A method for detecting dynamic interference of a spatial multi-bend catheter considering a springback effect and a related device

Abstract

The present application belongs to the technical field of pipe numerical control bending forming and intelligent manufacturing, and discloses a dynamic interference detection method for a space multi-bend guide pipe considering springback effect and a related device, which comprises the following steps: obtaining pipe basic parameters, bending process parameters and machine tool scene parameters of a space multi-bend guide pipe to be processed, and establishing a nominal processing path of the space multi-bend guide pipe and a parameterized machine tool scene model; for each bending unit, a springback prediction result of the bending unit is obtained through a pre-trained machine learning springback prediction model, and a dynamic pipe model considering the springback cumulative effect is constructed; the space multi-bend guide pipe is subjected to full-process dynamic scanning interference detection in the feeding, pipe rotating and bending stages, the interference risk of the current processing path is judged according to the effective safety clearance, and the final recommended processing path and the dynamic interference detection result are output. The present application significantly improves the engineering reliability of the processing path planning of the space multi-bend guide pipe.

Description

Technical Field

[0001] This invention belongs to the field of CNC bending forming and intelligent manufacturing technology of pipes, and relates to a dynamic interference detection method and related device for spatial multi-bend conduits that takes into account the springback effect. Background Technology

[0002] Multi-bend pipe fittings are widely used in aerospace, automotive, shipbuilding, construction machinery, energy equipment, hydraulic pipelines, and fluid transport systems. With the increasing demands for lightweight equipment structures, compact spatial layouts, and integrated piping systems, pipe fittings typically require multiple bends, spatial turns, and bending planes within a limited space. For complex multi-bend pipe fittings, the processing not only needs to ensure the final centerline shape meets design requirements but also ensures that each bending process does not interfere with the bending die, clamping blocks, pressure die, guide die, machine bed, ground, existing pipe sections, or non-adjacent pipe sections within the fitting itself during feeding, pipe rotation, and bending. Therefore, dynamic interference detection and safe path planning before processing are crucial aspects of CNC bending forming of multi-bend pipe fittings.

[0003] Currently, multi-bend pipe fittings are typically formed using CNC bending, push bending, or free bending methods. During CNC bending, the pipe fitting generally completes multiple bending units sequentially according to preset feeding length, rotation angle, bending angle, and bending radius. After each bending unit is completed, the previously formed bending segment continuously changes its spatial orientation as the pipe fitting undergoes subsequent feeding, rotation, and bending actions, participating in subsequent processing as part of the overall pipe fitting's spatial footprint. Due to the significant spatial extension length and complex orientation of the formed portion of multi-bend pipe fittings, they are prone to collisions with machine tool components, the machine bed, the ground, or other pipe segments during subsequent movements. Especially when there are many bending segments, large spatial rotation angles, small bending radii, or long total pipe lengths, relying solely on manual experience or final forming state verification often fails to accurately assess the dynamic interference risks during processing.

[0004] Existing methods for interfering with multi-bend pipe fittings typically establish a spatial model of the fitting based on a theoretical centerline or nominal machining path. This model is then dynamically scanned according to the feeding, rotation, and bending processes, and the minimum safe clearance between the fitting and the machine tool environment is calculated. Such methods can, to some extent, determine whether there is a collision risk along the nominal machining path. However, in actual bending processes, springback is inevitable after unloading. Springback causes the actual bending angle, bending radius, end position, and end posture to deviate from the nominal machining path, resulting in a difference between the spatial position of the actual formed pipe segment and the theoretical model. Therefore, if a nominal geometric model that does not consider springback is still used as the object of dynamic interference detection, the detection results cannot accurately reflect the spatial occupancy status of the fitting during the actual machining process.

[0005] For single-bend pipe fittings, springback mainly manifests as errors in the final bending angle, changes in the bending radius, or deviations in the end position. For multi-bend pipe fittings, however, springback exhibits significant transmissibility and cumulative effects. Springback in a preceding bending unit not only alters the post-spring shape of that unit but also changes the end position, local coordinate system, and overall spatial orientation of subsequent bending units at the start of processing. In other words, the springback of a previous bend affects the pipe's trajectory during feeding, rotation, and bending processes in subsequent bends. If dynamic interference detection still uses the nominal path, it may underestimate or overestimate the actual distances between the pipe fitting and the machine bed, ground, fixtures, or non-adjacent pipe sections during subsequent processing stages.

[0006] In existing interference detection methods, if the nominal centerline, which does not consider springback, is still used as the model for the formed pipe section, two types of problems may occur. First, the nominal path detection result may indicate safety, but in actual processing, springback from the preceding bending section may cause the pipe end to rise, sink, or deflect, causing it to approach the machine bed, ground, or machine tool components during subsequent pipe turning or bending, thus resulting in actual interference. Second, the nominal path detection result may indicate interference, but the actual spatial posture after springback may cause the pipe to move away from obstacles, leading to an overestimation of the interference risk. Both of these situations will result in inaccurate risk assessment before processing, increasing the number of trial bends, machine setup time, and scrap rate.

[0007] On the other hand, to improve the forming accuracy of pipe fittings, existing technologies have developed springback prediction methods based on empirical formulas, finite element simulations, analytical models, or machine learning models. Among these, machine learning methods can establish a nonlinear mapping relationship between inputs and springback amounts using historical processing data, material parameters, geometric parameters, and process parameters, enabling rapid prediction of springback under complex materials, complex cross-sections, or multi-parameter coupled conditions. Common models include neural networks, support vector regression, random forests, gradient boosting models, Gaussian process regression, and physically constrained machine learning models. These methods are typically used to predict bending angle springback, end position deviation, curvature changes, or final forming errors, and accordingly perform bending angle compensation or process parameter correction.

[0008] However, most existing machine learning springback prediction methods aim at compensating for the final forming accuracy, typically only outputting the springback angle, endpoint deviation, or compensation parameters of a single bending unit. They do not further embed the predicted post-springback centerline, post-springback end attitude, and post-springback spatial envelope model into the dynamic interference detection process for multi-bend pipe fittings. In other words, existing springback prediction and dynamic interference detection methods are usually separate: springback prediction is used for forming accuracy compensation, while interference detection is still mainly based on the nominal path or the path before compensation. This separate approach fails to reflect the coupling relationship between "machine learning springback prediction—post-springback geometry update—subsequent dynamic interference detection," and also makes it difficult to identify potential dynamic interference risks caused by springback.

[0009] Furthermore, springback compensation itself may introduce new interference risks. To counteract unloading springback, actual machining typically requires compensation corrections to bending angles, pipe turning angles, bending radii, or feeding parameters. For example, to obtain the target bending angle, the system may need to increase the actual bending angle; to correct the end-point spatial attitude, the system may need to adjust subsequent pipe turning angles or local bending sequences. Although the compensated machining path may improve the final geometric accuracy, its transient spatial attitude during machining differs from the original nominal path, potentially causing the fitting to approach the machine bed, fixture, ground, or other bends more closely during certain feeding, pipe turning, or bending stages. Therefore, generating compensation parameters solely based on springback prediction results without re-dynamically interferometry detection of the compensation path may still render the compensation path infeasible in actual machining.

