Digital-twin-based nc machine tool machining anomaly detection and correction method and system

Abstract

The application discloses a numerical control machine tool machining anomaly detection and correction method and system based on digital twinning, and belongs to the field of machine tool machining anomaly prediction. The method performs machine tool machining coupling noise adaptability monitoring on parameters for quantifying transient impact noise and low-frequency chatter noise interference intensity, so as to decide to start transient impact noise suppression processing and baseline drift correction processing. After the machine tool machining coupling noise adaptability monitoring ends, machine tool multi-source heterogeneous feature eligibility evaluation is performed, and based on the evaluation result, a decision is made to start digital twinning virtual-real mapping operation. After the digital twinning virtual-real mapping operation ends, machine tool machining anomaly detection is started, and based on the detection result, a decision is made to take machine tool machining anomaly correction, so as to realize the effect of improving the accuracy of numerical control machine tool machining anomaly detection and the reliability of anomaly correction.

Description

Technical Field

[0001] This invention relates to the field of machine tool machining anomaly prediction technology, and in particular to a method and system for detecting and correcting CNC machine tool machining anomalies based on digital twins. Background Technology

[0002] The manufacturing industry is currently undergoing a profound transformation towards high precision, high efficiency, and intelligent manufacturing. CNC machine tools, as core processing equipment in high-end manufacturing, widely empower key industrial chains such as aerospace and high-end automotive parts. Their processing accuracy and operational stability directly determine the quality of the final product. The core components of a CNC machine tool mainly include the spindle, cutting tool, guide rails, ball screw, and feed system. The spindle is responsible for driving the cutting tool or workpiece to rotate at high speed; its speed stability and rigidity directly affect cutting accuracy and processing efficiency. The ball screw enables the precise conversion of rotary motion to linear motion. The feed system, composed of servo motors and reducers, controls each axis to feed precisely along the programmed trajectory, determining the accuracy of the machining contour. Simultaneously, with the development of high-end manufacturing... The stringent requirements of the field for micron-level or even submicron-level machining accuracy further necessitate upgrades to the real-time monitoring and precise control capabilities of CNC machine tool machining processes. Digital twin technology, as a core supporting technology for intelligent manufacturing, can achieve real-time perception of physical states and synchronous evolution of virtual models by constructing a full-element, high-fidelity mapping model between physical entities and virtual spaces. This breaks through the limitations of traditional monitoring technologies, such as offline measurement or single-sensor monitoring. However, in actual machining conditions, various machining anomalies can easily occur due to the combined effects of multiple factors, such as tool wear and breakage, equipment thermal deformation, and guide rail wear. Traditional monitoring technologies cannot provide early warnings of potential anomalies, nor can they quickly locate the root cause of anomalies that have already occurred, leading to waste of raw materials and increased production costs.

[0003] The main implementation process for detecting machining anomalies in existing CNC machine tools is as follows: First, a high-fidelity digital twin (including a 3D geometric model, a multi-physics coupled physical model, and a machining operation behavior model) that accurately maps all elements of the physical CNC machine tool is constructed. Next, multi-source heterogeneous data of the machine tool during operation is collected (such as spindle vibration signals collected by vibration sensors installed at the spindle end, and temperature data collected by temperature sensors installed at the spindle bearings and ball screws). Then, the collected multi-source heterogeneous data is preprocessed (e.g., removing environmental noise and data standardization) to extract machine tool characteristic parameters strongly correlated with machining anomalies (such as vibration...). The standardized machine tool characteristic data is then input into the digital twin, which simultaneously outputs the theoretical prediction values ​​of the digital twin (such as theoretical vibration values ​​and theoretical cutting force values). Subsequently, the multi-dimensional deviation data between the theoretical prediction values ​​of the digital twin and the actual monitored values ​​of the CNC machine tool (such as vibration deviation and temperature deviation) is calculated. This multi-dimensional deviation data and the machine tool characteristic parameters are then input into the anomaly identification model (such as the random forest model), which outputs the abnormal results of CNC machine tool processing (such as tool wear, thermal deformation, geometric errors, vibration anomalies, etc.), and simultaneously locates the location of the anomaly. Finally, corrections are made based on the CNC machine tool processing problems and the location of the anomaly.

[0004] For example, the Chinese invention patent application CN120802844A discloses a digital twin-based program adjustment system for vertical CNC machine tools, which includes an association weight allocation module and a detection sequence division module. The association weight allocation module, combined with the type of the simulation code under test and historical adjustment control information, assigns association weights to each simulation code under test. The detection sequence division module, in conjunction with the association weights of the simulation codes under test, determines the corresponding detection sequence, performs probability prediction on multiple associated control codes that trigger abnormal features, and divides the subsequent maintenance sequence based on the trigger probability.

[0005] However, in the process of implementing the inventive technical solution in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems: In the process of detecting abnormalities in CNC machine tool processing, existing technologies determine the correlation degree of codes through weight allocation, divide the detection sequence according to the correlation weight, and then predict the probability of control codes that cause CNC machine tool processing abnormalities and plan the maintenance priority. However, such existing technical solutions do not design effective suppression mechanisms for coupled noise (such as transient impact noise, low-frequency noise, etc.), which may lead to deviations in subsequent allocation of correlation weights and prediction of abnormal code probabilities based on such distorted data. This can result in inaccurate positioning of abnormal control codes and unreasonable planning of maintenance sequences, ultimately affecting the accuracy of CNC machine tool processing abnormality detection and the reliability of subsequent correction operations.

[0006] When performing multi-source data acquisition under the condition of CNC machine tool performing multi-process continuous machining (such as alternating milling-drilling-boring processes) and machining large irregularly shaped workpieces (such as irregularly shaped frames of engineering machinery), the cutting parameters (such as speed, feed rate, etc.) frequently switch during the alternation of multiple processes, which will generate transient impact noise. In addition, the uneven rigidity distribution of large irregularly shaped workpieces and the inconsistent stress deformation during the machining process will cause low-frequency chatter noise. The transient impact noise and low-frequency chatter noise are coupled with each other, resulting in instantaneous spike disturbances and baseline drift in the sensor data, forming a complex noise field.

