GIL pipe manufacturing dynamic centering collaborative regulation method based on spatial pose perception

Abstract

This invention belongs to the field of pipe assembly and intelligent manufacturing technology, and relates to a dynamic centering collaborative control method for GIL pipe manufacturing based on spatial pose perception. The method includes: acquiring one-dimensional distance time-series data output by a laser ranging array at 5ms intervals and reconstructing it into a high-dimensional observation vector; calling the spatial state transition matrix to calculate a six-degree-of-freedom deviation state vector and performing low-pass filtering; extracting the first-order time-domain derivative and second-order time-domain cumulative features of the deviation state vector to calculate the time-varying damping correction coefficient; adjusting the feedback adjustment step size or switching to a discrete self-optimizing control mode based on the coefficient to decouple and obtain the adjustment amount; calculating the drive frequency control command based on the adjustment amount and performing feedforward superposition and saturation limiting before outputting it to the frequency converter. This invention implements step size adjustment and mode switching based on time-varying damping correction, blocks deformation time delay noise amplification, eliminates adjustment overshoot, and enhances the adaptive decoupling performance and precise centering convergence quality of the control system.

Description

Technical Field

[0001] This invention belongs to the field of pipe assembly and intelligent manufacturing technology, and relates to a dynamic centering and collaborative control method for GIL pipe manufacturing based on spatial pose perception. Background Technology

[0002] Currently, utilizing high-precision non-contact optical measurement and multi-degree-of-freedom digital attitude adjustment mechanisms to achieve automated intelligent docking of large-diameter pipes, such as GIS / GIL pipes, has become a core approach in the fields of power assembly and intelligent manufacturing. However, existing technologies still have limitations in handling three-dimensional dynamic space interference prevention and physical interference, high-frequency environmental disturbance immunity, and multi-dimensional trajectory closed-loop convergence throughout the docking process.

[0003] First, there is the contact-based relative pose measurement technology based on center-type circular rotation scanning. For example, Chinese invention patent application publication number CN108168432A discloses a pipe flange relative pose measurement system and method. By arranging a solid sensor rotation mechanism between two pipe flanges, a line-scanning laser rangefinder sensor is driven to perform circular motion to obtain information on the flange surfaces on both sides. Its shortcomings are: this scheme has strong physical occupancy in spatial architecture, and its measurement and decision-making logic lacks a rigid representation of the irreversible contraction of the central residence space of the pipe fitting during the final closing docking stage. It fails to map the inherent physical volume of the solid scanning mechanism into the dynamic boundary constraints of the subsequent docking solution space, which makes it easy for the guiding model to output movement commands that cannot be executed due to physical interference in the middle and later stages. In addition, it is highly dependent on the initial processing consistency of the discrete hole centers on the flange edge and lacks the ability to tolerate and dynamically calibrate topological geometric defects such as flange compression torsion and welding stress deformation in actual engineering.

[0004] Second, there is the large field-of-view attitude guidance technology based on macroscopic panoramic machine vision. For example, Chinese invention patent application publication number CN121546479A discloses a machine vision-based GIL pipe-assisted docking method and system. This method uses a panoramic camera deployed on a truss or robotic arm to construct a three-dimensional environment model for point cloud registration and combines historical pose differences for linear regression prediction compensation. However, this technology has the following drawbacks: it focuses on environmental state recognition and open-loop prediction under a macroscopic large field of view, and fails to establish a closed-loop correlation between the accuracy of feature points reconstructed by macroscopic vision and the microscopic bolt hole assembly tolerance inside the narrow gap docking surface. When facing complex construction environments, such as strong light variations, strong wind disturbances, and dust obstruction in substations and underground tunnels, it cannot map the physical compression of the control path by high-frequency Gaussian noise in real time. At the same time, the historical pose linear regression prediction model it uses is prone to cumulative drift and time lag in predicting trajectories when facing nonlinear and multi-frequency non-stationary swaying of suspended pipelines caused by cable inertia or sudden crosswinds. This makes it difficult to cope with the risk of equipment insulation medium damage or collision leakage under extreme disturbances.

[0005] Third, static coaxiality detection technology based on local discrete node collimation; for example, Chinese invention patent application publication number CN117606391A discloses a device and method for coaxiality detection of GIS or GIL straight segment units. By manually clamping scale units with photosensitive scales onto each connecting flange bolt, the eccentricity angle and displacement are displayed by observing the eccentric scale of the initial laser beam. Its drawback is that existing verification technologies of this kind usually regard the continuous docking guidance process as a static, discontinuous or fixed-step discrete node decision unit, ignoring the coupling effect of the spatial pose of the pipe fitting on the continuous spatiotemporal axis during dynamic travel and bolt pre-tightening. When facing dynamic assembly optimization, due to the lack of a mathematical solution mechanism to transform the local planar eccentricity parameters into a three-dimensional full-space six-degree-of-freedom 6-DOF relative pose geometric transformation matrix, if there is no dynamic forced shear control of physical safety boundaries, the attitude adjustment trajectory is prone to divergence or cumulative deviation, causing the output adjustment scheme to lose the value of fully closed-loop automated guidance due to exceeding the spatial attitude tolerance limit.

[0006] Therefore, the technical problem to be solved by this invention is how to construct an interference-free measurement architecture with full dynamic three-dimensional spatial pose perception and full closed-loop anti-disturbance capability, so that the pose guidance and calculation logic converges to the real microscopic assembly physical safety boundary throughout the entire docking cycle. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and solve the technical problems of the prior art in the whole process of pipe docking, which are caused by welding thermal distortion, self-weight deflection deformation and the intertwined coupling of actuator response hysteresis, resulting in the control path easily outputting physical interference commands, trajectory drift divergence and command overshoot chatter. The invention provides a dynamic centering and collaborative control method for GIL pipe manufacturing based on spatial pose perception.

