Numerical control machine tool control method and system based on real-time processing data feedback

Abstract

The invention provides a numerical control machine tool control method and system based on real-time machining data feedback, and relates to the technical field of machine tool control. Predicting control quality based on the multi-source data to obtain predicted quality; generating an online control constraint according to the predicted quality and the machine tool safety boundary; under the online control constraint, solving optimization is carried out by taking the machining cycle, the tool wear increment and the instantaneous power as targets, and target control parameters are obtained; and issuing and executing the target control parameters, executing parameter rollback when abnormal out-of-limit is detected, and performing incremental updating on the predictive control quality process by taking a rollback result as feedback. The technical problem that in the prior art, the control quality of a numerical control machine tool is poor can be solved, and the technical effect of improving the control quality of the numerical control machine tool is achieved.

Description

Technical Field

[0001] This application relates to the field of machine tool control technology, and in particular to a CNC machine tool control method and system based on real-time machining data feedback. Background Technology

[0002] With the widespread application of CNC machine tools in high-end manufacturing, aerospace, and precision automotive parts processing, the requirements for quality, efficiency, and stability during machining are becoming increasingly stringent. Traditional CNC machine tool control methods mostly rely on preset process parameters and empirical models to guide the machining process. While this method can achieve good results under stable operating conditions, it often falls short in complex and ever-changing actual machining environments.

[0003] Currently, existing research and technologies mostly focus on improvements in single aspects, such as tool condition monitoring based on spindle current or vibration signals, or thermal error compensation based on temperature field modeling. However, these methods often lack the fusion and unified processing of multi-source data, resulting in incomplete prediction results that cannot effectively reflect the actual operating status of machine tools under complex working conditions. Even though some studies have introduced real-time monitoring and prediction models, most of them are one-way monitoring, that is, they only perform anomaly detection without closed-loop linkage with control parameters, resulting in a disconnect between prediction and control.

[0004] In summary, existing technologies suffer from a lack of integration of multi-source real-time data and linkage of control links, resulting in single quality prediction results that are difficult to guide the dynamic optimization of machine tool parameters. This further affects the stability of the processing, the consistency of product quality, and the improvement of production efficiency. Summary of the Invention

[0005] The purpose of this application is to provide a CNC machine tool control method and system based on real-time machining data feedback, in order to solve the technical problems in the prior art where the lack of integration of multi-source real-time data and linkage of control links leads to single quality prediction results that are difficult to guide the dynamic optimization of machine tool parameters, which further affects the stability of the machining process, the consistency of product quality, and the improvement of production efficiency.

[0006] In view of the above problems, this application provides a CNC machine tool control method and system based on real-time machining data feedback.

[0007] Firstly, this application provides a CNC machine tool control method based on real-time machining data feedback, implemented through a CNC machine tool control system based on real-time machining data feedback. The method includes: collecting multi-source data from the CNC machine tool during machining; predicting control quality based on the multi-source data to obtain predicted quality; generating online control constraints based on the predicted quality and the machine tool safety boundary; under the online control constraints, optimizing the machining cycle, tool wear increment, and instantaneous power as objectives to obtain target control parameters; issuing and executing the target control parameters; when an abnormal limit is detected, the execution parameters are rolled back, and the rollback result is used as feedback to incrementally update the predicted control quality process.

[0008] Preferably, the CNC machine tool control method based on real-time machining data feedback further includes: acquiring multi-source real-time data of the CNC machine tool for time calibration and alignment, and outputting a time-calibrated and aligned multi-source data matrix, wherein the multi-source real-time data includes spindle current, feed current, vibration, acoustic emission, temperature, encoder displacement, and online measurement data; denoising and preprocessing the multi-source data matrix to output a multi-source signal sequence; dividing the multi-source signal sequence into a real-time sliding window, and extracting multi-source data according to the real-time sliding window.

[0009] Preferably, the CNC machine tool control method based on real-time machining data feedback further includes: extracting a real-time window of the multi-source data, and obtaining multiple consecutive historical scale windows of the real-time window; extracting multi-source sample data and control quality samples according to the multiple consecutive historical scale windows; training a control quality prediction channel using the multi-source sample data as the sample input set and the control quality samples as the sample output set to obtain a predicted control quality coefficient; and obtaining a control quality prediction channel when the predicted control quality coefficient meets a control quality coefficient threshold.

[0010] Preferably, the CNC machine tool control method based on real-time machining data feedback further includes: performing window adaptation on the multi-scale input samples and multi-scale output samples of the continuous multiple historical scale windows to obtain multi-source sample data and control quality samples; wherein, a dynamic window scale is configured based on the real-time window, and insufficient scale samples are filled and excess scale samples are clipped according to the dynamic window scale.