[0010] Furthermore, the springback prediction results from machine learning inherently possess a certain degree of uncertainty. This uncertainty may stem from factors such as material batch fluctuations, changes in friction conditions, differences in machine tool stiffness, variations in clamping states, insufficient training sample coverage, and model generalization errors. If springback prediction errors or model confidence are not considered in dynamic interferometric detection, risk misjudgments may occur under critical operating conditions where the minimum safety clearance is small. Especially for complex spatial multi-bend pipe fittings, when the clearance between a formed pipe section and an obstacle approaches the safety threshold, even a small springback prediction error may lead to a collision during actual processing. Therefore, it is necessary to incorporate the uncertainty or error compensation amount of machine learning predictions into the effective safety clearance criterion to improve the engineering reliability of interferometric detection.

[0011] In summary, existing dynamic interference detection methods for multi-bend pipe fittings mainly rely on scanning based on nominal processing paths, making it difficult to accurately reflect the actual spatial attitude changes caused by unloading springback. Existing machine learning springback prediction methods primarily serve final forming accuracy compensation and have not been effectively used to drive post-springback geometry updates and risk assessment during the interference detection process. Furthermore, existing springback compensation methods lack secondary verification of the processing safety of the compensation path. Therefore, there is an urgent need to propose a machine learning springback prediction-driven dynamic interference detection method for multi-bend pipe fittings. This method should be able to predict the springback deformation of each bending unit based on a machine learning model, dynamically update the post-springback centerline, end pose, and spatial envelope model of the formed bend, and perform dynamic interference detection, safety clearance judgment, and path correction based on the predicted post-springback geometry during subsequent feeding, pipe transfer, and bending processes. This would improve the accuracy of risk identification and safe path generation before processing multi-bend pipe fittings. Summary of the Invention

[0012] The purpose of this invention is to provide a dynamic interference detection method and related device for multi-bend conduits in space that considers the springback effect, thereby solving the problem that existing dynamic interference detection methods for multi-bend pipes are mainly based on nominal processing paths and have difficulty in considering the impact of unloading springback on the actual spatial attitude.

[0013] To achieve the above objectives, the present invention employs the following technical solution: A dynamic interferometric detection method for spatially curved ducts considering springback effect includes: Obtain the basic parameters, bending process parameters, and machine tool scene parameters of the spatial multi-bend conduit to be processed. Establish the nominal processing path of the spatial multi-bend conduit based on the bending process parameters, and establish a parametric machine tool scene model under a fixed machine head reference system based on the machine tool scene parameters. For each bending unit, the nominal bending process is completed using the nominal processing path. Based on the basic parameters of the bending unit, the springback prediction result of the bending unit is obtained through a pre-trained machine learning springback prediction model. The nominal geometric state of the formed bending segment is corrected to obtain the springback prediction geometric model after the bending unit is unloaded. The dynamic pipe fitting model considering the cumulative effect of springback is constructed. Based on dynamic pipe fitting models and parametric machine tool scene models, dynamic scanning interference detection is performed on the entire process of spatial multi-bend conduit during feeding, pipe turning and bending stages to calculate the effective safety clearance between the spatial multi-bend conduit and obstacles. The interference risk of the current processing path is judged based on the effective safety clearance. If the current processing path does not meet the safety requirements, the current processing path is compensated and corrected, and the dynamic interference detection is re-executed until the corrected or compensated processing path meets both the safety requirements and the post-formation accuracy requirements. Then, the final recommended processing path and dynamic interference detection results are output.

[0014] Furthermore, the basic parameters of the pipe fittings include geometric parameters and material parameters; Geometric parameters include the pipe's outer diameter, wall thickness, total length, number of bends, length of each straight segment, and target centerline information; Material parameters include elastic modulus, yield strength, hardening parameters, material batch information, and forming history information; The bending process parameters include the feeding length, tube angle, bending angle, and bending radius of each bending unit; Machine tool scene parameters include the dimensions, positions, and installation deviations of bending dies, clamping blocks, pressing dies, guide dies, bed, ground, and other machine tool components; The parametric machine tool scene model includes bending mold model, clamping block model, pressing mold model, guide mold model, bed model, ground model, and other obstacle models.

[0015] Furthermore, the machine learning rebound prediction model is one of the following, or a combination of the above models: neural network, support vector regression, random forest, gradient boosting model, Gaussian process regression, and physical constraint machine learning model; The machine learning springback prediction model takes the basic parameters of the pipe fitting, bending process parameters, machine tool scene parameters, and the forming state of the preceding bending unit as inputs, and takes the springback amount, end position deviation, end posture deviation, curvature change amount, or centerline deviation after the corresponding bending unit is unloaded as outputs. The machine learning springback prediction model is trained using historical processing data, finite element simulation data, online measurement data, or trial bending experimental data.

[0016] Furthermore, the post-missile prediction geometric model includes the post-missile centerline, the post-missile end coordinates, the post-missile local coordinate system, and the post-missile spatial envelope model.

[0017] Furthermore, the process of constructing the dynamic pipe fitting model includes: According to the processing sequence of the multi-bend pipe fitting, each bending unit is processed sequentially. For the first bending unit, the feeding, pipe turning and bending process is simulated according to its nominal processing path. After bending is completed, its springback deformation is predicted according to the machine learning model, and the springback prediction geometric model of the first bending unit is generated. For the second and subsequent bending units, the post-blowout predicted geometric model of all preceding bending units is used as the formed pipe segment, and the nominal path or compensation path of the current bending unit is used as the pipe segment to be processed, together forming the dynamic pipe fitting model at the current moment.

[0018] Furthermore, the entire dynamic scanning interferometry detection process includes: During the feeding stage, the position of the guide tube along the feeding direction is updated; during the tube turning stage, the rotational attitude of the guide tube around the machine head axis is updated; during the bending stage, the bending state of the current bending unit around the bending mold is updated; and in each scanning substep, the spatial attitude of the multi-bend tube is updated according to the current tube state, and the effective safety gap between the tube spatial envelope model after considering springback and the object being inspected is calculated. When calculating the effective safety clearance, a normal process contact exclusion zone is set for normal clamping, guiding or supporting contact between clamping blocks, pressing dies, bending dies and guide dies and pipe fittings. Contact within the normal process contact exclusion zone is not considered abnormal interference. A state that exceeds the normal process contact exclusion zone and the effective safety clearance is less than the preset safety clearance threshold is considered as an abnormal interference risk.

[0019] Furthermore, the method for compensating and correcting the current processing path includes... Based on the springback prediction results, the bending process parameters are compensated and corrected, and the feeding length, tube angle, bending angle, bending radius, bending sequence, fixture avoidance parameters, bed avoidance parameters, or combinations thereof are adjusted. If the springback prediction results indicate that the springback posture of a certain bending unit causes subsequent bending units to approach the bed, ground, or machine tool components, then the compensation parameters of the bending unit from which the springback originates, or the rotation angle, feeding length, or avoidance parameters of subsequent risky bending units, should be corrected first.

