Potential collision detection and optimization method for processing path trajectory simulation

Abstract

The invention provides a potential collision detection and optimization method for processing path trajectory simulation, which generates a virtual boundary through an ultrasonic dynamic sound field, and adjusts the shape of the boundary in real time to adapt to workpiece and environment changes. And a multi-path signal fusion and adaptive filtering technology is adopted, so that the detection precision is improved, and interference is reduced. Early warning and safety protection are realized through double-layer virtual boundary design; and the path is dynamically optimized by combining heuristic search and a fast path planning algorithm, and the collision avoidance requirement and the processing efficiency are balanced. Deep learning and reinforcement learning models are introduced, and self-adaptive optimization is performed on boundaries and paths based on historical data and real-time feedback. The method is suitable for multi-axis linkage machining and complex curved surface machining, and has high precision, strong anti-interference capability and wide applicability.

Description

Technical Field

[0001] The present invention belongs to the technical field of numerical control machining, and in particular relates to a potential collision detection and optimization method for machining path trajectory simulation. Background Art

[0002] In modern industrial machining, multi-axis machining and complex surface machining are becoming increasingly mainstream as product complexity and precision continue to increase. However, accurate simulation of machining paths and collision detection in these machining tasks remain important but challenging technical issues. Especially in multi-tool collaborative or dynamic machining scenarios, the potential collision risk between tools, workpieces, and fixtures increases significantly, making traditional path planning and collision avoidance methods difficult to meet real-time and accuracy requirements.

[0003] 1. Traditional methods often use predefined virtual boundaries for collision detection. These boundaries are typically based on the initial workpiece geometry and are difficult to adapt to workpiece deformation, fixture movement, and other environmental changes during machining. Because virtual boundaries are static, they are prone to false alarms or detection blind spots in complex machining tasks, impacting both efficiency and safety.

[0004] 2. Path planning is often based on fixed optimization algorithms (such as the shortest path algorithm), which cannot respond to dynamic changes in the machining environment in real time. In complex scenarios (such as multi-axis machining or collaborative machining), path planning algorithms often ignore the balance between collision avoidance priority and machining efficiency, which can lead to reduced efficiency or collision avoidance failure.

[0005] 3. In complex processing environments, ultrasonic signals are susceptible to interference from scattering, reflection, and ambient noise, resulting in reduced clarity of virtual boundaries. Current technologies often rely on simple filtering or signal enhancement methods, which cannot effectively solve the problems of multipath signal fusion and noise subtraction in complex environments.

[0006] 4. Existing technologies for virtual boundary adjustment and path planning are mostly based on rules or predefined parameters, lacking dynamic learning and optimization capabilities. The utilization rate of historical data is low, making it difficult to form a systematic adaptive optimization mechanism.

[0007] In summary, existing technologies have the following deficiencies in the fields of machining path simulation and collision detection:

[0008] The virtual boundary has weak dynamic adjustment capabilities: it is difficult to adapt to the dynamic changes of the workpiece and environment in complex processing tasks.

[0009] Insufficient real-time performance of path planning: Unable to balance collision avoidance requirements and machining efficiency, especially in highly dynamic multi-axis machining scenarios.

[0010] Limited anti-interference capability: Affected by environmental noise and signal interference, the accuracy of virtual boundary detection is insufficient.

[0011] Insufficient intelligence and adaptability: Lack of data-driven intelligent optimization mechanism, unable to cope with complex and changing processing requirements. Summary of the Invention

[0012] Therefore, the purpose of the present invention is to provide a potential collision detection and optimization method for machining path trajectory simulation, which significantly improves the accuracy, real-time nature and intelligence level of the machining path through dynamic virtual boundary adjustment, real-time path optimization and artificial intelligence-assisted optimization. It is particularly suitable for high-dynamic task scenarios such as multi-axis linkage machining, complex surface machining and collaborative machining.

[0013] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:

[0014] A potential collision detection and optimization method for machining path trajectory simulation realizes collision detection and avoidance in the machining path through virtual boundary construction based on ultrasonic dynamic sound field, real-time feedback control and path optimization algorithm. The method comprises the following steps:

[0015] S1. Generate a dynamic sound field in the processing area using an ultrasonic array to define a virtual boundary;

[0016] S2. The tool sensor module receives the ultrasonic signal in real time and calculates the relative distance between the tool and the virtual boundary;

[0017] S3. When the tool approaches or touches the virtual boundary, the path adjustment is triggered or the machining operation is paused;

[0018] S4. Dynamically adjust the virtual boundary shape based on real-time feedback to adapt to changes in the workpiece, fixture, and processing path.

[0019] The dynamic sound field is generated by superposition of multi-frequency ultrasonic waves, and the clarity of the virtual boundary and the ability to resist scattering interference are enhanced by adjusting the emission frequency, phase and intensity.

[0020] Among them, multi-path signal fusion technology is used to analyze the time delay, intensity attenuation and frequency characteristics of the ultrasonic echo signal, and to filter out the scattering and reflection interference, thereby improving the detection accuracy of the virtual boundary.

[0021] Among them, an adaptive filtering algorithm is used to dynamically optimize real-time signal processing, and the environmental noise signal is removed and the recognition ability of effective signals is enhanced by adjusting the filter parameters in real time.

[0022] Among them, sound-absorbing materials are arranged in the non-processing area to reduce the interference of multiple reflections of ultrasound and reduce the impact of complex processing environment on virtual boundary detection.

[0023] The virtual boundary includes two layers of boundaries:

[0024] The first layer boundary is used to detect the tool approaching the dangerous area and trigger path deceleration or adjustment;

[0025] The second layer boundary is used to detect the tool touching the boundary and trigger processing pause and alarm signal.

[0026] Wherein, a dynamic path optimization algorithm is adopted to dynamically re-plan the tool path according to the real-time distance between the tool and the virtual boundary, the motion trend and the processing strategy; the path optimization includes:

[0027] The tool slows down and adjusts the path direction when approaching the first layer boundary;

[0028] The tool triggers processing pause and generates a safe path around the collision area when entering the collision area.

[0029] Wherein, an environmental active compensation algorithm is used to generate a noise feature model before processing, recording the scattering characteristics of the workpiece, fixture and tool; during processing, the real-time signal is compared with the noise feature model, and the interference signal is deducted to improve the detection accuracy.