[0007] Existing technologies often employ general filtering algorithms with fixed parameters (such as mean filtering and simple low-pass filtering) to process data. They lack adaptive suppression mechanisms for coupled noise and fail to differentiate noise types for targeted filtering. This can lead to the inability to effectively filter out various types of coupled noise in multi-source heterogeneous machine tool data. Consequently, the extracted multi-source heterogeneous machine tool features (such as vibration signal kurtosis and cutting force fluctuation amplitude deviation) become distorted. When performing digital twin virtual-real mapping and real-time state simulation based on these distorted machine tool features, the theoretical prediction values ​​of the output digital twin (including theoretical vibration values ​​and theoretical cutting force values) show significant deviations. Ultimately, this results in low accuracy in CNC machine tool machining anomaly detection because the anomaly identification model cannot accurately capture real anomaly signals (easily misjudging noise superposition interference as machining anomalies, or masking real anomalies such as excessive tool wear due to signal distortion). Furthermore, when correcting anomalies based on low-accuracy anomaly detection results, problems such as mismatched correction strategies and inaccurate compensation amounts can easily occur, leading to low reliability in anomaly correction. This presents the technical challenges of low accuracy in CNC machine tool machining anomaly detection and low reliability in anomaly correction. Summary of the Invention

[0008] To address the technical problems of low accuracy in detecting and reliably correcting machining anomalies in existing CNC machine tools, this invention provides a method and system for detecting and correcting machining anomalies in CNC machine tools based on digital twins. The technical solution is as follows: On the one hand, a method for detecting and correcting machining anomalies in CNC machine tools based on digital twins is provided. This method includes: monitoring the machine tool machining coupling noise adaptability of parameters that quantify the interference intensity of transient impact noise and low-frequency chatter noise to determine whether to enable transient impact noise suppression processing and baseline drift correction processing; after the machine tool machining coupling noise adaptability monitoring is completed, initiating a machine tool multi-source heterogeneous feature qualification assessment, and determining whether to initiate digital twin virtual-real mapping operation based on the obtained assessment results; after the digital twin virtual-real mapping operation is completed, enabling machine tool machining anomaly detection, and determining whether to take machine tool machining anomaly correction measures based on the detection results.

[0009] On the other hand, a CNC machine tool machining anomaly detection and correction system based on digital twins is provided, including: a machine tool multi-source data noise monitoring module, a machine tool multi-source feature qualification monitoring module, and a machine tool machining anomaly monitoring module. The machine tool multi-source data noise monitoring module is used to monitor the machine tool machining coupling noise adaptability of parameters quantifying the intensity of transient impact noise and low-frequency chatter noise interference, in order to decide whether to enable transient impact noise suppression processing and baseline drift correction processing. The machine tool multi-source feature qualification monitoring module is used to initiate machine tool multi-source heterogeneous feature qualification assessment after the machine tool machining coupling noise adaptability monitoring is completed, and to determine whether to initiate digital twin virtual-real mapping operation based on the acquired assessment results. The machine tool machining anomaly monitoring module is used to enable machine tool machining anomaly detection after the digital twin virtual-real mapping operation is completed, and to decide whether to take machine tool machining anomaly correction measures based on the detection results.

[0010] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. The digital twin-based CNC machine tool machining anomaly detection and correction method provided by this invention monitors the machine tool machining coupling noise adaptability by quantifying the parameters of transient impact noise and low-frequency chatter noise interference intensity. This helps to filter transient impact noise and low-frequency chatter noise interference during machine tool machining, reduce noise-induced multi-source data distortion in machine tool machining, and provide a high-quality data foundation for subsequent machine tool machining feature extraction.

[0011] After the machine tool processing coupling noise adaptability monitoring is completed, the machine tool multi-source heterogeneous feature qualification assessment is initiated. Based on the obtained assessment results, it is determined whether to initiate the digital twin virtual-real mapping operation. This helps to identify the qualification level of the machine tool multi-source heterogeneous features, eliminate unqualified features with hidden deviations and parameter distortions, reduce the number of unqualified features entering the digital twin mapping process, and ensure the accuracy of the virtual-real mapping between the digital twin and the physical machine tool.

[0012] After the digital twin virtual-real mapping operation is completed, machine tool processing anomaly detection is enabled. Based on the detection results, a decision is made on whether to take corrective action for machine tool processing anomalies. This helps to capture various processing anomalies such as tool wear and thermal deformation during CNC machine tool processing, accurately locate the root cause and location of the anomalies, and effectively reduce workpiece scrap rate and equipment wear through targeted anomaly correction. This ensures the stability and continuity of the CNC machine tool processing process, improves workpiece processing accuracy and product qualification rate, and achieves the effect of improving the low accuracy of CNC machine tool processing anomaly detection and the reliability of anomaly correction. It effectively solves the problems of low accuracy of CNC machine tool processing anomaly detection and low reliability of anomaly correction in the existing technology.

[0013] 2. This invention, by selectively choosing machine tool machining coupling noise quantification parameters that include transient impact characteristic parameters and low-frequency chatter characteristic parameters, helps to capture the key features of two types of core coupling noise in multi-process machining of CNC machine tools. It achieves targeted quantification of transient impact noise and low-frequency chatter noise, solving the problems of broad and insufficiently targeted selection of coupling noise detection parameters in existing technologies, which cannot distinguish the core differences between the two types of noise, leading to ambiguous noise interference identification and high rates of missed and false judgments. This improves the targeting and effectiveness of coupling noise monitoring. The invention verifies whether the transient impact characteristic parameters meet the transient impact judgment conditions. If the verification result meets the transient impact judgment conditions, transient impact noise suppression processing is adopted; if the verification result does not meet the transient impact judgment conditions, low-frequency chatter noise discrimination based on low-frequency chatter characteristic parameters continues. This helps to achieve hierarchical and progressive accurate discrimination of the two types of coupling noise, prioritizing... This approach identifies transient impact noise with a more significant impact, avoiding confusion between the two types of noise. This reduces data spike distortion and equipment wear caused by transient impact noise, while also preventing ineffective processing that consumes system resources. It improves the targeting and efficiency of noise processing, addressing the shortcomings of existing technologies where noise discrimination lacks priority and processing is largely indiscriminate. Low-frequency flutter noise discrimination verifies whether the characteristic parameters of low-frequency flutter meet the criteria for low-frequency flutter. If the verification result meets the criteria, baseline drift correction is applied; if not, machine tool processing features are extracted. This helps to specifically address the baseline drift problem caused by low-frequency flutter, reducing the interference of low-frequency flutter noise with the accuracy of subsequent feature extraction. It overcomes the technical deficiencies of existing technologies that are inaccurate in identifying slowly varying noises like low-frequency flutter, which are easily overlooked, leading to data baseline drift and distortion in subsequent processing analysis.