[0008] To achieve the above-mentioned objectives, this invention provides a dynamic centering and collaborative control method for GIL (Gas Inertial Isolation System) pipe manufacturing based on spatial pose perception, comprising the following steps: Step S1: Acquire ranging data and combine high-dimensional observation vectors in 5ms cycle: Acquire one-dimensional distance time series data output by the laser ranging array distributed in the circumference of the pipe in a 5ms sampling cycle and combine them into a high-dimensional observation vector; Step S2, solve the six-degree-of-freedom deviation state vector and filter out high-frequency noise: call the spatial state transition matrix to solve the high-dimensional observation vector into a six-degree-of-freedom deviation state vector of the relative spatial position of the pipe mating surface, and use the sliding window median filter operator to filter out the high-frequency noise components in the six-degree-of-freedom deviation state vector. Step S3: Calculate the damping correction coefficient, adjust the feedback step size, or switch to discrete mode: Extract the first-order time-domain derivative and second-order time-domain cumulative characteristics of the filtered six-degree-of-freedom deviation state vector within the sliding time window. Calculate the time-varying damping correction coefficient within the control loop based on the first-order time-domain derivative and second-order time-domain cumulative characteristics. When the time-varying damping correction coefficient monotonically increases and is within the range of 1.25 to 1.55, monotonically reduce the feedback adjustment step size by 20%. When the time-varying damping correction coefficient is greater than 1.55, suspend the continuous deviation prediction operator and switch to a discrete self-optimization control mode with a fixed step size of 0.05mm to calculate the adjustment amount for each axis. Step S4: Calculate the frequency command, superimpose feedforward, limit and output to the frequency converter: Calculate the drive frequency control command based on the adjustment amount of each axis. When the calculated time domain transmission hysteresis parameter is greater than 15ms, superimpose the pulse feedforward control amount into the drive frequency control command, use the saturation cut-off valve to limit the highest output change rate of a single command, and output the limited drive frequency control command to the frequency converter.

[0009] In this invention, steps S1 to S4 form a closed-loop logic flow with parameter interlocking within the same programmable automation controller; in step S2, the spatial state transition matrix and sliding window median filter operator are used to correct the six-degree-of-freedom deviation state vector; in step S3, the step size is adjusted by shrinking feedback based on the first-order time-domain derivative and the second-order time-domain cumulative characteristic; in step S4, pulse-type feedforward control quantities are seamlessly superimposed to offset the commutation dead zone, so that the residual misalignment of the docking surface is stabilized below 0.30mm.

[0010] Step S1 of the present invention includes the following sub-steps: Step S11, using a laser ranging sensor group to obtain the spatial coordinates of multiple points on the end face of the pipe; Step S12, inputting the spatial coordinates of multiple points into the laser ranging array to convert them into one-dimensional distance time series data and constructing a high-dimensional observation vector.

[0011] Step S2 of the present invention further includes welding thermal strain offsetting processing implemented by a parallel-running thermal strain spatiotemporal delay compensation operator, including the following sub-steps: Step S21, calculating the local thermal deformation at the pipe joint based on the real-time acquired welding current signal and the cumulative timing of the heat input; Step S22, subtracting the local thermal deformation in situ from the six-degree-of-freedom deviation state vector to remove the pseudo-pose deviation component caused by thermal expansion, and inputting the corrected six-degree-of-freedom deviation state vector into step S3.

[0012] The calculation of the time-domain propagation hysteresis parameter in step S4 of the present invention includes the following sub-steps: Step S41, recording the system timestamp of sending the drive frequency control command from the time-domain response state observation unit; Step S42, obtaining the response timestamp of the pose sensor detecting the physical displacement of the pipe axis; Step S43, calculating the difference between the response timestamp and the system timestamp to obtain the time-domain propagation hysteresis parameter.

[0013] The seamless superposition of the pulse-type feedforward control quantity in step S4 of the present invention includes the following sub-steps: Step S44, when the calculated time-domain transmission hysteresis parameter is greater than 15ms, read the coupling static clearance parameter to determine the current transmission dead zone size; Step S45, generate a rectangular pulse signal with an amplitude positively correlated with the current transmission dead zone size and a duration equal to 10ms as the pulse-type feedforward control quantity; Step S46, seamlessly superimpose the rectangular pulse signal into the drive frequency control command to offset the commutation hysteresis caused by the transmission clearance.

[0014] The output to the frequency converter in step S4 of the present invention includes the following sub-steps: Step S47, converting the drive frequency control command containing the pulse feedforward control quantity into a pulse width modulation signal; Step S48, periodically sending the pulse width modulation signal to the frequency converter through the industrial Ethernet bus to adjust the motor speed of the multi-axis servo synchronous drive system.

[0015] Before the limited drive frequency control command is output to the frequency converter, step S4 of the present invention further includes a saturation cutoff limiting sub-step: step S49, the maximum output change rate of a single command is limited by the saturation cutoff valve, the amount of change exceeding the upper limit of the amplitude change rate is cut off and the change amplitude of the drive frequency control command is limited during adjacent cycles.

[0016] The present invention further includes a centering convergence determination step after the drive frequency control command is output to the frequency converter: Step S5, continuously acquire the six-degree-of-freedom deviation state vector, when the value of each degree of freedom of the six-degree-of-freedom deviation state vector is less than the first precision red line and the duration reaches 2s, stop sending the frequency control command to complete the automatic centering and assembly of the pipe.

[0017] Compared with the prior art, the present invention has at least the following beneficial effects: 1. In the dynamic alignment and coordinated control of GIL pipe manufacturing, a high-dimensional observation vector is reconstructed using circumferentially distributed ranging data. The six-degree-of-freedom deviation vector of the docking surface is solved using the spatial state transition matrix. A first-order low-pass sliding window median filter operator is used to eliminate measurement noise caused by arc reflection and sudden surface corrosion to provide a deterministic signal. A time-varying damping correction coefficient is constructed by combining the first-order time-domain derivative and the second-order time-domain cumulative characteristics of the deviation vector. When the coefficient is in the range of 1.25 to 1.55, the feedback adjustment step size is monotonically reduced by 20%. When it is greater than 1.55, continuous deviation prediction is suspended and the discrete step adjustment is switched to block the random amplification of noise by the differential term and eliminate servo overshoot and mechanical chatter caused by frequent alternation of control commands.

[0018] 2. This method collects welding current signals and heat input cumulative timing by using parallel-running thermal strain spatiotemporal delay compensation operators. It calculates the local thermal deformation at the pipe joint based on the deformation mechanism and subtracts the local thermal deformation in situ from the six-degree-of-freedom deviation state vector. This removes the pseudo-pose deviation component caused by thermal expansion at the front end of the feedback loop, so that the adaptive adjustment loop only makes a corrective response to the pure mechanical pose deviation caused by geometric misalignment. This breaks the strong coupling constraint of welding thermal distortion and geometric misalignment at the mechanism level and avoids the divergence of the control loop caused by the pseudo-signal of thermal expansion.

[0019] 3. This method utilizes the actuator time-domain response state observation unit to compare the moment the drive control command is issued with the initial moment when the physical response is captured by the ranging end, and calculates the current dynamic transmission hysteresis parameter. When the parameter is greater than the 15ms threshold, a pulse-type dead-zone feedforward control quantity is seamlessly superimposed at the control output to offset the commutation dead zone. This avoids excessive accumulation of integral gain in the control closed loop due to transmission time delay at the critical point of heavy-load start-up, eliminates spatial pose overshoot and directional oscillation caused by the drive mechanism crossing the dead zone, and enables the control system to have adaptive dead-zone self-calibration capability under non-ideal high-load inertial conditions, thereby maintaining the action synchronization of the multi-axis linkage mechanism in the continuous control process. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the control flow of the GIL pipe dynamic alignment and coordinated control method of the present invention; Figure 2 This is a schematic diagram of the state transition of the control loop step size and mode switching in this invention. Detailed Implementation

[0021] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments and accompanying drawings.

