

[0234] It should be noted that the admittance controller is pre-established based on a second-order differential model, and the specifically established second-order differential model is as follows:

[0235] (20)

[0236] (21)

[0237] (22)

[0238] (23)

[0239] (24)

[0240] (25)

[0241] (26)

[0242] (27)

[0243] (28)

[0244] (29)

[0245] (30)

[0246] (31)

[0247] In formulas (20) to (25), represents the actual position of the moving platform in the x, y, and z directions. represents the execution speed (the first derivative of the position) of the moving platform in the x, y, and z directions. represents the actual external force acting in the x, y, and z directions. represents the desired external force (target value) in the x, y, and z directions. represents the damping coefficient in the x, y, and z directions (used to suppress the speed). represents the stiffness coefficient in the x, y, and z directions (used to suppress the position deviation). represents the virtual mass (inertia parameter) in the x, y, and z directions. represents the desired position (target value) in the x, y, and z directions.

[0248] In formulas (26) to (31), represents the actual angle (or angular displacement) of the moving platform about the x, y, and z axes. represents the execution angular velocity (the first derivative of the angle) of the moving platform about the x, y, and z axes. represents the actual external torque acting about the x, y, and z axes. Denote the expected external torque (target value) about the x, y, and z axes. Denote the damping coefficients about the x, y, and z axes (used to suppress the angular velocity). Denote the stiffness coefficients about the x, y, and z axes (used to suppress the angular deviation). Denote the virtual moments of inertia about the x, y, and z axes (inertial parameters). Denote the expected angles (target values) about the x, y, and z axes.

[0249] In formulas (20) to (31), dx(1), dx(3), dx(5), dx(7), dx(9), dx(11) denote the execution speed / angular velocity. dx(2), dx(4), dx(6), dx(8), dx(10), dx(12) denote the execution acceleration / angular acceleration. The superscript d (e.g., ) represents the preset value of the admittance controller.

[0250] Based on formulas (20) to (31), it can be determined that:

[0251] The actual position of the six-axis platform during movement, which is fed back in real time is expressed in the form of an ordinary differential equation system as shown in formula (32). The actual speed of the six-axis platform during movement is expressed in the form of an ordinary differential equation system as shown in formula (33). The expected position is expressed in the form of an ordinary differential equation system as shown in formula (34).

[0252] It should be noted that the expected position is the approximate position of the preset pipe groove surface relative to the positioner. The expected position can restrict the range within which the positioner searches for the pipe groove surface. When the distance to the pipe groove surface is too far, the six-axis movement will gradually stop to avoid collision interference of non-positioner components.

[0253] Specifically, the thrust of a single chain predicted by the prediction model, and the actual contact force obtained by subtracting the magnitude of the current six-axis load force calculated by the real-time dynamics model from the magnitude of the six-axis load force before the butt welding operation (the gravity compensation amount of the moving platform on the cone head) is expressed in the form of an ordinary differential equation system as shown in formula (35).

[0254] The expected contact force is the target contact force generated by the preset extrusion of the positioner and the pipe groove. The expected contact force is expressed in the form of an ordinary differential equation system as shown in formula (36).

[0255] The target stiffness The expression form of the ordinary differential equation system of is shown in Formula (37). Target damping The expression form of the ordinary differential equation system of

[0256] It should be noted that the target stiffness , target damping and target inertia are parameters that control the dynamic response of the six-axis platform during the butt-jointing process. Through these parameters, the butt-jointing speed and the peak value of the contact force during the butt-jointing process can be adjusted, etc.

[0257] Actual position (32)

[0258] Actual speed (33)

[0259] Desired position (34)

[0260] Actual contact force (35)

[0261] Desired contact force (36)

[0262] Target stiffness (37)

[0263] Target damping (38)

[0264] Target inertia (39)

[0265] Among them, the meanings of the various parameters in Formulas (32) to (39) can be seen in Formulas (20) to (31) shown, and will not be elaborated here.

[0266] It can be understood that during the entire butt-jointing process from when the butt-jointing device and the pipe groove have never been in contact to when they come into contact and are squeezed, the load on the six-axis moving platform will change accordingly. In order to ensure that the butt-jointing device and the pipe groove maintain stable contact, it is necessary to control the magnitude of the load force received by the six-axis platform.

[0267] Therefore, the embodiment of the present invention adopts a second-order differential model to guide the change relationship between the contact situation of the butt-jointing device and the pipe groove and the magnitude of the six-axis load force, so as to achieve the goal of close and stable contact between the butt-jointing device and the pipe groove surface during the change process of the two.