[0014] 3. When CNC machine tools are used to process high-precision thin-walled irregular parts (such as thin-walled frames for aerospace applications, thin-walled cavities for precision instruments, etc.), micro-deformation is easily generated during the machining process, and the cutting load is prone to slight fluctuations. Simultaneously, the surface quality requirements of the workpiece are extremely high, which may lead to hidden deviations in some machine tool multi-source heterogeneous features (such as slight deviations in feature values ​​from theoretical values, or minor fluctuations in time-series features). The single machine tool feature compliance rate cannot identify such hidden anomalies, easily leading to misjudgments and affecting the accuracy of subsequent digital twin mapping. By specifically acquiring quantitative parameters for machine tool feature accuracy, including the relative deviation rate of machine tool features and the time-series fluctuation coefficient of machine tool features, it helps to solve the technical shortcomings of existing technologies where the single machine tool feature compliance rate can only determine whether a feature is within a preset threshold range, and cannot identify hidden anomalies such as slight deviations in feature values ​​and minor fluctuations in time-series features. This addresses the problem of missed or misjudged hidden deviations in high-precision machining scenarios, providing a more comprehensive and refined quantitative dimension for feature accuracy assessment. The result of weighted coupling of the machine tool feature accuracy quantification parameter and the corresponding machine tool feature accuracy influencing parameter serves as the machine tool feature accuracy index. This helps to balance the deviation of feature values ​​with temporal stability, while adapting the influence weight of different features on the twin mapping accuracy. It avoids the one-sidedness of single parameter evaluation, achieving comprehensive and accurate quantification of the accuracy of multi-source heterogeneous features of the machine tool. It determines whether the machine tool feature accuracy index is greater than the preset accuracy threshold. If so, the corresponding multi-source heterogeneous feature of the machine tool is marked as a qualified multi-source heterogeneous feature of the machine tool, and the virtual-real mapping operation of the digital twin is performed based on the qualified multi-source heterogeneous feature of the machine tool. Otherwise, a distortion prompt of multi-source heterogeneous feature of the machine tool is sent. This helps to reduce the input of distorted features into the digital twin from the source, ensure the reliability of the input basis of virtual-real mapping, improve the state synchronization accuracy of the physical machine tool and the digital twin, reduce the distortion of the twin model and the misjudgment of machining anomaly detection caused by implicit deviations, and thus ensure the machining quality and surface accuracy of high-precision thin-walled irregular parts. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart of a digital twin-based CNC machine tool machining anomaly detection and correction method provided in an embodiment of this application; Figure 2 A flowchart outlining the general overview of the digital twin-based CNC machine tool machining anomaly detection and correction method provided in this application embodiment; Figure 3A logic diagram for transient impact noise suppression processing in the digital twin-based CNC machine tool machining anomaly detection and correction method provided in the embodiments of this application; Figure 4 A baseline drift correction processing logic diagram for the CNC machine tool machining anomaly detection and correction method based on digital twin provided in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of a digital twin-based CNC machine tool machining anomaly detection and correction system provided in an embodiment of this application. Detailed Implementation

[0017] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0018] Example 1, as Figure 1 The diagram shows a flowchart of a digital twin-based CNC machine tool machining anomaly detection and correction method provided in Embodiment 1 of this application. The method includes the following steps: machine tool multi-source data noise monitoring. During the acquisition of multi-source machine tool machining data, parameters for quantifying the intensity of transient impact noise and low-frequency chatter noise interference are monitored for machine tool machining coupling noise adaptability to identify transient impact noise and low-frequency chatter noise interference during CNC machine tool machining. This monitoring determines whether to enable transient impact noise suppression processing and baseline drift correction processing. Transient impact noise suppression processing is used to suppress instantaneous spike disturbances caused by cutting parameter switching, reducing spike interference mixed into the multi-source machine tool machining data. Baseline drift correction processing is used to correct baseline drift interference caused by uneven workpiece rigidity during machine tool machining, correcting distorted data. Through multi-source machine tool data noise monitoring, it helps to filter out two types of core coupling noise interference from the source, reducing noise-induced machine tool machining multi-source data distortion and baseline drift problems, ensuring the integrity and authenticity of the original multi-source machine tool machining data, and providing high-quality data support for subsequent multi-source machine tool feature extraction.

[0019] Machine tool multi-source feature qualification monitoring involves conducting a multi-source heterogeneous feature qualification assessment after the machine tool machining coupling noise adaptability monitoring is completed. This assessment is used to identify the qualification of feature extraction. Based on the obtained assessment results, it is determined whether to initiate the digital twin virtual-physical mapping operation to achieve precise synchronization between the physical machine tool and the digital twin state and support the prediction of machining state. Through multi-source feature qualification monitoring, the effectiveness and accuracy of multi-source heterogeneous features of the machine tool can be identified, and unqualified features with hidden deviations and parameter distortions can be eliminated. This reduces the number of unqualified features entering the digital twin mapping process and ensures the reliability of the input basis for virtual-physical mapping.

[0020] Machine tool machining anomaly monitoring involves detecting and identifying abnormal machine tool machining states, types, and locations after the digital twin virtual-real mapping operation is completed. Based on the detection results, a decision is made on whether to implement machine tool machining anomaly correction. This correction is used to rectify various machining anomalies (such as tool wear, thermal deformation, geometric errors, and vibration anomalies) that occur during CNC machine tool machining, restoring the machine tool to its normal machining state. Through machine tool machining anomaly monitoring, the advantages of virtual-real collaboration of digital twins can be leveraged to accurately capture various machining anomalies, precisely locate their root causes and locations, and solve the problems of low anomaly detection accuracy and insufficient correction specificity in existing technologies. This ensures the stability of the CNC machine tool machining process and improves workpiece machining accuracy and product qualification rate.

