An adaptive surface finishing method and device for additive manufacturing curved surface components

Abstract

The application discloses a kind of adaptive surface finishing methods and devices for additive manufacturing curved surface component, it is related to additive manufacturing post-data processing technical field, including the following steps: mechanical structure, for multi-degree-of-freedom motion platform and end effector;Wherein, multi-degree-of-freedom motion platform is 6-axis robot, for adapting the spatial posture adjustment of complex curved surface;End effector integrates finishing tool, sensor and adaptive adjusting mechanism, finishing tool includes grinding head, polishing wheel and laser finishing head, sensor includes force sensor and vision sensor, and adaptive adjusting mechanism is pressure compensation device;Control system has hardware and software, hardware includes industrial controller, sensor data acquisition card and servo drive system;Software includes feature recognition algorithm library, finishing parameter decision software and real-time monitoring interface;The application solves the problems of low efficiency, poor adaptability and lack of adaptive adjustment capability of existing automatic finishing of traditional finishing.

Description

Technical Field

[0001] This invention relates to the field of data processing technology after additive manufacturing, and specifically to an adaptive surface trimming method and apparatus for additively manufactured curved components. Background Technology

[0002] Additive manufacturing technology, due to its high degree of design freedom and ability to form complex structures, is widely used in high-end manufacturing fields such as aero-engine blades, medical implants, and mold cavities. These curved surface components typically have extremely high requirements for surface quality, not only needing to meet strict dimensional accuracy and surface roughness standards, but also needing to ensure good mechanical properties to guarantee service reliability.

[0003] However, surface defects such as lamination, protrusions, and pits are inevitably generated during additive manufacturing. Therefore, surface finishing has become a key process in the post-processing stage of additive manufacturing. Current mainstream surface finishing methods for curved components have significant shortcomings: traditional manual finishing is inefficient, relies heavily on operator experience, and is prone to mechanical damage to components, making it difficult to guarantee consistent finishing quality; traditional mechanical finishing methods use fixed-parameter processing modes, which cannot adapt to the geometrical variations of complex curved surfaces, resulting in limited finishing accuracy; existing automated finishing technologies mostly rigidly execute preset programs, lacking the ability to adaptively adjust to individual component differences (such as geometric deviations and material property fluctuations), leading to poor finishing results, and even over-finishing or under-finishing, severely restricting the industrial application of additively manufactured curved components. Summary of the Invention

[0004] This invention provides an adaptive surface finishing method and apparatus for additively manufactured curved surface components, addressing the technical problem of poor surface finishing quality in existing additively manufactured curved surface components. By combining a deep learning segmentation model, it accurately identifies free-form and regular surface types, as well as defect types such as protrusions, pits, and delamination, while precisely capturing material hardness and toughness characteristics, providing comprehensive and high-precision basic data support for finishing. Based on a reinforcement learning-based surface feature-finishing parameter mapping model, with surface accuracy achievement rate and component damage rate as core optimization indicators, it dynamically outputs highly adaptable tool path, pressure, speed, and tool type parameters, achieving precise finishing of complex curved surface components.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0006] An adaptive surface trimming method for additively manufactured curved surface components includes the following steps:

[0007] Surface feature perception: Acquire three-dimensional geometric data of additively manufactured curved components through visual scanning or tactile feedback. The three-dimensional geometric data includes curvature, contour, and surface defect distribution. Automatically identify surface type, defect type, and material properties based on machine learning algorithms. Surface types include free-form surfaces and regular surfaces. Defect types include protrusions, pits, and layering. Material properties include hardness and toughness.

[0008] Trimming strategy generation: Establish a surface feature-trimming parameter mapping model. Based on the identified surface features and material properties, dynamically output trimming tool path, pressure, speed and tool type parameters to achieve multi-objective optimization of surface accuracy assurance, trimming time minimization and component damage risk minimization.

[0009] Real-time feedback control: Force and displacement sensors monitor force and displacement signals during the trimming process to evaluate the trimming effect on surface roughness and dimensional accuracy in real time; if a trimming deviation is detected, the trimming parameters are automatically adjusted iteratively to form a closed-loop control of perception-decision-execution-feedback.