[0268] It can be understood that in step S403, the specific implementation process of calculating the pose adjustment amount is as follows:

[0269] First, calculate the difference between the load force and the load moment minus the gravity compensation amount of the cone head on the moving platform to obtain the actual contact resultant force of several pairs of guiding blocks in the cone head mechanism in contact with the groove. Secondly, obtain the actual positions and actual velocities of each branch chain during the alignment process. Then, input the preset desired contact resultant force, actual contact resultant force, actual position, and actual velocity into the admittance controller. Finally, combine the ordinary differential equations in the admittance controller and perform iterative solution according to the time step of the control period to obtain the pose adjustment amount of the pose adjustment mechanism and output it.

[0270] It can be understood that in step S403, the specific implementation process of using the position controller to adjust the lengths of each branch chain during the alignment process according to the pose adjustment amount is as follows:

[0271] First, use the position controller to calculate the alignment output force of the pose adjustment mechanism according to the pose adjustment amount. Secondly, adjust the lengths of each branch chain during the alignment process through the alignment output force.

[0272] It should be noted that the position controller is pre-established based on a proportional-derivative controller, and the expression form of the position controller is shown in formula (40).

[0273] (40)

[0274] In formula (40), represents the actual position; represents the pose adjustment amount; represents the derivative of the pose adjustment mechanism; F is the alignment output force of the pose adjustment mechanism; and represent the positive gain parameters of the position controller.

[0275] In some preferred embodiments, during the entire alignment operation, in order to avoid the influence of the cone head gravity (the mass of the cone head is close to 1 ton) on the alignment, it is necessary to correct the influence of the cone head gravity on the alignment output force of the pose adjustment mechanism and adjust the influence of the alignment output force. For this purpose, the magnitude of the load force is pre-loaded onto the driving forces of each branch chain of the pose adjustment mechanism, thereby compensating for the influence of gravity on the control performance.

[0276] It should be noted that by compensating for the influence of gravity on the control performance, the system can be effectively ensured to perform tasks stably and accurately.

[0277] It should be noted that during the butt-welding process of the internal welding machine, the contact between the butt-welding device and the pipe groove requires extremely high stability and precision. Therefore, in the force control system of the attitude adjustment mechanism in the embodiments of the present invention, an outer-loop position control (i.e., a position controller and a admittance controller) is added. The pose adjustment amount is calculated through the load force and the load torque, and the pose adjustment amount is used to adjust the attitude adjustment mechanism. The pose adjustment can effectively stabilize the attitude change process, ensuring that a stable attitude is beneficial to estimating the magnitude of the external contact force through the drive current of the attitude adjustment mechanism. Although the movement during this butt-welding process is relatively slow, through this optimization, the butt-welding precision and the stability of the contact process are significantly improved, greatly enhancing the overall welding quality and stability.

[0278] Step S304: After completing the butt-welding operation, raise the rear expansion shoe to fix the relative position of the cone head mechanism and the groove, and retract the butt-welding guide block.

[0279] It should be particularly noted that after completing the butt-welding operation, the pulling force output by the attitude adjustment mechanism can also keep a certain pre-tightening force between the butt-welding device and the pipe groove. At this time, the welding torch in the cone head mechanism is just aligned with the pipe groove (actually the limiting position of the butt-welding guide block), meeting the welding conditions.

[0280] It can be understood that raising the rear expansion shoe to fix the relative position of the cone head mechanism and the groove in the internal welding machine makes the cone head mechanism no longer move relative to the pipe, which helps to ensure that the movement of the welding torch during the welding process is not disturbed.

[0281] Step S305: When the groove of another pipe coincides with the groove, raise the front expansion shoe to fix the relative position of the cone head mechanism and the groove of the other pipe, so that the welding torch in the cone head mechanism welds the two pipes.

[0282] In the specific process of implementing Step S305, when the groove of another pipe coincides with the current groove to form a weld seam, raise the front expansion shoe in the tightening mechanism of the cone head mechanism to fix the relative position of the cone head mechanism and the groove of the other pipe, and make the welding torch in the cone head mechanism weld the weld seam, thereby completing the welding of the two pipes.

