Dynamic data synchronization optimization method for multi-axis machining path

Abstract

The present application relates to the technical fields of multi-axis linkage machining, and discloses a dynamic data synchronization optimization method for multi-axis linkage machining path. The method comprises collecting multiple sensor data in multi-axis linkage machining to obtain machining dynamic parameters; combining historical machining data to predict machining state and obtain predicted state parameters; collecting machining surface images, identifying actual state parameters through multiple state classifiers containing multiple classification paths according to the predicted state parameters; predicting machining path deformation according to the actual state parameters to obtain predicted path deformation parameters; performing dynamic window analysis based on the parameters to obtain machining characteristic parameters; and adjusting machining path moving step length according to the machining characteristic parameters to obtain an optimization result. This method enhances the adaptability and accuracy of path optimization through dynamic data synchronization and multi-link processing, which is beneficial to improving machining effect.

Description

Technical Field

[0001] This invention relates to the field of multi-axis linkage machining technology, specifically a method for dynamic data synchronization optimization of multi-axis linkage machining paths. Background Technology

[0002] In modern manufacturing, multi-axis machining technology, with its ability to machine complex curved surfaces and high-precision parts, is widely used in high-end manufacturing industries such as aerospace, automotive manufacturing, and precision mold making. As the structures of machined parts become increasingly complex, the requirements for machining accuracy and efficiency continue to rise, making path optimization in multi-axis machining a key factor affecting machining quality. Currently, the control of multi-axis linkage machining paths largely relies on preset programs. These programs are based on theoretical models and static parameters, making it difficult to cope with dynamic changes that occur during machining. During machining, factors such as tool wear, uneven material hardness, and fluctuations in cutting force can cause deviations between the actual machining path and the preset path, thereby affecting the machining accuracy and surface quality of the parts. To address the aforementioned issues, existing technologies have proposed several machining path optimization methods. These methods typically adjust the machining path by collecting some parameters from the machining process, such as cutting speed and feed rate. However, these methods have significant limitations: the types of parameters collected are limited, making it difficult to comprehensively reflect the dynamic changes in the machining process; they lack accurate prediction of the machining state and real-time evaluation of the machined surface quality, resulting in insufficient timeliness and accuracy of path optimization.

[0003] Existing machining path optimization methods often neglect the coordination between axes when dealing with complex paths in multi-axis simultaneous machining, easily leading to incoordination between axes and further exacerbating machining path deviations. Furthermore, due to numerous dynamic interference factors during machining, traditional static path optimization methods struggle to adapt to dynamic changes in the machining environment and cannot achieve real-time dynamic optimization of the machining path. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic data synchronization optimization method for multi-axis linkage machining paths to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a dynamic data synchronization optimization method for multi-axis linkage machining paths, the method comprising: Collect data from multiple sensors during multi-axis linkage machining to obtain dynamic machining parameters; Based on the aforementioned processing dynamic parameters and historical processing data, processing status prediction is performed to obtain predicted status parameters; The processing surface image is acquired, and the processing surface image is input into multiple state classifiers according to the predicted state parameters to identify and obtain the actual state parameters, wherein each state classifier includes multiple classification paths; Based on the actual state parameters, the deformation of the processing path is predicted to obtain the predicted path deformation parameters. Based on the predicted path deformation parameters, dynamic window analysis is performed to obtain processing feature parameters; Based on the processing characteristic parameters, the step size of the processing path is adjusted to obtain the optimized processing path result.

[0006] Preferably, data from multiple sensors are collected during the multi-axis linkage machining process to obtain dynamic machining parameters, including: Real-time monitoring of axis position data, axis speed data, and load data of multi-axis machining equipment; Collect the cumulative running time of the multi-axis machining equipment to obtain running time parameters; By integrating the axis position data, axis speed data, load data, and running time parameters, the machining dynamic parameters are calculated.

[0007] Preferably, based on the processing dynamic parameters and historical processing data, processing status prediction is performed to obtain predicted status parameters, including: Obtain historical maintenance data of multi-axis machining equipment of the same model, and extract sample processing dynamic parameter sets and sample status parameter sets; A processing state predictor is constructed, and the sample processing dynamic parameter set and sample state parameter set are used as training data for supervised training and testing. The processing dynamic parameters are input into the processing state predictor, and the predicted state parameters are output.