[0011] Preferably, the CNC machine tool control method based on real-time machining data feedback further includes: introducing the machine tool safety boundary, comparing the predicted quality with the machine tool safety boundary, adjusting the machine tool safety boundary to generate a feasible range; mapping the feasible range to machine tool adjustable parameters to generate online control constraints.

[0012] Preferably, the CNC machine tool control method based on real-time machining data feedback further includes: inputting the online control constraints into an optimization objective function and outputting first optimized control parameters, wherein the optimization objective function includes optimization objectives for the machining cycle, the tool wear increment, and the instantaneous power; calculating machining performance indicators based on the first optimized control parameters, solving for the optimized control parameters that minimize the machining performance indicator values ​​under the online control constraints, and obtaining the target control parameters.

[0013] Preferably, the CNC machine tool control method based on real-time machining data feedback further includes: collecting multi-source real-time data to determine whether abnormal limits have been exceeded, and obtaining anomaly detection results; executing a rollback strategy based on the anomaly detection results to obtain safety control parameters after rollback; calculating the error between the safety control parameters and the target control parameters, and performing incremental updates.

[0014] Preferably, the CNC machine tool control method based on real-time machining data feedback further includes: the predicted quality includes tool wear, surface roughness, thermal geometry error and vibration stability margin.

[0015] Preferably, the CNC machine tool control method based on real-time machining data feedback further includes: the target control parameters include spindle speed, feed rate, depth of cut, width of cut, and trajectory acceleration / deceleration.

[0016] Secondly, this application also provides a CNC machine tool control system based on real-time machining data feedback, used to execute the CNC machine tool control method based on real-time machining data feedback as described in the first aspect, including: a multi-source data collection module for collecting multi-source data of the CNC machine tool during machining; a predicted quality acquisition module for predicting control quality based on the multi-source data to obtain predicted quality; a constraint generation module for generating online control constraints based on the predicted quality and machine tool safety boundaries; a parameter acquisition module for solving and optimizing the target control parameters under the online control constraints, with machining cycle, tool wear increment, and instantaneous power as objectives; and an execution module for issuing the target control parameters for execution, and when an abnormal limit is detected, the execution parameters are rolled back, and the rollback result is used as feedback to incrementally update the predicted control quality process.

[0017] The technical solution provided in this application has at least the following technical effects or advantages: by achieving the technical goal of constructing a predictive control closed-loop system based on multi-source real-time data feedback and dynamically generating and optimizing control parameters during machine tool processing, it achieves the technical effects of improving the accuracy of machining quality prediction, realizing adaptive adjustment of machine tool control, extending tool life, and ensuring a steady increase in production efficiency.

[0018] The above description is merely an overview of the technical solution of this application. To enable a clearer understanding of the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the CNC machine tool control method based on real-time machining data feedback proposed in this application.

[0021] Figure 2 This is a schematic diagram of the CNC machine tool control system based on real-time machining data feedback in this application.

[0022] Figure labeling: Multi-source dataset module 1, Prediction quality acquisition module 2, Constraint generation module 3, Parameter acquisition module 4, Execution module 5. Detailed Implementation

[0023] This application provides a CNC machine tool control method and system based on real-time machining data feedback. It addresses the technical problems in existing technologies where the lack of multi-source real-time data fusion and control linkage leads to single-source quality prediction results that are difficult to guide dynamic optimization of machine tool parameters, further impacting machining process stability, product quality consistency, and production efficiency. The application achieves the technical goal of constructing a predictive control closed-loop system based on multi-source real-time data feedback, dynamically generating and optimizing control parameters during machine tool machining. This results in improved accuracy of machining quality prediction, adaptive adjustment of machine tool control, extended tool life, and a steady increase in production efficiency.

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0025] Example 1, please refer to the appendix. Figure 1 This application provides a CNC machine tool control method based on real-time machining data feedback, applied to a CNC machine tool control system based on real-time machining data feedback, specifically including: S1: Collect multi-source data during the machining process of CNC machine tools.

[0026] Furthermore, this application also includes: collecting multi-source real-time data of CNC machine tools for time calibration and alignment, and outputting a time-calibrated and aligned multi-source data matrix, wherein the multi-source real-time data includes spindle current, feed current, vibration, acoustic emission, temperature, encoder displacement and online measurement data; denoising and preprocessing the multi-source data matrix to output a multi-source signal sequence; dividing the multi-source signal sequence into a real-time sliding window, and extracting multi-source data according to the real-time sliding window.