[0021] In this embodiment, by monitoring machine tool multi-source data noise, machine tool multi-source feature conformity, and machine tool machining anomaly, a progressive monitoring system can be constructed, forming a complete closed loop from data purification to feature screening and anomaly handling. This improves the accuracy of CNC machine tool machining anomaly detection and the reliability of anomaly correction. Specifically, machine tool multi-source data noise monitoring provides a basic guarantee for machine tool multi-source feature conformity monitoring, improves the accuracy of feature extraction, and reduces the interference of noise residue on feature evaluation. Machine tool multi-source feature conformity monitoring builds a reliable bridge for machine tool machining anomaly monitoring, provides accurate twin model support for anomaly detection, and reduces misjudgments of anomalies caused by feature distortion.

[0022] It should be explained that the implementation and effectiveness of the digital twin-based CNC machine tool machining anomaly detection and correction method provided in this application embodiment rely on a pre-built dedicated digital twin data support system. This system integrates and archives full-dimensional machine tool machining data resources. The content includes core configuration parameters calibrated by machine tool process experts and maintenance teams, such as preset cutting parameter change rate thresholds, preset vibration acceleration peak thresholds, and preset cutting force mutation amplitude thresholds. At the same time, this data system includes standardized CNC machine tool historical machining data (covering vibration acceleration, cutting force, and displacement time series data under different workpiece types and different machining processes, as well as corresponding anomaly correction result records) and model iteration data (such as multi-round feature evaluation, anomaly detection model adjustment). The system includes parameter optimization and adjustment logs during the testing and twin calibration process. This data support system employs a hierarchical hybrid storage architecture. It manages structured parameters (such as cutting process parameter tables, noise processing weight coefficients, and anomaly correction priority rules) through a relational database, and stores unstructured data, such as raw sensor waveform sequences, machine tool acoustic data, and workpiece deformation image data, using a non-relational database. Technical personnel can periodically verify and dynamically optimize the preset parameters stored in the data system based on real-time collected machine tool processing data and anomaly correction feedback. This ensures that the data system can continuously adapt to the complex and ever-changing processing scenarios of CNC machine tools, such as high-precision thin-walled irregular parts processing, multi-process alternation, and gradual tool wear.

[0023] It should also be explained that the baseline drift correction method for CNC machine tool machining anomaly detection and correction based on digital twins provided in this application relies on a preset baseline compensation value mapping table. The logic, data support, and parameter association rules of this mapping table need further clarification. This mapping table, as the core carrier for accurately acquiring vibration and displacement baseline compensation values, can provide suitable compensation parameters for baseline drift correction under different machining conditions. The parameter correspondences stored within it are calibrated by technicians using CNC machine tool measured data and process experience, and then archived into the digital twin data support system, laying the foundation for the accuracy of baseline correction.

[0024] From the perspective of data composition, the construction of this baseline compensation value mapping table integrates the quantitative analysis results of multi-dimensional machine tool machining parameter combinations and baseline drift characteristics. The data sources include core drift parameters such as guide rail vibration acceleration baseline drift trend slope and displacement baseline drift accumulation, as well as working condition influencing parameters such as cutting load fluctuation, workpiece material stiffness, and machining feed rate, and multi-dimensional combination data of the two types of parameters. By systematically analyzing the influence of each parameter combination on the degree of baseline drift, it ensures that suitable compensation parameters can be found under different drift scenarios.

[0025] During the process of establishing the correlation, technicians will configure differentiated adaptation weights based on the degree of influence of parameter combinations on baseline stability. For example, when the cumulative displacement drift caused by micro-deformation of the workpiece exceeds the standard and the cutting load fluctuates drastically, a higher amplitude displacement baseline compensation value will be matched to offset the drift effect under the combined working conditions. At the same time, through statistical verification of long-term processing monitoring data, abnormal calibration data caused by non-working condition factors such as temporary zero drift of sensors and fluctuations in equipment power supply will be eliminated to reduce the interference of invalid data on the mapping relationship. This ensures that the correlation between each parameter combination and compensation value in the mapping table has engineering adaptability and statistical reliability, thereby ensuring the adaptability and correction accuracy of baseline drift correction processing in complex processing scenarios such as high precision and multiple processes.

[0026] like Figure 2 The diagram shown is a general flowchart of the CNC machine tool machining anomaly detection and correction method based on digital twin provided in this application embodiment. Figure 2 It is known that: machine tool processing coupling noise adaptability monitoring is performed, and transient impact characteristic parameters and low-frequency chatter characteristic parameters are obtained. It is determined whether the transient impact characteristic parameters meet the transient impact judgment conditions. If they do, transient impact noise suppression processing is adopted; otherwise, low-frequency chatter noise discrimination is performed, and the low-frequency chatter characteristic parameters meet the low-frequency chatter judgment conditions. If they do, baseline drift correction processing is adopted; otherwise, machine tool processing feature extraction is performed. After the machine tool processing feature extraction is completed, the machine tool multi-source heterogeneous feature qualification assessment is initiated, and the machine tool feature compliance rate is obtained. It is determined whether the machine tool feature compliance rate is greater than the preset feature compliance rate threshold. If not, a machine tool multi-source heterogeneous feature distortion prompt is sent; otherwise, a digital twin virtual-real mapping operation is performed. After the digital twin virtual-real mapping operation is completed, machine tool processing anomaly detection is enabled, and CNC machine tool processing anomaly results are obtained. It is determined whether the CNC machine tool processing anomaly result is non-empty. If so, targeted machine tool processing anomaly correction is performed; otherwise, the processing operation continues.