[0010] Optionally, visual scanning uses 3D laser scanning technology, machine learning algorithm is deep learning segmentation model, and tactile feedback is acquired by collecting tactile data of component surface through contact pressure sensor array.

[0011] Optionally, the surface feature-trimming parameter mapping model is a reinforcement learning-based decision model. The reinforcement learning decision model uses the surface accuracy achievement rate, trimming time reduction ratio, and component damage rate of the reinforcement learning reward function as the core optimization indicators.

[0012] Optionally, the criteria for determining the trimming deviation are: the surface roughness exceeds the preset threshold of ±0.2μm, or the dimensional accuracy error is greater than ±0.02mm, and the adjustment range of the parameters in the iterative adjustment is linearly related to the magnitude of the deviation.

[0013] An adaptive surface trimming apparatus for additively manufactured curved surface components includes:

[0014] The mechanical structure consists of a multi-degree-of-freedom motion platform and an end effector. The multi-degree-of-freedom motion platform is a 6-axis robot used to adapt to the spatial posture adjustment of complex curved surfaces. The end effector integrates dressing tools, sensors, and an adaptive adjustment mechanism. The dressing tools include a grinding head, a polishing wheel, and a laser dressing head. The sensors include force sensors and vision sensors. The adaptive adjustment mechanism is a pressure compensation device.

[0015] The control system comprises hardware and software. The hardware includes an industrial controller, sensor data acquisition cards, and a servo drive system. The software includes a feature recognition algorithm library, trimming parameter decision software, and a real-time monitoring interface. It supports data interaction with additive manufacturing equipment and can read component design models.

[0016] Optionally, the positioning accuracy of the multi-degree-of-freedom motion platform is ≤ ±0.01mm, and the motion speed is ≥500mm / s.

[0017] Optionally, the end effector's sensors also include a temperature sensor to monitor the surface temperature of the component in real time during the trimming process. When the temperature exceeds a preset threshold of 80°C, the trimming parameters are automatically adjusted or the trimming is paused.

[0018] Optionally, the trimming tool can be quickly switched via a quick-change interface with a switching time of ≤30s. The quick-change interface has a self-locking function with a positioning accuracy of ≤±0.005mm.

[0019] Optionally, the control system software also includes a data storage module for recording surface feature data, trimming parameters, trimming effect detection data, and equipment operating status data.

[0020] Optionally, the control system supports remote monitoring and parameter adjustment, and data transmission is achieved through industrial Ethernet or 5G communication modules with a transmission latency of ≤10ms.

[0021] The beneficial effects of this invention are:

[0022] 1. This invention utilizes a multi-source sensing approach that integrates 3D laser scanning and a contact pressure sensor array, combined with a deep learning segmentation model. This allows for precise identification of free-form and regular curved surfaces, as well as defect types such as protrusions, pits, and layering. Simultaneously, it accurately captures material hardness and toughness characteristics, providing comprehensive and high-precision basic data support for trimming. Based on a reinforcement learning-based surface feature-trimming parameter mapping model, with surface accuracy achievement rate and component damage rate as core optimization indicators, it dynamically outputs highly adaptable toolpath, pressure, speed, and tool type parameters. Combined with ±0.2μm surface roughness deviation thresholds and ±0.02mm dimensional accuracy deviation thresholds, it achieves precise trimming of complex curved surface components. Compared to the experience-dependent nature of traditional manual trimming and the insufficient adaptability of fixed-parameter mechanical trimming, this invention effectively avoids over-trimming, under-trimming, and mechanical damage, significantly improving the consistency of component surface quality and dimensional accuracy.

[0023] 2. The multi-degree-of-freedom motion platform of this invention employs a 6-axis robot with a positioning accuracy of ≤±0.01mm and a movement speed of ≥500mm / s. It flexibly adapts to the spatial posture adjustment of complex curved surfaces and, combined with an optimal trimming path based on dynamic programming of surface features, avoids ineffective work strokes. The trimming tool achieves rapid switching within ≤30 seconds via a quick-change interface, with a positioning accuracy of ≤±0.005mm, eliminating the need for frequent manual adjustments and significantly reducing downtime during tool changes. Simultaneously, the reinforcement learning decision model incorporates the trimming time reduction ratio into the core optimization index, achieving efficient matching of trimming parameters while ensuring accuracy and safety, avoiding the time-consuming problem of repeated trial and error in traditional trimming. Furthermore, the control system supports data interaction with additive manufacturing equipment, directly reading the component design model, eliminating intermediate data conversion and remodeling steps, further improving overall operational efficiency and effectively shortening the post-processing production cycle of additively manufactured components. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of the device structure of the present invention;