[0022] Example 1: This embodiment discloses a dynamic centering and collaborative control method for GIL pipe manufacturing based on spatial pose perception, including the following steps: Step S1: Acquire ranging data and combine high-dimensional observation vectors in 5ms cycle: Acquire one-dimensional distance time series data output by the laser ranging array distributed in the circumference of the pipe in a 5ms sampling cycle and combine them into a high-dimensional observation vector; Step S2, solve the six-degree-of-freedom deviation state vector and filter out high-frequency noise: call the spatial state transition matrix to solve the high-dimensional observation vector into a six-degree-of-freedom deviation state vector of the relative spatial position of the pipe mating surface, and use the sliding window median filter operator to filter out the high-frequency noise components in the six-degree-of-freedom deviation state vector. Step S3: Calculate the damping correction coefficient, adjust the feedback step size, or switch to discrete mode: Extract the first-order time-domain derivative and second-order time-domain cumulative characteristics of the filtered six-degree-of-freedom deviation state vector within the sliding time window. Calculate the time-varying damping correction coefficient within the control loop based on the first-order time-domain derivative and second-order time-domain cumulative characteristics. When the time-varying damping correction coefficient monotonically increases and is within the range of 1.25 to 1.55, monotonically reduce the feedback adjustment step size by 20%. When the time-varying damping correction coefficient is greater than 1.55, suspend the continuous deviation prediction operator and switch to a discrete self-optimization control mode with a fixed step size of 0.05mm to calculate the adjustment amount for each axis. Step S4: Calculate the frequency command, superimpose feedforward, limit and output to the frequency converter: Calculate the drive frequency control command based on the adjustment amount of each axis. When the calculated time domain transmission hysteresis parameter is greater than 15ms, superimpose the pulse feedforward control amount into the drive frequency control command, use the saturation cut-off valve to limit the highest output change rate of a single command, and output the limited drive frequency control command to the frequency converter.

[0023] In this embodiment, steps S1 to S4 form a closed-loop logic flow with parameter interlocking within the same programmable automation controller; in step S2, the spatial state transition matrix and sliding window median filter operator are used to correct the six-degree-of-freedom deviation state vector; in step S3, the step size is adjusted by shrinking feedback based on the first-order time-domain derivative and the second-order time-domain cumulative characteristic; in step S4, the pulse-type feedforward control quantity is seamlessly superimposed to offset the commutation dead zone, so that the residual misalignment of the docking surface is stabilized below 0.30mm.

[0024] Step S1 in this embodiment includes the following sub-steps: Step S11, using a laser ranging sensor group to obtain the spatial coordinates of multiple points on the pipe end face; Step S12, inputting the spatial coordinates of multiple points into the laser ranging array to convert them into one-dimensional distance time series data and construct a high-dimensional observation vector.

[0025] Step S2 in this embodiment also includes welding thermal strain offsetting processing implemented by a parallel-running thermal strain spatiotemporal delay compensation operator, including the following sub-steps: Step S21, calculate the local thermal deformation at the pipe joint based on the real-time collected welding current signal and the cumulative timing of the heat input; Step S22, subtract the local thermal deformation in situ from the six-degree-of-freedom deviation state vector to remove the pseudo-pose deviation component caused by thermal expansion, and input the corrected six-degree-of-freedom deviation state vector to step S3.

[0026] In this embodiment, step S4, calculating the time-domain propagation hysteresis parameter, includes the following sub-steps: Step S41, recording the system timestamp of sending the drive frequency control command from the time-domain response state observation unit; Step S42, obtaining the response timestamp of the pose sensor detecting the physical displacement of the pipe axis; Step S43, calculating the difference between the response timestamp and the system timestamp to obtain the time-domain propagation hysteresis parameter.

[0027] In this embodiment, step S4, which involves seamlessly superimposing the pulse-type feedforward control quantity, includes the following sub-steps: Step S44, when the calculated time-domain transmission hysteresis parameter is greater than 15ms, the static clearance parameter of the coupling is read to determine the current transmission dead zone size; Step S45, a rectangular pulse signal with an amplitude positively correlated with the current transmission dead zone size and a duration equal to 10ms is generated as the pulse-type feedforward control quantity; Step S46, the rectangular pulse signal is seamlessly superimposed onto the drive frequency control command to offset the commutation hysteresis caused by the transmission clearance.

[0028] In this embodiment, step S4, outputting to the frequency converter includes the following sub-steps: Step S47, converting the drive frequency control command containing pulse feedforward control quantity into a pulse width modulation signal; Step S48, periodically sending the pulse width modulation signal to the frequency converter via the industrial Ethernet bus to adjust the motor speed of the multi-axis servo synchronous drive system.

[0029] In this embodiment, before the limited drive frequency control command is output to the inverter, step S4 further includes a saturation cutoff limiting sub-step: step S49, the highest output change rate of a single command is limited by the saturation cutoff valve, the amount of change exceeding the upper limit of the amplitude change rate is cut off and the change amplitude of the drive frequency control command is limited during adjacent cycles.

[0030] This embodiment includes a centering convergence determination step after the drive frequency control command is output to the frequency converter: Step S5, continuously acquire the six-degree-of-freedom deviation state vector, when the value of each degree of freedom of the six-degree-of-freedom deviation state vector is less than the first precision red line and the duration reaches 2s, stop sending the frequency control command to complete the automatic centering and assembly of the pipe.