[0283] In the embodiments of the present invention, by adjusting the contact state between the internal welding machine and the pipeline in real time, the adaptability and stability of the internal welding machine are significantly improved, the influence of harsh environments on the construction process is effectively reduced, the high quality of welding operations is ensured, and the overall work efficiency is enhanced. By precisely utilizing the drive current of the attitude adjustment structure and combining deep neural network technology with a dynamics model, the contact force between the cone head mechanism and the groove is predicted, and the attitude adjustment control amount during the pipe bending and butt joint processes is calculated. This method not only avoids the use and maintenance costs of traditional force sensors but also ensures the stability and accuracy of the welding process by adjusting the attitude based on the feedback contact force. This method has strong robustness, can flexibly respond to different pipeline welding requirements, achieve high-precision automatic butt joint, and further improve the welding quality and efficiency.

[0284] To further illustrate the overall process of the automatic butt joint method of the internal welding machine in more detail, refer to Figure 5 which shows the control flow block diagram provided by the embodiments of the present invention.

[0285] Combined with Figure 5 the content shown:

[0286] First, when the internal welding machine travels to a bent pipe, the attitude adjustment mechanism of the internal welding machine inputs the actual position into the impedance controller to calculate the driving force of the branch chain in the attitude adjustment mechanism, thereby driving the internal welding machine to bend.

[0287] Secondly, the butt joint process is divided into three parts:

[0288] The first part: The internal welding machine traverses the bent pipeline to reach near the groove position, and inputs the branch chain motion data and current data in the attitude adjustment mechanism into the prediction model to predict the current actual thrust of each branch chain.

[0289] According to the current actual thrust of each branch chain through the real-time dynamics forward solution model, calculate the load force and load torque received by the moving platform in the attitude adjustment mechanism. Then input the load force and load torque into the admittance controller to calculate the pose adjustment amount x d . Finally, use the position controller to control the butt joint guiding block of the internal welding machine to reach the groove limit position based on the actual position x and the pose adjustment amount x d of the attitude adjustment mechanism to complete the butt joint operation.

[0290] It should be noted that during the pipe bending and butt joint processes, according to the included angle between the branch chain and gravity, the deflection data and offset data of the branch chain, and the weight and centroid coordinates of the cone head mechanism in the current attitude of the attitude adjustment mechanism, calculate the driving force of multiple branch chains in the attitude adjustment mechanism in the current attitude, and load the driving force onto each branch chain to compensate for gravity.

[0291] Part 2: The welding gun of the welding unit on the cone head mechanism is just aimed at the groove (before welding the joint, the welding area needs to be polished and shaped, for example: shaping the right end face of pipe A to form a groove, shaping the left end face of pipe B to form another groove, the purpose of the alignment is to combine the two grooves to form a weld and the preset movement trajectory of the welding gun is perfectly overlapped with the weld). At this time, the tensioning mechanism on the cone head extends out to support the pipe wall so that the cone head mechanism is fixed to the pipe.

[0292] Part 3: Next, the groove of the other "new pipe" is overlapped with the groove of the "parent pipe" and the front expansion shoe at the cone head is raised to fix the new pipe. The scene at this time is: the internal welding machine passes through the welded "parent pipe" to the groove position, then expands and fixes it with the parent pipe, the groove of the new pipe overlaps with the groove of the parent pipe, then expands and fixes it with the new pipe, and the welding gun is aimed at the weld.

[0293] The role of the branch chain of the posture adjustment mechanism in the above-mentioned docking process is to adjust the posture of the cone head mechanism by providing push-pull force to meet the posture change requirements of turning and matching the groove.

[0294] Preferably, see the embodiments of the present invention Figure 6 The data curve diagram of the control process of the automatic matching method of the internal welding machine is shown.

[0295] Figure 6 The data sequence of displacement (curve 1), speed (curve 2), and current (curve 3) of a single branch chain in the six-axis attitude adjustment mechanism of the internal welding machine is collected, and the thrust (curve 5) data sequence of the branch chain is predicted using the LSTM neural network model. Curve 4 represents the thrust data calculated by the dynamic model. When curve 4 is close to curve 5, it means that the thrust data predicted by the neural network is consistent with the actual thrust data.

[0296] It should be noted that the vertical axis value ranges of curves 4 and 5 are different, but their values ​​are actually close. Since it is not possible to enlarge the curves to show the similarity of the values, this point is not highlighted.

[0297] Therefore, from Figure 6 The content shown can determine that the advantage of using the LSTM neural network model for thrust prediction is that it can effectively handle complex nonlinear relationships, especially the complex changes between current and torque when the speed is reversed. The neural network provides more accurate predictions than traditional models by learning patterns in historical data, especially in situations that traditional models cannot describe. In addition, the neural network has strong adaptability and can automatically adjust to different loads and working conditions without the need to manually adjust parameters. It also reduces the reliance on complex physical modeling, simplifies the modeling process, and improves the accuracy and generalization of predictions.