[0030] Wherein, artificial intelligence algorithm is used to learn and optimize the processing history data, including a virtual boundary precision optimization model based on deep learning and a dynamic path planning model based on reinforcement learning, to improve processing efficiency and collision avoidance capability.

[0031] Wherein, the method is suitable for multi-axis linkage processing scene and complex curved surface processing, and through real-time reconstruction of virtual boundary and path adjustment strategy, the accuracy and efficiency of the processing path are guaranteed, and the processing path is dynamically optimized in the multi-tool or multi-robot cooperation environment to avoid path interference.

[0032] Compared with the prior art, the present application has the beneficial effects that:

[0033] 1. Dynamic adjustment of virtual boundary

[0034] Through dynamic sound field generation technology, the virtual boundary is dynamically adjusted according to real-time feedback data, adapting to workpiece deformation, fixture movement and environmental changes. Precise fitting and real-time updating of the virtual boundary are realized without manual intervention.

[0035] The first layer boundary realizes early warning function, triggers path deceleration or adjustment in advance, and improves the reaction speed of the system. The second layer boundary is used for collision area detection to ensure safe shutdown and alarm function when the tool touches the boundary. The accuracy and stability of the virtual boundary are improved to adapt to the dynamic processing environment. A layered protection mechanism is provided to balance the processing efficiency and safety.

[0036] 2. Dynamic path optimization improves processing efficiency and safety

[0037] Utilizing heuristic search and fast random tree algorithms, the optimal collision avoidance path is quickly generated in complex machining environments. This achieves high path dynamics and enables real-time adjustment of tool direction and speed.

[0038] Through innovative optimization formulas, the system comprehensively considers multiple factors, including modular design, material properties, path efficiency, and processing costs. It adjusts optimization target weights based on real-time feedback, balancing collision avoidance requirements with processing efficiency. This intelligently optimizes the processing path, ensuring collision avoidance while maximizing processing efficiency. It demonstrates excellent adaptability in complex multi-axis machining and dynamic collaboration scenarios.

[0039] 3. Intelligent virtual boundary and path planning optimization

[0040] Using deep learning models, we predict virtual boundary adjustments, improving boundary definition accuracy in complex scenarios. Boundary adjustments are intelligent and automated, reducing the need for manual intervention.

[0041] The reinforcement learning model dynamically optimizes path planning strategies based on real-time machining data. It learns and iterates on collision avoidance experience during machining to optimize the success rate and response speed of path planning.

[0042] Virtual boundaries and path planning achieve intelligent and adaptive capabilities, significantly reducing machining errors and demonstrating higher reliability and flexibility in dynamic and complex machining scenarios.

[0043] 4. Enhanced anti-interference ability

[0044] Multipath signal fusion technology is used, combined with an adaptive filtering algorithm, to effectively filter scattered signals and environmental noise. An active environmental compensation algorithm is used to reduce noise interference and improve the detection accuracy of virtual boundaries.

[0045] Sound-absorbing materials are placed in non-processing areas to reduce multiple reflection interference and optimize the sound field distribution. Dynamic adjustment of the emission frequency and phase further improves the stability of the sound field.

[0046] The clarity of virtual boundaries and anti-interference capabilities are significantly improved. The system can adapt to complex processing environments with high noise and high vibration, and its operation is more stable.

[0047] 5. Wide applicability

[0048] Supports complex surface processing, dynamically adjusts virtual boundaries to fit the workpiece shape, and achieves high-precision multi-axis linkage processing path planning.

[0049] In multi-robot or multi-tool collaborative processing, path planning is dynamically optimized to avoid path conflicts and resource waste.

[0050] The system is suitable for various processing scenarios, including simple path processing, complex curved surface processing, and dynamic cooperative processing. It has high expansibility and compatibility in a wide range of applications.

[0051] 6. Improved processing safety

[0052] The first layer of boundaries enables early warning and triggers deceleration and path adjustment to avoid potential collisions. The second layer of boundaries ensures precise protection of collision areas and timely shutdown with alarms. When the tool enters the collision range, path re-planning is triggered to quickly generate a safe path. This avoids equipment damage or workpiece scrap due to abnormal processing.

[0053] The processing process is safer and more reliable, reducing the probability of accidents. It reduces equipment wear and tear and workpiece scrap rate, improving processing quality.

[0054] 7. Efficient data utilization and system optimization

[0055] Through deep learning training of processing history data, collision avoidance and optimization experience are extracted to improve the efficiency and accuracy of future processing tasks.

[0056] The system dynamically adjusts virtual boundaries and path optimization strategies based on real-time feedback, achieving intelligent closed-loop control of the processing process.

[0057] The data-driven optimization mechanism improves the intelligence and efficiency of the system. The system can continuously learn and optimize itself, maintaining high efficiency for a long time.

[0058] In summary, the present invention achieves the following significant benefits through dynamic sound field generation, real-time virtual boundary adjustment, path optimization algorithms, and artificial intelligence assisted optimization:

[0059] Improves the accuracy and adaptability of virtual boundaries.

[0060] Achieves intelligent optimization of path planning, balancing collision avoidance requirements and processing efficiency.

[0061] Enhances anti-interference ability and environmental adaptability.

[0062] Provides a safe and efficient processing protection mechanism.

[0063] Has wide applicability, supporting complex processing scenarios.

[0064] Achieves the intelligence and adaptability of the system, enabling continuous optimization of processing performance.

[0065] These effects make the present invention suitable for high dynamic and complex processing tasks, with important technical value and application prospects. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be derived from these drawings without inventive effort. Among them:

[0067] Figure 1 Schematic diagram of the method flow of the present invention;

[0068] Figure 2 Schematic diagram of information flow in the method of the present invention. DETAILED DESCRIPTION

[0069] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0070] See also Figure 1 and 2 A potential collision detection and optimization method for machining path trajectory simulation implements collision detection and avoidance within the machining path through virtual boundary construction based on ultrasonic dynamic sound fields, real-time feedback control, and a path optimization algorithm. This method is suitable for multi-axis machining scenarios and complex surface machining. By reconstructing virtual boundaries and implementing path adjustment strategies in real time, it ensures machining path accuracy and efficiency. Furthermore, it dynamically optimizes machining paths in multi-tool or multi-robot collaborative environments to avoid path interference.