[0031] Example 2: In this embodiment, when the system faces the continuous processing conditions of automated manufacturing and assembly welding of large-diameter long pipes for ultra-high voltage gas-insulated metal-enclosed transmission lines, the aluminum alloy outer shell pipe with a length of 12m and a diameter of 800mm is dynamically moved at a continuous propulsion speed of 50mm / s accompanied by a rotation of 2rpm. Due to the continuous self-weight deflection and bending deformation of the large-span metal pipe as the cantilever span elongates, and the asymmetric transient heat input generated by continuous welding at the joint causes non-uniform thermal expansion of the metal material, the mechanical deformation signal accompanying the welding temperature rise evolution mechanism is mixed into the spatial ranging channel. This leads to a strongly coupled nonlinear noise signal, including material elastic deformation and self-weight sagging, being input into the control unit feedback loop. In this case, if a traditional negative feedback loop is used... The multi-axis hydraulic bracket for position correction control lacks an internal decoupling mechanism for the intertwined and coupled signals of geometric misalignment, gravitational deformation fluctuations, and actuator response hysteresis. Control commands typically lag behind the actual deformation rate of the pipe, leading to saturation divergence in the regulating loop under continuous accumulation of integral gain. This results in frequent commutation commands from the actuator, causing control overshoot and command chatter along the machining axis, leading to mechanical transmission losses. To address this dynamic instability of the spatial axis, the control unit periodically collects one-dimensional distance time-series data from a laser ranging array distributed around the pipe mating surface at a fixed sampling period of 5ms via a field industrial Ethernet bus. The control unit internally combines this raw distance time-series data into a high-dimensional observation vector. The control unit invokes a preset spatial state transition equation to transform the high-dimensional observation vector. The mapping solution is a spatial six-degree-of-freedom deviation state vector representing the relative spatial position of the mating surfaces of the two pipe sections. Based on this, to address the pose drift caused by asymmetric heat input from the welding power source, a parallel-running thermal strain spatiotemporal delay compensation operator calculates the local thermal deformation at the pipe joint based on the real-time acquired welding current signal and the cumulative time sequence of heat input, and then calculates the spatial six-degree-of-freedom deviation state vector. The local thermal deformation is subtracted from the in-situ, thereby isolating the deformation component caused by thermal expansion. This allows the adjustment loop to only correct mechanical orientation deviations caused by geometric misalignment. At the same time, a first-order low-pass sliding window median filter operator is used to filter the corrected deviation state vector to remove high-frequency noise components and eliminate measurement noise caused by arc reflection and sudden surface corrosion.

[0032] Obtain the filtered spatial six-degree-of-freedom deviation state vector Subsequently, the collaborative control module operating inside the control unit extracts the spatial six-degree-of-freedom deviation state vector. Within the sliding time window, the first-order time-domain derivative and the second-order time-domain cumulative characteristics are used. The first-order time-domain derivative is used to characterize the transient evolution trend of the cantilever deflection due to the self-weight of the pipe, and the second-order time-domain cumulative characteristics are used to deduce the transient transmission hysteresis factor of the multi-axis actuator. Together, these factors are used to construct the time-varying damping correction coefficient within the control loop. Time-varying damping correction factor Satisfy the following formula: ,in, This is the time-varying damping correction coefficient within the control loop. The first weighted feature constant is preset. , is the first-order differential vector of the pose state, used to characterize the first-order time-domain derivative of the cantilever deflection due to the self-weight of the pipe; The second weighted feature constant is preset. is a finite time-domain integral vector of the pose state, used to characterize the second-order time-domain cumulative feature of the pose state; To calibrate the inherent stiffness of the control loop system and eliminate the inconsistency in physical dimensions between the first-order differential vector and the finite-time integral vector when directly summing them, the first-order differential vector is used in the actual data processing flow within the control loop. With finite-time integral vector Multiply by the characteristic constant and Previously, each quantity was pre-processed by using a built-in normalization operator to strip away its corresponding physical units, and the differential term was divided by the normalization operator in the control kernel. The reference velocity scalar, divide the integral term by The reference spatiotemporal cumulative scalar is converted into a purely numerical dimensionless scalar before being added and composed. This ensures that the time-varying damping correction coefficient obtained after the composite accumulation is obtained. It possesses completely dimensionless properties, achieving absolute self-consistency in mathematical logic and physical meaning of multidimensional control loops.

[0033] When the time-domain transfer hysteresis parameter calculated by the control unit exceeds the preset time threshold of 15ms, the control unit feeds forward a pulse-type dead-time control quantity into the drive frequency control command to compensate for the commutation delay of the servo adjustment mechanism, while simultaneously adjusting the time-varying damping correction coefficient. When the preset safety damping critical threshold of 1.25 is exceeded, the collaborative control module automatically triggers the conditional control flow graded cutoff rule, if the time-varying damping correction coefficient... Monotonically increasing and within the threshold range of 1.25 to 1.55, the control unit monotonically reduces the feedback adjustment step size of the control loop by 20%, while the time-varying damping correction coefficient... When the deviation exceeds a threshold of 1.55, the control unit suspends the deviation prediction operator and switches to a discrete self-optimization control mode with a fixed step size of 0.05mm. This prevents the random amplification of noise by the differential term. Specifically, the objective function of the discrete self-optimization control mode is set to minimize the Euclidean norm of the relative spatial six-degree-of-freedom deviation state vector of the pipe mating surface. After suspending the continuous deviation prediction operator, the control unit uses the six-degree-of-freedom deviation state vector of the current cycle as the reference input and applies a magnitude of [value missing] in turn to each independent control axis of the multi-axis linkage servo system. The system employs exploratory fixed-step stair-stepping and monitors the changing trend of the six-degree-of-freedom deviation state vector norm in real time. If the deviation norm monotonically decreases after stepping along a certain axis, the stepping direction along that axis is maintained and the deviation norm is output. The adjustment amount is adjusted, and vice versa, until the deviation values ​​of all dimensions converge to within the preset accuracy red line. In this way, by disconnecting the derivative feedback of continuous prediction, deterministic multi-axis cooperative decoupling adjustment is completed in discrete space. The control unit generates corresponding drive frequency control commands based on the adjustment amounts of each axis obtained from the above adaptive decoupling calculation. The control unit limits the maximum output change rate of a single command through a saturation limiting valve, cutting off changes exceeding the upper limit of the amplitude change rate, and ultimately controlling the drive frequency command. The power frequency conversion unit, which periodically outputs power to the servo bracket, drives the multi-axis actuator to complete adaptive alignment, reducing the residual misalignment of the mating surface from the initial 1.24mm to 0.11mm. This value is stable below the 0.30mm pipe manufacturing specification red line. The multi-axis linkage mechanism maintains motion synchronization under load inertia conditions. The control loop generates decoupling adjustment between time-varying gravity deflection deformation, actuator motion hysteresis, and geometric misalignment through data signal conversion. The control system relies on the dynamic correlation adjustment of internal multi-dimensional parameters to eliminate the degradation of control quality caused by physical gaps and environmental interference.

[0034] Example 3: In this embodiment, when the system faces the verification of the dynamic alignment control accuracy of the spatial geometric axis in the manufacturing of large and heavy-duty pipes, the test platform is based on the load hydraulic synchronous alignment mechanical assembly. The input data is provided by a circumferentially arranged laser ranging array and a variable frequency power system. The measurement accuracy of the laser ranging array is 0.01mm and the sampling frequency is 200Hz. Regarding the sampling period control parameter, the centrifugal vibration frequency of the pipe rotation, the response bandwidth of the servo motor, and the bus transmission rate impose a limit on the value of the sampling period. In order to balance the real-time tracking accuracy of the signal and the calculation load of the control unit, a parameter constraint relationship is established according to the sampling theorem. When the maximum welding interference cutoff frequency of the pipe joint is within the range of 100Hz, in order to avoid high-frequency aliasing of the control loop signal and capture the pose change caused by asymmetric thermal input, the sampling period tends to the upper limit cutoff time domain of the system calculation time window. The parameter constraint relationship is applied to determine that the set value of this sampling period is 5ms.