[0008] Preferably, an image of the processed surface is acquired, and based on the predicted state parameters, the processed surface image is input into multiple state classifiers to identify and obtain the actual state parameters, including: Real-time images of the processed surface are captured using a vision sensor; Based on the predicted state parameters, select the multiple state levels that are closest to the predicted state parameters; Select multiple state classifiers corresponding to the multiple state levels, and each state classifier includes multiple classification paths based on feature matching; The processed surface image is input into multiple classification paths within the multiple state classifiers, and multiple sets of state classification results are output. Analyze the multiple state classification result sets, calculate the state level probability, and output the state level with the highest probability as the actual state parameter.

[0009] Preferably, based on the actual state parameters, a processing path deformation prediction is performed to obtain predicted path deformation parameters, including: Obtain historical maintenance data of multi-axis machining equipment of the same model, and extract the sample actual state parameter set and sample path deformation parameter set; A path deformation predictor is constructed, and supervised training and testing are performed using the set of actual state parameters of the samples and the set of path deformation parameters of the samples as training data. The actual state parameters are input into the path deformation predictor, and the predicted path deformation parameters are output.

[0010] Preferably, based on the predicted path deformation parameters, dynamic window analysis is performed to obtain processing feature parameters, including: A fixed-length time window is set, and the predicted path deformation parameters are slide-segmented to obtain multiple data segments; Analyze each data segment and extract time series features; Calculate the average value and probability distribution of the time series features to obtain the processing feature parameters.

[0011] Preferably, adjusting the shift step size of the processing path based on the processing feature parameters to obtain an optimized processing path result includes: Based on the aforementioned processing characteristic parameters, calculate the rate of change of the processing path; Based on the rate of change, determine the adjustment ratio of the shift step size; By applying the aforementioned adjustment ratio, the step size of the processing path is modified, and the optimized processing path result is output.

[0012] Preferably, adjusting the step size of the processing path includes: Set boundary limits for the push step size; When the processing characteristic parameter indicates a high rate of change, reduce the shift step size; When the processing characteristic parameter indicates a low rate of change, increase the shift step size; A redundant buffer layer is used to handle transient out-of-bounds shift steps, ensuring the stability of step size adjustments.

[0013] Preferably, handling transient out-of-bounds shift steps using a redundant buffer layer includes: Monitor the deviation of the adjustment value of the push step size and the boundary limit; If the deviation value is less than the buffer threshold, temporary tolerance is allowed; If the deviation value continues to exceed the buffer threshold, the push step size will be reset to the safe default value.

[0014] Preferably, after obtaining the optimized machining path result, the parameters of the multi-axis machining equipment are synchronously updated, including: The optimized processing path results are converted into control commands; The control commands are transmitted in real time to the controller of the multi-axis machining equipment to update the machining path parameters.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By acquiring dynamic machining parameters from multiple sensors, various dynamic information during the machining process can be comprehensively captured, covering aspects such as tool status, cutting force, and vibration. This provides rich and comprehensive basic data for subsequent path optimization. Based on these dynamic parameters and historical machining data, machining status prediction can be performed to anticipate potential changes during machining, making path optimization more forward-looking and avoiding path deviations caused by sudden changes in machining status. By acquiring images of the machined surface and combining them with predicted state parameters, the actual state parameters are obtained through multiple state classifiers. Each classifier contains multiple classification paths. This multi-dimensional, multi-path recognition method improves the accuracy of judging the actual state of the machined surface, effectively reduces the errors that may be caused by a single recognition method, and thus more accurately reflects the machining quality of the part. Based on the actual state parameters, machining path deformation prediction can be performed to specifically predict possible deformations of the path, providing a clear direction for path adjustment. Dynamic window analysis based on predicted path deformation parameters yields processing feature parameters, enabling real-time extraction of key features related to path optimization during dynamically changing processing, ensuring the timeliness and relevance of the feature parameters. Adjusting the processing path's step size according to these feature parameters allows for dynamic path adjustment, enabling the path to flexibly adapt to actual processing conditions and better accommodate dynamic disturbances and complex operating situations. This method fully considers the coordination of motion between axes in multi-axis linkage machining during the optimization process. Through synchronous processing of dynamic data, it ensures the coordination of motion between axes and reduces path deviations caused by incoordination between axes. This dynamic data synchronization optimization method enables the machining path to always match the actual state of the machining process, thereby improving the machining accuracy and surface quality of parts and enhancing the stability and reliability of the machining process. This method does not rely on a fixed static model and can adapt to different processing materials, tool types and processing conditions. It has strong versatility and adaptability and can be widely used in various multi-axis linkage processing scenarios to meet the processing quality and efficiency requirements of different industries. Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating the working principle of the dynamic data synchronization optimization method for multi-axis linkage machining paths described in this invention. Figure 2A flowchart for collecting dynamic processing parameters; Figure 3 A flowchart for identifying actual state parameters; Figure 4 A flowchart for predicting deformation along the processing path; Figure 5 This is a flowchart for adjusting the step size and handling boundaries. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 This invention provides a method for dynamic data synchronization optimization of multi-axis linkage machining paths, the method comprising: By acquiring sensor data in real time during the machining process, combining historical machining data with visual inspection, the machining path is dynamically adjusted to improve machining accuracy and efficiency. This method first collects data on axis position, speed, load, and runtime during multi-axis linkage machining, integrating this data to obtain dynamic machining parameters. Based on these dynamic parameters and historical data, a trained machining state predictor predicts the current machining state and outputs predicted state parameters. Simultaneously, a visual sensor captures images of the machined surface, and an appropriate state classifier is selected based on the predicted state parameters. The images are then analyzed to identify actual state parameters. These actual state parameters are input into a path deformation predictor to predict potential deformations of the machining path, obtaining predicted path deformation parameters. Based on these predicted deformation parameters, time-series features are extracted through dynamic window analysis to calculate machining feature parameters. Finally, the machining path's step size is adjusted according to the machining feature parameters, outputting the optimized machining path result, and the control commands of the multi-axis machining equipment are updated synchronously.