[0071] Specific steps:

[0072] S1. Dynamic Sound Field Generation and Virtual Boundary Definition

[0073] Dynamic Acoustic Field Construction: An ultrasonic array generates a dynamic acoustic field within the processing area, defining the virtual boundaries of the processing path. This dynamic acoustic field is generated by superimposing multiple ultrasonic frequencies. By adjusting the emission frequency, phase, and intensity, the clarity of the virtual boundaries and the ability to resist scattering interference are enhanced.

[0074] Interference Signal Filtering and Environmental Optimization: Multipath signal fusion technology analyzes the time delay, intensity attenuation, and frequency characteristics of ultrasonic echo signals to filter out scattered and reflected interference, thereby improving virtual boundary detection accuracy. An adaptive filtering algorithm dynamically optimizes real-time signal processing, removing environmental noise signals through real-time adjustment of filter parameters to enhance the ability to identify valid signals. An active environmental compensation algorithm generates a noise signature model before processing, recording the scattering characteristics of the workpiece, fixture, and tool. During processing, the real-time signal is compared with the noise signature model to remove interference signals and improve detection accuracy.

[0075] Physical environment optimization: In the non-processing area, arrange sound-absorbing materials to reduce the interference of multiple reflections of ultrasonic waves and reduce the impact of complex processing environment on virtual boundary detection.

[0076] Hierarchical design of virtual boundaries:

[0077] First layer boundary: Trigger path deceleration or adjustment when detecting the proximity of the tool to the dangerous area;

[0078] Second layer boundary: Trigger processing pause and send alarm signal when detecting the tool touching the boundary.

[0079] Further, the specific implementation method of S1 is:

[0080] S1.1 Generation of dynamic sound field

[0081] ① Arrangement and control of ultrasonic array

[0082] Array arrangement: Arrange ultrasonic transmitter and receiver arrays at key positions in the processing area (such as workpiece edges and tool path around). According to the complexity of the processing task, two-dimensional or three-dimensional array layout can be selected.

[0083] Transmitting parameter design: Set the working frequency of ultrasonic waves (20-100 kHz), and select the appropriate frequency range according to the workpiece material and processing environment.

[0084] Phase control: Use phase adjustment to form directional beams to improve the coverage range and energy concentration of the sound field.

[0085] ② Multi-frequency ultrasonic wave superposition to generate sound field

[0086] Use multi-group frequency superposition technology to make ultrasonic waves of different frequencies form multiple independent sound fields in space. Each frequency corresponds to a sub-sound field, which is used to improve the signal redundancy and detection accuracy. Signal decoupling is achieved through frequency separation technology for subsequent processing.

[0087] ③ Dynamic adjustment of sound field distribution

[0088] According to the three-dimensional model of the workpiece and the fixture, use digital signal processor (DSP) to adjust the transmission frequency, phase and intensity of ultrasonic waves in real time, to ensure the accurate matching of the sound field and the processing area.

[0089] The parameters of dynamic adjustment include: transmission angle, covering complex curved surfaces or edges. Intensity gain, adapting to the reflection characteristics of different materials.

[0090] S1.2 Definition and hierarchy of virtual boundaries

[0091] ① Preliminary construction of virtual boundaries

[0092] Define the scope of the virtual boundary based on the sound field distribution and processing tasks:

[0093] Inner boundary (first layer): It is a warning line that marks the dangerous area. When the tool approaches this boundary, it triggers deceleration or path adjustment.

[0094] Outer boundary (second layer): identifies the collision area. When the tool touches this boundary, the machining is paused.

[0095] ②Boundary layer optimization

[0096] The roles of boundary layers are dynamically assigned based on machining complexity and precision requirements: the first boundary layer is suitable for rapid collision avoidance and early warning. The second boundary layer provides hard constraints to ensure the safety of equipment and workpieces.

[0097] Define boundary parameters: First layer distance threshold: set according to processing speed and path complexity, usually 1-2 times of tool diameter. Second layer distance threshold: slightly smaller than the virtual boundary trigger condition, 0.5-0.8 times.

[0098] S1.3 Filtering of interference signals

[0099] ①Multi-path signal fusion technology

[0100] Receive ultrasonic signals from multiple receivers and analyze path differences. By calculating time delays and comparing intensities, the system extracts valid echo signals and filters out scattered and multi-reflected signals.

[0101] ②Real-time filter optimization

[0102] An adaptive filtering algorithm is applied to adjust the filter parameters based on the real-time acquired signal. Initially, the filter weights are set based on ambient noise and echo characteristics. Dynamically, the filter parameters are updated in real time to eliminate new interference during the machining process.

[0103] S1.4 Active Environmental Compensation and Physical Optimization

[0104] ①Environmental active compensation algorithm

[0105] Run the noise learning mode before machining to record the scattered signal characteristics of the workpiece, fixture, and machining environment. This generates a noise signature model that includes the frequency and intensity distribution of interfering signals. During machining, the real-time acquired signal is compared with the noise signature model to automatically deduct interference components.

[0106] ②Arrangement of sound-absorbing materials

[0107] Place sound-absorbing materials on non-working surfaces in the machining area (such as the back of fixtures and machine tool walls) to absorb multiple reflected ultrasonic signals. Optimize the layout and thickness of the sound-absorbing materials to reduce noise propagation paths.

[0108] S1.5 Self-correction and real-time optimization of the sound field

[0109] ① Real-time self-correction mechanism

[0110] Based on ultrasonic sensor data, the system analyzes changes in the sound field in real time. If uneven sound fields or areas of interference are detected, the system dynamically adjusts the power and direction of the transmitter to ensure that the virtual boundary remains stable.

[0111] ②Reconstruction of virtual boundaries

[0112] During machining, the acoustic field distribution is reconstructed in real time based on workpiece deformation or fixture movement. Reconstruction steps: rescan the workpiece surface, update the acoustic field distribution parameters, and adjust the boundary shape in real time.

[0113] The implementation of step S1 combines dynamic sound field generation, signal processing, and virtual boundary construction to form a high-precision virtual boundary system that adjusts in real time, providing reliable basic data input for subsequent tool monitoring and path optimization.