[0035] The control unit operates under low disturbance intensity gradient conditions, with the initial alignment geometric deviation of the pipe set at 0.50 mm and the welding current at 120 A. The distance data acquired by the laser ranging array is superimposed with Gaussian white noise with a signal-to-noise ratio of 20 dB and power frequency interference harmonics at a frequency of 50 Hz. Test data shows that the spatial six-degree-of-freedom deviation state vector output by the control group without time-varying damping correction and thermal strain spatiotemporal delay compensation... Axial oscillation occurs, and the control command reversing frequency reaches 15.3Hz, resulting in a residual misalignment of 0.18mm at the mating surface. However, the prototype of this invention utilizes a first-order low-pass sliding window median filter operator to eliminate high-frequency noise components and a thermal strain time-space delay compensation operator to isolate thermal expansion interference. The residual misalignment at the mating surface converges to 0.04mm, and the output command reversing chatter frequency of the transmission mechanism decreases to 0Hz. Under moderate disturbance gradient conditions, the control unit adjusts the initial alignment geometric deviation of the pipe to 1.50mm and increases the welding current to 240A. At this time, the control group experiences divergence due to the accumulation of integral gain in the adjustment loop, increasing the residual misalignment to 0.45mm, which exceeds the 0.30mm process safety threshold. Simultaneously, the partially missing control group, with the thermal strain time-space delay compensation operator removed, is subjected to non-uniform thermal expansion interference signals from the pipe. The control unit calculates the spatial six-degree-of-freedom deviation state vector. Measurement drift occurs, and the multi-axis hydraulic bracket outputs an offset correction command. The residual misalignment at the mating surface is 0.32mm. Meanwhile, the sample of this invention, employing a complete control strategy, has a time-varying damping correction coefficient. When the value is in the range of 1.25 to 1.55, the collaborative control module reduces the feedback adjustment step size of the control loop by 20% according to the graded cutoff rule to offset the coupling interference between the self-weight deflection and the mechanical clearance, and the residual misalignment of the mating surface converges to 0.11mm.

[0036] The control unit operates under extreme conditions with high disturbance intensity gradients. The initial alignment geometric deviation of the pipe is set to 3.00 mm, and the welding current is adjusted to 360 A. The first out-of-range control group sets the critical threshold for safety damping to 1.10. Because the threshold is set below the lower limit of 1.25, the feedback adjustment step size reduction action is triggered prematurely. The adjustment speed of the multi-axis hydraulic bracket is lower than the deformation evolution speed of the cantilever deflection, and the residual misalignment at the butt joint remains at 0.38 mm. The second out-of-range control group sets the critical threshold for safety damping to 1.70. Because the threshold is set above the upper limit of 1.55, the control unit adjusts the time-varying damping correction coefficient. While maintaining continuous operation of the deviation prediction operator, the differential term amplifies the noise signal in the spatial ranging channel. The power frequency conversion unit of the transmission system outputs a commutation drive signal, and the control command flutter frequency of the transmission system increases to 38.5Hz. Wear occurs in the mechanical transmission components, and the residual misalignment at the mating surface increases to 0.52mm. Meanwhile, the time-varying damping correction coefficient of this invention... When the threshold value of 1.55 is exceeded, the collaborative control module shuts down the continuous predictive control program and switches to a discrete self-optimizing control mode with a fixed step size of 0.05mm. The action response curve of the multi-axis actuator enters the stable region after crossing the safety damping critical threshold. The control loop blocks the loop divergence caused by thermal stress impact, and the residual misalignment at the mating surface remains at 0.19mm. The performance deviation data of the first and second out-of-range control groups show that the safety damping critical threshold range of 1.25 to 1.55 provides the parameter boundary for the convergence of the multi-axis servo closed-loop adaptive control loop. The multi-dimensional comparison data obtained in this experiment show that by reconstructing the high-dimensional observation vector within a 5ms sampling period... And solve for the spatial six-degree-of-freedom deviation state vector. Combined with time-varying damping correction coefficients based on the first-order time-domain derivative and the second-order time-domain cumulative characteristics. Adjusting the feedback adjustment step size and switching the discrete self-optimization control mode keeps the residual misalignment of the mating surface within the process limit of 0.30mm. The control command reversal chatter of the multi-axis linkage mechanism converges to 0Hz under the nonlinear thermal distortion interference environment.

[0037] Example 4: This embodiment combines the embodiments. Figures 1 to 2 This section describes a method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception, such as... Figure 1As shown, in step S1, ranging data is acquired and combined into a high-dimensional observation vector in a 5ms cycle. Then, in step S2, the six-degree-of-freedom deviation state vector is calculated and high-frequency noise is filtered out. Next, in step S3, the damping correction coefficient is calculated to adjust the feedback step size or switch the discrete mode. Finally, in step S4, the frequency command is calculated, superimposed on the feedforward, and limited before being output to the frequency converter. Figure 2 As shown, the system is in the system initialization zero-point calibration state. In this state, the pre-zero-point calibration program runs, locks the physical zero point and corrects the matrix. Following the conditional path of locking the physical zero point and loading the corrected spatial state transition matrix, it enters the continuous feedback step size adjustment state. In this state, the continuous deviation prediction operator runs, extracting the time-domain derivative and cumulative features. When the trigger condition of a 20% monotonically decreasing feedback adjustment step size when the time-varying damping correction coefficient is in the range of 1.25 to 1.55 is met, it also points to and maintains this continuous feedback step size adjustment state. From this continuous feedback step size adjustment state, two independent conditional decision paths branch out. When the condition of a time-varying damping correction coefficient greater than 1.55 is met... When the system enters the discrete self-optimization control state along this path, the continuous deviation prediction operator is suspended and a 0.05mm step-by-step process is executed. When the condition that the values ​​of all degrees of freedom are less than the first accuracy red line and the duration reaches 2s is met, the system directly enters the automated alignment and pairing completion state along another path. In this state, the accuracy red line is met for 2s and the frequency control command is stopped. In addition, when the system is in the aforementioned discrete self-optimization control state, if the bottom jump condition that the values ​​of all degrees of freedom are less than the first accuracy red line and the duration reaches 2s is met, the system will traverse laterally along the path of this condition and directly switch to the aforementioned automated alignment and pairing completion state.