[0298] Preferably, refer to the embodiments of the present invention. Figure 7 Schematic diagram showing the change of the pose of the six-axis platform over time in the static platform coordinate system (b system).

[0299] It should be noted that the degrees of freedom of the six-axis platform include the forward and backward translations along the X-axis, Y-axis, and Z-axis, as well as the rotations around the X-axis, Y-axis, and Z-axis. And before discussing the problem, the directions of these X, Y, and Z axes have been preset.

[0300] Combined with Figure 7 As can be seen from the content shown, during the butt joint operation, the six-axis platform first retreats along the Z-axis. When a certain butt joint guiding block touches the groove, the six-axis senses the resistance and adjusts the pose to avoid excessive extrusion of the butt joint device. The butt joint devices that have not touched the groove will continue to approach the groove under the adjustment of the six-axis pose until the limiting positions of the three butt joint guiding blocks in the taper head mechanism all touch the groove. At this time, the resultant force of the three contact forces will be perpendicular to the groove surface and collinear, and the magnitude of the resultant force meets the preset contact force requirement. The six-axis movement stops at this time.

[0301] Furthermore, refer to the embodiments of the present invention. Figure 8 Schematic diagram showing the change of the contact force of the six-axis platform over time in the moving platform coordinate system (w system).

[0302] In Figure 8 the curve shown, the resultant force of the contact force and the contact moment can be decomposed into components in different axial directions. Specifically, Fx, Fy, and Fz respectively represent the components of the contact resultant force on the x-axis, y-axis, and z-axis; Mx, My, and Mz represent the components of the contact resultant moment around the x-axis, y-axis, and z-axis.

[0303] From Figure 8 it can be seen that at time t = 0s, both the contact force and the moment are zero, meaning that the three butt joint guiding blocks have not yet contacted the groove. When the six-axis platform starts to retreat, the contact force gradually generates. As the adjustment of the six-axis pose is completed, the final contact resultant force acts only in the z-axis direction, and the contact moment is zero. This indicates that at this moment, the contact forces of the three butt joint guiding blocks are equal, thus achieving the goal of butt joint control.

[0304] More specifically, refer to the embodiments of the present invention. Figure 9 Schematic diagram showing the relationship between the distance between the butt joint device and the groove and time, and the embodiments of the present invention Figure 10 Schematic diagram showing the relationship between the magnitude of the contact force between the butt joint device and the groove and time.

[0305] In the embodiments of the present invention, the internal welding machine is equipped with three butt joint devices, and the goal of its butt joint operation is: all three butt joint devices contact the groove, and the extrusion forces generated are equal.

[0306] InFigure 9 and Figure 10 In Figure 10 , at t = 0s, the butt-welding process starts. The three butt-welding devices have not yet contacted the groove, so the contact force between the three butt-welding devices and the groove is zero at this time. Moreover, the distances between the three butt-welding devices and the groove are large and inconsistent, marking the initial state of butt-welding control.

[0307] As the six-axis platform drives the cone head mechanism to retreat, some of the butt-welding devices start to contact the groove and generate contact force, and at the same time the distance between them and the groove shortens, while the other butt-welding devices have not yet contacted, so the contact force remains zero. At this time, the six-axis platform adjusts its attitude to gradually stabilize the contact force until the contact force increases to 1000 Newtons and finally remains at this value, indicating that the six-axis movement has been completed, the goal of controlling the contact force has been achieved and it remains stationary. At this time, the distance between the three butt-welding devices and the groove is 0.

[0308] It should be noted specifically that in Figure 10 the content shown, when t > 110s, the three contact forces do not fully reach the target of 1000N, which is mainly due to the static error caused by the control accuracy, but the error is within the acceptable range.

[0309] Corresponding to an automatic butt-welding method for an internal welding machine provided by an embodiment of the present invention, referring to Figure 3 Figure 11 , a structural block diagram of an automatic butt-welding device for an internal welding machine provided by an embodiment of the present invention is shown. The device includes: a first calculation unit 1101, a driving unit 1102, a butt-welding unit 1103, a first fixing unit 1104, and a second fixing unit 1105. Figure 11

[0310] The first calculation unit 1101 is configured to, when the internal welding machine travels to a bent pipe inside the pipeline through the fuselage movement system, use an impedance controller to calculate the driving forces of multiple branch chains in the attitude adjustment mechanism according to the actual position of the attitude adjustment mechanism; the control strategy of the impedance controller is pre-created according to the target inertia, target stiffness value, and target damping.