[0026] Figure 2 This is a flowchart illustrating one embodiment of the device of the present invention;

[0027] Figure 3 This is a flowchart illustrating another embodiment of the device of the present invention. Detailed Implementation

[0028] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0029] Example 1

[0030] like Figure 1 As shown, this embodiment provides an adaptive surface trimming device for additively manufactured curved surface components, comprising:

[0031] The mechanical structure consists of a multi-degree-of-freedom motion platform and an end effector. The multi-degree-of-freedom motion platform is a 6-axis robot used to adapt to the spatial posture adjustment of complex curved surfaces. The end effector integrates dressing tools, sensors, and an adaptive adjustment mechanism. The dressing tools include a grinding head, a polishing wheel, and a laser dressing head. The sensors include force sensors and vision sensors. The adaptive adjustment mechanism is a pressure compensation device.

[0032] The mechanical structure is the core execution part for realizing surface trimming operations. It mainly consists of two parts: a multi-degree-of-freedom motion platform and an end effector. The two work together to complete the precise positioning and trimming operation of complex curved surface components.

[0033] Employing a 6-axis robot as a multi-degree-of-freedom motion platform, its core function is to achieve spatial posture adjustment during finishing operations, precisely adapting to complex curved surface components with different curvatures and shapes. The 6-axis robot possesses multi-dimensional flexible motion capabilities, enabling the end effector to complete finishing actions at any angle and position in three-dimensional space, ensuring that the finishing tool and the component surface always maintain the optimal contact posture.

[0034] With the existing configuration, the positioning accuracy of the multi-degree-of-freedom motion platform is ≤ ±0.01mm and the motion speed is ≥500mm / s, which can meet the requirements of high-precision and high-efficiency finishing operations and is suitable for the processing of precision curved surface components with strict dimensional accuracy requirements.

[0035] The end effector is a core working unit that integrates multiple functional modules, mainly including dressing tools, sensors, and adaptive adjustment mechanisms. These modules work together to achieve efficient and precise surface dressing and provide real-time feedback of dressing process data.

[0036] The dressing tool system integrates three core dressing tools: a grinding head, a polishing wheel, and a laser dressing head. Different tools can be selected based on the component's surface material and dressing requirements: the grinding head is suitable for rough dressing operations such as surface roughness removal and burr cleaning; the polishing wheel is suitable for fine dressing operations such as improving surface smoothness; and the laser dressing head is suitable for high-precision, micro-structure surface dressing operations. In the optional configuration, the dressing tools can be quickly switched via a quick-change interface with a switching time of ≤30 seconds. The quick-change interface has a self-locking function with a positioning accuracy of ≤±0.005mm, ensuring accurate positioning and stable connection after tool switching, and preventing dressing accuracy from being affected by tool misalignment.

[0037] The core sensor configuration consists of a force sensor and a vision sensor, with a temperature sensor optional. The force sensor collects real-time contact force data between the tool and the component surface during the trimming process, providing pressure feedback for adaptive adjustment. The vision sensor acquires image information of the component surface, assisting in surface feature recognition and trimming effect detection. The optional temperature sensor monitors the component surface temperature in real-time during trimming. When the temperature exceeds a preset threshold of 80℃, it automatically triggers trimming parameter adjustments (e.g., reducing trimming speed, decreasing contact pressure) or pauses trimming to prevent high temperatures from affecting the component material and performance. The end effector also includes a temperature sensor to monitor the component surface temperature in real-time during trimming. When the temperature exceeds a preset threshold of 80℃, it automatically triggers trimming parameter adjustments or pauses trimming.