[0020] A dynamic interferometric detection system for spatially curved ducts considering springback effect, comprising: The parameter module is used to obtain the basic parameters of the pipe fitting, bending process parameters and machine tool scene parameters of the spatial multi-bend conduit to be processed, establish the nominal processing path of the spatial multi-bend conduit based on the bending process parameters, and establish a parameterized machine tool scene model under the fixed machine head reference system based on the machine tool scene parameters. The modeling module is used to complete the nominal bending process for each bending unit using the nominal processing path. Based on the pipe basic parameters of the bending unit, the springback prediction result of the bending unit is obtained through a pre-trained machine learning springback prediction model. The nominal geometric state of the formed bending segment is corrected to obtain the springback prediction geometric model after the bending unit is unloaded. The dynamic pipe model considering the cumulative effect of springback is used to construct the dynamic pipe model. The calculation module is used to perform dynamic scanning interference detection on the spatial multi-bend conduit during the feeding, pipe turning and bending stages based on the dynamic pipe fitting model and the parameterized machine tool scene model, and to calculate the effective safety clearance between the spatial multi-bend conduit and obstacles. The output module is used to determine the interference risk of the current processing path based on the effective safety clearance. If the current processing path does not meet the safety requirements, the current processing path is compensated and corrected, and dynamic interference detection is re-executed until the corrected or compensated processing path meets both the safety requirements and the post-formation accuracy requirements. Then, the final recommended processing path and dynamic interference detection results are output.

[0021] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method.

[0022] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method.

[0023] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a dynamic interference detection method for spatial multi-bend ducts considering springback effects. By acquiring the basic parameters of the duct, bending process parameters, and machine tool scene parameters, and establishing a nominal processing path and a parameterized machine tool scene model, a standardized data foundation and static spatial reference system are provided for all subsequent analyses. This ensures that the entire dynamic interference detection process has clear initial boundary conditions, transforming the complex physical processing environment into a computable digital model, laying a solid foundation for subsequent kinematic simulation of the duct. For each bending unit, the traditional method of directly using the nominal path as the basis for subsequent detection is abandoned. Instead, a pre-trained machine learning springback prediction model is creatively used to obtain the unloading springback amount of the bending unit immediately after the nominal bending is completed, and the nominal geometric state of the formed bending segment is corrected to obtain the post-springback prediction geometric model. Simultaneously, by constructing a dynamic pipe model considering the cumulative effect of springback, this step, for the first time, directly embeds machine learning prediction results into the bend-by-bend update of the pipe's spatial attitude. This transforms the geometric shape of the pipe involved in subsequent interferometry detection from theoretical design values ​​to actual predicted values ​​considering physical springback, effectively overcoming the deficiency of traditional methods in describing the cumulative effect of preceding springback on the spatial attitude of subsequent bends in multi-bend ducts. Based on the updated dynamic pipe model and a static parametric machine tool scenario model, dynamic scanning interferometry detection is performed on the duct throughout the three process stages of feeding, pipe rotation, and bending, and the effective safety clearance between the duct and various obstacles is calculated. The outstanding effect of this step is that it places the springback-corrected non-rigid, variable-attitude pipe within a space composed of rigid obstacles such as the machine tool, bed, and ground, simulating each instantaneous state in actual processing with time-sequential scanning substeps. This makes interferometry detection no longer a static verification of the final formed part, but a dynamic risk assessment of the entire processing lifecycle. A closed-loop feedback mechanism based on safety judgment results is introduced: when the current path does not meet safety requirements, it is not directly abandoned or blindly adjusted, but rather compensated and corrected before dynamic interference detection is re-executed until the corrected processing path simultaneously crosses both the dynamic safety clearance and post-spring forming accuracy thresholds, at which point the recommended path and detection results are finally output. This invention ensures that the final output processing path is safe not only in the static nominal state, but also in the actual processing transient state considering springback, and that the geometric accuracy of the formed conduit meets design requirements. This organically unifies forming accuracy control with processing safety, avoiding risk omissions caused by relying solely on nominal path detection, and improving the safety and reliability of processing path planning for multi-bend pipe fittings. It is applicable to the feeding, pipe turning, and bending processes of multi-bend pipe fittings, updating the post-spring geometry of the formed bend based on machine learning springback prediction results, and further conducting dynamic interference detection, safety clearance judgment, and processing path correction.

[0024] Furthermore, coupling machine learning springback prediction results with the dynamic interference detection process is used not only to predict final forming errors or compensate for bending angles, but also to update the space occupancy status of multi-bend pipe fittings during processing. This enables the identification of risks of interference after springback even when the nominal path is safe, as well as new interference risks that may be introduced during processing by the springback compensation path.

[0025] Furthermore, this invention introduces machine learning springback prediction results into the dynamic interference detection process of multi-bend pipe fittings, so that interference detection is no longer based solely on the nominal processing path, but on the springback prediction geometric model after considering springback, which can more accurately reflect the space occupancy status during the actual processing of pipe fittings.

[0026] Furthermore, this invention can describe the cumulative effect of springback from preceding bending units on the spatial attitude of subsequent bending units in multi-bend pipe fittings. For multi-bend pipe fittings, springback from preceding bends not only affects the final shape but also alters the fitting's attitude during subsequent feeding, pipe rotation, and bending processes. This invention improves the reliability of dynamic interference detection for complex multi-bend pipe fittings by updating the springback prediction geometric model for each bend.

[0027] Furthermore, the present invention can identify the risk of springback-induced interference. When the nominal path detection result shows safety, but the predicted path after springback has insufficient safety clearance or even physical interference with the bed, ground, machine tool components or non-adjacent pipe sections, it can be determined as springback-induced interference, thereby avoiding the missed judgment caused by traditional nominal path detection. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a flowchart of the dynamic interference detection process between the nominal path and the rebound path in Embodiment 1 of the present invention.

[0030] Figure 2 This is a graph showing the minimum safety gap variation when springback prediction is not enabled in Embodiment 1 of the present invention.

[0031] Figure 3 This is a graph showing the change in minimum safety gap after enabling machine learning rebound prediction in Embodiment 1 of the present invention.

[0032] Figure 4 This is a graph showing the change in minimum safety gap after enabling machine learning rebound prediction correction in Embodiment 1 of the present invention.

[0033] Figure 5 This is a flowchart of the dynamic interference verification process for the machine learning rebound compensation path in Embodiment 2 of the present invention.

[0034] Figure 6 This is a graph showing the change in the minimum safety gap of the nominal path in Embodiment 2 of the present invention.

[0035] Figure 7 This is a graph showing the change in minimum safety gap after springback compensation in Embodiment 2 of the present invention.

[0036] Figure 8 This is a schematic diagram of a dynamic interference detection system for a spatially curved duct considering the springback effect, representing a preferred embodiment of the present invention.