[0027] Preferably, the parameters for quantifying the interference intensity of transient impact noise and low-frequency chatter noise include transient impact characteristic parameters for determining transient impact noise interference and low-frequency chatter characteristic parameters for determining low-frequency chatter noise interference. The transient impact characteristic parameters include the rate of change of cutting parameters, the peak value of vibration acceleration, and the amplitude of sudden changes in cutting force. The specific acquisition process is as follows: By monitoring the values ​​of cutting parameters (such as cutting speed, feed rate, and depth of cut) before and after switching during multi-process alternating machining, the starting time of cutting parameter switching and the switching interval time are recorded. The ratio of the difference in cutting parameters before and after switching to the switching interval time is calculated and used as the rate of change of cutting parameters to characterize the severity of cutting parameter switching. Cutting parameters represent key parameters used to describe the core process state during CNC machine tool cutting, including process parameters that directly affect cutting load and machining stability, such as cutting speed, feed rate, and depth of cut. For example, suppose the steady-state value of a certain cutting parameter before switching is X1 (X1 represents the average value of the cutting parameter in the stable machining stage before switching), and the value after switching is X2 (X2 represents the value of the cutting parameter after switching), and the switching interval time is Δt (Δt represents the value from the time the cutting parameter switches to the time the ... The time interval from the issuance of the command to the switching of cutting parameters is calculated. The rate of change of the cutting parameter is obtained by calculating the ratio of the difference in cutting parameters before and after the switching to the switching interval, i.e., the rate of change of the cutting parameter = |X1-X2| / Δt. Different cutting parameters (cutting speed, feed rate, depth of cut) are calculated in the same way to obtain the corresponding rate of change. Vibration sensors installed at the end of the CNC machine tool spindle collect vibration acceleration data of the machine tool guideway within a preset time period before and after the start of the cutting parameter switching. Based on time-domain extremum extraction algorithms (such as sliding window extremum algorithm, peak search...),... The system uses algorithms such as the oscilloscope to perform time-domain analysis on the collected vibration acceleration data, extracts the maximum value of the vibration acceleration data within the time period, and uses it as the peak value of vibration acceleration to characterize the instantaneous impact intensity of the spindle. By using a cutting force sensor installed at the tool holder of the CNC machine tool, the cutting force data within a preset time period before and after the start time of the cutting parameter switching is collected. The difference between the maximum value of the cutting force after the switching and the steady-state value of the cutting force before the switching is calculated and used as the amplitude of the cutting force mutation to characterize the degree of the cutting load mutation. The steady-state value of the cutting force represents the average value of the cutting force data continuously collected by the cutting force sensor before the cutting parameter switching.

[0028] Low-frequency chatter characteristic parameters include the amplitude of the low-frequency component of vibration acceleration and the amplitude of displacement fluctuation. The specific acquisition process is as follows: Vibration sensors installed on the machine tool guideway collect vibration acceleration data of the guideway within a preset time period. Frequency domain decomposition processing is performed on the collected vibration acceleration data based on frequency domain decomposition algorithms (such as Fourier transform and wavelet decomposition algorithms). The maximum value of the vibration acceleration signal amplitude within the preset low-frequency band is extracted and used as the amplitude of the low-frequency component of vibration acceleration to characterize the low-frequency chatter intensity. The preset low-frequency band is pre-set by designated personnel to define the frequency range corresponding to low-frequency chatter noise during CNC machine tool machining, ensuring that the extracted characteristic parameters accurately reflect the low-frequency chatter. The actual intensity of the vibration; by using a displacement sensor installed on the machine tool table, the machining displacement data of the machine tool table relative to the workpiece is collected within a preset time period. The difference between the maximum and minimum values ​​of the machining displacement data within the preset time period is calculated and used as the displacement fluctuation amplitude to characterize the degree of displacement fluctuation caused by the deformation of the workpiece under stress; (the core object of this displacement is the workpiece to be processed, and the collected data is the machining position displacement of the workpiece driven by the machine tool table) Transient impact noise is judged based on transient impact characteristic parameters. The transient impact noise judgment indicates whether the transient impact characteristic parameters meet the transient impact judgment conditions. If the verification result meets the transient impact judgment conditions, it indicates that the frequent switching of cutting parameters leads to a sudden change in cutting load. The increased impact between the tool and the workpiece creates a potential for instantaneous spike disturbances. Therefore, transient impact noise suppression is implemented to precisely eliminate these spike disturbances, preventing them from infiltrating multi-source machine tool data and ensuring the reliability of subsequent data processing and feature extraction. If the verification result does not meet the transient impact judgment criteria, low-frequency chatter noise discrimination based on low-frequency chatter characteristic parameters continues. Specifically, the transient impact judgment criteria require that the rate of change of cutting parameters be greater than a preset cutting parameter rate of change threshold (e.g., the rate of change of cutting speed be greater than a preset cutting speed rate of change threshold), the peak value of vibration acceleration be greater than a preset peak value of vibration acceleration threshold, and the amplitude of the sudden change in cutting force be greater than a preset amplitude of the sudden change in cutting force threshold. For a transient impact judgment condition to be considered met, all three conditions mentioned above must be met simultaneously; otherwise, it is considered not to meet the transient impact judgment condition. Among them, the preset cutting speed change rate threshold is represented by the average value of the cutting speed change rate over a historical time period, the preset vibration acceleration peak value threshold is represented by the average value of the vibration acceleration peak value over a historical time period, the preset cutting force mutation amplitude threshold is represented by the average value of the cutting force mutation amplitude over a historical time period, and the machine tool machining multi-source data represents a collection of various types of monitoring data collected by various sensors during CNC machine tool machining, reflecting different machining state dimensions, including vibration acceleration data of machine tool guideways, cutting force data of spindles, and machining displacement data of worktables, etc.

[0029] Low-frequency chatter noise is identified based on low-frequency chatter characteristic parameters. The low-frequency chatter noise identification involves checking whether the low-frequency chatter characteristic parameters meet the low-frequency chatter judgment criteria. If the check result meets the low-frequency chatter judgment criteria, it indicates that the uneven rigidity distribution and inconsistent stress deformation of the large, irregularly shaped workpiece lead to fluctuations in the contact stiffness between the tool and the workpiece, resulting in baseline drift. Baseline drift correction is then implemented to reduce baseline drift caused by low-frequency chatter noise and decrease the distortion rate of multi-source machine tool machining data due to drift interference. If the check result does not meet the low-frequency chatter judgment criteria, the corresponding multi-source machine tool machining data is marked as qualified. Multi-source data is processed, and machine tool processing features are extracted based on the multi-source data processed by qualified machine tools. The low-frequency chatter judgment condition specifically refers to the amplitude of the low-frequency component of vibration acceleration being greater than the preset low-frequency acceleration component threshold, and the displacement fluctuation amplitude being greater than the preset displacement fluctuation amplitude threshold. Both of the above conditions must be met simultaneously to be considered as meeting the low-frequency chatter judgment condition; otherwise, it is considered as not meeting the low-frequency chatter judgment condition. The preset low-frequency acceleration component threshold is represented by the average value of the amplitude of the low-frequency component of vibration acceleration over a historical time period, and the preset displacement fluctuation amplitude threshold is represented by the average value of the displacement fluctuation amplitude over a historical time period.