[0027] Specifically, time calibration and alignment of multi-source real-time data from CNC machine tools refers to acquiring real-time signals from different parts of the machine tool through various sensors during the machining process, and unifying the time base of these real-time signals so that data from different sources can be compared and analyzed on the same time axis. Multi-source real-time data includes spindle current, feed current, vibration, acoustic emission, temperature, encoder displacement, and online measurement data. Spindle current, the current consumed to drive the spindle rotation, reflects the cutting load; feed current, the current driving the feed system, reflects the load state of the tool movement; vibration is the mechanical vibration signal generated by the machine tool during machining, which can be used to detect the workpiece and tool status; acoustic emission is the high-frequency sound signal generated during machining, which can reflect material fracture or cutting abnormalities; temperature is used to monitor the thermal state of the spindle, tool, and workpiece to avoid thermal deformation affecting machining accuracy; encoder displacement records the positional changes of each axis of the machine tool; and online measurement data are workpiece dimensions or surface quality indicators measured in real time during machining. After time calibration and alignment, a multi-source data matrix is ​​output, with each row corresponding to a time point and each column corresponding to a sensor signal.

[0028] For multi-source data matrices, denoising and preprocessing are performed to eliminate environmental noise and sensor interference through methods such as filtering, outlier removal, and smoothing, making the signals more stable and reliable. The denoising process may include using bandpass filtering to remove low-frequency and high-frequency noise from vibration signals, performing moving averages on current signals to smooth abrupt changes, using low-pass filtering on temperature signals to remove transient fluctuations, and using thresholding or median filtering to process abnormal peaks, thereby obtaining a clear multi-source signal sequence and providing a reliable foundation for subsequent analysis.

[0029] Dividing a multi-source signal sequence into real-time sliding windows involves segmenting continuous multi-source signals according to a fixed time length or number of sampling points, allowing for feature extraction and analysis within each window. The real-time sliding window employs a preset overlap ratio, such as a window length of 1 second and a sliding interval of 0.5 seconds, enabling continuous windows to share some data, achieving both continuity and real-time performance. Extracting multi-source data from the real-time sliding window involves calculating the characteristic values ​​of each sensor within each window, such as the mean of the spindle current, the root mean square of vibration, the peak value of acoustic emission, and the rate of temperature change, thus forming a feature vector for each time period, which is convenient for quality prediction and control optimization.

[0030] S2: Based on the multi-source data, predict the control quality to obtain the predicted quality.

[0031] Furthermore, this application also includes: the predicted quality includes tool wear, surface roughness, thermal geometry error, and vibration stability margin.

[0032] Specifically, multi-source data-based predictive control quality utilizes multi-source signals collected during machining, such as spindle current, feed current, vibration, acoustic emission, temperature, encoder displacement, and online measurement data, as inputs. A predictive model is then established to estimate the quality performance during machining. Multi-source data refers to signals from different sensors that collectively reflect the machine tool's operating status and machining conditions. Control quality refers to the performance indicators of the workpiece's machining results, including tool wear, surface roughness, dimensional accuracy, and thermal geometric errors. These multiple indicators determine whether the workpiece meets the expected machining requirements. Prediction involves using historical and real-time data, through a trained control quality prediction channel, to estimate the possible state of machining quality in advance.

[0033] Predicted quality refers to the estimated machining quality obtained after calculation by a predictive model, reflecting the potential machining level of the workpiece under current multi-source data conditions. Predicted quality is a multi-dimensional composite index; for example, tool wear is predicted to be 0.2 mm, surface roughness 0.8 μm, thermal geometry error 0.05 mm, and vibration stability margin 0.3 mm / s. Predicted quality can determine whether the machining process is safe and efficient, and provides a basis for subsequent online control constraints and optimization.

[0034] Furthermore, predicted quality includes tool wear, surface roughness, thermal geometry error, and vibration stability margin. Tool wear refers to the prediction of the degree of wear on the cutting edge of the tool during machining using multi-source data. It manifests as dulling of the cutting edge or surface damage, and its increase directly affects machining accuracy and workpiece surface quality. Surface roughness reflects the microscopic geometric characteristics of the workpiece surface texture. Surface roughness is numerically measured to measure the smoothness of the machined surface. For example, a predicted surface roughness of 0.8 micrometers indicates a relatively smooth surface, while a roughness exceeding 1.2 micrometers may not meet precision machining requirements. Predicting changes in surface roughness can help optimize cutting speed and feed rate, ensuring the workpiece meets design requirements while avoiding excessive energy consumption. Thermal geometry error refers to the structural thermal deformation caused by temperature rise during machining, resulting in deviations in the relative position between the tool and the workpiece. Thermal geometry error can lead to a decrease in dimensional accuracy. For example, during prolonged cutting, a 10-degree Celsius increase in spindle temperature may cause thermal expansion of 0.05 millimeters, causing the part to exceed tolerances. By predicting thermal geometry error, cooling or compensation measures can be taken in advance to ensure the stability of machining accuracy. Vibration stability margin refers to the stability allowance for a machine tool to avoid chatter during the cutting process. Chatter is a self-excited vibration phenomenon that seriously affects surface quality and tool life. The larger the vibration stability margin, the stronger the machine tool's anti-chatter capability under the current operating conditions.