[0037] Figure 9 This is a schematic diagram of the electronic device structure according to a preferred embodiment of the present invention. Detailed Implementation

[0038] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0039] Obviously, the described embodiments are only some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0040] It should be noted that the terminals involved in the embodiments of this application may include, but are not limited to, mobile phones, personal digital assistants (PDAs), wireless handheld devices, tablet computers, personal computers (PCs), MP3 players, MP4 players, wearable devices (e.g., smart glasses, smartwatches, smart bracelets), smart home devices, and other smart devices.

[0041] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0042] The present invention will now be described in further detail with reference to the accompanying drawings: This invention provides a dynamic interferometric detection method for spatial multi-bend ducts considering springback effect, specifically including the following steps: S1. Obtain the basic parameters, bending process parameters, and machine tool scene parameters of the multi-bend conduit to be processed: The basic parameters of the conduit include geometric parameters and material parameters. The geometric parameters include the outer diameter, wall thickness, total length, number of bends, length of each straight segment, and target centerline information of the conduit. The material parameters include the elastic modulus, yield strength, hardening parameters, material batch information, and forming history information. The bending process parameters include the feeding length, pipe turning angle, bending angle, and bending radius of each bending unit. The machine tool scene parameters include the dimensions, positions, and installation deviation information of the bending die, clamping block, pressing die, guide die, bed, ground, and other machine tool components.

[0043] S2. Establish the nominal machining path and parametric machine tool scene model for multi-bend pipe fittings: Based on the bending process parameters, generate the nominal machining path for the multi-bend pipe fittings. The nominal machining path includes the theoretical spatial state of each bending unit during the feeding stage, pipe turning stage, and bending stage. Establish a parametric machine tool scene model under a fixed machine head reference frame. The fixed machine head reference frame uses the feeding direction, the main bending direction, and the secondary bending direction as the basic directions. The parametric machine tool scene model includes a bending die model, a clamping block model, a pressing die model, a guide die model, a bed model, a ground model, and other obstacle models. The above machine tool scene model remains under the fixed machine head reference frame during dynamic scanning to simulate the relative motion relationship where the pipe fitting moves while the machine tool body does not rotate with the pipe fitting during actual machining.

[0044] S3. Establish a machine learning springback prediction model: Establish a machine learning springback prediction model for predicting the springback deformation of each bending unit in a multi-bend pipe fitting. The machine learning springback prediction model takes the geometric parameters, material parameters, process parameters, machine tool status parameters, and forming state of the preceding bending unit as inputs, and takes the springback amount, end position deviation, end posture deviation, curvature change, or centerline deviation after the corresponding bending unit is unloaded as outputs.

[0045] The machine learning springback prediction model can be a neural network, support vector regression, random forest, gradient boosting model, Gaussian process regression, physical constraint machine learning model, or a combination of the above models. The machine learning springback prediction model can be trained using historical processing data, finite element simulation data, online measurement data, or trial bending experimental data.

[0046] S4. Reconstruct the post-bending pipe geometry model based on the machine learning springback prediction results: For each bending unit, after its nominal bending process is completed, input the geometric parameters, material parameters, process parameters and current pipe state of the bending unit into the machine learning springback prediction model to obtain the springback prediction results of the bending unit.

[0047] Based on the springback prediction results, the nominal centerline, bending angle, bending radius, end position, and end attitude of the bending unit are corrected to obtain the post-springback prediction geometric model after unloading. The post-springback prediction geometric model includes the post-springback centerline, post-springback end coordinates, post-springback local coordinate system, and post-springback spatial envelope model. In the subsequent bending unit scanning process of multi-bend pipe fittings, the previously formed bends no longer use the nominal geometric model for interference detection. Instead, the post-springback prediction geometric model, corrected by machine learning springback prediction, is used to participate in the dynamic interference judgment of subsequent feeding, pipe transfer, and bending stages.

[0048] S5. Construct a dynamic pipe fitting model considering the springback accumulation effect: Process each bending unit sequentially according to the processing order of the multi-bend pipe fitting. For the first bending unit, simulate the feeding, pipe turning, and bending process according to its nominal processing path, and predict its springback deformation based on a machine learning model after bending to generate the springback prediction geometry model of the first bending unit. For the second and subsequent bending units, use the springback prediction geometry models of all preceding bending units as the formed pipe segments, and the nominal path or compensation path of the current bending unit as the pipe segment to be processed, together forming the dynamic pipe fitting model at the current moment. Thus, the springback effect of the preceding bending units can be gradually transmitted to the spatial attitude of the subsequent bending units, thereby reflecting the springback accumulation effect in the processing of multi-bend pipe fittings.

[0049] S6. Perform full-process dynamic interference scanning on multi-bend pipe fittings considering springback: For each bending unit, perform dynamic scanning according to the feeding stage, pipe turning stage, and bending stage. Each stage is divided into several scanning sub-steps. In each scanning sub-step, update the spatial attitude of the multi-bend pipe fitting according to the current pipe fitting status, and calculate the minimum safe clearance between the spatial envelope model of the pipe fitting considering springback and machine tool components, bed, ground, fixture, guide mold, formed pipe segment, and its own non-adjacent pipe segment. In the feeding stage, update the position of the pipe fitting along the feeding direction; in the pipe turning stage, update the rotational attitude of the pipe fitting around the machine head axis; in the bending stage, update the bending state of the current bending unit around the bending mold. The previously formed bending segment always participates in the space occupancy calculation of the current stage with the predicted state after springback. For normal clamping, guiding, or supporting contact between the clamping block, pressure mold, bending mold, and guide mold and the pipe fitting, set a normal process contact exclusion area. Contact within the normal process contact exclusion area is not treated as abnormal interference; a state outside the normal process contact exclusion area and with insufficient safe clearance is judged as an abnormal interference risk.

[0050] S7. Calculate the effective safety gap for springback sensing and determine the risk of interference: Determine whether the current processing path meets the safety requirements based on the minimum safety gap corresponding to the predicted geometry model after springback. When the minimum safety gap is greater than or equal to the preset safety gap threshold, the current path is determined to meet the safety requirements; when the minimum safety gap is less than the preset safety gap threshold but no physical overlap has occurred, the current path is determined to have insufficient safety margin; when the minimum safety gap is less than zero, the current path is determined to have physical interference.

[0051] When the nominal path meets the safety requirements, but the rebound path updated based on machine learning rebound prediction does not meet the safety requirements, the risk is determined to be either rebound-induced interference risk or rebound-induced insufficient safety margin risk. The risk information includes the rebound source bending unit, the risk occurrence bending unit, the risk occurrence stage, the risk object, the minimum safety gap of the nominal path, the minimum safety gap of the rebound path, and the rebound prediction amount.

[0052] S8. Correcting the machining path based on springback prediction and interference detection results: When a springback deviation is predicted in the bending unit according to the machine learning model, the bending process parameters are compensated and corrected based on the springback prediction results. The compensation and correction include adjusting the feed length, tube angle, bending angle, bending radius, bending sequence, fixture avoidance parameters, bed avoidance parameters, or combinations thereof.