[0038] Example 5: In this embodiment, when the system faces multi-axis linkage hydraulic correction conditions with intertwined interference from variable loads and asymmetric welding thermal fields, a 12m long and 800mm diameter aluminum alloy outer shell tube undergoes dynamic changes during continuous propulsion at a speed of 50mm / s accompanied by 2rpm rotation. This is due to the self-weight deflection and bending deformation of the large-span metal tube as the cantilever span extends, and the non-uniform thermal expansion of the metal material caused by asymmetric transient heat input induced by continuous welding at the joint. Consequently, mechanical deformation signals accompanying welding temperature rise fluctuations are mixed into the spatial ranging channel, including... The nonlinear noise signal coupled with the elastic deformation of the material and the downward pressure due to gravity is input to the feedback loop of the control unit. If a traditional negative feedback position correction control multi-axis hydraulic bracket is selected at this time, the control system lacks an internal decoupling mechanism for the intertwined and coupled signals of geometric misalignment, gravitational deformation fluctuation, and actuator response hysteresis. The control command lags behind the actual deformation speed of the pipe. Under the state of continuous accumulation of integral gain, the adjustment loop saturates and diverges, the multi-axis hydraulic bracket reverses frequently, and mechanical transmission losses of control overshoot and command chatter are caused on the machining axis.

[0039] To address mechanical transmission losses, the control unit uses a field industrial Ethernet bus to periodically acquire one-dimensional distance time-series data from a laser ranging array distributed around the pipe mating surface at a fixed sampling period of 5ms. The control unit's internal dynamic memory allocates a data buffer register with a length of 5 and a circular queue structure for each ranging channel in the laser ranging array. Within each 5ms sampling period, the control unit reads the single-channel sampled value output by the corresponding laser sensor and writes it to the tail address space in the data buffer register. Simultaneously, it removes historical data from the head position to complete the sliding shift update of the circular queue. The five discrete sampled values ​​stored in the data buffer register are sent to a temporary stack in ascending order. The processor reads the third value located at the middle physical storage address as the filtered output value for the current sampling period, eliminating pulse noise caused by arc flicker. The processor combines the filtered distance physical quantities from each channel into a high-dimensional observation vector. Then, multiply by a preset spatial state transition matrix, which is constructed based on the static topological geometric mapping relationship between the multi-axis bracket servo drive axis and the pipe center axis, to calculate the spatial six-degree-of-freedom deviation state vector characterizing the relative spatial position of the two pipe sections' mating surfaces. Based on this, the parallel-running thermal strain spatiotemporal delay compensation operator reads the real-time current sequence from the current sensor of the welding power source, and calculates the local thermal deformation caused by the current heating of the docking interface according to the thermal conduction conservation model. This calculation is then applied to the spatial six-degree-of-freedom deviation state vector. The local thermal deformation is subtracted in situ from the axial component to isolate the pose drift caused by non-uniform heating. When the thermal strain time-space delay compensation operator is running, the control unit collects the real-time welding current signal through the current sensor connected to the welding power supply output terminal. The transient thermal input timing sequence is constructed by multiplying the square of the welding current signal with the preset butt thermal resistance and input to the first-order low-pass thermal conduction inertial link to calculate the local thermal deformation at the pipe joint. Among them, the thermal distortion time delay constant of the first-order low-pass thermal conduction inertial link is set to 12s based on the thermal conductivity of 800mm pipe wall under the processing temperature rise state, and the thermomechanical strain coupling coefficient is set to 0.02μm / J. The processor will deviate the local thermal deformation in the six-degree-of-freedom spatial state vector. The corresponding axial component is subtracted in situ, and the pseudo pose deviation component caused by non-uniform heating is isolated at the front end of the feedback control, so that the control command output by the collaborative control module only corrects the mechanical pose deviation caused by geometric misalignment.

[0040] To establish the critical safety damping thresholds of 1.25 and 1.55 in the conditional control flow graded cutoff rule, the control unit runs a standardized engineering calibration program during the system initialization phase. The control unit drives a multi-axis hydraulic bracket to apply a step displacement excitation signal of 1.0 mm to the pipe for 50 ms, and uses a laser ranging array at a sampling frequency of 200 Hz to track the dynamic pose response curve of the pipe mating surface. The processor then calibrates the system based on the inherent stiffness values. An evolution model for the damping criterion is established, and transmission time delay parameters are used. The time-varying damping correction coefficients involved in the damping criterion evolution model are calibrated to define its critical threshold boundary. Satisfy mathematical expression in, This is the time-varying damping correction coefficient within the control loop. The first weighted feature constant is preset. , is the first-order differential vector of the pose state, used to characterize the first-order time-domain derivative of the cantilever deflection due to the self-weight of the pipe; The second weighted feature constant is preset. is a finite time-domain integral vector of the pose state, used to characterize the second-order time-domain cumulative feature of the pose state; For the system's inherent stiffness calibration value of the control loop, the damping criterion evolution model satisfies the following constraint rule: when the system experiences self-weight cantilever deflection bending and the time-varying damping correction coefficient... When the system is within the lower critical range, the control unit suppresses energy accumulation by reducing the adjustment step size. Conversely, when the system's dynamic impedance fluctuates and the time-varying damping correction coefficient... When crossing the upper critical boundary, loop stability is restored by truncating the continuous predictive control flow, and the measured transmission time delay parameters are used. The threshold condition used for calculating the damping criterion evolution model yields a corresponding transmission time delay of 15ms and the system's inherent stiffness calibration value. The lower critical value when the value is 1.0 is 1.25, which corresponds to the initial inflection point of damped divergence in the system. The calculated upper instability value is 1.55, which corresponds to the failure point where the differential term noise amplification exceeds the system damping limit. This provides a closed-loop digital criterion for the conditional control flow tiered truncation rule for the control unit's operation. In more detail, the determination of the above critical values ​​is based on the analytical calculation of the frequency domain characteristics of the transmission system. When the input transmission time delay parameter... The value is 15ms and the stiffness calibration value. When the value is 1.0, the control unit internally calculates the system's phase margin using the closed-loop frequency response function, and the time-varying damping correction coefficient is calculated based on the control gain combination. When the value reaches 1.25, the phase margin of the feedback loop first drops from the nominal 60° to the critical equilibrium point of 45°. At this point, the system begins to exhibit an underdamped oscillating tendency, and this is defined as the lower limit trigger point for monotonically reducing the adjustment step size to suppress gain accumulation; while when When the value is further increased and exceeds 1.55, the phase margin of the closed-loop system will be reduced to near 1.55. The system's dynamic impedance balance is disrupted, and the differential term amplifies random noise into severe mechanical chatter. Therefore, it is defined as the upper limit of the absolute instability boundary forcibly suspending the continuous predictive control flow. The control unit processes the spatial six-degree-of-freedom deviation state vector. At that time, the three translation axis components and three rotation axis components are extracted. The first-order time-domain derivative and finite-time-domain integral of each component within the current 5ms sampling period are multiplied by the preset gain coefficient of the corresponding degree of freedom. The results are then converted into the scalar form of the magnitude of the first-order derivative vector of the pose state and the magnitude of the finite-time-domain integral vector of the pose state through the square root operation. The time-varying damping correction coefficient, which serves as the dimensionless control gain correction factor, is then calculated. Among them, the first weighted characteristic constant Set to 5, the second weighted characteristic constant. Set to 2, the system's inherent stiffness calibration value. The value is set to 1. This value is determined by measuring the static topological geometric mapping relationship of the drive shaft under rated lifting conditions using an aluminum alloy outer shell tube with a length of 12m and a diameter of 800mm. This value is used to adjust the time-varying damping correction coefficient. This serves as a digital criterion for directly participating in the judgment feedback adjustment step size monotonically decreasing by 20% or switching to a discrete self-optimizing control mode with a fixed step size of 0.05mm.