[0035] S3: Generate online control constraints based on the predicted quality and machine tool safety boundary.

[0036] Furthermore, this application also includes: introducing the machine tool safety boundary, comparing the predicted quality with the machine tool safety boundary, adjusting the machine tool safety boundary to generate a feasible interval; mapping the feasible interval to machine tool adjustable parameters to generate online control constraints.

[0037] Specifically, introducing machine tool safety boundaries refers to the constraints set during CNC machining to ensure that the machine tool and workpiece operate within safe limits. Machine tool safety boundaries can be determined by limit parameters provided by the equipment manufacturer, industry standards, or historical operating data. Examples include a maximum spindle speed of no more than 8000 rpm, a feed rate of no more than 500 mm / s, tool wear of no more than 0.3 mm, and a maximum temperature of 80 degrees Celsius. These safety boundaries serve as protective measures to prevent equipment damage or workpiece failure due to overload, overheating, or excessive wear.

[0038] By comparing predicted quality with the machine tool's safety boundary, adjusting the machine tool's safety boundary to generate a feasible range involves comparing and analyzing the predicted machining quality indicators with the machine tool's safety boundary, and dynamically adjusting the safety boundary based on the actual machining conditions. For example, if the predicted tool wear is 0.25 mm and the safety boundary is 0.3 mm, the feasible range is between 0 and 0.25 mm, leaving a certain safety margin. If the predicted temperature is 75 degrees Celsius and the safety boundary is 80 degrees Celsius, the feasible range is between 0 and 75 degrees Celsius. Dynamic adjustment ensures that the machine's performance limits are approached while avoiding the risk of failure.

[0039] Mapping feasible intervals to adjustable machine tool parameters to generate online control constraints involves transforming safe and feasible intervals into parameter ranges that the machine tool can directly adjust, such as spindle speed, feed rate, depth of cut, cutting width, and trajectory acceleration / deceleration. The mapping process converts predicted quality and feasible intervals into control conditions. For example, if tool wear is predicted to be close to its upper limit, the feed rate is reduced by 10 mm / s to extend tool life; if temperature is predicted to approach its boundary, the depth of cut is reduced by 0.1 mm to reduce heat generation. The resulting online control constraints can then serve as the basis for subsequent optimization and execution.

[0040] S4: Under the online control constraints, the target control parameters are obtained by solving the optimization problem with the machining cycle, tool wear increment and instantaneous power as objectives.

[0041] Furthermore, this application also includes: inputting the online control constraints into an optimization objective function and outputting first optimization control parameters, wherein the optimization objective function includes optimization objectives for the machining cycle, the tool wear increment, and the instantaneous power; calculating machining performance indicators based on the first optimization control parameters, solving for the optimization control parameters that minimize the machining performance indicator values ​​under the online control constraints, and obtaining the target control parameters.

[0042] Furthermore, this application also includes: the target control parameters include spindle speed, feed rate, depth of cut, width of cut, and trajectory acceleration / deceleration.

[0043] Specifically, inputting online control constraints into the optimization objective function and outputting the first optimized control parameters means incorporating the online control constraints as input conditions into the mathematical optimization model and obtaining the initial control parameters through optimization. Online control constraints are the parameter ranges formed by the machine tool under the combined effects of safety boundaries and predicted quality. For example, the spindle speed is between 6000 and 7500 rpm, the feed rate is between 300 and 400 mm / s, and the depth of cut cannot exceed 1.2 mm. The optimization objective function is a mathematical expression used to measure the machining effect, including three optimization objectives: machining cycle, tool wear increment, and instantaneous power. The machining cycle represents the time required to complete the machining task, the tool wear increment represents the increase in tool wear during machining, and the instantaneous power represents the energy consumption level of the machine tool during cutting. By integrating the optimization objectives into the function, a set of reasonable control parameters, i.e., the first optimized control parameters, can be found under the condition of satisfying the constraints. The optimization objective function is as follows: , For processing cycle, For tool wear increment, Instantaneous power , , This is a weighting factor used to balance the importance of different objectives.