[0030] As described above, quantifying the parameters of transient impact noise and low-frequency chatter noise interference intensity, as well as the discrimination of transient impact noise and low-frequency chatter noise, helps to capture the differentiated characteristics of the two types of core coupled noise during CNC machine tool machining. This enables targeted identification and hierarchical progressive judgment of transient impact noise and low-frequency chatter noise, effectively overcoming the technical shortcomings of existing technologies, such as the broad selection of coupled noise quantification parameters, lack of distinction between the two types of noise, and high rates of missed and false judgments. It reduces the blind processing or untimely handling of noise interference caused by noise confusion, ensuring the integrity and authenticity of the original multi-source machine tool data. At the same time, it reduces the consumption of system resources and the impact on machining efficiency caused by invalid noise processing, providing high-quality data support for subsequent multi-source heterogeneous feature extraction of machine tools, and improving the accuracy, adaptability, and engineering reliability of the entire anomaly detection and correction scheme from the source.

[0031] like Figure 3 The diagram shown is a logic diagram for transient impact noise suppression in the digital twin-based CNC machine tool machining anomaly detection and correction method provided in this application embodiment. Figure 3It is known that transient impact noise suppression processing is performed, and it is determined whether the peak value of vibration acceleration is greater than the preset vibration peak rejection threshold and whether the amplitude of the sudden change in cutting force is greater than the preset cutting force peak rejection threshold. If not, machine tool processing feature extraction is performed; otherwise, peak data time synchronization discrimination is performed, and it is determined whether the peak-parameter switching time deviation is less than the preset time deviation threshold. If not, machine tool processing feature extraction is performed; otherwise, peak disturbance suppression is performed. After the peak disturbance suppression is completed, it is determined whether the signal-to-noise ratio of the machine tool processing data is greater than the preset signal-to-noise ratio threshold. If so, low-frequency chatter noise discrimination is performed; otherwise, a transient impact noise suppression processing failure prompt is sent.

[0032] Preferably, the specific process for transient impact noise suppression is as follows: Accurately locate and represent instantaneous peak data; extract multi-source machine tool machining data (including vibration acceleration data and cutting force data) within a preset time period before and after the start of the cutting parameter switching; determine whether the peak vibration acceleration is greater than a preset vibration peak rejection threshold, and whether the sudden change in cutting force amplitude is greater than a preset cutting force peak rejection threshold. The preset vibration peak rejection threshold is used to define the standard amplitude of transient impact peaks caused by cutting parameter switching in the vibration acceleration data, distinguishing between normal machining vibration fluctuations and transient impact abnormal peaks, providing a quantitative basis for the identification and rejection of vibration-related peak data. According to reports, the preset cutting force peak rejection threshold is used to define the standard amplitude of instantaneous impact peaks in cutting force data caused by cutting parameter switching. This effectively distinguishes between normal cutting load fluctuations and abnormal instantaneous impact peaks, providing a quantitative basis for the identification and rejection of cutting force peak data. Both the preset vibration peak rejection threshold and the preset cutting force peak rejection threshold are set in advance by the preset personnel. If so, it indicates that there are potential instantaneous peak disturbances in the multi-source data of machine tool processing. The abnormal situation where the amplitude of vibration acceleration data and the amplitude of cutting force data simultaneously exceed the limits could be due to peak disturbances caused by transient impact noise resulting from cutting parameter switching, or it could be due to random interference in the processing environment or external factors. Amplitude anomalies caused by non-transient impact factors such as electromagnetic interference cannot be directly identified as peak data corresponding to transient impact noise. To avoid misidentifying non-transient impact interference as peak data and ensure positioning accuracy, it is necessary to obtain the occurrence time of the corresponding multi-source machine tool processing data and perform peak data time synchronization discrimination based on the occurrence time of the multi-source machine tool processing data. Conversely, the vibration acceleration data and cutting force data corresponding to the multi-source machine tool processing data are judged as qualified multi-source machine tool processing data, and machine tool processing features are extracted based on the qualified multi-source machine tool processing data. Peak data time synchronization discrimination refers to judging the peak-parameter switching time deviation. If the time deviation is less than a preset threshold, the vibration acceleration data and cutting force data corresponding to the multi-source data of machine tool processing are identified as instantaneous spike data caused by transient impact noise, thus completing the accurate positioning of the instantaneous spike data. Otherwise, the corresponding vibration acceleration data and cutting force data are identified as qualified multi-source data of machine tool processing, and machine tool processing features are extracted based on the qualified multi-source data of machine tool processing. The preset time deviation threshold is represented by the average value of the spike-parameter switching time deviation over a historical time period. The spike-parameter switching time deviation is represented by the result of the difference calculation between the time when the instantaneous spike data appears and the time when the cutting parameter switching starts.

[0033] Specifically, peak disturbance suppression based on instantaneous peak data is performed as follows: First, instantaneous peak data is removed from the multi-source data (vibration acceleration data, cutting force data) of machine tool machining to avoid residual peak interference. Second, a data segment with a preset interpolation compensation window duration before and after the peak data is selected. A linear interpolation algorithm is used to compensate and fill the gaps in the multi-source data (vibration acceleration data, cutting force data) of machine tool machining after peak removal, ensuring data continuity and integrity and avoiding data loss affecting subsequent analysis. The data segment represents the vibration acceleration data of the machine tool guideway and the cutting force data of the machine tool spindle that are not significantly affected by transient impact noise within the preset interpolation compensation window duration before and after the peak data occurrence. The preset interpolation compensation window duration is set in advance by the operator. After the transient impact noise suppression processing is completed, data is obtained to characterize the degree of residual data interference. The signal-to-noise ratio (SNR) of machine tool processing data is represented by the ratio of the effective signal amplitude of machine tool processing to the original machine tool processing signal amplitude before peak disturbance suppression. The effective signal amplitude is represented by averaging the amplitudes of multi-source machine tool processing data (vibration acceleration data, cutting force data) collected for a preset effect evaluation period after peak disturbance suppression. This average is used to uniformly quantify the overall strength of the effective signal after suppression. The preset effect evaluation period is pre-set by the designated personnel. The system determines whether the SNR of the machine tool processing data is greater than a preset SNR threshold. If so, it continues to perform low-frequency chatter noise discrimination based on low-frequency chatter characteristic parameters; otherwise, it sends a transient impact noise suppression failure warning. The preset SNR threshold is represented by the average value of the SNR of machine tool processing data over a historical time period.