[0311] The driving unit 1102 is configured to drive the cone head mechanism in the internal welding machine to deflect according to the driving force so that the internal welding machine passes through the bent pipe.

[0312] The butt-welding unit 1103 is configured to, when the internal welding machine travels to the pipe orifice end, raise the butt-welding guiding block and the groove of the pipe orifice end, and perform butt-welding operations based on a prediction unit, a second calculation unit, and an adjustment unit.

[0313] The prediction unit is configured to use a prediction model to predict the current actual thrust of each branch chain based on the branch chain movement data and current data in the attitude adjustment mechanism; the prediction model is pre-trained by using a six-axis dynamic model for a neural network model; the six-axis dynamic model is pre-constructed according to a six-axis platform.

[0314] A second calculation unit, configured to calculate the load force and load torque in the moving platform coordinate system according to the current actual thrust of each branch chain through a real-time kinematic forward solution model; the six-axis platform includes a moving platform and a static platform; the moving platform coordinate system is a coordinate system constructed based on the moving platform.

[0315] An adjustment unit, configured to calculate the pose adjustment amount of the pose adjustment mechanism according to the load force and load torque by using an admittance controller, and adjust the lengths of the respective branch chains during the alignment process so that the alignment guide block reaches the groove limit position; the admittance controller is pre-established based on a second-order differential model.

[0316] A first fixing unit 1104, configured to raise the rear expansion shoe after the alignment operation is completed to fix the relative position of the cone head mechanism and the groove, and retract the alignment guide block.

[0317] A second fixing unit 1105, configured to raise the front expansion shoe when the groove of another pipe coincides with the groove to fix the relative position of the cone head mechanism and the groove of the other pipe, so that the welding torch in the cone head mechanism welds the two pipes.

[0318] In an embodiment of the present invention, by adjusting the contact state between the internal welding machine and the pipe in real time, the adaptability and stability of the internal welding machine are significantly improved, the influence of the harsh environment on the construction process is effectively reduced, and at the same time, the high quality of the welding operation is ensured, and the overall working efficiency is improved. By accurately utilizing the drive current of the pose adjustment structure, and combining the deep neural network technology and the dynamic model, the contact force between the cone head mechanism and the groove is predicted, and the pose adjustment control amount during the bending and alignment processes is calculated. By adjusting the pose by feedback of the contact force, the stability and accuracy of the welding process are ensured.

[0319] Combined with Figure 11 As shown in the content, the device further includes: a definition unit, a first creation unit, a second creation unit, and a construction unit.

[0320] A definition unit, configured to define the structure of the six-axis platform, and the structure at least includes a static platform, a moving platform, the cylinder of the branch chain, and the piston rod of the branch chain.

[0321] A first creation unit, configured to define the kinematic physical quantities based on the static platform coordinate system, and create the Jacobian matrix of the moving platform, the Jacobian matrix of the cylinder of the branch chain, and the Jacobian matrix of the piston rod of the branch chain in combination with the force data of the six-axis platform; the static platform coordinate system is a coordinate system constructed based on the static platform.

[0322] A second creation unit, configured to construct the inertia matrix of the cone head mechanism, the inertia matrix of the cylinder of the branch chain, and the inertia matrix of the piston rod of the branch chain.

[0323] A construction unit, configured to construct a six-axis dynamic model in combination with all the Jacobian matrices and all the inertia matrices.

[0324] Combined with Figure 11 As shown in the content, the device further includes: a first acquisition unit, a preprocessing unit, an output unit, a third calculation unit, a determination unit, and a parameter adjustment unit.

[0325] The first acquisition unit is configured to sample a plurality of loads from a preset load space, and for each load, acquire corresponding motion sequence data from a preset position space and a preset dynamic space.

[0326] The preprocessing unit is configured to input all the loads and all the motion sequence data into a six-axis dynamics model to obtain initial training data including branch chain motion data, current data, and branch chain thrust data, and preprocess the initial training data to obtain training data.

[0327] The output unit is configured to input the training data into a long short-term memory deep neural network model and output a predicted value of the branch chain thrust.

[0328] The third calculation unit is configured to calculate a loss function of the long short-term memory deep neural network model according to the predicted value.

[0329] The determination unit is configured to, if the loss function converges, determine the long short-term memory deep neural network model as a trained prediction model.

[0330] The parameter adjustment unit is configured to, if the loss function does not converge, adjust the parameters of the long short-term memory deep neural network model and return to execute the output unit.