[0044] The processing performance index is calculated by weighting the first optimal control parameter. The optimal control parameter that minimizes the processing performance index value under online control constraints is then obtained. This involves substituting the initially obtained control parameters into the calculation formula for the processing performance index and balancing different optimization objectives using a weighted approach. (Processing performance index) It is a comprehensive evaluation value used to measure the overall level of machining efficiency, tool life, and energy consumption, where the weighting factors are... , , The weighting of machining performance is determined by the machining requirements. For example, in high-speed machining, the machining cycle time is more important, so time has a greater weight; in precision machining, tool life is more important, so tool wear has a greater weight. The smaller the machining performance index obtained after weighted calculation, the better the machining scheme. Among all possible combinations of control parameters, the set of parameters that minimizes the machining performance index is selected as the final target control parameters. For example, if the final spindle speed is determined to be 6800 rpm, the feed rate is 350 mm / s, and the depth of cut is 1 mm, the machining cycle time can be controlled to around 30 seconds while ensuring that the tool wear increment does not exceed 0.2 mm and the instantaneous power is kept below 9 kW.

[0045] Furthermore, the target control parameters include spindle speed, feed rate, depth of cut, width of cut, and trajectory acceleration / deceleration. Spindle speed refers to the rotational speed of the machine tool spindle per unit time. Spindle speed directly affects the cutting speed, thus affecting tool wear and surface roughness. For example, when the spindle speed increases from 6000 rpm to 7000 rpm, cutting efficiency increases, but tool wear also accelerates; therefore, a trade-off must be made between efficiency and tool life. Feed rate refers to the linear velocity of the relative movement of the tool or workpiece during machining. Higher feed rates result in higher machining efficiency, but cutting forces and surface roughness are also affected. For example, increasing the feed rate from 300 mm / s to 350 mm / s shortens machining time, but the surface roughness may increase from 0.7 μm to 0.9 μm; therefore, a reasonable selection needs to be made based on machining requirements. Depth of cut refers to the depth to which the tool penetrates the workpiece material in a single pass. The depth of cut determines the amount of material removed in a single cut. For example, increasing the depth of cut from 0.5 mm to 0.8 mm increases the removal efficiency by 60%, but also increases spindle load and vibration risk. Therefore, optimization requires balancing efficiency and machine tool safety boundaries. Cut width refers to the width of the contact between the tool and the workpiece, i.e., the dimension of the tool's cutting action in the transverse direction. Increasing the cut width significantly improves the material removal rate, but also leads to increased power consumption and thermal errors. For example, increasing the cut width from 1 mm to 2 mm doubles the removal rate, but the spindle power may increase from 2 kW to 3.5 kW. Track acceleration and deceleration refer to the dynamic adjustment of speed by the machine tool according to changes in the path during movement, used to ensure machining accuracy and equipment safety. Improper acceleration and deceleration control can cause track errors or vibrations. For example, insufficient deceleration at corners may lead to a track deviation of 0.02 mm, while a reasonable acceleration and deceleration strategy can control the deviation within 0.01 mm.

[0046] S5: The target control parameters are sent out for execution. When an abnormal limit is detected, the execution parameters are rolled back, and the rollback result is used as feedback to incrementally update the predictive control quality process.

[0047] Furthermore, this application also includes: collecting multi-source real-time data to determine whether abnormal limits have been exceeded, and obtaining an anomaly detection result; executing a rollback strategy based on the anomaly detection result to obtain the safety control parameters after rollback; calculating the error between the safety control parameters and the target control parameters, and performing incremental updates.

[0048] Specifically, collecting multi-source real-time data to determine whether abnormal limits have been exceeded and obtaining anomaly detection results means that multiple signals of machine tool operation are continuously collected by sensors during the machining process, including spindle current, feed current, vibration, acoustic emission, temperature, displacement and online measurement results, and compared with pre-set safety thresholds or predicted quality boundaries. If the signal exceeds the allowable range, it is determined to be an abnormal limit exceedance, and anomaly detection results are obtained, thereby triggering subsequent processing.

[0049] The rollback strategy, implemented based on anomaly detection results, determines the safety control parameters after rollback. This means that when an anomaly is detected exceeding limits, the machine tool's control parameters are automatically adjusted to a more conservative range according to a preset strategy to ensure machining safety. The rollback strategy is a pre-set emergency measure, such as reducing the spindle speed by 10% when the spindle current abnormally increases, or reducing the depth of cut by 0.1 mm when tool wear approaches its limit. The safety control parameters after rollback are the adjusted actual execution values. For example, the target control parameters might have a feed rate of 350 mm / s, but after an anomaly is triggered, it rolls back to 320 mm / s to reduce load and risk.