[0114] S2. Calculation of the distance between the tool and the virtual boundary

[0115] The tool sensor module receives ultrasonic signals in real time and calculates the relative distance between the tool and the virtual boundary. Based on the distance between the tool and the boundary, the risk level is dynamically determined and different collision avoidance strategies are triggered.

[0116] Furthermore, the specific implementation method of S2 is:

[0117] S2.1 Ultrasonic signal acquisition and fusion

[0118] ① Operation of ultrasonic array

[0119] An ultrasonic array (transmitter and receiver) is arranged in the machining area to determine the position of the tool relative to the virtual boundary by transmitting ultrasonic signals and receiving echo signals.

[0120] The sensor on the tool receives the echo signal in real time and uses it to calculate the distance from the boundary.

[0121] The propagation time difference of the ultrasonic signal (time delay method) is used to preliminarily calculate the tool position. This is based on conventional formulas for ultrasonic propagation (such as d = v·Δt, which is common knowledge and will not be elaborated on).

[0122] ②Multi-path signal fusion

[0123] Fusion of signals collected by multiple receivers eliminates interference caused by scattering or reflections. Valid signals are filtered by signal strength attenuation and path time difference. Multipath fusion technology combines signals from different receivers using a weighted average method to improve tool position accuracy.

[0124] ③Signal processing and real-time update

[0125] Use dynamic filters to adjust filter parameters in real time according to the noise environment and processing conditions: filter out noise and invalid signals, and retain valid echo information.

[0126] Combined with the active environmental compensation algorithm, a noise characteristic model is generated before processing; during processing, the echo signal and noise characteristics are compared in real time to eliminate interference.

[0127] S2.2 Distance calculation and risk level assessment

[0128] ① Calculation of the distance between the tool and the virtual boundary

[0129] After the tool position is determined, the distance between it and each point on the virtual boundary is calculated in real time, and the closest distance is selected:

[0130] d min =min{d1,d2,…,d n}; where d min : The shortest distance between the tool and the virtual boundary. d1,d2,…,d n : The distance from the tool to each point of the virtual boundary.

[0131] ② Dynamic classification of risk levels

[0132] According to d min The layered design of virtual boundaries is divided into the following risk levels:

[0133] Safety range: When d min >D1, the tool is away from the first layer boundary, and the system maintains normal processing.

[0134] Warning range: When D2 <d min ≤D1, the tool approaches the first layer boundary, triggering path deceleration or adjustment.

[0135] Collision range: When d min ≤D2, the tool enters the second layer boundary area, the system suspends processing and alarms.

[0136] ③ Dynamic update of risk assessment During the processing, the risk level is updated in real time, and the boundary parameters (such as distance thresholds D1 and D2) are adjusted according to changes in the processing environment to ensure that the boundary always adapts to the current task requirements.

[0137] S2.3 Optimization formula Q o Dynamic application of pt

[0138] ①Introduction of optimization formula

[0139] Based on the real-time distance calculation, the optimization formula Q is introduced opt, conduct a comprehensive assessment of the tool's operating status:

[0140] Where M is the modular design coefficient, a dimensionless value that represents the contribution of modular design to machining path flexibility and is calculated based on machining parameters (such as the number and size of workpiece modules). α is the path sensitivity, a dimensionless value that represents the impact of path design on machining efficiency and is typically related to path complexity and machining direction. β is the material property influence factor, a dimensionless value that represents the influence of material properties (such as strength and elasticity) on tool path planning. E is the material elastic modulus, expressed in Pascals (Pa), extracted from the material database and used to describe the material's resistance to deformation. n is the nonlinear loading adjustment factor, a dimensionless value that represents the correction of material properties under nonlinear stress loading conditions. V is the machining area volume, expressed in cubic meters (m3), which is the three-dimensional volume of the machining area calculated in conjunction with the tool path. γ is the path efficiency factor, a dimensionless value that represents the dynamic changes in path planning efficiency. C is the unit machining cost, expressed in monetary units (e.g., yuan / meter). The cost of tool path consumption is calculated in real time. τ is the remaining tool life, expressed in hours (h), which is estimated based on the tool's accumulated operating time and current load state. m: Life sensitivity index, dimensionless, indicating the sensitivity of tool performance to changes in remaining life. η: Friction loss factor, dimensionless, indicating the energy loss during machining.

[0141] The role of the optimization formula: Dynamically evaluate tool performance, including comprehensive optimization of path planning, machining efficiency and material properties.

[0142] ③Dynamic application of optimization formula

[0143] During the processing, the real-time parameters are substituted into the formula to calculate Q opt . Combined with Q opt The calculation results are used for path optimization:

[0144] If Q opt ≥Q target , maintain the current path.

[0145] If Q opt target , trigger path adjustment to optimize tool running direction and speed.

[0146] S2.4 System Feedback and Path Adjustment

[0147] ① Feedback mechanism

[0148] Risk level and Q opt The calculation results are fed back to the path optimization module, triggering corresponding adjustments:

[0149] Safety margin: No adjustment.

[0150] ​Warning range: Slow down or adjust direction.

[0151] Collision range: Pause processing and replan the path.

[0152] ② Dynamic adjustment of virtual boundaries

[0153] According to the relative position of the tool and the boundary and Q opt The virtual boundary shape is dynamically adjusted to adapt to changes in the workpiece, fixture and processing path.

[0154] Step S2 is achieved by ultrasonic signal acquisition, distance calculation and optimization formula Q opt The dynamic application of the system enables real-time assessment of tool status and path adjustment. This step, combined with risk classification and system feedback mechanisms, provides precise input data for subsequent path optimization.

[0155] S3. Dynamic Path Optimization and Collision Avoidance

[0156] Path adjustment mechanism: When the tool approaches the virtual boundary, the path adjustment is triggered or the machining operation is paused.

[0157] Path optimization includes the following two situations:

[0158] The tool approaches the first layer boundary: the system triggers the path to slow down or adjust the path direction;

[0159] Tool enters collision zone: Triggers a machining pause and generates a safe path around the collision zone.