[0038] The adaptive adjustment mechanism employs a pressure compensation device. Its core function is to automatically adjust the contact pressure between the dressing tool and the component surface based on contact force data fed back from the force sensor. When changes in the curvature of the component surface cause abnormal contact pressure, the pressure compensation device can respond quickly. Through elastic adjustment of the mechanical structure or active adjustment of the drive components, it maintains the contact pressure within a preset reasonable range, ensuring uniform dressing results and avoiding localized over- or under-dressing issues.

[0039] The control system comprises hardware and software. The hardware includes an industrial controller, sensor data acquisition cards, and a servo drive system. The software includes a feature recognition algorithm library, trimming parameter decision software, and a real-time monitoring interface. It supports data interaction with additive manufacturing equipment and can read component design models.

[0040] Hardware component: The industrial controller, as the core control unit, is responsible for receiving and processing various data signals, executing control algorithms, sending control commands to various execution components, and ensuring the coordinated operation of the entire device.

[0041] Sensor data acquisition cards are used to collect real-time data from various sensors, including force, vision, and temperature sensors, and transmit the data to industrial controllers for processing and analysis, providing data support for control decisions.

[0042] The servo drive system is paired with the actuators of the multi-degree-of-freedom motion platform and the adaptive adjustment mechanism. It receives instructions from the industrial controller and drives the corresponding components to complete precise movements, ensuring motion accuracy and response speed.

[0043] Software component: The feature recognition algorithm library has built-in multiple surface feature recognition algorithms, which can automatically identify key features of the component surface curvature changes, defect locations and dimensional deviations based on image data collected by the vision sensor and the component design model read, providing a basis for adjustment parameter decisions.

[0044] The trimming parameter decision software automatically determines and generates optimal trimming parameters (such as trimming tool type, movement speed, contact pressure, and trimming path) based on feature recognition results, sensor feedback data, and preset trimming requirements. It also has a parameter adaptive adjustment function, which can dynamically optimize trimming parameters based on real-time monitoring data.

[0045] The real-time monitoring interface provides a visual operation and monitoring interface, allowing operators to view key data during the trimming process in real time (such as contact pressure, component temperature, movement trajectory, and equipment operating status). At the same time, operators can manually set parameters and start / pause trimming operations through the interface.

[0046] The data storage module (optional) is used to record surface feature data, trimming parameters, trimming effect detection data, and equipment operating status data. The data can be stored for a long time, which is convenient for subsequent traceability, process optimization, and equipment maintenance.

[0047] Remote monitoring and parameter adjustment module (optional): Supports remote data transmission and control via industrial Ethernet or 5G communication module. Operators can monitor the device's operating status in real time on a remote terminal and adjust parameters remotely as needed, improving the convenience of operation and maintenance; transmission latency ≤10ms ensures the real-time performance and reliability of remote control.

[0048] like Figure 2 As shown, the workflow of this device is based on data-driven, precise positioning, adaptive trimming, and real-time feedback optimization. The specific principles are as follows:

[0049] Data interaction and model import: By establishing a data connection between the control system and the additive manufacturing equipment, the design model (such as a CAD model) of the component to be modified is read to obtain the basic information of the component's surface features and dimensional parameters.

[0050] Surface feature recognition: The vision sensor collects actual image data of the component surface, combines it with the imported design model, and automatically identifies the curvature distribution, defect location, and deviation of the actual size from the design model of the component surface through the feature recognition algorithm library, and transmits the recognition results to the industrial controller.

[0051] Trimming parameter decision: Based on feature recognition results and combined with preset trimming quality requirements (such as surface roughness and dimensional accuracy), the trimming parameter decision software automatically selects the appropriate trimming tool, generates the optimal trimming path, movement speed, and contact pressure parameters, and sends the parameter commands to the servo drive system and end effector.

[0052] Precise positioning and trimming operations: The multi-degree-of-freedom motion platform (6-axis robot) drives the end effector to move to the designated position according to the instructions and adjusts it to the optimal working posture; the end effector starts the trimming tool according to the preset parameters, and at the same time, the adaptive adjustment mechanism adjusts the contact state between the tool and the component surface in real time according to the contact pressure data fed back by the force sensor to ensure the stability of the trimming process.