[0050] Incremental updates calculate the error between the safety control parameters and the target control parameters. This involves comparing the rolled-back safety parameters with the originally optimized target parameters, calculating the error, and using this error to correct the prediction model. The error refers to the deviation between the target parameters and the actual executed parameters. For example, if the target spindle speed is 6800 rpm, and the rolled-back safety value is 6500 rpm, the difference is 300 rpm. Incremental updates correct the prediction model through error correction, enabling the next optimization to more accurately consider abnormal situations, thereby improving the model's adaptability and robustness to complex machining environments.

[0051] Furthermore, this application also includes: extracting a real-time window of the multi-source data, and obtaining multiple consecutive historical scale windows of the real-time window; extracting multi-source sample data and control quality samples based on the multiple consecutive historical scale windows; training a control quality prediction channel using the multi-source sample data as the sample input set and the control quality samples as the sample output set to obtain a predicted control quality coefficient; and obtaining a control quality prediction channel when the predicted control quality coefficient meets a control quality coefficient threshold.

[0052] Furthermore, this application also includes: performing window adaptation on the multi-scale input samples and multi-scale output samples of the continuous multiple historical scale windows to obtain multi-source sample data and control quality samples; wherein, a dynamic window scale is configured based on the real-time window, and insufficient scale samples are filled and excess scale samples are pruned according to the dynamic window scale.

[0053] Specifically, continuously acquired multi-source signals are divided into current analysis windows according to preset time lengths. Simultaneously, multiple historical data segments of varying lengths are selected to form historical scale windows, used to capture short-term and long-term data change characteristics. The real-time window refers to the data set within a fixed time period used for analysis at the current moment, while the historical scale window refers to data from several consecutive time periods preceding the real-time window.

[0054] Window adaptation of multi-scale input and output samples across multiple historical scale windows to obtain multi-source sample data and control quality samples involves uniformly processing historical input signals of different lengths and time spans with corresponding output machining quality data, ensuring that each sample matches in both the time and feature dimensions. Multi-scale input samples refer to sensor features extracted from multiple historical windows over different time periods, such as spindle current, feed current, vibration, and temperature. Multi-scale output samples refer to machining quality indicators corresponding to the input samples, such as tool wear, surface roughness, and dimensional deviations. Through window adaptation, multi-scale samples are integrated into multi-source sample data and control quality samples in a unified format, facilitating model training and predictive analysis.

[0055] The dynamic window scale configuration based on the real-time window refers to dynamically adjusting the length or scale of the historical window according to the characteristics of the current real-time processing window, such as processing speed, tool wear rate, or workpiece material hardness, so that it can more accurately reflect the status of the processing process. The imputation of insufficient scale samples and the pruning of excess scale samples based on the dynamic window scale mean that when the historical window length is insufficient, missing data points are filled through interpolation or resampling; when the historical window length is too long or contains redundant data, the amount of data is reduced through pruning or downsampling to ensure the consistency and comparability of input and output samples in feature dimensions. For example, if the current real-time window length is 1 second and the historical window length is 0.8 seconds, linear interpolation is used to imput insufficient scale samples; if the historical window length is 1.5 seconds, the data exceeding 0.5 seconds is pruned to match the real-time window length.

[0056] Using multi-source sample data as the input set and control quality samples as the output set, a control quality prediction channel is trained to obtain the predicted control quality coefficients. In other words, the control quality prediction channel is trained using both input and output samples, enabling it to predict processing quality based on the input multi-source data. The control quality prediction channel refers to the prediction model channel, which can output processing quality estimates in real time. The predicted control quality coefficients are performance indicators obtained after model training, such as prediction error or confidence level, used to measure the model's predictive ability and reliability. For example, a root mean square error (RMSE) of less than 0.1 micrometers indicates high prediction accuracy.

[0057] When the predicted control quality coefficient meets the control quality coefficient threshold, the control quality prediction channel is obtained. This means that after training, if the predicted quality coefficient reaches the preset reliability standard, such as the error threshold or confidence threshold, the control quality prediction channel is considered reliable and can be used for quality prediction and control decisions in subsequent real-time processing. This ensures that the control quality prediction channel has sufficient accuracy and stability in actual use and avoids low-precision prediction from affecting the processing control effect.

[0058] In summary, the CNC machine tool control method based on real-time machining data feedback provided in this application has the following technical effects: by realizing the technical goal of constructing a predictive control closed-loop system based on multi-source real-time data feedback and dynamically generating and optimizing control parameters during machine tool machining, it achieves the technical effects of improving the accuracy of machining quality prediction, realizing adaptive adjustment of machine tool control, extending tool life, and ensuring a steady increase in production efficiency.