[0160] Dynamic Path Optimization Algorithm: A lightweight dynamic path optimization algorithm, combined with a heuristic search algorithm and fast path planning, reduces computational complexity. Based on the tool's real-time distance from the virtual boundary, motion trends, and machining strategies, it dynamically generates the optimal path in high-risk areas.

[0161] Furthermore, the specific implementation method of S3 is:

[0162] S3.1 Triggering conditions for path adjustment

[0163] ① Real-time monitoring and judgment

[0164] According to the minimum distance d between the tool and the virtual boundary calculated in step S2 min ,Determine the conditions for triggering path adjustment or collision avoidance operation:

[0165] Safety margin: d min >D1, the tool is away from the boundary and no path adjustment is required.

[0166] Warning range: D2 <d min ≤D1, the tool approaches the first layer boundary, triggering deceleration or path adjustment.

[0167] Collision range: d min ≤ D2, the tool enters the second layer boundary area, triggering a processing pause and re-planning the path.

[0168] ② Dynamic risk assessment

[0169] At each monitoring period, the current path risk is assessed based on the relative relationship between the tool position and the virtual boundary, combined with path complexity, processing speed, and environmental changes.

[0170] Adjust the priority of path optimization dynamically through risk level.

[0171] S3.2 Dynamic path optimization algorithm

[0172] ① Path optimization principles

[0173] Dynamic path optimization needs to consider the following principles:

[0174] Collision avoidance priority: when the tool approaches the virtual boundary, prioritize path adjustment to ensure safety.

[0175] Efficiency balance: while avoiding collisions, try to maintain the efficiency of the path as much as possible (such as the shortest path, lowest energy consumption).

[0176] Real-time: the optimization process needs to meet real-time response requirements to ensure that the system quickly adapts to changes during processing.

[0177] ② Implementation method of path optimization

[0178] (1) Heuristic search algorithm

[0179] Use the optimized A*(A-star) algorithm to quickly generate a new path that meets the collision avoidance conditions:

[0180] Define the path cost function f(n): f(n) = g(n) + h(n); where g(n): the cumulative cost of the current path. h(n): estimated cost from the current node to the target node (heuristic function).

[0181] During the search process, dynamically increase the collision avoidance priority weight to ensure that the path is away from the high-risk area of the virtual boundary.

[0182] (2) Rapid path planning (RRT)

[0183] In complex multi-axis machining scenarios, use the Rapidly-exploring Random Tree (RRT) algorithm to generate a collision avoidance path: randomly sample path points, expand the tree structure, and find a safe path from the starting point to the target point. Optimize path smoothness to ensure that the path meets the processing accuracy requirements.

[0184] (3) Dynamic constraint adjustment

[0185] Dynamically adjust path optimization constraints based on real-time feedback, such as safety range within the processing area, tool speed limit, etc.

[0186] ③ Combined with the optimization formula Q opt

[0187] In the path optimization process, the optimization formula Q opt Dynamic introduction to evaluate the comprehensive performance of path adjustment;

[0188] Optimization goal: select Q opt The largest path plan ensures a balance between collision avoidance and efficiency.

[0189] S3.3 Path Adjustment Strategy

[0190] ①Route adjustment within the warning range

[0191] Deceleration processing: When the tool enters the warning range, the system dynamically adjusts the tool feed speed to reduce the risk of collision during processing.

[0192] Fine-tune direction: Through the path fine-tuning algorithm, the tool's running direction is optimized to stay as far away from the virtual boundary as possible while maintaining the continuity of the machining path.

[0193] ②Path adjustment within the collision range

[0194] Pause operation: When the tool enters the collision range, the processing is immediately paused to ensure the safety of the workpiece and equipment.

[0195] Path Replanning: The system triggers the path replanning module to generate a new, safe path based on the current tool position and machining target. Path replanning combines heuristic search algorithms (such as A*) and random sampling (such as RRT) to quickly find a collision-avoiding path.

[0196] ③ Selection of multi-path solutions

[0197] During the path adjustment or replanning process, the system generates multiple candidate paths and opt The calculation result of the formula selects the path with the best comprehensive performance.

[0198] S3.4 System feedback and closed-loop adjustment

[0199] ①Real-time feedback mechanism

[0200] After the path adjustment, the tool position, operating status, and other information are fed back to the virtual boundary adjustment module in real time (step S4) to further optimize the shape and distribution of the boundary. If high-risk areas still exist after the path adjustment, the next round of optimization is triggered.

[0201] ② Closed-loop adjustment mechanism

[0202] During the machining process, the path optimization parameters (such as tool speed, path curvature limit, etc.) are dynamically adjusted to improve the adaptability of the collision avoidance operation. The key data during the adjustment process is recorded and used for subsequent training of the reinforcement learning model (see step S5).

[0203] Step S3 implements closed-loop control of dynamic path optimization and collision avoidance operations, from trigger condition determination to path adjustment and replanning, and then to system feedback and closed-loop adjustment, ensuring safe and efficient tool operation and providing a highly real-time and highly adaptable solution for collision avoidance optimization in complex machining environments.

[0204] S4. Dynamic Adjustment of Real-time Virtual Boundaries

[0205] The virtual boundary shape is dynamically adjusted based on real-time feedback during the machining process to adapt to changes in the workpiece, fixtures, and machining path. The intersection of the tool path and the virtual boundary is predicted in advance, and the virtual boundary distribution of the machining area is dynamically optimized through the boundary reconstruction mechanism.

[0206] Furthermore, the specific implementation method of S4 is:

[0207] S4.1 Real-time feedback and virtual boundary adjustment trigger

[0208] ①Feedback data collection

[0209] The real-time data obtained from steps S2 and S3 include:

[0210] The closest distance d between the tool and the virtual boundary min .

[0211] The offset of the tool path.

[0212] Environmental parameters (such as temperature, vibration, processing speed).

[0213] The feedback data is used to determine whether the virtual boundary needs to be adjusted.

[0214] ② Trigger condition determination

[0215] Static error detection: Triggers adjustments when the tool deviates from the preset path, or when the boundary shape does not match the actual position of the workpiece / fixture.

[0216] Dynamic environment changes: When workpiece deformation, fixture movement, or external environment changes (such as increased noise) during machining exceed the set threshold, dynamic boundary reconstruction is triggered.