[0041] Under the shift data flow of the data buffer register and the closed-loop digital criterion control, the coordinated control module will adjust the time-varying damping correction coefficient. Compared with a closed-loop digital criterion, the time-varying damping correction coefficient is triggered when the system is under processing deformation conditions. When the value monotonically increases to 1.32, the feedback adjustment step size is automatically reduced by 20% to slow down the outward divergence trend of the control gain. This is achieved when the nonlinear abrupt load causes a time-varying damping correction coefficient. When the value increases to 1.58, the conditional control flow hierarchical truncation rule suspends the deviation prediction operator and switches to a discrete self-optimization control mode with a fixed step size of 0.05mm. The multi-axis hydraulic bracket calculates and outputs the corresponding drive frequency control command based on the adjustment amount. The control unit limits the maximum output rate of change of a single command through a saturation limiting valve and periodically outputs the power frequency conversion unit of the servo bracket to drive the multi-axis hydraulic bracket for centering. The residual misalignment of the pipe mating surface converges from the initial 1.24mm to 0.11mm, and the residual misalignment remains below the pipe manufacturing specification red line of 0.30mm. The mechanical transmission chatter frequency of the multi-axis linkage servo system drops to 0Hz, eliminating the risk of control command overshoot caused by differential term amplification noise in the control loop. The frequency conversion drive signal maintains convergence under nonlinear thermal distortion interference. The multi-axis hydraulic bracket relies on the dynamic correlation adjustment of internal parameters to reduce the degradation of control quality caused by physical gaps and environmental interference. The continuous deviation prediction operator adopts a first-order Taylor time-domain extrapolation algorithm in the unsuspended state, based on the current spatial six-degree-of-freedom deviation state vector. The corresponding first-order time-domain derivative is used to predict the forward pose prediction value for the next 5ms sampling period; when the time-varying damping correction coefficient is used... When the value exceeds 1.55, the control unit cuts off the differential extrapolation term in the first-order Taylor time-domain extrapolation algorithm to suspend the operator. The first accuracy red line in the convergence determination step is set according to the manufacturing specifications for gas-insulated metal-enclosed transmission lines. The allowable misalignment of the three translation axis components is set to be less than or equal to 0.15 mm, and the angular residual of the three rotation axis components is set to be less than or equal to 0.002 rad. The deviation state vector in six degrees of freedom in space... When all degrees of freedom values ​​are within the first precision red line and remain within the limit for 400 consecutive control cycles, the control unit stops sending drive frequency control commands, completing the automated assembly and docking of the pipes.

[0042] Example 6: In this embodiment, when the system faces initial pose offset caused by installation deviations of ranging hardware from different production lines and zero offset of the valve port of the multi-axis hydraulic bracket, the processor runs a pre-zero calibration program to obtain the initial geometric dimensions. Under static conditions, the control unit sends a zero-adjustment pulse to the multi-axis hydraulic bracket and reads the relative displacement of the piston rod fed back by the displacement sensor. When the relative displacement deviation of the multi-axis hydraulic bracket is less than 0.01mm for 30s, the physical zero position is locked, and the standard calibration loop with a roundness deviation of less than 0.02mm is placed at the working axis position. The control unit reads the initial distance sequence output by the laser ranging array and performs least squares fitting calculation with the standard model to solve the spatial coordinates and angular deviation gain of each laser sensor relative to the transmission axis of the multi-axis hydraulic bracket, and corrects the initial reference term in the spatial state transition matrix. When the system faces the debugging condition of loading the corrected spatial state transition matrix into the controller's computing kernel, the control unit runs the control program during continuous processing with a pipe feeding speed of 50mm / s and a rotation of 2rpm. The control system solves the spatial six-degree-of-freedom deviation state vector within a 5ms cycle. The pose change of the long pipe under its own weight deflection is characterized by a time-varying damping correction coefficient. Adjusting the feedback adjustment step size, the multi-axis hydraulic bracket output position correction adjustment amount, the residual misalignment of the pipe mating surface converges from the initial 1.24mm to 0.09mm, and the control command reversal chatter frequency of the multi-axis linkage mechanism converges to 0Hz.

[0043] Example 7: In this embodiment, when the system faces the manufacturing condition of multiple specifications of gas-insulated metal-enclosed transmission line pipes being assembled and welded along the axis, where the bending stiffness changes abruptly due to different pipe wall thicknesses and the propulsion stroke extends, the pipe's self-weight increases with time-varying torque as the cantilever span extends outward from 2400 kg in the initial propulsion state to 12 m. The processor inside the control unit, based on the pipe's outer diameter of 800 mm and nominal wall thickness data entered in the current processing parameters, reads the initial stiffness reference vector matching the pipe's outer diameter from memory. When the high-dimensional observation vector... The spatial six-degree-of-freedom bias state vector is calculated by solving the spatial state transition matrix. Furthermore, when the pipe advance stroke crosses a stiffness change node by 6m, the adaptive matrix correction operator uses a differential variable scalar algorithm to extract the pressure feedback increment of the multi-axis hydraulic bracket within the previous sampling period. The pressure feedback increment is then compared with the nominal inertial moment to calculate and correct the system's inherent stiffness calibration value. The correction method satisfies the system's inherent stiffness calibration value. Measurement value of current axial advance stroke The linear transformation function relationship that increases and decreases satisfies the following expression: ,in, This is the corrected system natural stiffness calibration value. To statically calibrate the initial stiffness value, The preset stiffness reduction factor, This is the current measured value of the axial propulsion stroke.