[0059] Example 2: Based on the same inventive concept as the CNC machine tool control method based on real-time machining data feedback in the foregoing examples, this application also provides a CNC machine tool control system based on real-time machining data feedback. Please refer to the appendix. Figure 2 The system includes: a multi-source dataset module 1, used to collect multi-source data from the CNC machine tool during the machining process; a predicted quality acquisition module 2, used to predict control quality based on the multi-source data; a constraint generation module 3, used to generate online control constraints based on the predicted quality and the machine tool safety boundary; a parameter acquisition module 4, used to optimize the target control parameters under the online control constraints, with machining cycle, tool wear increment, and instantaneous power as objectives; and an execution module 5, used to send the target control parameters for execution. When an abnormal limit is detected, the execution parameters are rolled back, and the rollback result is used as feedback to incrementally update the predicted control quality process.

[0060] Furthermore, the CNC machine tool control system based on real-time machining data feedback is also used for: collecting multi-source real-time data of the CNC machine tool for time calibration and alignment, and outputting a time-calibrated and aligned multi-source data matrix, wherein the multi-source real-time data includes spindle current, feed current, vibration, acoustic emission, temperature, encoder displacement, and online measurement data; denoising and preprocessing the multi-source data matrix to output a multi-source signal sequence; dividing the multi-source signal sequence into a real-time sliding window, and extracting multi-source data according to the real-time sliding window.

[0061] Furthermore, the CNC machine tool control system based on real-time machining data feedback is also used for: extracting the real-time window of the multi-source data, and obtaining multiple consecutive historical scale windows of the real-time window; extracting multi-source sample data and control quality samples according to the multiple consecutive historical scale windows; training the control quality prediction channel with the multi-source sample data as the sample input set and the control quality samples as the sample output set to obtain the predicted control quality coefficient; and obtaining the control quality prediction channel when the predicted control quality coefficient meets the control quality coefficient threshold.

[0062] Furthermore, the CNC machine tool control system based on real-time machining data feedback is also used to: perform window adaptation on the multi-scale input samples and multi-scale output samples of the continuous multiple historical scale windows to obtain multi-source sample data and control quality samples; wherein, a dynamic window scale is configured based on the real-time window, and insufficient scale samples are filled and excess scale samples are clipped according to the dynamic window scale.

[0063] Furthermore, the CNC machine tool control system based on real-time machining data feedback is also used to: introduce the machine tool safety boundary, compare the predicted quality with the machine tool safety boundary, adjust the machine tool safety boundary to generate a feasible range; and map the feasible range to the machine tool adjustable parameters to generate online control constraints.

[0064] Furthermore, the CNC machine tool control system based on real-time machining data feedback is also used to: input the online control constraints into an optimization objective function and output a first optimization control parameter, wherein the optimization objective function includes optimization objectives for the machining cycle, the tool wear increment, and the instantaneous power; calculate machining performance indicators based on the first optimization control parameter, solve for the optimization control parameter that minimizes the machining performance indicator value under the online control constraints, and obtain the target control parameter.

[0065] Furthermore, the CNC machine tool control system based on real-time machining data feedback is also used for: collecting multi-source real-time data to determine whether abnormal limits have been exceeded, and obtaining abnormal detection results; executing a rollback strategy based on the abnormal detection results to obtain the safety control parameters after rollback; calculating the error between the safety control parameters and the target control parameters, and performing incremental updates.

[0066] Furthermore, the CNC machine tool control system based on real-time machining data feedback is also used for: the predicted quality includes tool wear, surface roughness, thermal geometry error and vibration stability margin.

[0067] Furthermore, the CNC machine tool control system based on real-time machining data feedback is also used for: the target control parameters include spindle speed, feed rate, depth of cut, width of cut, and trajectory acceleration / deceleration.

[0068] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The CNC machine tool control method and specific examples based on real-time machining data feedback in the foregoing embodiment one are also applicable to the CNC machine tool control system based on real-time machining data feedback in this embodiment. Through the foregoing detailed description of the CNC machine tool control method based on real-time machining data feedback, those skilled in the art can clearly understand the CNC machine tool control system based on real-time machining data feedback in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0070] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A CNC machine tool control method based on real-time machining data feedback, characterized in that, include: Collect multi-source data during the machining process of CNC machine tools; Based on the multi-source data, the predicted control quality is obtained; Online control constraints are generated based on the predicted quality and machine tool safety boundaries; Under the online control constraints, the target control parameters are obtained by solving the optimization problem with the machining cycle, tool wear increment and instantaneous power as objectives; The target control parameters are sent out for execution. When an abnormality is detected exceeding the limit, the execution parameters are rolled back. The rollback result is used as feedback to incrementally update the predictive control quality process.