[0019] Example 1: See Figure 2When collecting sensor data during multi-axis linkage machining, the system monitors the axis position, axis speed, and load data of the equipment in real time. Axis position data is acquired through high-precision encoders or linear scales installed on each moving axis, recording the real-time coordinate changes of each axis in three-dimensional space. Axis speed data is obtained through the built-in feedback unit of the servo motor or an external speed measuring device. If the equipment does not directly provide a speed signal, the instantaneous speed is derived by calculating the difference in axis position between two consecutive time points and dividing by the time interval. Load data is acquired through current and torque sensors integrated into the spindle or feed axis drive chain, reflecting the cutting force, frictional resistance, or mechanical stress during machining in real time. Simultaneously, the equipment control system's runtime timing module continuously records the cumulative running time since startup, generating runtime parameters.

[0020] When integrating the above data, timestamp alignment technology is employed to ensure that data collected by different sensors are synchronized under the same time base. For example, through hardware trigger signals or software time synchronization protocols, axis position, speed, load, and running time data are aligned according to millisecond-level time windows. Subsequent data preprocessing involves filtering and noise reduction of axis position and speed data to eliminate signal fluctuations caused by mechanical vibration; dimensional transformation of load data is performed, such as converting current values ​​to standard mechanical units (Newtons or Newton-meters); and segmented normalization of running time parameters eliminates numerical overflow issues caused by long-term operation. Finally, a weighted fusion algorithm merges the four types of data into a multi-dimensional vector form of machining dynamic parameters, which comprehensively characterize the real-time motion state and mechanical load of the equipment.

[0021] When predicting machining status based on dynamic machining parameters and historical machining data, record data of the same model of multi-axis machining equipment is extracted from the equipment manufacturer's historical database or locally stored maintenance logs. Historical maintenance data includes operating records throughout the equipment's entire lifecycle, from which sample sets of dynamic machining parameters and sample sets of status parameters are selected. The sample set of dynamic machining parameters consists of historical axis position, speed, load, and operating time data, stored in time series. The sample set of status parameters consists of status labels marked during equipment maintenance, such as assessment results based on vibration analysis, lubrication oil testing, or component wear measurement. Labels include discrete levels or continuous values ​​such as "normal," "slight thermal deformation," "bearing wear," and "increased guideway clearance."

[0022] The model uses a set of dynamic processing parameters as input features and a set of sample state parameters as the target output, dividing the model into training and testing sets. The predictor model can be structured using random forests, gradient boosting trees, or fully connected neural networks. During training, the model learns the mapping relationship from multi-dimensional dynamic parameters to state labels, such as the correlation between abnormal shaft speed fluctuations and bearing wear, or the correlation between sudden load increases and guide rail deformation. After training, cross-validation is performed to evaluate the model's generalization ability on unseen data. During deployment, the real-time collected processing dynamic parameters are input into the predictor, and the model outputs predicted state parameters. These parameters represent the current potential state of the equipment in the form of a probability distribution or discrete values, such as "62% probability of thermal deformation" or "wear level 3".

[0023] The model is deployed via embedded processors or edge computing devices, employing quantization compression technology to reduce computational complexity. Input data is processed using a sliding window mechanism, with each inference based on a dynamic parameter sequence from the most recent N time points, capturing the temporal dependencies of state changes. Output results are cached in a circular buffer for subsequent modules to access. If the predicted state parameters conflict with the device's current alarm threshold, an anomaly warning process is triggered.