[0053] Real-time monitoring and feedback optimization: During the trimming process, each sensor (force, vision, and temperature) collects data in real time and transmits it to the control system. The real-time monitoring interface displays the relevant data synchronously. If the temperature sensor detects that the surface temperature of the component exceeds 80°C, or the force sensor detects abnormal contact pressure, the system will automatically trigger the adjustment of trimming parameters or suspend trimming. The vision sensor detects the trimming effect in real time. If the preset requirements are not met, the system will re-optimize the trimming parameters until the quality standards are met.

[0054] Data recording and job completion: After the trimming job is completed, the data storage module automatically saves the surface feature data, trimming parameters and trimming effect detection data of this job; the device is reset, completing the entire trimming process.

[0055] Example 2

[0056] Based on Example 1, such as Figure 3 As shown, this embodiment provides an adaptive surface trimming method for additively manufactured curved surface components, including the following steps:

[0057] Surface feature perception: Acquire three-dimensional geometric data of additively manufactured curved components through visual scanning or tactile feedback. The three-dimensional geometric data includes curvature, contour, and surface defect distribution. Automatically identify surface type, defect type, and material properties based on machine learning algorithms. Surface types include free-form surfaces and regular surfaces. Defect types include protrusions, pits, and layering. Material properties include hardness and toughness.

[0058] To obtain comprehensive and accurate key geometric and material information of additively manufactured curved surface components, a data foundation is provided for the formulation of subsequent trimming strategies, as detailed below:

[0059] The data acquisition method employs a multi-source perception fusion approach combining visual scanning and tactile feedback. Visual scanning enables non-contact acquisition of large-scale curved surface geometric data, suitable for quickly obtaining overall component contour information. Tactile feedback, on the other hand, acquires microscopic geometric and mechanical data of the component surface through tactile interaction, compensating for the shortcomings of visual scanning in perceiving surface micro-defects and local material properties. The combination of the two ensures the comprehensiveness and accuracy of the perceived data.

[0060] The core sensing data clearly defines three key three-dimensional geometric data, specifically including: ① Curvature: a core geometric parameter reflecting the degree of local bending of the surface, which is an important basis for distinguishing different surface types and formulating differentiated repair strategies; ② Contour: the overall shape trajectory data of the component surface, which determines the macroscopic scope of repair and the basis for path planning; ③ Surface defect distribution: covering the location, size and quantity information of various abnormal shapes on the component surface, which is the core target of repair. Automatic feature recognition is based on machine learning algorithms to intelligently analyze perceived data and automatically identify three core features: ① Surface type classification: Surfaces are divided into free-form surfaces (without fixed analytical equations and complex, irregular shapes, such as the surface of aero-engine blades) and regular surfaces (with explicit analytical equations, such as cylindrical, conical, and spherical surfaces). The difficulty of trimming and the path planning logic of the two types of surfaces differ significantly, and classification and recognition are prerequisites for achieving accurate trimming; ② Defect type recognition: Properly distinguishes three typical defects: protrusions (protrusions formed by excess material on the surface), pits (depressions formed by missing surface material), and layering (periodic surface texture defects formed by layering during additive manufacturing). The trimming methods and required parameters for different defects vary greatly, and accurate recognition ensures targeted trimming; ③ Material property recognition: Focusing on core mechanical properties closely related to the trimming process, namely hardness (the material's ability to resist local deformation, which determines the amount of pressure required for trimming) and toughness (the material's ability to resist fracture, which affects the assessment of component damage risk during trimming), providing a material basis for the safe adaptation of trimming parameters.

[0061] Trimming strategy generation: Establish a surface feature-trimming parameter mapping model. Based on the identified surface features and material properties, dynamically output trimming tool path, pressure, speed and tool type parameters to achieve multi-objective optimization of surface accuracy assurance, trimming time minimization and component damage risk minimization.

[0062] Based on the surface features and material properties obtained through perception, it outputs the optimal trimming parameters to achieve multi-objective optimization, as detailed below:

[0063] Establishing a "surface feature-trimming parameter mapping model" is the core carrier for achieving adaptive decision-making. Through pre-training or iterative optimization, a precise correlation is constructed between surface features (surface type, defect type, curvature and contour), material properties (hardness, toughness) and trimming parameters, ensuring the scientific nature and adaptability of parameter output.