[0331] Combined with Figure 11 As shown in the content, the device further includes: a second acquisition unit, a fourth calculation unit, and a fifth calculation unit.

[0332] The second acquisition unit is configured to acquire branch chain driving force data in multiple different postures of the attitude adjustment mechanism.

[0333] The fourth calculation unit is configured to calculate the load force and load torque in the moving platform coordinate system according to the multiple groups of branch chain driving force data through a real-time dynamics forward solution model; the moving platform coordinate system is a coordinate system established according to the moving platform in the attitude adjustment mechanism.

[0334] The establishment unit is configured to establish a linear equation between the load force, load torque, cone head weight, and center of gravity coordinates received by the moving platform of the attitude adjustment mechanism.

[0335] The fifth calculation unit is configured to calculate the cone head weight and center of gravity coordinates according to the least squares method and the linear equation.

[0336] Correspondingly, the first calculation unit 1101 is specifically configured to: obtain the actual position of the attitude adjustment mechanism, where the actual position includes the angle between each branch chain of the attitude adjustment mechanism and gravity, the deflection data and offset data of the branch chain; use the real-time dynamics forward solution model to calculate the gravity compensation amount of the moving platform on the cone head in the coordinate system of the moving platform according to the angle, deflection data, offset data, cone head weight, and center of gravity coordinates; through the real-time dynamics inverse solution model, convert the gravity compensation amount of the moving platform on the cone head into the gravity compensation amount of the branch chain on the cone head in the current attitude; the gravity compensation amount of the branch chain on the cone head is the force exerted by the cone head gravity on each branch chain; input the angle, deflection data, offset data, and gravity compensation amount of the branch chain on the cone head into the impedance controller to calculate the driving forces of multiple branch chains in the attitude adjustment mechanism.

[0337] Combined with Figure 11 the content shown, the adjustment unit includes a calculation module and an adjustment module.

[0338] The calculation module is configured to calculate the pose adjustment amount of the attitude adjustment mechanism by using the admittance controller according to the load force, load torque, and gravity compensation amount of the moving platform on the cone head.

[0339] The adjustment module is configured to adjust the lengths of each branch chain during the butt joint process by using the position controller according to the pose adjustment amount, so that the butt joint guiding block reaches the groove limit position; the position controller is pre-established based on the proportional-derivative controller.

[0340] The adjustment module is specifically configured to: calculate the butt joint output force of the attitude adjustment mechanism by using the position controller according to the pose adjustment amount; adjust the lengths of each branch chain during the butt joint process through the butt joint output force.

[0341] Combined with Figure 11 the content shown, the calculation module includes a calculation sub-module, an acquisition sub-module, an input sub-module, and an iterative solution sub-module.

[0342] The calculation sub-module is configured to calculate the difference between the load force and load torque minus the gravity compensation amount of the moving platform on the cone head to obtain the actual contact resultant force of several butt joint guiding blocks in the cone head mechanism in contact with the groove.

[0343] The acquisition sub-module is configured to acquire the actual positions and actual speeds of each branch chain during the butt joint process.

[0344] The input sub-module is configured to input the preset expected contact resultant force, actual contact resultant force, actual position, and actual speed into the admittance controller.

[0345] The iterative solution sub-module is configured to combine the ordinary differential equations in the admittance controller and perform iterative solution according to the time step of the control period to obtain the pose adjustment amount of the attitude adjustment mechanism.

[0346] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the differences between each embodiment and other embodiments are emphasized. In particular, for a system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the description of the method embodiment. The systems and system embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative work.

[0347] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0348] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An automatic alignment method for an internal welding machine, characterized in that: The method comprises: When the internal welding machine moves to the elbow inside the pipeline through the fuselage motion system, the driving force of multiple branches in the attitude adjustment mechanism is calculated according to the actual position of the attitude adjustment mechanism by using the impedance controller; the control strategy of the impedance controller is pre-created according to the target inertia, target stiffness value and target damping; Driving the cone head mechanism in the internal welding machine to deflect according to the driving force, so that the internal welding machine passes through the bent pipe; When the internal welding machine moves to the pipe end, the alignment guide block is raised to perform an alignment operation with the groove of the pipe end, and the alignment operation includes: The prediction model is used to predict the current actual thrust of each branch chain based on the branch chain motion data and current data in the posture adjustment mechanism; the prediction model is obtained by training the neural network model in advance using the six-axis dynamics model; the six-axis dynamics model is constructed in advance according to the six-axis platform; The load force and load moment in the moving platform coordinate system are calculated according to the current actual thrust of each branch chain through a real-time dynamics forward solution model; the six-axis platform includes a moving platform and a static platform; the moving platform coordinate system is a coordinate system constructed according to the moving platform; The admittance controller is used to calculate the posture adjustment amount of the posture adjustment mechanism according to the load force and the load torque, and the length of each branch chain is adjusted during the docking process so that the docking guide block reaches the groove limit position; the admittance controller is preliminarily established based on the second-order differential model; After the matching operation is completed, the rear expansion shoe is raised to fix the relative position of the cone head mechanism and the groove, and the matching guide block is retracted; When the groove of another pipeline overlaps with the groove, the front expansion shoe is raised to fix the relative position of the cone head mechanism and the groove of the other pipeline, so that the welding gun in the cone head mechanism welds the two pipelines.