[0024] Example 2: See Figure 3 When acquiring images of the machined surface, a high-resolution industrial camera or line-scan imaging system is used. The camera is mounted above or to the side of the working area of ​​the multi-axis machining equipment and adjusted to the optimal observation angle using a robotic arm or fixed bracket. The imaging system is equipped with a coaxial light source or a ring light source, and an appropriate wavelength is selected according to the reflection characteristics of the machined material. For example, blue LED light sources are used to reduce thermal interference in metal processing, while diffuse white light sources are used for composite materials. The camera trigger signal is synchronized with the spindle rotation or feed motion of the equipment, and the exposure timing is controlled by encoder pulses or PLC signals to ensure that the image is captured the instant the tool leaves the cutting area, avoiding motion blur. The image resolution is set to no less than 5 million pixels, and the frame rate is dynamically adjusted to adapt to the machining speed. During high-speed cutting, a region clipping mode is enabled to improve acquisition efficiency. The captured raw image is transmitted to the image processing unit for preprocessing operations: adaptive histogram equalization is applied to enhance surface texture contrast, median filtering is used to suppress cutting fluid splash noise, and perspective transformation is performed to correct geometric distortion caused by viewing angle tilt.

[0025] When filtering state levels based on predicted state parameters, the probability distribution or discrete labels output by the predictor are analyzed. If the predicted state parameters are presented as probability vectors, such as [Normal: 0.2, Slight Wear: 0.5, Severe Wear: 0.3], all levels with probability values ​​exceeding a set threshold are selected. If they are discrete labels, adjacent levels are expanded according to a preset level mapping table. For example, if the predicted label is "Moderate Thermal Deformation," the mapping table defines its adjacent levels as including "Slight Thermal Deformation" and "Severe Thermal Deformation." Finally, 3 to 5 candidate state levels are selected to form a set of state classifier calls. Each state level corresponds to an independently trained classifier model, which is trained offline on a massive amount of labeled images.

[0026] After selecting the classifier corresponding to the candidate state level, its internal multi-classification path configuration is loaded. Each classifier contains three heterogeneous analysis paths: The first path is based on traditional machine vision algorithms, performing gray-level co-occurrence matrix texture feature extraction, calculating 14-dimensional feature vectors such as contrast, correlation, and entropy, and inputting them into a support vector machine for roughness classification; the second path uses edge geometry analysis, detecting surface contours through the Canny operator, extracting topological features such as edge length distribution and curvature change rate, and combining them with a random forest regressor to predict shape deviations; the third path is based on a convolutional neural network architecture, using an improved ResNet-34 backbone network, directly outputting defect detection results end-to-end after data augmentation of the input image. The three paths run in parallel within the classifier, sharing the same input image but performing different feature transformations.

[0027] The preprocessed surface image is simultaneously input into the paths of multiple state classifiers. Each classifier's three paths independently process the image and generate intermediate results: the texture path outputs the surface roughness level (e.g., a five-level scale of Ra0.4 and Ra3.2); the geometry path outputs the contour deviation index (a continuous value of 0 to 00); and the deep learning path outputs a defect probability map and a comprehensive state score (0 to 1.0). After standardization, the results from each path are uniformly converted to a state level probability vector. For example, a contour deviation value of 85 from the geometry path is converted into a probability distribution of [normal: 0.1, slight: 0.3, moderate: 0.4, severe: 0.2] by looking up a table.

[0028] When analyzing the outputs of multiple classifiers, a hierarchical fusion strategy is employed. First, path results are aggregated within each classifier: the probability vectors of the three paths are fused using a weighted average, with weights dynamically allocated based on the historical accuracy of each path on the validation set. For example, the texture path has a weight of 0.3, the geometry path 0.4, and the deep learning path 0.3. This fusion yields the state probability distribution for a single classifier. Then, cross-classifier integration is performed: for the same state level, the results from different classifiers (e.g., the probability values ​​of the "moderate wear" level across three candidate classifiers) are averaged as the final confidence level. Finally, the probability values ​​of all candidate levels are compared, and the level label with the highest confidence level is selected as the actual state parameter.

[0029] The actual state parameters are output in a structured data format, including primary level labels, confidence scores, and auxiliary information. The auxiliary information records key feature values ​​for each classification path, such as the entropy value of the texture path, the coordinates of the maximum contour deviation location of the geometric path, and the storage path of the defect heatmap for the deep learning path. The output data is written to a shared memory area for real-time reading by the subsequent path deformation prediction module. When the confidence score falls below a preset threshold, an image re-acquisition mechanism is triggered; when conflicts between different path results exceed the tolerance limit, a manual review flag is activated. The entire processing flow is completed within 500 milliseconds, utilizing FPGA to accelerate convolution operations and feature extraction, meeting the real-time requirements of the processing.