[0217] S4.2 Virtual Boundary Adjustment Mechanism

[0218] ① Preliminary shape correction

[0219] Adjust the shape of the virtual boundary based on real-time data:

[0220] Geometric correction: Redefine boundary positions based on the latest 3D model of the workpiece or fixture.

[0221] Layered optimization: The distance thresholds between the first layer (safety boundary) and the second layer (collision boundary) are dynamically updated to adapt to the current processing environment.

[0222] ②Dynamic reconstruction algorithm

[0223] Boundary reconstruction formula:

[0224] The new virtual boundary shape is determined by the following formula: new (x,y,z)=B init (x,y,z)+ΔB(x,y,z); where, B new (x,y,z): The adjusted boundary shape. B init (x,y,z): Initial boundary shape. ΔB(x,y,z): Adjustment amount, determined by real-time feedback data.

[0225] Calculation of adjustment ΔB: ΔB is calculated by combining the relative position offset between the tool and the boundary and the environmental change: ΔB = κ·d min +ψ·ΔE; where κ is the sensitivity coefficient of position offset. min : Minimum distance between the tool and the boundary. ψ: Weight coefficient for environmental change. ΔE: Amount of environmental change (e.g., vibration amplitude, temperature change).

[0226] ③Multi-layer boundary adjustment

[0227] First layer boundary (safety boundary): The distance threshold D1 is dynamically adjusted according to the processing speed to avoid excessive false alarms.

[0228] Second layer boundary (collision boundary): The threshold D2 is updated in real time according to the workpiece deformation or fixture movement to ensure accurate triggering when the tool contacts.

[0229] S4.3 Reinforcement Learning-Assisted Dynamic Adjustment

[0230] ①Introduction of reinforcement learning model

[0231] Optimize virtual boundary adjustments through reinforcement learning algorithms (such as Deep Q Network, DQN):

[0232] Input: Real-time feedback data (tool position, boundary offset, environmental changes).

[0233] Output: Boundary adjustment amount ΔB.

[0234] Learning objectives: Minimize the deviation between the virtual boundary and the actual processing area. Optimize the efficiency of boundary shape adjustment.

[0235] ②Learning mechanism

[0236] Training data: extracted from historical processing paths and adjustment records, including boundary adjustment data under different workpiece shapes and environmental conditions.

[0237] Real-time learning: Combines feedback data from the current processing task to continuously optimize boundary adjustment strategies.

[0238] S4.4 Adaptive Virtual Boundary Optimization

[0239] ① Parameter adaptive adjustment

[0240] Adaptively optimize the weight coefficients κ and ψ in the adjustment formula:

[0241] When the tool is running stably, reduce the weight of κ to reduce unnecessary adjustments.

[0242] When the environment changes significantly, the weight of ψ is increased to give priority to responding to environmental disturbances.

[0243] ②Boundary dynamic update frequency

[0244] Dynamically adjust the update frequency according to the complexity of the processing task:

[0245] High-complexity tasks: The frequency is increased (e.g., 10 updates per second).

[0246] Low-complexity tasks: The frequency is reduced (such as 2 updates per second).

[0247] S4.5 System Feedback and Closed-Loop Control

[0248] ① Feedback mechanism

[0249] After the adjustment is completed, the new virtual boundary shape is fed back to the path optimization module (step S3). The path optimization solution is re-evaluated based on the updated boundary to ensure the matching of the boundary and the path.

[0250] ② Closed-loop control

[0251] Achieve closed-loop control of virtual boundary adjustment and path optimization: The virtual boundary adjustment results directly affect the weight distribution of path optimization. Path optimization feedback data is used to further optimize the boundary shape.

[0252] Step S4 implements dynamic adjustment of the virtual boundary. It optimizes the boundary shape through real-time feedback, dynamic reconstruction formulas, and reinforcement learning models. Combined with adaptive parameter adjustment and closed-loop control mechanisms, it provides high-precision, high-dynamic boundary definition and adjustment capabilities for complex processing scenarios, ensuring the safety and efficiency of the processing path.

[0253] S5. AI-assisted optimization

[0254] Through artificial intelligence algorithms, we learn and optimize historical processing data to further improve the accuracy of virtual boundary detection and path planning efficiency:

[0255] A deep learning-based virtual boundary accuracy optimization model for dynamic error compensation and noise filtering;

[0256] A dynamic path planning model based on reinforcement learning is used for real-time optimization of path collision avoidance strategies.

[0257] Establish a data collection and model iteration mechanism to dynamically update the artificial intelligence model during the processing.

[0258] Furthermore, the specific implementation method of step S5 is:

[0259] S5.1 Data acquisition and preprocessing

[0260] ①Data collection

[0261] The following data is collected during the machining process: the relative distance between the tool position and the virtual boundary; virtual boundary adjustment records, including the initial boundary shape and the dynamically adjusted shape; optimization records during the path planning process, including trigger conditions, optimized path, and time consumption; dynamic changes in environmental parameters (such as vibration, temperature, and noise interference); and machining result evaluation indicators (such as collision avoidance success rate, machining efficiency, and energy consumption).

[0262] ②Data classification and labeling

[0263] The collected data is classified according to the type of processing task, such as simple path processing, multi-axis complex surface processing, collaborative robot processing, etc.

[0264] Label key events in the data, such as:

[0265] Trigger collision avoidance adjustment: Record when the tool approaches the virtual boundary.

[0266] Path optimization completed: recorded when the optimized path is successfully generated.

[0267] Boundary adjustment completed: the shape change record after the virtual boundary is reconstructed.

[0268] ③Data preprocessing

[0269] Remove abnormal data, such as unreasonable records caused by noise interference or sensor errors. Normalization processing standardizes different physical quantities (such as distance, time, and energy consumption) to adapt to the artificial intelligence model.

[0270] S5.2 Training and Optimization of Deep Learning Models

[0271] ① Deep learning model selection

[0272] Use lightweight neural network structure for virtual boundary optimization.

[0273] Input: Machining history data, including tool positions, boundary adjustment records and environmental parameters.

[0274] Output: Boundary adjustment ΔB(x,y,z) or boundary reconstruction suggestion.