[0053] If the springback prediction results indicate that the springback posture of a certain bending unit causes subsequent bending units to approach the bed, ground, or machine tool components, then the compensation parameters of the bending unit from which the springback originates, or the rotation angle, feeding length, or avoidance parameters of subsequent risky bending units, should be corrected first.

[0054] After compensation and correction are completed, a new compensated machining path is generated, and dynamic interference scanning of the entire feeding, tube transfer, and bending process is performed again on the compensated machining path. Only when the compensated machining path meets the safety clearance requirements throughout the entire machining process, and the predicted geometry after compensation meets the target forming accuracy requirements, is the compensated machining path output as the final recommended machining path.

[0055] S9. Output dynamic interference detection results and path correction results: Output dynamic interference detection results considering machine learning springback prediction. The output results include the springback prediction amount of each bending unit, the centerline after springback, the pose of the rear end of the springback, the minimum safety gap, the first interference object, the first interference stage, the maximum interference amount, the type of springback-induced interference, the safety judgment result of the compensation path, and the final recommended processing path.

[0056] The present invention will be further described in detail below through specific embodiments: Example 1: A dynamic detection method for nominal safe paths transforming into insufficient safety gap interference paths due to machine learning bounce prediction, such as... Figure 1 As shown.

[0057] This embodiment illustrates that during the processing of multi-bend pipe fittings, if dynamic interference detection is performed solely based on the nominal processing path, the processing path meets the preset safety clearance requirements. However, after enabling a machine learning springback prediction model and updating the post-springback geometry of the formed bend based on the prediction results, the same processing path exhibits a safety clearance-deficient interference risk during subsequent pipe rotation and bending processes. This embodiment demonstrates that the present invention can identify springback-induced dynamic interference risks that are difficult to detect using traditional nominal path detection methods, thus proving the necessity of introducing machine learning springback prediction results into the interference detection process.

[0058] This embodiment uses a three-bend spatial pipe as an example. The pipe is made of 6061 aluminum alloy, with an outer diameter of 16.0 mm and a wall thickness of 2.0 mm. A preset safety clearance threshold of 4 mm is set, and a parametric machine tool scene model is established under a fixed machine head reference system. The parametric machine tool scene model includes a bending die, clamping block, pressing die, guide die, machine bed, and ground. In this embodiment, the ground and machine bed are the primary risk detection objects, while the bending die, clamping block, pressing die, guide die, and non-adjacent pipe sections of the pipe itself are secondary detection objects. The preset safety clearance threshold is 4 mm.

[0059] In this embodiment, the nominal bending process parameters of the multi-bend pipe fitting are as follows: the feeding length of the first bending unit is 80.000 mm, the pipe turning angle is 0.000°, the bending angle is 90.000°, and the bending radius is 70.000 mm; the feeding length of the second bending unit is 130.000 mm, the pipe turning angle is 80.000°, the bending angle is 107.000°, and the bending radius is 60.000 mm; the feeding length of the third bending unit is 90.000 mm, the pipe turning angle is 60.000°, the bending angle is 72.000°, and the bending radius is 70.000 mm.

[0060] First, the nominal processing path of the three-bend pipe fitting is generated based on the nominal bending process parameters mentioned above. For each bending unit, a full-process dynamic scan of the feeding stage, pipe transfer stage, and bending stage is performed sequentially. In each scan substep, the spatial orientation of the pipe fitting is updated, and the effective safety clearance between the equivalent spatial envelope model of the pipe fitting and the bed, ground, bending die, clamping block, pressing die, guide die, and non-adjacent pipe segments of the pipe fitting itself is calculated.

[0061] Without enabling a machine learning springback prediction model, the three-bend pipe fitting is dynamically scanned according to the nominal centerline, nominal bending angle, and nominal bending radius. Figure 2As shown, the scanning results indicate that no interference risk occurred during the feeding, pipe transfer, and bending processes of the nominal processing path in the first, second, and third bending units. The minimum effective safety clearance for the entire path is 4.433 mm, corresponding to the machine bed; simultaneously, the minimum effective safety clearance between the pipe and the ground is approximately 4.9 mm. Since all of the above minimum effective safety clearances are greater than the preset safety clearance threshold of 4 mm, if only the nominal processing path is used for judgment, the processing path of this three-bend pipe fitting would be determined to meet the safety processing requirements.

[0062] Subsequently, a machine learning springback prediction model was activated to predict the unloading springback after each bending unit was completed. The machine learning springback prediction model takes material, pipe outer diameter, wall thickness, bending radius, bending angle, and historical springback data as input, and outputs the springback correction amount corresponding to each bending unit. For this embodiment, the springback correction results output by the machine learning springback prediction model are as follows: the predicted springback angle for the first bending unit is 2.861°, the predicted springback angle for the second bending unit is 3.068°, and the predicted springback angle for the third bending unit is 2.699°.

[0063] Based on the aforementioned machine learning springback prediction results, the post-springback geometry of the formed bends after each bending unit is updated. Specifically, the post-springback bending angle, post-springback centerline, end coordinates, and end local coordinate system of the formed bends are corrected according to the predicted springback angle, and the corrected post-springback prediction geometry model is used as the space occupancy model in subsequent feeding, pipe rotation, and bending processes. Thus, the springback effect of the preceding bending unit is transferred to the dynamic scanning process of the subsequent bending unit.

[0064] After enabling the machine learning springback prediction model, a full-process dynamic interference detection was re-performed on the three-bend pipe fitting. The detection process included 144 scanning substeps and 145 dynamic states, such as... Figure 3 As shown, the rescan results indicate that during the pipe rotation stage of the third bending unit, the spatial attitude of the formed pipe segment after springback correction changes, causing the local spatial envelope of the pipe to move closer to the ground during the rotation process. When the pipe rotation progress of the third bending unit reaches 60.000, the effective safety clearance between the pipe and the ground decreases to 2.345 mm. This value is less than the preset safety clearance threshold of 4 mm, therefore, this state is determined to have an interference risk due to insufficient safety clearance, and it is recorded as the first interference risk state.

[0065] During the continued scanning of the bending stage of the third bending unit, in multiple scanning substeps with bending progresses of 3.000, 6.000, 9.000, 12.000, 15.000, 18.000, 21.000, and 24.000, the effective safe clearance between the pipe and the ground was consistently maintained at approximately 2.345 mm, continuously below the preset safe clearance threshold of 4 mm. A total of 33 interference risk substeps were identified during this detection process. The first interference risk occurred during the pipe rotation stage of the third bending unit, with the ground as the risk object, and the minimum effective safe clearance being 2.345 mm.