2. The method according to claim 1, characterized in that The process of building a six-axis dynamic model based on the six-axis platform in advance includes: Define the structure of the six-axis platform, which at least includes a static platform, a dynamic platform, a cylinder barrel of a branch chain, and a piston rod of the branch chain; Based on the static platform coordinate system, the motion physical quantity is defined, and the Jacobian matrix of the dynamic platform, the Jacobian matrix of the cylinder barrel of the branch chain, and the Jacobian matrix of the piston rod of the branch chain are created in combination with the force data of the six-axis platform; the static platform coordinate system is a coordinate system constructed according to the static platform; Construct the inertia matrix of the cone head mechanism, the inertia matrix of the cylinder barrel of the branch chain, and the inertia matrix of the piston rod of the branch chain; Combine all Jacobian matrices and all inertia matrices to build a six-axis dynamic model.

3. The method according to claim 1, characterized in that The process of pre-training the neural network model using the six-axis dynamics model to obtain the prediction model includes: Sampling a plurality of loads from a preset load space, and for each of the loads, acquiring corresponding motion sequence data from a preset position space and a preset dynamic space; Inputting all loads and all motion sequence data into the six-axis dynamics model to obtain initial training data including branch chain motion data, current data and branch chain thrust data, and preprocessing the initial training data to obtain training data; Inputting the training data into a long short-term memory deep neural network model, and outputting a predicted value of the branch chain thrust; Calculating a loss function of the long short-term memory deep neural network model according to the predicted value; If the loss function converges, determining the long short-term memory deep neural network model as a trained prediction model; If the loss function does not converge, the parameters of the long short-term memory deep neural network model are adjusted, and the process returns to the step of inputting the training data into the long short-term memory deep neural network model and outputting the predicted value of the branch chain thrust.

4. The method according to claim 1, characterized in that: The method further comprises: Acquiring driving force data of multiple groups of branches in the posture adjustment mechanism under different postures; According to the multiple sets of branch chain driving force data, the load force and load moment in the moving platform coordinate system are calculated by a real-time dynamics forward solution model; the moving platform coordinate system is a coordinate system established according to the moving platform in the posture adjustment mechanism; Establishing a linear equation of the load force, load moment, cone head weight and center of gravity coordinates on the moving platform of the posture adjustment mechanism; Calculate the cone head weight and the center of gravity coordinates according to the least square method and the linear equation; Correspondingly, the use of the impedance controller to calculate the driving force of the plurality of branches in the posture adjustment mechanism according to the actual position of the posture adjustment mechanism includes: Acquiring the actual position of the attitude adjustment mechanism, wherein the actual position includes the angle between each branch chain in the attitude adjustment mechanism and gravity, deflection data and offset data of the branch chain; Using a real-time dynamics forward solution model, the gravity compensation amount of the cone head by the moving platform in the coordinate system below the moving platform is calculated according to the angle, the deflection data, the offset data, the cone head weight and the center of gravity coordinates; The gravity compensation amount of the moving platform on the cone head is converted into the gravity compensation amount of the branch chain on the cone head in the current posture through the real-time dynamic inverse solution model; the gravity compensation amount of the branch chain on the cone head is the force applied by the gravity of the cone head to each branch chain; The angle, the deflection data, the offset data and the amount of gravity compensation of the branch chain on the cone head are input into an impedance controller to calculate the driving force of multiple branch chains in the posture adjustment mechanism.

5. The method according to claim 4, characterized in that The method of calculating the posture adjustment amount of the posture adjustment mechanism according to the load force and the load torque by using the admittance controller and adjusting the length of each branch chain during the docking process so that the docking guide block reaches the groove limit position includes: Calculating the posture adjustment amount of the posture adjustment mechanism according to the load force and the load moment, and the gravity compensation amount of the cone head by the moving platform using the admittance controller; A position controller is used to adjust the length of each branch chain during the docking process according to the posture adjustment amount, so that the docking guide block reaches the groove limit position; the position controller is pre-established based on a proportional-differential controller.