[0030] Example 3: See Figure 4 When predicting machining path deformation based on actual state parameters, historical maintenance records are extracted from the distributed database of multi-axis machining equipment. These records contain actual state parameters and corresponding path deformation data collected by the same model of equipment during past operating cycles. Actual state parameters are stored in time-series format, including structured fields such as surface roughness level, geometric accuracy deviation index, and thermal deformation coefficient; path deformation parameters are recorded as position offsets, attitude angle errors, and trajectory following delay values ​​for the six-dimensional motion axes. The sample dataset undergoes cleaning and screening: abnormal records during sensor failure periods are removed, measurement drift caused by changes in environmental temperature and humidity is compensated for, and non-uniformly sampled data is resampled to a fixed frequency using cubic spline interpolation. The filtered set of sample actual state parameters and sample path deformation parameters are strictly aligned by timestamp and divided into training and validation subsets.

[0031] The number of nodes in the network input layer matches the dimension of the actual state parameters. The hidden layer contains two 64-unit GRU layers, and the output layer corresponds to the six-dimensional continuous values ​​of the path deformation parameters. During training, the input sequence is truncated into time windows, with each window containing a vector of actual state parameters for T consecutive time points. The label data is the ground truth value of the path deformation parameters at the end of the window. The Huber loss function is used, and the optimizer employs the Nesterov momentum-accelerated Adam algorithm. After training, the model weights are deployed to the real-time inference engine, and quantized to 8-bit integers to reduce computational latency. During inference, the T actual state parameters before the current time are input into the predictor, and the predicted path deformation parameters are output. These parameters include the predicted offsets of the motion axes (δx, δy, δz) and the predicted biases of the attitude angles (δα, δβ, δγ) within the next time interval Δt.

[0032] When performing dynamic window analysis based on predicted path deformation parameters, a sliding window of length W is set. The window advances along the time axis with a fixed step size S, covering newly generated predicted deformation data with each slide. Each data segment within the window contains a path deformation sequence of K consecutive sampling points, where K is determined by the window length W and the sampling frequency. Three types of time series features are extracted during the analysis: the first type is statistical features, including the mean, variance, and kurtosis of the offsets of each motion axis; the second type is trend features, calculated by summing the absolute values ​​of the first differences of the sequences to characterize the intensity of change; and the third type is morphological features, calculated using the following formula to determine the amplitude skewness of each data point within the window. : in: This represents the comprehensive evaluation value of the path deformation at the t-th sampling point within the window. This evaluation value is generated by weighted fusion of six-dimensional deformation parameters. Indicates all within the window The arithmetic mean; Window length (in seconds); Output a dimensionless measure of morphological asymmetry.

[0033] After feature extraction, the probability distribution is calculated: the cumulative probability density of the historical distribution of the variance term in the statistical features is calculated, and exponential smoothing is applied to trend features. The values ​​are converted into probability values ​​using the Sigmoid function. The final processed feature parameters are output as triples. ,in A vector of six-dimensional axis offsets. The geometric mean of the probabilities of each characteristic distribution is given. for The probability of value transformation.

[0034] A dual-buffer mechanism is employed for real-time processing: the front buffer receives the predicted deformation parameter stream, while the back buffer performs window analysis. As the window slides, new data overwrites the oldest data, and eigenvalues ​​are incrementally updated to reduce computational overhead. Feature parameters are updated after each window analysis, accessible via shared memory mapping to the step size adjustment module. The window length W is dynamically adjusted based on the equipment's dynamic response performance, shortening to 0.5 seconds for high-speed machining and extending to 5 seconds for precision machining, adaptively configured by the equipment control system's real-time performance monitoring module.

[0035] Example 4: See Figure 5 When adjusting the step size of the machining path based on the machining feature parameters, the system first parses the feature parameter triples. The meaning of each component in the text. The vector contains the mean offset of the six axes of motion, reflecting the average positional deviation of each axis within the window period; The value represents the overall fluctuation probability of path deformation; the higher the value, the more unstable the deformation trend. The value, as the probability of morphological asymmetry, indicates the degree of risk of sudden deformation. Based on these three components, the path change rate index is calculated: The Euclidean norm of a vector and The product of the values ​​is used as the basic rate of change, which is then multiplied by... The reciprocal of the value yields the normalized overall rate of change R. The range of this rate of change R is mapped to the interval 0 to 1, and is converted into a step size adjustment ratio through a piecewise linear function.