[0066] Therefore, without enabling the machine learning springback prediction model, the minimum effective safety clearance for the entire nominal processing path of this three-bend pipe fitting is 4.433 mm, which is greater than the preset safety clearance threshold of 4 mm, and is thus determined to be a safe path. However, after enabling the machine learning springback prediction model and incorporating the post-springback geometry into subsequent dynamic scans, the same processing path exhibits a risk of insufficient ground-direction safety clearance during the pipe-turning stage of the third bending unit and subsequent bending stages. The minimum effective safety clearance decreases to 2.345 mm, which is lower than the preset safety clearance threshold of 4 mm. Therefore, this risk is not one that can be detected by nominal path detection, but rather a springback-induced interference risk that manifests during subsequent processing after the post-springback spatial attitude update of the formed bend segment caused by the machine learning springback prediction results.

[0067] The risk type was identified as rebound-induced insufficient safety clearance interference risk, and the corresponding risk information was recorded. This risk information includes: the risk source is post-rebound geometry update caused by machine learning rebound prediction; the risk occurred in the third bending unit; the initial risk occurrence stage was the pipe rotation stage; the initial risk occurrence progress was 60.000; the risk object was the ground; the nominal path minimum effective safety clearance was 4.433 mm; the minimum effective safety clearance after enabling rebound prediction was 2.345 mm; the preset safety clearance threshold was 4 mm; the number of interference risk sub-steps was 33; and the total number of detection sub-steps was 144.

[0068] Furthermore, to address the aforementioned interference risk caused by insufficient safety clearance due to springback, the parameters of the original multi-bend path were modified. Since the risk first occurs during the pipe rotation stage of the third bending unit, and the risk object is the ground, it is determined that this risk is mainly related to the spatial attitude of the formed pipe segment before the start of processing in the third bending unit, the pipe rotation attitude of the third bending unit, and the subsequent bending sweep direction. Therefore, the process parameters of the second and third bending units are modified first to improve the safety clearance between the springback path and the ground.

[0069] The revised process parameters for the multi-bend pipe fitting are as follows: For the first bending unit, the feeding length is 80.000 mm, the turning angle is 0.000°, the bending angle is 90.000°, and the bending radius is 70.000 mm; for the second bending unit, the feeding length is adjusted from 130.000 mm to 120.000 mm, the turning angle remains at 80.000°, the bending angle remains at 107.000°, and the bending radius remains at 60.000 mm; for the third bending unit, the feeding length is adjusted from 90.000 mm to 70.000 mm, the turning angle is adjusted from 60.000° to 90.000°, the bending angle is adjusted from 72.000° to 60.000°, and the bending radius is adjusted from 70.000 mm to 55.000 mm.

[0070] like Figure 4 As shown, the springback-sensing dynamic interference detection results of the corrected path indicate that the path contains 144 scanning substeps and 145 dynamic states. During the entire process of feeding, tube transfer, and bending, no interference risk substeps with clearances smaller than the preset safety clearance threshold were detected, and the initial interference risk was "none". The current minimum effective safety clearance of the corrected path is 10.751 mm, significantly greater than the preset safety clearance threshold of 4 mm; specifically, the minimum effective safety clearance between the pipe and the ground is approximately 10.75 mm, and the minimum effective safety clearance between the pipe and the machine bed is also greater than the safety threshold. Therefore, the corrected springback-sensing machining path is determined to meet the safety machining requirements.

[0071] As can be seen from this embodiment, if only the nominal processing path is used for dynamic interference detection, the three-bend pipe fitting will be judged to meet the safe processing requirements. However, by using the method of the present invention, the springback spatial attitude of the formed bend can be dynamically updated using machine learning springback prediction results, and the interference risk of insufficient safety clearance in the ground direction can be identified during the pipe rotation and bending process of the third bending unit. Therefore, the present invention can effectively avoid the omission of springback-induced risks by the traditional nominal path interference detection method, and improve the accuracy and reliability of dynamic interference detection before processing of multi-bend pipe fittings.

[0072] Example 2: Dynamic Interference Verification and Safe Output Method for Machine Learning Bounce Compensation Paths, such as Figure 5 As shown.

[0073] This embodiment illustrates that in the processing of multi-bend pipe fittings, machine learning springback prediction results can not only be used for bending angle compensation and forming accuracy correction, but must also be further involved in the dynamic interference verification of the compensation path. If compensation process parameters are generated solely based on the springback prediction results without safety checks on the entire process of feeding, pipe transfer, and bending after compensation, a path may be obtained that improves the forming accuracy after springback but leaves the dynamic safety clearance unknown. This embodiment demonstrates that the present invention can perform dynamic interference detection and safety screening on the machine learning springback compensation path, thereby obtaining a processing path that simultaneously meets the requirements of forming accuracy after springback and dynamic safety clearance.

[0074] This embodiment uses a three-bend spatial pipe as the object. The pipe is made of 6061 aluminum alloy, with an outer diameter of 16.0 mm and a wall thickness of 2.0 mm. A preset safety clearance threshold of 4 mm is set, and a parametric machine tool scene model is established under a fixed machine head reference system. The machine tool scene model includes a bending die, clamping block, pressing die, guide die, machine bed, and ground. The machine bed, ground, bending die, clamping block, pressing die, guide die, and non-adjacent pipe sections of the pipe itself are all used as objects for dynamic interference detection.

[0075] In this embodiment, the original nominal bending process parameters are as follows: the feeding length of the first bending unit is 95.000 mm, the tube angle is 0.000°, the bending angle is 80.000°, and the bending radius is 75.000 mm; the feeding length of the second bending unit is 125.000 mm, the tube angle is 75.000°, the bending angle is 92.000°, and the bending radius is 65.000 mm; the feeding length of the third bending unit is 85.000 mm, the tube angle is 45.000°, the bending angle is 68.000°, and the bending radius is 70.000 mm.

[0076] First, based on the aforementioned original nominal process parameters, a nominal processing path for the multi-bend pipe fitting is generated. Then, dynamic interference detection is performed on this nominal processing path throughout the entire process of feeding, pipe transfer, and bending. Figure 6 As shown, the test results indicate that the original nominal path did not exhibit any interference risk below the preset safety gap threshold during the entire processing process. Therefore, from the perspective of processing safety, the original nominal path meets the requirements for dynamic interference detection.

[0077] For the original nominal path in this embodiment, the machine learning springback prediction model outputs the following results: the predicted springback angle for the first bending unit is 2.730°, the predicted springback angle for the second bending unit is 2.874°, and the predicted springback angle for the third bending unit is 2.676°. If processing is still performed according to the original nominal bending angle, the equivalent bending angle after springback for the first bending unit is approximately 77.270°, the equivalent bending angle after springback for the second bending unit is approximately 89.126°, and the equivalent bending angle after springback for the third bending unit is approximately 65.324°, all of which are smaller than the corresponding target bending angle. Therefore, without compensation, the final multi-bend pipe fitting will have insufficient springback forming angle.

[0078] To improve the post-spring forming accuracy, a direct springback compensation path is generated based on the predicted springback angle using machine learning. Specifically, the bending angle of the first bending unit is compensated from 80.000° to 82.730°, the bending angle of the second bending unit is compensated from 92.000° to 94.874°, and the bending angle of the third bending unit is compensated from 68.000° to 70.676°; the feeding length, tube angle, and bending radius remain unchanged in the first round of compensation.