[0064] The output trimming parameters clearly define four key trimming parameters, covering all elements of trimming execution: ① Trimming tool path: the movement trajectory of the trimming tool, which needs to be dynamically planned according to the surface contour, curvature changes, and defect distribution to ensure coverage of all areas to be trimmed and the optimal path; ② Pressure: the pressure exerted by the trimming tool on the component surface, which needs to be adapted to the material hardness and defect type (e.g., harder materials and protruding defects require greater pressure, while tough materials require controlled pressure to avoid damage); ③ Speed: the movement speed of the trimming tool, which affects trimming efficiency and surface quality (too fast a speed can easily lead to surface roughness, while too slow a speed will result in excessive trimming time); ④ Tool type: selecting an appropriate tool according to the surface type, defect size, and material characteristics (e.g., rigid tools are suitable for regular curved surfaces and high-hardness materials, while flexible tools are suitable for free-form surfaces and easily damaged materials).

[0065] The multi-objective optimization objectives clearly define three core optimization goals to achieve a balance between repair effect, efficiency, and safety: ① Surface accuracy assurance: Ensure that the surface roughness and dimensional accuracy of the repaired component meet the preset requirements, which is the core quality objective of repair; ② Minimize repair time: By optimizing tool path and speed parameters, maximize repair efficiency and reduce production costs while ensuring quality; ③ Minimize component damage risk: By adjusting parameters to suit material properties, avoid generating new defects (such as cracks and deformation) during the repair process, ensuring that the component performance is not affected.

[0066] Real-time feedback control: Force and displacement sensors monitor force and displacement signals during the trimming process to evaluate the trimming effect on surface roughness and dimensional accuracy in real time; if a trimming deviation is detected, the trimming parameters are automatically adjusted iteratively to form a closed-loop control of perception-decision-execution-feedback.

[0067] By monitoring in real time and adjusting dynamically, the stability of the trimming process and the consistency of the trimming effect are ensured, forming a complete closed-loop control, as detailed below:

[0068] Real-time monitoring utilizes two core sensor types—force sensors and displacement sensors—to collect key physical signals during the finishing process: ① Force signal: reflects the interaction force between the finishing tool and the component surface, indirectly characterizing the actual magnitude of the finishing pressure, the difficulty of material removal, and the presence of abnormal resistance (such as encountering unidentified hard defects); ② Displacement signal: reflects the actual displacement of the finishing tool and the deformation of the component surface, indirectly assessing real-time changes in dimensional accuracy. Based on these two types of signals, the core performance indicators of surface roughness and dimensional accuracy after finishing are calculated and evaluated in real time.

[0069] Deviation detection and adjustment: If real-time evaluation finds that the trimming effect does not meet expectations (i.e., there is a trimming deviation), the system will automatically start an iterative adjustment mechanism: dynamically correct the trimming parameters (such as path, pressure and speed) according to the type and magnitude of the deviation, and re-execute the trimming process.

[0070] The closed-loop control is formed through a full-process closed-loop design of "perception (acquiring features) - decision (generating parameters) - execution (implementing adjustments) - feedback (monitoring and evaluation)" to ensure that the adjustment process can dynamically adapt to the actual feature differences of the components and the uncertainties in the adjustment process, and finally stably output high-quality curved surface components that meet the requirements.

[0071] Specifically, visual scanning uses 3D laser scanning technology, the machine learning algorithm is a deep learning segmentation model, and tactile feedback is achieved by collecting tactile data of the component surface through a contact pressure sensor array.

[0072] Visual scanning technology explicitly adopts 3D laser scanning technology. 3D laser scanning technology has the advantages of high scanning accuracy, fast speed and non-contact measurement. It can quickly acquire high-precision three-dimensional point cloud data of component surfaces, accurately restore the curvature and contour information of the surface, and is especially suitable for geometric data acquisition of complex free-form surfaces, providing a high-quality visual data foundation for subsequent feature recognition.

[0073] Machine learning algorithms are specifically defined as deep learning segmentation models. Deep learning segmentation models (such as U-Net and Mask R-CNN) have excellent performance in image / point cloud segmentation tasks. They can accurately identify regions in 3D geometric data that belong to different surface types (free-form surfaces / regular surfaces) and different defect types (protrusions / pits / layers), achieving automatic feature classification and localization. Compared with traditional machine learning algorithms (such as SVM and random forest), they have higher recognition accuracy and anti-interference ability, and are especially suitable for scenarios involving the simultaneous recognition of multiple features on complex surfaces.