[0036] The baseline step size is predefined in the process documentation and represents the single interpolation displacement of the equipment under ideal conditions. Hard boundary limits are set during adjustment: the maximum step size does not exceed 150% of the baseline value, and the minimum step size is not less than 30% of the baseline value. When the overall rate of change R exceeds 0.7, it is considered a high rate of change state, and the step size decreases by 5% per cycle; when R is below 0.3, it is considered a low rate of change state, and the step size increases by 3% per cycle. During adjustment, the load current of each axis is monitored in real time. If the current of any axis exceeds the safety threshold, the step size increase operation is immediately paused.

[0037] When handling step size exceeding limits in the redundant buffer layer, a circular buffer of length 5 is maintained to record the most recent step size adjustments. After each adjustment, the percentage deviation between the current step size and the boundary is calculated. If the deviation exceeds the buffer threshold (upper limit deviation 10% or lower limit deviation 15%) for three consecutive times, the step size reset mechanism is triggered. The reset value is the median of the three most recent valid step sizes in the buffer. If the limit is still exceeded, the step size is rolled back to 80% of the baseline step size.

[0038] Table 1: Example record of dynamic adjustment of processing path step size.

[0039] Referring to Table 1, this table records the step size adjustment process over 6 seconds. The initial step size of 1.20 mm was continuously reduced due to a persistently high rate of change. By 10:05:05, the cumulative deviation exceeded the lower limit threshold of 12%, triggering the reset mechanism to restore the step size to 1.05 mm. The buffer counter column displays the number of cycles in which the out-of-limit state persisted; the counter returned to zero after the reset.

[0040] During implementation, step size adjustment commands are sent to each axis driver via a real-time bus. The X / Y / Z linear axes use absolute displacement commands, while the A / B / C rotary axes use angular increment commands. After each adjustment, the system detects the deviation between the actual position and the commanded position of each axis. If the deviation fails to converge for two consecutive cycles, it automatically switches to conservative step size mode. In conservative mode, the step size adjustment range is halved, while the gain parameter of the servo control loop is increased.

[0041] When the spindle speed exceeds 8000 rpm, the buffer threshold is relaxed to an upper limit of 15% and a lower limit of 20%; when machining soft materials such as aluminum alloys, the lower threshold is tightened to 10%. Threshold adjustments are based on dynamic calculations using elastic modulus parameters from the material database and the spindle power curve. During the over-limit tolerance period, the system records additional monitoring indicators including: peak-to-peak value of each axis's following error, driver temperature rise slope, and frequency domain energy distribution of the cutting force. This data is used for subsequent analysis of the impact of step size adjustments on machining quality.

[0042] When the rate of change R shows a monotonically increasing trend over five consecutive cycles, the step size reduction plan is activated in advance; when the R value exhibits high-frequency oscillations, the step size adjustment is frozen and the vibration suppression algorithm is activated. The plan parameters are stored in the process knowledge base, and the optimal adjustment strategy is preset according to different tool-material combinations. For example, when milling stainless steel, the step size reduction plan is set to 7% per cycle; while when machining composite materials, a gradual adjustment of 3% is used.

[0043] The final optimized machining path is output as a time-space coordinate sequence. Each coordinate point includes six-axis position commands, recommended feed rates, and step size confidence markers. Confidence markers are divided into three levels: high confidence (stable operation for more than 10 seconds after step size adjustment), medium confidence (step size is in the adjustment transition period), and low confidence (step size has just been reset or is in conservative mode). The path optimization module encapsulates this data into standard CNC commands, which are then transmitted to the motion controller for execution after safety verification.

[0044] High-priority tasks are handled with boundary monitoring and emergency resets, while low-priority tasks are executed with trend prediction and knowledge base queries. All adjustment records are written to non-volatile memory, forming a closed-loop optimization historical data chain. When the same tool is detected repeatedly triggering step size resets under similar machining parameters, a tool wear warning report is automatically generated.