[0079] Therefore, the process parameters for the direct springback compensation path are as follows: the feeding length of the first bending unit is 95.000 mm, the tube angle is 0.000°, the bending angle is 82.730°, and the bending radius is 75.000 mm; the feeding length of the second bending unit is 125.000 mm, the tube angle is 75.000°, the bending angle is 94.874°, and the bending radius is 65.000 mm; the feeding length of the third bending unit is 85.000 mm, the tube angle is 45.000°, the bending angle is 70.676°, and the bending radius is 70.000 mm.

[0080] The test results show that the direct springback compensation path improves the post-spring forming accuracy without introducing new dynamic interference risks. A dynamic scan of the entire process of feeding, tube rotation, and bending was performed on the compensated three-bend path, including 144 scanning sub-steps and 145 dynamic states. Figure 7 As shown, the test results indicate that no interference substeps smaller than the preset safety clearance threshold were found throughout the entire process, and the first interference was zero. The current minimum effective safety clearance of the compensation path is 16.612 mm, which is significantly greater than the preset safety clearance threshold of 4 mm; among them, the minimum effective safety clearance between the fitting and the machine bed is approximately 54.67 mm, and the minimum effective safety clearance between the fitting and the ground is approximately 187.0 mm, both of which meet the requirements for safe processing.

[0081] This embodiment demonstrates that the machine learning springback prediction results can not only be used to determine the springback compensation amount for each bending unit, but also combined with the dynamic interference detection process to perform safety checks on the compensated processing path. In this embodiment, after compensating the bending angles of the three bends based on the predicted springback angle, the dynamic interference detection of the entire springback sensing process is re-executed. The detection results show that the minimum effective safety clearance of the compensated path is 16.612 mm, which is greater than the preset safety clearance threshold of 4 mm, and no interference risk was found throughout the process. Therefore, this compensated path can simultaneously meet the requirements for post-spring forming accuracy and dynamic safety clearance, and can be used as the final processing path output.

[0082] Example 3: This invention also provides a dynamic interferometric detection system for spatial multi-bend ducts that considers the springback effect, such as... Figure 8 As shown, the system includes: a parameter module, a modeling module, a calculation module, and an output module.

[0083] The parameter module is used to obtain the basic parameters of the pipe fitting, bending process parameters and machine tool scene parameters of the spatial multi-bend conduit to be processed, establish the nominal processing path of the spatial multi-bend conduit based on the bending process parameters, and establish a parameterized machine tool scene model under the fixed machine head reference system based on the machine tool scene parameters. The modeling module is used to complete the nominal bending process for each bending unit using the nominal processing path. Based on the pipe basic parameters of the bending unit, the springback prediction result of the bending unit is obtained through a pre-trained machine learning springback prediction model. The nominal geometric state of the formed bending segment is corrected to obtain the springback prediction geometric model after the bending unit is unloaded. The dynamic pipe model considering the cumulative effect of springback is used to construct the dynamic pipe model. The calculation module is used to perform dynamic scanning interference detection on the spatial multi-bend conduit during the feeding, pipe turning and bending stages based on the dynamic pipe fitting model and the parameterized machine tool scene model, and to calculate the effective safety clearance between the spatial multi-bend conduit and obstacles. The output module is used to determine the interference risk of the current processing path based on the effective safety clearance. If the current processing path does not meet the safety requirements, the current processing path is compensated and corrected, and dynamic interference detection is re-executed until the corrected or compensated processing path meets both the safety requirements and the post-formation accuracy requirements. Then, the final recommended processing path and dynamic interference detection results are output.

[0084] It is understood that the dynamic interference detection system for spatial multi-bend ducts considering the springback effect provided by the present invention corresponds to the dynamic interference detection method for spatial multi-bend ducts considering the springback effect provided in the foregoing embodiments. The relevant technical features of the dynamic interference detection system for spatial multi-bend ducts considering the springback effect can be referred to the relevant technical features of the dynamic interference detection method for spatial multi-bend ducts considering the springback effect, and will not be repeated here.

[0085] Another object of the present invention is to provide an electronic device, such as... Figure 9 As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor performing the steps of the dynamic interference detection method for spatially curved ducts that takes into account springback effects.

[0086] The dynamic interference detection method for spatially curved ducts considering springback effect includes the following steps: Obtain the basic parameters, bending process parameters, and machine tool scene parameters of the spatial multi-bend conduit to be processed. Establish the nominal processing path of the spatial multi-bend conduit based on the bending process parameters, and establish a parametric machine tool scene model under a fixed machine head reference system based on the machine tool scene parameters. For each bending unit, the nominal bending process is completed using the nominal processing path. Based on the basic parameters of the bending unit, the springback prediction result of the bending unit is obtained through a pre-trained machine learning springback prediction model. The nominal geometric state of the formed bending segment is corrected to obtain the springback prediction geometric model after the bending unit is unloaded. The dynamic pipe fitting model considering the cumulative effect of springback is constructed. Based on dynamic pipe fitting models and parametric machine tool scene models, dynamic scanning interference detection is performed on the entire process of spatial multi-bend conduit during feeding, pipe turning and bending stages to calculate the effective safety clearance between the spatial multi-bend conduit and obstacles. The interference risk of the current processing path is judged based on the effective safety clearance. If the current processing path does not meet the safety requirements, the current processing path is compensated and corrected, and the dynamic interference detection is re-executed until the corrected or compensated processing path meets both the safety requirements and the post-formation accuracy requirements. Then, the final recommended processing path and dynamic interference detection results are output.

[0087] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the dynamic interference detection method for spatially curved ducts considering springback effects.

[0088] The dynamic interference detection method for spatially curved ducts considering springback effect includes the following steps: Obtain the basic parameters, bending process parameters, and machine tool scene parameters of the spatial multi-bend conduit to be processed. Establish the nominal processing path of the spatial multi-bend conduit based on the bending process parameters, and establish a parametric machine tool scene model under a fixed machine head reference system based on the machine tool scene parameters. For each bending unit, the nominal bending process is completed using the nominal processing path. Based on the basic parameters of the bending unit, the springback prediction result of the bending unit is obtained through a pre-trained machine learning springback prediction model. The nominal geometric state of the formed bending segment is corrected to obtain the springback prediction geometric model after the bending unit is unloaded. The dynamic pipe fitting model considering the cumulative effect of springback is constructed. Based on dynamic pipe fitting models and parametric machine tool scene models, dynamic scanning interference detection is performed on the entire process of spatial multi-bend conduit during feeding, pipe turning and bending stages to calculate the effective safety clearance between the spatial multi-bend conduit and obstacles. The interference risk of the current processing path is judged based on the effective safety clearance. If the current processing path does not meet the safety requirements, the current processing path is compensated and corrected, and the dynamic interference detection is re-executed until the corrected or compensated processing path meets both the safety requirements and the post-formation accuracy requirements. Then, the final recommended processing path and dynamic interference detection results are output.