[0044] The adaptive matrix correction operator calibrates the system's inherent stiffness. In the update state, the coordinated control module will update the system's inherent stiffness calibration value. Real-time input of time-varying damping correction coefficient The solution loop is based on the current axial propulsion travel measurement value. The increase in self-weight cantilever bending deflection due to increased time-varying damping correction factor The scaling compensation is obtained, and the control unit uses the latest time-varying damping correction coefficient calculated in the middle. Adjusting the drive frequency control command of the multi-axis hydraulic bracket power frequency conversion unit When the pipe is advanced to its maximum stroke of 12m and the end sags to the point of deflection, the system's inherent stiffness calibration value is... The value is reduced from the initial 1.0 to 0.76, thereby increasing the calculated time-varying damping correction factor. Accurately crossing the critical threshold of 1.55 for safety damping and activating the conditional control flow hierarchical cutoff rule, the control loop suspends the deviation prediction operator and switches to a discrete self-optimizing control mode with a fixed step size of 0.05mm. The residual misalignment of the mating surface smoothly converges from 1.24mm to 0.08mm under variable torque load disturbance, and the commutation chatter frequency of the multi-axis hydraulic bracket drops to 0Hz, thereby eliminating the control dead zone time delay overshoot under long-span cantilever conditions.

Claims

1. A method for dynamic centering coordination and regulation of GIL pipe manufacturing based on spatial pose perception, characterized in that, Includes the following steps: Step S1: Acquire ranging data and combine high-dimensional observation vectors in 5ms cycle: Acquire one-dimensional distance time series data output by the laser ranging array distributed in the circumference of the pipe in a 5ms sampling cycle and combine them into a high-dimensional observation vector; Step S2, solve the six-degree-of-freedom deviation state vector and filter out high-frequency noise: call the spatial state transition matrix to solve the high-dimensional observation vector into a six-degree-of-freedom deviation state vector of the relative spatial position of the pipe mating surface, and use the sliding window median filter operator to filter out the high-frequency noise components in the six-degree-of-freedom deviation state vector. Step S3: Calculate the damping correction coefficient, adjust the feedback step size, or switch to discrete mode: Extract the first-order time-domain derivative and second-order time-domain cumulative characteristics of the filtered six-degree-of-freedom deviation state vector within the sliding time window. Calculate the time-varying damping correction coefficient within the control loop based on the first-order time-domain derivative and second-order time-domain cumulative characteristics. When the time-varying damping correction coefficient monotonically increases and is within the range of 1.25 to 1.55, monotonically reduce the feedback adjustment step size by 20%. When the time-varying damping correction coefficient is greater than 1.55, suspend the continuous deviation prediction operator and switch to a discrete self-optimization control mode with a fixed step size of 0.05mm to calculate the adjustment amount for each axis. Step S4: Calculate the frequency command, superimpose feedforward, limit and output to the frequency converter: Calculate the drive frequency control command based on the adjustment amount of each axis. When the calculated time domain transmission hysteresis parameter is greater than 15ms, superimpose the pulse feedforward control amount into the drive frequency control command, use the saturation cut-off valve to limit the highest output change rate of a single command, and output the limited drive frequency control command to the frequency converter.

2. The method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception according to claim 1, characterized in that, Steps S1 to S4 form a closed-loop logic flow with parameter interlocking within the same programmable automation controller; in step S2, the spatial state transition matrix and sliding window median filter operator are used to correct the six-degree-of-freedom deviation state vector; in step S3, the step size is adjusted by shrinking feedback based on the first-order time-domain derivative and the second-order time-domain cumulative characteristic; in step S4, the pulse-type feedforward control quantity is seamlessly superimposed to offset the commutation dead zone, so that the residual misalignment of the docking surface is stabilized below 0.30mm.

3. The method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception according to claim 1, characterized in that, Step S1 includes the following sub-steps: Step S11, using a laser ranging sensor group to obtain the spatial coordinates of multiple points on the pipe end face; Step S12, inputting the spatial coordinates of multiple points into the laser ranging array to convert them into one-dimensional distance time series data and construct a high-dimensional observation vector.

4. The method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception according to claim 1, characterized in that, Step S2 also includes welding thermal strain offsetting processing implemented by the parallel-running thermal strain spatiotemporal delay compensation operator, including the following sub-steps: Step S21, calculate the local thermal deformation at the pipe joint based on the real-time acquired welding current signal and the cumulative time sequence of heat input; Step S22, subtract the local thermal deformation in situ from the six-degree-of-freedom deviation state vector to remove the pseudo-pose deviation component caused by thermal expansion, and input the corrected six-degree-of-freedom deviation state vector to step S3.

5. The method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception according to claim 1, characterized in that, Step S4 involves calculating the time-domain propagation hysteresis parameters, which includes the following sub-steps: Step S41, recording the system timestamp of the drive frequency control command sent from the time-domain response state observation unit; Step S42: Obtain the response timestamp of the pose sensor detecting the physical displacement of the pipe axis; Step S43: Calculate the difference between the response timestamp and the system timestamp to obtain the time-domain propagation hysteresis parameter.

6. The method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception according to claim 1, characterized in that, The seamless superposition of the pulse-type feedforward control quantity in step S4 includes the following sub-steps: Step S44, when the calculated time-domain transmission hysteresis parameter is greater than 15ms, read the coupling static clearance parameter to determine the current transmission dead zone size; Step S45, generate a rectangular pulse signal with an amplitude positively correlated with the current transmission dead zone size and a duration equal to 10ms as the pulse-type feedforward control quantity; Step S46, seamlessly superimpose the rectangular pulse signal into the drive frequency control command to offset the commutation hysteresis caused by the transmission clearance.

7. The method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception according to claim 1, characterized in that, The output to the frequency converter in step S4 includes the following sub-steps: Step S47, converting the drive frequency control command containing the pulse feedforward control quantity into a pulse width modulation signal; Step S48, periodically sending the pulse width modulation signal to the frequency converter through the industrial Ethernet bus to adjust the motor speed of the multi-axis servo synchronous drive system.

8. The method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception according to claim 1, characterized in that, Before the limited drive frequency control command is output to the inverter, step S4 also includes a saturation cutoff limiting sub-step: step S49, the maximum output change rate of a single command is limited by the saturation cutoff valve, the amount of change exceeding the upper limit of the amplitude change rate is cut off and the change amplitude of the drive frequency control command is limited between adjacent cycles.

9. The method for dynamic centering and collaborative control of GIL pipe manufacturing based on spatial pose perception according to claim 1, characterized in that, After the drive frequency control command is output to the frequency converter, the centering convergence determination step is also included: Step S5, continuously acquire the six-degree-of-freedom deviation state vector. When the value of each degree of freedom of the six-degree-of-freedom deviation state vector is less than the first precision red line and the duration reaches 2s, stop sending the frequency control command to complete the automatic centering and assembly of the pipe.

Images