6. The method according to claim 5, characterized in that The method of calculating the posture adjustment amount of the posture adjustment mechanism by using the admittance controller according to the load force and the load moment, and the gravity compensation amount of the cone head by the moving platform, comprises: Calculate the difference between the load force and the load moment minus the compensation amount of the moving platform on the cone head gravity to obtain the actual contact force between the plurality of matching guide blocks and the groove in the cone head mechanism; Obtaining the actual position and actual speed of each branch chain during the docking process; Inputting the preset expected contact resultant force, the actual contact resultant force, the actual position and the actual speed into the admittance controller; Combined with the ordinary differential equation group in the admittance controller, an iterative solution is performed according to the time step of the control cycle to obtain the posture adjustment amount of the posture adjustment mechanism.

7. The method according to claim 6, characterized in that The step of adjusting the length of each branch chain during the docking process according to the posture adjustment amount by using the position controller includes: Calculating the corresponding output force of the posture adjustment mechanism according to the posture adjustment amount using a position controller; The length of each branch chain during the docking process is adjusted by the docking output force.

8. An automatic docking device for an internal welding machine, characterized in that: The device comprises: The first calculation unit is used to calculate the driving force of multiple branches in the attitude adjustment mechanism according to the actual position of the attitude adjustment mechanism by using the impedance controller when the internal welding machine moves to the elbow inside the pipeline through the fuselage motion system; the control strategy of the impedance controller is pre-created according to the target inertia, the target stiffness value and the target damping; A driving unit, used for driving a cone head mechanism in the internal welding machine to deflect according to the driving force, so that the internal welding machine passes through the bent pipe; An alignment unit, used for raising the alignment guide block and the groove of the pipe end when the internal welding machine moves to the pipe end, and performing an alignment operation based on the prediction unit, the second calculation unit and the adjustment unit; A prediction unit, used for predicting the current actual thrust of each branch chain based on the branch chain motion data and current data in the posture adjustment mechanism by using a prediction model; the prediction model is obtained by training a neural network model in advance using a six-axis dynamics model; the six-axis dynamics model is constructed in advance according to a six-axis platform; The second calculation unit is used to calculate the load force and load moment in the dynamic platform coordinate system according to the current actual thrust of each branch chain through a real-time dynamics forward solution model; the six-axis platform includes a dynamic platform and a static platform; an adjustment unit, configured to calculate the posture adjustment amount of the posture adjustment mechanism according to the load force and the load torque by using an admittance controller, and adjust the length of each branch chain during the docking process so that the docking guide block reaches the groove limit position; the admittance controller is preliminarily established based on a second-order differential model; A first fixing unit is used to raise the rear expansion shoe to fix the relative position of the cone head mechanism and the groove after the matching operation is completed, and to retract the matching guide block; The second fixing unit is used to raise the front expansion shoe to fix the relative position of the cone head mechanism and the groove of the other pipe when the groove of the other pipe overlaps with the groove, so that the welding gun in the cone head mechanism can weld the two pipes.

9. The device according to claim 8, characterized in that The device also includes: A definition unit is used to define the structure of the six-axis platform, wherein the structure at least includes a static platform, a dynamic platform, a cylinder barrel of a branch chain, and a piston rod of the branch chain; A first creation unit is used to define the motion physical quantity based on the static platform coordinate system, and to create the Jacobian matrix of the dynamic platform, the Jacobian matrix of the cylinder barrel of the branch chain, and the Jacobian matrix of the piston rod of the branch chain in combination with the force data of the six-axis platform; the static platform coordinate system is a coordinate system constructed according to the static platform; The second creation unit is used to construct the inertia matrix of the cone head mechanism, the inertia matrix of the cylinder barrel of the branch chain, and the inertia matrix of the piston rod of the branch chain; The construction unit is used to combine all Jacobian matrices and all inertia matrices to construct a six-axis dynamic model.

10. An internal welding machine, characterized in that: It includes a cone head mechanism, a body motion system, an attitude adjustment mechanism, a memory and a processor; The cone head mechanism includes a matching guide block; The attitude adjustment mechanism is located between the cone head mechanism and the fuselage motion system; the attitude adjustment mechanism at least includes a plurality of branch chains and an attitude guide wheel; A computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of the automatic alignment method for an internal welding machine as claimed in any one of claims 1 to 7 are implemented.

Images