[0034] As described above, transient impact noise suppression processing helps to eliminate vibrations and instantaneous peak data of cutting force caused by switching cutting parameters, effectively filtering the impact of peak interference on multi-source machine tool processing data, reducing the distortion rate of original data caused by the mixing of peak data, the risk of misjudgment in subsequent multi-source heterogeneous feature extraction of machine tools, and the impact wear of transient impacts on key components such as machine tool spindles and worktables. It also improves the integrity and authenticity of multi-source machine tool processing data, the accuracy of feature extraction, and the reliability of data support, achieving efficient purification and quality improvement of machine tool processing data, and providing a high-quality data foundation for subsequent virtual-real mapping of digital twins.

[0035] like Figure 4 The diagram shown is a baseline drift correction processing logic diagram of the CNC machine tool machining anomaly detection and correction method based on digital twin provided in this application embodiment. Figure 4It is known that: baseline drift correction is performed, and the correlation coefficient of low-frequency chatter is obtained. It is determined whether the correlation coefficient of low-frequency chatter is greater than the preset correlation coefficient threshold. If not, machine tool processing features are extracted; otherwise, low-frequency chatter adaptive baseline correction is performed. After completion, the low-frequency chatter noise correction result is verified, and the low-frequency interference suppression rate and baseline stability coefficient are obtained. It is determined whether the low-frequency interference suppression rate and baseline stability coefficient meet the low-frequency correction qualification conditions. If yes, machine tool processing features are extracted; otherwise, a low-frequency chatter noise correction failure prompt is sent.

[0036] Preferably, the specific process of baseline drift correction is as follows: Step 1, extract the time-domain baseline of vibration acceleration data within a preset low-frequency band, and the time-domain baseline of machining displacement data within a preset time period; the time-domain baseline represents the slow shift trend of data over time in the time domain dimension, characterized by a linear trend fitting line; the method for extracting the time-domain baseline of vibration acceleration data is as follows: sort the vibration acceleration data within the preset low-frequency band according to the acquisition time sequence, and generate the overall trend fitting line corresponding to the vibration acceleration data based on a linear fitting algorithm (such as least squares method, univariate linear regression method, etc.). The overall trend fitting line corresponding to the vibration acceleration data represents the slow shift pattern of machine tool guideway vibration acceleration data within the preset low-frequency band as the acquisition time progresses, which can quantify the degree of baseline drift caused by low-frequency flutter. The acquisition time sequence indicates the sequential time order of data acquisition by the machine tool guideway vibration acceleration sensor, used to clarify the temporal correlation of data points and avoid fitting deviations caused by temporal disorder; the method for extracting the time-domain baseline of machining displacement data is as follows: sort the machining displacement data within the preset time period according to the machining time sequence, and generate the overall trend fitting line corresponding to the vibration acceleration data based on a linear fitting algorithm. The process involves generating an overall trend fitting line corresponding to the machining displacement data. This line represents the slow shift of the machining displacement data of the machine tool table relative to the workpiece within a preset time period as the machining process progresses. It quantifies the baseline drift caused by factors such as workpiece micro-deformation and guide rail creep. The machining sequence indicates the order in which the CNC machine tool executes machining operations, clarifying the correspondence between the machining displacement data and the actual machining steps. Step two involves obtaining the baseline drift trend slope, which characterizes the rate of change of baseline drift, and the drift accumulation, which characterizes the overall degree of baseline drift. The baseline drift trend slope is represented by the slope value of the corresponding time-domain baseline, including the guide rail vibration acceleration baseline drift trend slope and the machining displacement baseline drift trend slope. The larger the absolute value of the slope, the faster the baseline drift rate. The drift accumulation is represented by the absolute value of the difference between the fitted value at the start and end of the data acquisition in the corresponding time-domain baseline, including the guide rail vibration acceleration drift accumulation and the machining displacement drift accumulation. The larger the absolute value of the difference, the more significant the overall baseline drift.

[0037] Step 3: Calculate the low-frequency flutter correlation coefficient between the baseline drift trend slope, cumulative drift, and low-frequency flutter characteristic parameters (including the amplitude of the low-frequency component of vibration acceleration and the fluctuation amplitude of displacement) using multivariate correlation analysis methods (such as multivariate Pearson correlation analysis). Step 4: Determine if the low-frequency flutter correlation coefficient is greater than a preset correlation coefficient threshold. If it is, it is determined to be a baseline drift characteristic caused by low-frequency flutter noise, and low-frequency flutter adaptive baseline correction is performed. Otherwise, the corresponding machine tool machining multi-source data is marked as qualified machine tool machining multi-source data, and based on qualified machine tool machining... Multi-source data is used to extract machine tool processing features. The preset correlation coefficient threshold is represented by the average correlation coefficient of low-frequency chatter over a historical time period. Step five involves low-frequency chatter-adaptive baseline correction. Specifically, the baseline drift trend slope and cumulative drift amount are input into a preset baseline compensation value mapping table. The table then queries and retrieves the vibration acceleration baseline compensation value used to compensate for guide rail vibration acceleration baseline drift and the displacement baseline compensation value used to compensate for machine tool processing displacement baseline drift. The preset baseline compensation value mapping table stores the positive and negative values ​​of different baseline drift trend slopes and the corresponding values ​​for different cumulative drift amounts. The compensation values ​​are positive and negative. The slope of the baseline drift trend corresponds to the drift direction, and the cumulative drift corresponds to the compensation amplitude. A positive slope indicates that the multi-source data of machine tool processing has a positive shift over time, i.e., positive drift; a negative slope indicates that the multi-source data of machine tool processing has a negative shift over time, i.e., negative drift. According to the acquisition sequence, the original amplitude of each vibration acceleration data in the preset low-frequency band and the corresponding vibration acceleration baseline compensation value are calculated to correct the baseline drift of the vibration acceleration data, thus obtaining the corrected vibration acceleration data. Step 6: Based on the processing sequence, perform difference calculation on the original amplitude of each processing displacement data within the preset time period and the corresponding displacement baseline compensation value to complete the baseline drift correction of the processing displacement data and obtain the corresponding corrected processing displacement data; Step 7: Based on the sliding window smoothing algorithm, smooth the multi-source data of machine tool processing after low-frequency chatter adaptation baseline correction (including corrected vibration acceleration data and corrected processing displacement data) to eliminate local fluctuations and retain the overall temporal characteristics and effective processing signals of the data; Step 8: Verify the low-frequency chatter noise correction results.