[0089] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0090] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0091] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0092] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A dynamic interferometric detection method for spatial multi-bend ducts considering springback effect, characterized in that, include: Obtain the basic parameters, bending process parameters, and machine tool scene parameters of the spatial multi-bend conduit to be processed. Establish the nominal processing path of the spatial multi-bend conduit based on the bending process parameters, and establish a parametric machine tool scene model under a fixed machine head reference system based on the machine tool scene parameters. For each bending unit, the nominal bending process is completed using the nominal processing path. Based on the basic parameters of the bending unit, the springback prediction result of the bending unit is obtained through a pre-trained machine learning springback prediction model. The nominal geometric state of the formed bending segment is corrected to obtain the springback prediction geometric model after the bending unit is unloaded. The dynamic pipe fitting model considering the cumulative effect of springback is constructed. Based on dynamic pipe fitting models and parametric machine tool scene models, dynamic scanning interference detection is performed on the entire process of spatial multi-bend conduit during feeding, pipe turning and bending stages to calculate the effective safety clearance between the spatial multi-bend conduit and obstacles. The interference risk of the current processing path is judged based on the effective safety clearance. If the current processing path does not meet the safety requirements, the current processing path is compensated and corrected, and the dynamic interference detection is re-executed until the corrected or compensated processing path meets both the safety requirements and the post-formation accuracy requirements. Then, the final recommended processing path and dynamic interference detection results are output.

2. The dynamic interference detection method for spatial multi-bend ducts considering springback effect according to claim 1, characterized in that, The basic parameters of pipe fittings include geometric parameters and material parameters; Geometric parameters include the pipe's outer diameter, wall thickness, total length, number of bends, length of each straight segment, and target centerline information; Material parameters include elastic modulus, yield strength, hardening parameters, material batch information, and forming history information; The bending process parameters include the feeding length, tube angle, bending angle, and bending radius of each bending unit; Machine tool scene parameters include the dimensions, positions, and installation deviations of bending dies, clamping blocks, pressing dies, guide dies, bed, ground, and other machine tool components; The parametric machine tool scene model includes bending mold model, clamping block model, pressing mold model, guide mold model, bed model, ground model, and other obstacle models.

3. The dynamic interference detection method for spatial multi-bend ducts considering springback effect according to claim 1, characterized in that, The machine learning rebound prediction model is one of the following, or a combination of the above models: neural network, support vector regression, random forest, gradient boosting model, Gaussian process regression, and physically constrained machine learning model; The machine learning springback prediction model takes the basic parameters of the pipe fitting, bending process parameters, machine tool scene parameters, and the forming state of the preceding bending unit as inputs, and takes the springback amount, end position deviation, end posture deviation, curvature change amount, or centerline deviation after the corresponding bending unit is unloaded as outputs. The machine learning springback prediction model is trained using historical processing data, finite element simulation data, online measurement data, or trial bending experimental data.

4. The dynamic interference detection method for spatial multi-bend ducts considering springback effect according to claim 1, characterized in that, The post-missile prediction geometric model includes the post-missile centerline, post-missile end coordinates, post-missile local coordinate system, and post-missile spatial envelope model.

5. The dynamic interference detection method for spatial multi-bend ducts considering springback effect according to claim 1, characterized in that, The process of constructing a dynamic pipe fitting model includes: According to the processing sequence of the multi-bend pipe fitting, each bending unit is processed sequentially. For the first bending unit, the feeding, pipe turning and bending process is simulated according to its nominal processing path. After bending is completed, its springback deformation is predicted according to the machine learning model, and the springback prediction geometric model of the first bending unit is generated. For the second and subsequent bending units, the post-blowout predicted geometric model of all preceding bending units is used as the formed pipe segment, and the nominal path or compensation path of the current bending unit is used as the pipe segment to be processed, together forming the dynamic pipe fitting model at the current moment.

6. The dynamic interference detection method for spatial multi-bend ducts considering springback effect according to claim 1, characterized in that, The entire process of dynamic scanning interferometry detection includes: During the feeding stage, the position of the guide tube along the feeding direction is updated; during the tube turning stage, the rotational attitude of the guide tube around the machine head axis is updated; during the bending stage, the bending state of the current bending unit around the bending mold is updated; and in each scanning substep, the spatial attitude of the multi-bend tube is updated according to the current tube state, and the effective safety gap between the tube spatial envelope model after considering springback and the object being inspected is calculated. When calculating the effective safety clearance, a normal process contact exclusion zone is set for normal clamping, guiding or supporting contact between clamping blocks, pressing dies, bending dies and guide dies and pipe fittings. Contact within the normal process contact exclusion zone is not considered abnormal interference. A state that exceeds the normal process contact exclusion zone and the effective safety clearance is less than the preset safety clearance threshold is considered as an abnormal interference risk.

7. The dynamic interference detection method for spatial multi-bend ducts considering springback effect according to claim 1, characterized in that, Methods for compensating and correcting the current processing path include Based on the springback prediction results, the bending process parameters are compensated and corrected, and the feeding length, tube angle, bending angle, bending radius, bending sequence, fixture avoidance parameters, bed avoidance parameters, or combinations thereof are adjusted. If the springback prediction results indicate that the springback posture of a certain bending unit causes subsequent bending units to approach the bed, ground, or machine tool components, then the compensation parameters of the bending unit from which the springback originates, or the rotation angle, feeding length, or avoidance parameters of subsequent risky bending units, should be corrected first.

8. A dynamic interferometric detection system for spatial multi-bend ducts considering springback effect, characterized in that, include: The parameter module is used to obtain the basic parameters of the pipe fitting, bending process parameters and machine tool scene parameters of the spatial multi-bend conduit to be processed, establish the nominal processing path of the spatial multi-bend conduit based on the bending process parameters, and establish a parameterized machine tool scene model under the fixed machine head reference system based on the machine tool scene parameters. The modeling module is used to complete the nominal bending process for each bending unit using the nominal processing path. Based on the pipe basic parameters of the bending unit, the springback prediction result of the bending unit is obtained through a pre-trained machine learning springback prediction model. The nominal geometric state of the formed bending segment is corrected to obtain the springback prediction geometric model after the bending unit is unloaded. The dynamic pipe model considering the cumulative effect of springback is used to construct the dynamic pipe model. The calculation module is used to perform dynamic scanning interference detection on the spatial multi-bend conduit during the feeding, pipe turning and bending stages based on the dynamic pipe fitting model and the parameterized machine tool scene model, and to calculate the effective safety clearance between the spatial multi-bend conduit and obstacles. The output module is used to determine the interference risk of the current processing path based on the effective safety clearance. If the current processing path does not meet the safety requirements, the current processing path is compensated and corrected, and dynamic interference detection is re-executed until the corrected or compensated processing path meets both the safety requirements and the post-formation accuracy requirements. Then, the final recommended processing path and dynamic interference detection results are output.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images