[0045] Example 5: When the push step size exceeds the limit, the system establishes a real-time monitoring mechanism to track the deviation between the step size adjustment value and the preset boundary through a redundant buffer layer. The boundary value is dynamically configured according to the equipment model. The upper limit of the linear axis step size is usually 1.5 times the reference value, and the upper limit of the rotary axis step size is 1.8 times the reference value; the lower limit is uniformly set to 30% of the reference value. The deviation is calculated as a percentage. When the current step size exceeds the upper limit, a positive deviation is recorded, and when it is below the lower limit, a negative deviation is recorded. The buffer threshold is set in a hierarchical structure: the short-term tolerance threshold is set to ±5% of the boundary value, the medium-term warning threshold is set to ±8%, and the emergency intervention threshold is set to ±12%. The monitoring data is refreshed every 10 milliseconds and stored in a circular queue to retain the most recent 50 sampling points. When the detected deviation value is less than the short-term tolerance threshold, the system maintains the current step size but activates the observation mode. In this mode, the sensor sampling frequency is increased to twice the normal value, focusing on monitoring the spindle vibration acceleration, the following error of each axis, and the cutting temperature gradient. If the deviation does not increase within three cycles, the observation mode is exited. If the deviation persists but does not reach the warning threshold, the observation period is extended to 10 cycles, and the load fluctuation spectrum characteristics are recorded. When the deviation value continuously exceeds the warning threshold, a buffer compensation strategy is activated. The system automatically creates a virtual buffer space, allowing the actual step size to temporarily exceed the boundary value, while dynamically adjusting the servo control parameters: the position loop proportional gain is reduced by 20%, and the speed loop integral time constant is increased by 15%. During the compensation period, the deviation change rate is calculated each cycle. If the change rate shows a decreasing trend, the compensation strategy is maintained; if the change rate continues to rise, a pre-reset check is triggered. The pre-reset check analyzes the step size adjustment records of the most recent 20 cycles to identify whether there is a periodic fluctuation pattern, and selects gradual rollback or step adjustment based on the pattern characteristics.

[0046] When the deviation value reaches the emergency intervention threshold or the number of consecutive over-limit cycles reaches 3, a step size reset operation is performed. The reset value is dynamically determined based on the equipment's operating status: 80% of the reference step size under normal operating conditions; 70% of the reference step size under high-speed cutting conditions; and 90% of the reference step size under precision machining conditions. The reset process uses a ramp transition method, linearly transitioning from the current step size to the target value within 5 control cycles. After the reset is completed, a forced cooling period of 30 cycles is implemented, during which the step size adjustment function is locked, allowing only manual intervention. Each reset event generates a detailed log, including the trigger time, over-limit axis identifier, step size values ​​before and after the reset, and environmental parameter snapshots.

[0047] After obtaining the optimized machining path, the parameter synchronization update process is initiated. The path optimization module converts the spatiotemporal coordinate sequence into a control instruction set that the equipment can resolve. For linear axis motion, a G01 linear interpolation instruction is generated with an F feed rate parameter; for rotary axis motion, a G02 / G03 circular arc instruction is generated with a C-axis rotation angle parameter. Inverse kinematics verification is performed during instruction conversion to ensure that the displacement of each axis is within the physical travel range. Smooth transition codes are inserted for critical instructions such as reversal points and corner transition points to avoid sudden speed changes.

[0048] Control commands are transmitted to the device controller via real-time Ethernet. The communication protocol employs a deterministic transmission mechanism, with each command appended with a 16-bit CRC checksum and an 8-bit sequence number. The transmission period is strictly synchronized with the device interpolation period, with a default setting of 2 milliseconds. Upon receiving a command, the controller performs three levels of verification: syntax verification checks the integrity of the command format, range verification confirms that the parameters are within the device's allowable range, and correlation verification analyzes the coordination of multi-axis motion. After successful verification, the command is stored in a double-buffered queue and takes effect at the start of the next interpolation period.

[0049] When the communication packet loss rate exceeds 0.1%, the system automatically switches to a redundant communication channel. When three consecutive command verifications fail, processing is paused and the system reverts to a safe position. When the controller detects a discontinuous command sequence, it initiates a missing command reconstruction algorithm. This algorithm predicts the coordinates of missing points based on historical path data and uses cubic spline interpolation to generate a transition path. All abnormal events are pushed to the monitoring interface in real time, while the current processing status is frozen for operator confirmation. The tracking error of each axis is calculated by comparing the command position with the actual feedback position. When the tracking error exceeds the allowable value of 50%, the feed rate for the next three cycles is automatically compensated. When the error continues to increase, the system rolls back to the previous valid path parameters. Execution data is fed back to the path optimization module to form a closed loop, used to correct the parameter weights of the step size adjustment model. The entire synchronous update process is completed under uninterrupted equipment operation conditions, with the end-to-end latency from command generation to execution effective within 5 milliseconds.

[0050] After each workpiece is processed, the system analyzes the deviation distribution between the actual path and the optimized path, generating a path matching report. The report includes indicators such as maximum positional deviation, average following error, and corner overcut. Long-term operational data is used to optimize communication parameter configurations, such as dynamically adjusting data packet size and retransmission timeout. When the matching degree consistently falls below the set standard, the system automatically initiates a retraining process for the path optimization module.