[0074] The specific method of tactile feedback is to collect tactile data from the component surface through a contact pressure sensor array. This contact pressure sensor array enables multi-point synchronous pressure sensing, accurately acquiring pressure distribution data during the contact between the tool and the component surface. This not only indirectly infers the microscopic geometry of the component surface (e.g., lower pressure in concave areas and higher pressure in convex areas), but also senses the material's hardness distribution characteristics through pressure changes, providing direct data support for material property identification and subsequent pressure parameter optimization.

[0075] The surface feature-trimming parameter mapping model is a decision model based on reinforcement learning. The core optimization indicators of the reinforcement learning decision model are the surface accuracy achievement rate, trimming time reduction ratio, and component damage rate of the reinforcement learning reward function.

[0076] The specific type of mapping model is the surface feature-adjustment parameter mapping model, which is a decision model based on reinforcement learning. Reinforcement learning is a machine learning method that optimizes decisions through agent-environment interaction, reward acquisition, and iterative optimization. It is suitable for dynamic, uncertain, and complex decision-making scenarios. Applying it to the adjustment parameter mapping model allows the model to continuously learn the adaptation rules between surface features and parameters through continuous adjustment practice, gradually optimizing the parameter output strategy. Compared with traditional static mapping models (such as empirical formulas and neural network fitting), it has stronger dynamic adaptability and long-term optimization capabilities.

[0077] The core optimization metrics of the reinforcement learning model are defined as follows: the reward function of reinforcement learning takes "surface accuracy compliance rate, repair time reduction ratio, and component damage rate" as the core optimization metrics.

[0078] The reward function is the core guiding principle of a reinforcement learning model, directly determining the model's optimization direction: ① Surface accuracy achievement rate: As a core quality indicator, the higher the achievement rate, the higher the reward value, ensuring that the model prioritizes outputting parameters that guarantee trimming quality; ② Trimming time reduction ratio: As an efficiency indicator, the higher the ratio (i.e., the shorter the trimming time), the higher the reward value, guiding the model to optimize efficiency while ensuring quality; ③ Component damage rate: As a safety indicator, the lower the damage rate (i.e., the less likely the component is to be damaged), the higher the reward value, preventing the model from sacrificing component safety in pursuit of quality or efficiency. These three factors work together to form the reward function, ensuring that the model achieves a multi-objective balanced optimization of quality, efficiency, and safety.

[0079] The criteria for determining the trimming deviation are: the surface roughness exceeds the preset threshold of ±0.2μm, or the dimensional accuracy error is greater than ±0.02mm. The adjustment range of the parameters in the iterative adjustment is linearly related to the magnitude of the deviation.

[0080] The quantitative judgment criteria for trimming deviation clearly define the deviation thresholds for trimming effect indicators, enabling precise and quantifiable judgment of deviations: ① Surface roughness deviation threshold: When the surface roughness after trimming exceeds the preset threshold ±0.2μm, it is judged that a trimming deviation exists; ② Dimensional accuracy deviation threshold: When the dimensional accuracy error after trimming is greater than ±0.02mm, it is judged that a trimming deviation exists. This quantitative standard provides a clear judgment basis for real-time monitoring, avoids the uncertainty of subjective judgment, and ensures the consistency and accuracy of deviation identification.

[0081] The correlation rule for parameter iterative adjustment is that the adjustment range of parameters is linearly correlated with the magnitude of the deviation value. That is, the larger the deviation value (e.g., surface roughness exceeding a preset threshold of 0.5 μm), the larger the parameter adjustment range (e.g., appropriately increasing the dressing pressure and decreasing the dressing speed); the smaller the deviation value (e.g., surface roughness exceeding a preset threshold of 0.1 μm), the smaller the parameter adjustment range. This linear correlation adjustment rule ensures the stability and rationality of parameter adjustment, avoiding new deviations or component damage due to excessive adjustment range, while also ensuring adjustment efficiency, allowing the dressing effect to quickly return to the acceptable range.