[0275] ②Model training

[0276] Training objectives:

[0277] Minimize virtual boundary deviation:

[0278] Among them, B pred : The boundary shape predicted by the model. B true : The actual adjusted boundary shape. N: The number of training samples.

[0279] Training data: The boundary adjustment data recorded during the processing is used as the training set. Simulated noise and random interference are added to enhance the robustness of the model.

[0280] Training method: Use GPU acceleration for batch training, and the optimization algorithm uses the Adam optimizer.

[0281] ③Model optimization and verification

[0282] After optimizing the model on the training set, the validation set data is used to test the model performance to ensure its adaptability in different processing scenarios.

[0283] Model optimization direction: Improve prediction accuracy in complex geometric shape processing. Reduce response time during boundary adjustment.

[0284] S5.3 Design and implementation of reinforcement learning model

[0285] ① Reinforcement learning framework

[0286] Use deep reinforcement learning (such as DDPG or PPO) to optimize the path planning policy.

[0287] State space: includes tool position, virtual boundary shape, current path direction and environmental parameters.

[0288] Action space: including path direction adjustment, speed adjustment or re-planning of the path.

[0289] Reward function:

[0290] Reward for successful collision avoidance: R success=+10;

[0291] Punish collision: R collision =-20;

[0292] Reward Path Efficiency: Where, Δt: the actual time of path planning. target : Target planning time.

[0293] ②Training and updating

[0294] Simulation environment: Use virtual machining scenarios (such as complex surfaces and dynamic fixtures) to simulate path planning and collision avoidance processes under different conditions.

[0295] Training process: In the initial stage, the action space is explored using a random strategy. During the training process, the strategy is continuously updated to maximize the reward function.

[0296] ③Real-time application

[0297] After the reinforcement learning model is deployed, it receives tool position and boundary data in real time and predicts path adjustment actions. The model continuously updates its strategy based on actual machining data, enhancing its adaptability to dynamic environments.

[0298] S5.4 Adaptive Optimization Mechanism

[0299] ①Virtual boundary adaptive optimization

[0300] During the machining process, the input weights of the deep learning model are adjusted based on real-time feedback: the boundary deviation weight is increased to improve the adaptability in complex environments, and the influence of irrelevant parameters (such as noisy data in low-vibration environments) is reduced.

[0301] ②Dynamic priority adjustment of path planning

[0302] A reinforcement learning model is used to dynamically adjust the collision avoidance priority and path efficiency weights: when the tool approaches the boundary, the collision avoidance priority is increased; when the tool moves away from the boundary, the path efficiency is optimized.

[0303] ③Parameter update and model iteration

[0304] Based on historical processing data and current task performance, deep learning and reinforcement learning models are regularly updated to optimize virtual boundary adjustment rules and improve path planning response speed and accuracy.

[0305] S5.5 System Feedback and Evaluation

[0306] ① Feedback mechanism

[0307] The optimization results are fed back to the virtual boundary adjustment module (step S4) and the path planning module (step S3) to ensure that the modules operate in coordination.

[0308] ②System performance evaluation

[0309] The following metrics are used to evaluate the effectiveness of AI optimization: virtual boundary deviation rate, path planning success rate and efficiency, and overall system processing time and energy consumption.

[0310] The solution of the present invention was used for testing, and representative results are summarized as follows:

[0311] Case 1: Dynamic collision avoidance in complex surface machining

[0312] An irregularly shaped metal workpiece with a complex curved surface requires multi-axis machining. Traditional path planning methods fail to detect potential collisions between the tool and fixture during machining, resulting in workpiece failure.

[0313] Implementation of the present invention:

[0314] A virtual boundary is generated through the ultrasonic dynamic sound field, and the boundary dynamically fits the surface shape, monitoring the distance between the tool and the workpiece surface in real time.

[0315] When the tool approaches the first layer of the boundary, the system triggers the path to slow down and adjust the direction; when entering the second layer of the boundary, the processing is suspended and the path is replanned.

[0316] Use deep learning models to optimize virtual boundary adjustments to avoid false alarms and misoperations.

[0317] Effect: No collision, complete processing, working hours reduced by 20%, and processing accuracy improved by 15%.

[0318] Case 2: Path optimization in multi-tool collaborative machining

[0319] In a multi-tool collaborative machining task, two tools need to process different areas simultaneously, and the paths have intersections. Traditional path planning methods lack a dynamic collision avoidance mechanism, which leads to interference between tools.

[0320] Implementation of the present invention:

[0321] Construct independent virtual boundaries for multiple tools and monitor the relative distances between tools in real time.

[0322] The reinforcement learning model dynamically optimizes the path, reallocates the tool task order, and reduces path intersections.

[0323] The double-layer boundary design ensures that the tools can maintain high efficiency while avoiding collision during collaboration.

[0324] Effect: Tool collaboration efficiency is increased by 30%, and path conflicts are completely avoided.

[0325] Case 3: Complex processing in a high-noise environment

[0326] Due to high-frequency vibration and environmental noise in the processing workshop, the ultrasonic signal is seriously interfered with. The traditional virtual boundary method cannot stably detect the tool position and the virtual boundary is blurred.

[0327] Implementation of the present invention:

[0328] Multi-path signal fusion and adaptive filtering algorithms are used to remove noise interference and enhance signal effectiveness.

[0329] Arrange sound-absorbing materials in non-processing areas to further reduce reflection interference.

[0330] The active environmental compensation algorithm is used to dynamically adjust the boundary accuracy to ensure clear virtual boundaries.

[0331] Effect: The virtual boundary is clear, the false alarm rate is reduced to 5%, and the processing efficiency is increased by 20%.

[0332] Case 4: Adaptive machining in a dynamic fixture system

[0333] During machining, the fixture is designed to be flexible and needs to be dynamically adjusted to secure the workpiece. Traditional virtual boundaries cannot adapt to changes in fixture position, which can easily lead to collisions.

[0334] Implementation of the present invention:

[0335] The virtual boundary tracks the fixture position in real time, predicts the fixture movement trend through the reinforcement learning model, and adjusts the boundary shape in advance.

[0336] Add fixture motion parameters to path planning to dynamically optimize tool paths and avoid fixture interference areas.

[0337] Effect: No downtime during the entire processing, processing time is shortened by 10%, and the adaptability of the fixture is significantly improved.