[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0052] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic data synchronization optimization of multi-axis linkage machining paths, characterized in that, Includes the following steps: Collect data from multiple sensors during multi-axis linkage machining to obtain dynamic machining parameters; Based on the aforementioned processing dynamic parameters and historical processing data, processing status prediction is performed to obtain predicted status parameters; The processing surface image is acquired, and the processing surface image is input into multiple state classifiers according to the predicted state parameters to identify and obtain the actual state parameters, wherein each state classifier includes multiple classification paths; Based on the actual state parameters, the deformation of the processing path is predicted to obtain the predicted path deformation parameters. Based on the predicted path deformation parameters, dynamic window analysis is performed to obtain processing feature parameters; Based on the processing characteristic parameters, the step size of the processing path is adjusted to obtain the optimized processing path result.

2. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 1, characterized in that, Data from multiple sensors during multi-axis linkage machining is collected to obtain dynamic machining parameters, including: Real-time monitoring of axis position data, axis speed data, and load data of multi-axis machining equipment; Collect the cumulative running time of the multi-axis machining equipment to obtain running time parameters; By integrating the axis position data, axis speed data, load data, and running time parameters, the machining dynamic parameters are calculated.

3. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 2, characterized in that, Based on the aforementioned processing dynamic parameters and historical processing data, processing status prediction is performed to obtain predicted status parameters, including: Obtain historical maintenance data of multi-axis machining equipment of the same model, and extract sample processing dynamic parameter sets and sample status parameter sets; A processing state predictor is constructed, and the sample processing dynamic parameter set and sample state parameter set are used as training data for supervised training and testing. The processing dynamic parameters are input into the processing state predictor, and the predicted state parameters are output.

4. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 3, characterized in that, Acquire images of the processed surface; input these images into multiple state classifiers based on the predicted state parameters to identify and obtain the actual state parameters, including: Real-time images of the processed surface are captured using a vision sensor; Based on the predicted state parameters, select the multiple state levels that are closest to the predicted state parameters; Select multiple state classifiers corresponding to the multiple state levels, and each state classifier includes multiple classification paths based on feature matching; The processed surface image is input into multiple classification paths within the multiple state classifiers, and multiple sets of state classification results are output. Analyze the multiple state classification result sets, calculate the state level probability, and output the state level with the highest probability as the actual state parameter.

5. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 4, characterized in that, Based on the actual state parameters, a processing path deformation prediction is performed to obtain the predicted path deformation parameters, including: Obtain historical maintenance data of multi-axis machining equipment of the same model, and extract the sample actual state parameter set and sample path deformation parameter set; A path deformation predictor is constructed, and supervised training and testing are performed using the set of actual state parameters of the samples and the set of path deformation parameters of the samples as training data. The actual state parameters are input into the path deformation predictor, and the predicted path deformation parameters are output.

6. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 5, characterized in that, Based on the predicted path deformation parameters, dynamic window analysis is performed to obtain processing feature parameters, including: A fixed-length time window is set, and the predicted path deformation parameters are slide-segmented to obtain multiple data segments; Analyze each data segment and extract time series features; Calculate the average value and probability distribution of the time series features to obtain the processing feature parameters.

7. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 6, characterized in that, Based on the aforementioned processing characteristic parameters, the step size of the processing path is adjusted to obtain an optimized processing path result, including: Based on the aforementioned processing characteristic parameters, calculate the rate of change of the processing path; Based on the rate of change, determine the adjustment ratio of the shift step size; By applying the aforementioned adjustment ratio, the step size of the processing path is modified, and the optimized processing path result is output.

8. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 7, characterized in that, Adjusting the step size of the machining path includes: Set boundary limits for the push step size; When the processing characteristic parameter indicates a high rate of change, reduce the shift step size; When the processing characteristic parameter indicates a low rate of change, increase the shift step size; A redundant buffer layer is used to handle transient out-of-bounds shift steps, ensuring the stability of step size adjustments.

9. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 8, characterized in that, Handling transient out-of-bounds shift steps using redundant buffer layers includes: Monitor the deviation of the adjustment value of the push step size and the boundary limit; If the deviation value is less than the buffer threshold, temporary tolerance is allowed; If the deviation value continues to exceed the buffer threshold, the push step size will be reset to the safe default value.

10. The dynamic data synchronization optimization method for multi-axis linkage machining paths as described in claim 9, characterized in that, After obtaining the optimized machining path results, the parameters of the multi-axis machining equipment are synchronously updated, including: The optimized processing path results are converted into control commands; The control commands are transmitted in real time to the controller of the multi-axis machining equipment to update the machining path parameters.

Images