[0082] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope described in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An adaptive surface trimming method for additively manufactured curved surface components, characterized in that, Includes the following steps: Curved surface feature perception: three-dimensional geometric data of additively manufactured curved surface components are acquired through visual scanning or tactile feedback. The three-dimensional geometric data includes curvature, contour and surface defect distribution. Based on machine learning algorithms, the surface type, defect type and material properties are automatically identified. The surface type includes free-form surface and regular surface, the defect type includes protrusion, pit and layer texture, and the material properties include hardness and toughness. Trimming strategy generation: Establish a surface feature-trimming parameter mapping model. Based on the identified surface features and material properties, dynamically output trimming tool path, pressure, speed and tool type parameters to achieve multi-objective optimization of surface accuracy assurance, trimming time minimization and component damage risk minimization. Real-time feedback control: Force and displacement sensors monitor force and displacement signals during the trimming process to evaluate the trimming effect on surface roughness and dimensional accuracy in real time; if a trimming deviation is detected, the trimming parameters are automatically adjusted iteratively to form a closed-loop control of perception-decision-execution-feedback.

2. The adaptive surface trimming method for additively manufactured curved surface components according to claim 1, characterized in that, The visual scanning uses 3D laser scanning technology, the machine learning algorithm is a deep learning segmentation model, and the tactile feedback collects tactile data of the component surface through a contact pressure sensor array.

3. The adaptive surface trimming method for additively manufactured curved surface components according to claim 1, characterized in that, The surface feature-trimming parameter mapping model is a reinforcement learning-based decision model. The core optimization indicators of the reinforcement learning reward function are the surface accuracy achievement rate, trimming time reduction ratio, and component damage rate.

4. The adaptive surface trimming method for additively manufactured curved surface components according to claim 1, characterized in that, The criteria for determining the trimming deviation are: the surface roughness exceeds the preset threshold of ±0.2μm, or the dimensional accuracy error is greater than ±0.02mm. The adjustment range of the parameters in the iterative adjustment is linearly related to the magnitude of the deviation.

5. An adaptive surface trimming apparatus for additively manufactured curved surface components, used to execute an adaptive surface trimming method for additively manufactured curved surface components according to any one of claims 1-4, characterized in that, include: The mechanical structure is for a multi-degree-of-freedom motion platform and an end effector. The multi-degree-of-freedom motion platform is a 6-axis robot used to adapt to the spatial posture adjustment of complex curved surfaces; the end effector integrates dressing tools, sensors and adaptive adjustment mechanisms. The dressing tools include grinding heads, polishing wheels and laser dressing heads. The sensors include force sensors and vision sensors. The adaptive adjustment mechanism is a pressure compensation device. The control system comprises hardware and software. The hardware includes an industrial controller, a sensor data acquisition card, and a servo drive system. The software includes a feature recognition algorithm library, trimming parameter decision software, and a real-time monitoring interface, which supports data interaction with additive manufacturing equipment and can read component design models.

6. The adaptive surface trimming device for additive manufacturing curved surface components according to claim 5, characterized in that, The positioning accuracy of the multi-degree-of-freedom motion platform is ≤ ±0.01 mm, and the motion speed is ≥ 500 mm / s.

7. The adaptive surface trimming device for additive manufacturing curved surface components according to claim 5, characterized in that, The end effector also includes a temperature sensor for real-time monitoring of the component surface temperature during the trimming process. When the temperature exceeds a preset threshold of 80°C, the trimming parameters are automatically adjusted or the trimming is paused.

8. The adaptive surface trimming device for additive manufacturing curved surface components according to claim 5, characterized in that, The trimming tool can be quickly switched via a quick-change interface with a switching time of ≤30s. The quick-change interface has a self-locking function with a positioning accuracy of ≤±0.005mm.

9. The adaptive surface trimming device for additive manufacturing curved surface components according to claim 5, characterized in that, The software of the control system also includes a data storage module for recording surface feature data, trimming parameters, trimming effect detection data, and equipment operating status data.

10. An adaptive surface trimming device for additive manufacturing curved surface components according to claim 5, characterized in that, The control system supports remote monitoring and parameter adjustment, and data transmission is achieved through industrial Ethernet or 5G communication modules with a transmission delay of ≤10ms.

Images