[0338] Case 5: Improving path efficiency in complex tasks

[0339] A complex machining task involving multiple processes needs to be completed efficiently, and traditional path planning methods cannot optimize path efficiency while avoiding collisions.

[0340] Implementation of the present invention:

[0341] Optimization formulas are used to comprehensively evaluate path efficiency and collision avoidance priority, optimizing path length and time while maintaining safety.

[0342] Combined with reinforcement learning strategies, the path direction and speed are dynamically adjusted to improve the overall processing efficiency.

[0343] Results: Processing time was shortened to 4 hours, path optimization reduced redundant moves by 30%, and overall efficiency increased by 25%.

[0344] The above cases demonstrate that the present invention achieves the following in complex processing scenarios:

[0345] Efficient and accurate dynamic adjustment of virtual boundaries.

[0346] Real-time path optimization and collision avoidance capabilities.

[0347] Excellent anti-interference performance and environmental adaptability.

[0348] Intelligent performance in multi-tool collaboration and dynamic machining environments.

[0349] These cases prove that the present invention has significant technical advantages and application value.

[0350] The applicable scenarios of the present invention are extended:

[0351] Multi-axis linkage processing scenario: The method can adapt to complex processing paths and ensure processing accuracy and efficiency by adjusting virtual boundaries and path optimization strategies in real time.

[0352] Complex surface processing: Dynamic sound field and multi-path signal fusion technology ensure that virtual boundaries can fit complex surfaces, achieving high-precision collision avoidance.

[0353] Multi-tool or multi-robot collaborative machining: In a multi-robot or multi-tool collaborative environment, tool paths are dynamically optimized to avoid path interference, ensuring safe and efficient collaborative machining.

[0354] Technical advantages of the present invention:

[0355] High-precision collision detection: Dynamic sound field and multi-frequency ultrasonic superposition technology combine to improve the detection accuracy of virtual boundaries. Multipath signal fusion and adaptive filtering algorithms effectively filter out interference signals.

[0356] Combining real-time performance with high efficiency: A lightweight path optimization algorithm reduces computational burden. A dynamic priority adjustment model balances collision avoidance requirements with processing efficiency.

[0357] Enhanced robustness and adaptability: Active environmental compensation algorithms and the use of sound-absorbing materials reduce interference from complex environments. Real-time boundary adjustment and path prediction ensure the system can adapt to changing processing environments.

[0358] Intelligent upgrade: The artificial intelligence model improves path optimization and boundary accuracy, and dynamically learns optimization strategies during the machining process.

[0359] In general, the present invention realizes closed-loop control from detection to collision avoidance, from the dynamic construction of virtual boundaries to path optimization strategies, and then to the learning application of artificial intelligence. It meets the high-precision requirements of multi-axis linkage, complex surface processing and collaborative processing environments, has strong real-time and adaptability, and is suitable for high-end intelligent processing scenarios.

[0360] Although the present invention has been described above with reference to embodiments, various modifications may be made thereto and equivalent components may be substituted without departing from the scope of the present invention. In particular, as long as there are no structural conflicts, the various features of the embodiments disclosed herein may be combined with each other in any manner, and the omission of an exhaustive description of such combinations in this specification is solely for the sake of space and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A potential collision detection and optimization method for machining path trajectory simulation, characterized in that: Collision detection and avoidance in a machining path are achieved through virtual boundary construction, real-time feedback control, and path optimization algorithms based on an ultrasonic dynamic sound field. The method comprises the following steps: S1. Generate a dynamic sound field in the processing area using an ultrasonic array to define a virtual boundary; S2. The tool sensor module receives the ultrasonic signal in real time and calculates the relative distance between the tool and the virtual boundary; S3. When the tool approaches or touches the virtual boundary, the path adjustment is triggered or the machining operation is paused; S4. Dynamically adjust the virtual boundary shape based on real-time feedback to adapt to changes in the workpiece, fixture, and processing path.

2. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: The dynamic sound field is generated by superposition of multi-frequency ultrasonic waves, and the clarity of the virtual boundary and the ability to resist scattering interference are enhanced by adjusting the emission frequency, phase and intensity.

3. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: Multipath signal fusion technology is used to analyze the time delay, intensity attenuation and frequency characteristics of the ultrasonic echo signal, and to filter out scattering and reflection interference, thereby improving the detection accuracy of virtual boundaries.

4. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: Adaptive filtering algorithm is used to dynamically optimize real-time signal processing, and the environmental noise signal is removed and the recognition ability of effective signals is enhanced by adjusting the filter parameters in real time.

5. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: Sound-absorbing materials are arranged in the non-processing area to reduce the interference of multiple ultrasonic reflections and reduce the impact of complex processing environment on virtual boundary detection.

6. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: The virtual boundary includes two layers of boundaries: The first layer of boundaries is used to detect when the tool approaches a dangerous area and trigger path deceleration or adjustment; The second layer of boundary is used to detect when the tool touches the boundary, triggering a processing pause and issuing an alarm signal.

7. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: A dynamic path optimization algorithm is used to dynamically replan the tool path based on the real-time distance between the tool and the virtual boundary, motion trends, and machining strategies. The path optimization includes: When the tool approaches the first layer boundary, it slows down and adjusts the path direction; When the tool enters the collision zone, the machining is paused and a safe path is generated to bypass the collision zone.

8. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: An active environmental compensation algorithm is used to generate a noise characteristic model before processing, recording the scattering characteristics of the workpiece, fixture, and tool. During processing, the real-time signal is compared with the noise characteristic model, and interference signals are subtracted to improve detection accuracy.

9. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: The processing history data is learned and optimized through artificial intelligence algorithms, including a virtual boundary accuracy optimization model based on deep learning and a dynamic path planning model based on reinforcement learning, to improve processing efficiency and collision avoidance capabilities.

10. The potential collision detection and optimization method for machining path trajectory simulation according to claim 1, characterized in that: The method is suitable for multi-axis linkage machining scenarios and complex surface machining. It ensures the accuracy and efficiency of the machining path by reconstructing virtual boundaries and path adjustment strategies in real time, and dynamically optimizes the machining path in a multi-tool or multi-robot collaborative environment to avoid path interference.

Images