2. The CNC machine tool control method based on real-time machining data feedback as described in claim 1, characterized in that, Collect multi-source data during the machining process of CNC machine tools, including: The system collects multi-source real-time data from CNC machine tools, performs time calibration and alignment, and outputs a time-calibrated and aligned multi-source data matrix. The multi-source real-time data includes spindle current, feed current, vibration, acoustic emission, temperature, encoder displacement, and online measurement data. The multi-source data matrix is ​​denoised and preprocessed to output a multi-source signal sequence; The multi-source signal sequence is divided into real-time sliding windows, and multi-source data is extracted based on the real-time sliding windows.

3. The CNC machine tool control method based on real-time machining data feedback as described in claim 1, characterized in that, Based on the multi-source data, predictive control quality is obtained, including the following: Extract the real-time window of the multi-source data, and obtain multiple consecutive historical scale windows of the real-time window; Based on the aforementioned multiple consecutive historical scale windows, extract multi-source sample data and control quality samples; Using the multi-source sample data as the sample input set and the control quality samples as the sample output set, the control quality prediction channel is trained to obtain the predicted control quality coefficients. When the predicted control quality coefficient meets the control quality coefficient threshold, the control quality prediction channel is obtained.

4. The CNC machine tool control method based on real-time machining data feedback as described in claim 3, characterized in that, Based on the aforementioned multiple consecutive historical scale windows, multi-source sample data and control quality samples are extracted, including: Window adaptation is performed on the multi-scale input samples and multi-scale output samples of the continuous multiple historical scale windows to obtain multi-source sample data and control quality samples. Specifically, a dynamic window scale is configured based on the real-time window, and insufficient scale samples are filled and excessive scale samples are clipped according to the dynamic window scale.

5. The CNC machine tool control method based on real-time machining data feedback as described in claim 1, characterized in that, Based on the predicted quality and machine tool safety boundary, online control constraints are generated, including: Introducing the machine tool safety boundary, and combining the predicted quality with the machine tool safety boundary, adjusting the machine tool safety boundary to generate a feasible range; The feasible range is mapped to the adjustable parameters of the machine tool to generate online control constraints.

6. The CNC machine tool control method based on real-time machining data feedback as described in claim 1, characterized in that, Under the online control constraints, the target control parameters are obtained by solving the optimization problem with the machining cycle, tool wear increment, and instantaneous power as objectives, including: The online control constraints are input into the optimization objective function, and the first optimized control parameters are output. The optimization objective function includes the optimization objectives of the machining cycle, the tool wear increment, and the instantaneous power. The processing performance index is calculated by weighting the first optimized control parameters, and the optimized control parameters that minimize the processing performance index value under the online control constraints are solved to obtain the target control parameters.

7. The CNC machine tool control method based on real-time machining data feedback as described in claim 1, characterized in that, When an abnormal limit is detected, parameter rollback is performed, and the rollback result is used as feedback to incrementally update the predictive control quality process, including: Collect real-time data from multiple sources to determine whether any abnormal limits have been exceeded, and obtain anomaly detection results; Based on the anomaly detection results, a rollback strategy is executed to obtain the safety control parameters after rollback. Calculate the error between the safety control parameters and the target control parameters, and perform incremental updates.

8. The CNC machine tool control method based on real-time machining data feedback as described in claim 1, characterized in that, The predicted quality includes tool wear, surface roughness, thermal geometry error, and vibration stability margin.

9. The CNC machine tool control method based on real-time machining data feedback as described in claim 1, characterized in that, The target control parameters include spindle speed, feed rate, depth of cut, width of cut, and trajectory acceleration / deceleration.

10. A CNC machine tool control system based on real-time machining data feedback, characterized in that, The steps for implementing the CNC machine tool control method based on real-time machining data feedback as described in any one of claims 1 to 9 include: The multi-source dataset module is used to collect multi-source data from CNC machine tools during the machining process; The prediction quality acquisition module is used to predict control quality based on the multi-source data and obtain the predicted quality. The constraint generation module is used to generate online control constraints based on the predicted quality and machine tool safety boundary. The parameter acquisition module is used to solve for and optimize the target control parameters under the online control constraints, with the machining cycle, tool wear increment and instantaneous power as the objectives. The execution module is used to send out the target control parameters for execution. When an abnormal limit is detected, the execution parameters are rolled back, and the rollback result is used as feedback to incrementally update the predictive control quality process.

Images