[0038] Specifically, the verification process for low-frequency flutter noise correction results is as follows: The low-frequency interference suppression rate, used to characterize the thoroughness of low-frequency flutter noise suppression, and the baseline stability coefficient, used to characterize the stability of the corrected data baseline, are obtained. The low-frequency interference suppression rate is represented by the ratio of the corrected low-frequency vibration acceleration amplitude to the original low-frequency vibration acceleration amplitude; the smaller the ratio, the more thorough the low-frequency flutter noise suppression. The baseline stability coefficient includes the acceleration baseline stability coefficient and the machining displacement baseline stability coefficient. The acceleration baseline stability coefficient is represented by the variance of the corrected vibration acceleration data baseline, specifically by extracting the variance of the corrected vibration acceleration data baseline. The time-domain baseline of velocity data is calculated by performing deviation calculations on the vibration acceleration data and the fitted values ​​on the corresponding time-series baseline fitting line. The variance of all deviation calculations is then calculated, and the smaller the variance, the more stable the baseline. The stability coefficient of the machining displacement baseline is represented by the fluctuation variance of the corrected machining displacement data baseline. Specifically, the time-domain baseline of the corrected machining displacement data is extracted, the deviation calculations are performed on the machining displacement data and the fitted values ​​on the corresponding time-series baseline fitting line, and the variance of all deviation calculations is then calculated. The low-frequency interference suppression rate and baseline stability coefficient are then assessed to determine if they conform to the low-frequency correction. If the conditions are met, the corresponding machine tool machining multi-source data will be marked as qualified machine tool machining multi-source data, and machine tool machining features will be extracted based on the qualified machine tool machining multi-source data. Otherwise, a low-frequency chatter noise correction failure prompt will be sent. The qualified conditions for low-frequency correction are that the low-frequency interference suppression rate is less than the preset suppression rate threshold, the acceleration baseline stability coefficient is less than the preset vibration baseline stability threshold, and the machining displacement baseline stability coefficient is less than the preset displacement baseline stability threshold. Among them, the preset suppression rate threshold is represented by the average low-frequency interference suppression rate over a historical period, and the preset vibration baseline stability threshold is represented by the acceleration baseline stability coefficient over a historical period. The average value of the coefficients represents the machine tool processing feature extraction. Machine tool processing feature extraction is based on time-domain statistical analysis methods (such as univariate linear regression, time-domain extreme value statistics, etc.) to extract machine tool processing features from multi-source data of qualified machine tool processing, obtaining multi-source heterogeneous features of the machine tool (such as vibration signal kurtosis, cutting force fluctuation amplitude, displacement positioning deviation, etc.). Based on the extracted multi-source heterogeneous features of the machine tool, a qualification assessment of the machine tool multi-source heterogeneous features is initiated. The multi-source heterogeneous features of the machine tool represent a set of non-homogeneous features extracted from multi-source data of qualified machine tool processing that can characterize the processing state of CNC machine tools from different dimensions, including vibration signal kurtosis, cutting force fluctuation amplitude, displacement positioning deviation, etc.

[0039] As described above, baseline drift correction helps reduce baseline drift interference caused by uneven workpiece rigidity and low-frequency chatter during machine tool processing. It restores the true trend of multi-source data such as machine tool vibration acceleration and processing displacement, reduces data distortion caused by baseline offset, reduces the risk of deviation in subsequent multi-source heterogeneous feature extraction of machine tools, and reduces the accuracy loss of virtual-real mapping of digital twins. It improves the reliability and stability of multi-source data in machine tool processing, the accuracy of feature parameter representation, and the accuracy of the twin model in replicating the physical machine tool state. It also reduces the problems of misjudgment and correction lag of processing anomalies caused by data baseline drift, and provides stable and accurate data support for feature qualification assessment and anomaly detection in high-precision processing scenarios.

[0040] Preferably, the specific process for evaluating the compliance of machine tool multi-source heterogeneous features is as follows: Obtain the machine tool feature compliance rate, which characterizes the compliance level of the machine tool multi-source heterogeneous features; the machine tool feature compliance rate is represented by the ratio of the total number of machine tool multi-source heterogeneous features monitored by the data volume monitor to the preset total number of feature extractions, wherein the preset total number of feature extractions is set in advance by preset personnel; determine whether the machine tool feature compliance rate is greater than the preset feature compliance rate threshold. If so, mark the corresponding machine tool multi-source heterogeneous feature as a qualified machine tool multi-source heterogeneous feature, and perform a digital twin virtual-real mapping operation based on the qualified machine tool multi-source heterogeneous feature; otherwise, send a machine tool multi-source heterogeneous feature distortion prompt, wherein the preset feature compliance rate threshold is represented by the average value of the machine tool feature compliance rate over a historical time period.

[0041] Specifically, the process of mapping the virtual and real worlds in a digital twin is as follows: The multi-source heterogeneous features of a qualified machine tool are input into the digital twin. Simultaneously, real-time operating data of the physical machine tool (such as real-time spindle speed, real-time cutting force of the tool holder, real-time vibration amplitude of the guide rail, and real-time displacement of the worktable) is collected. The theoretical prediction values ​​output by the twin (such as predicted cutting force, predicted vibration amplitude, and predicted displacement positioning value) are calculated, along with a multi-dimensional deviation sequence of the machine tool's machining from the real-time operating data of the physical machine tool. Based on these multi-dimensional deviations, machine tool machining anomaly detection is enabled. The multi-dimensional deviation sequence represents an ordered dataset that quantifies the degree of deviation between the theoretical prediction values ​​of the digital twin and the real-time operating data of the physical machine tool from multiple key machining dimensions. The system encompasses multi-dimensional deviation data directly related to the machine tool's machining state and accuracy, such as cutting force deviation and vibration amplitude deviation. It comprehensively reflects the matching degree between the digital twin and the physical machine tool. The multi-dimensional deviations in machine tool machining are represented by calculating the absolute value of the difference between the corresponding theoretical prediction value output by the twin and the value of the same type of parameter in the real-time operating data of the physical machine tool. For example, the cutting force deviation in machine tool machining is represented by calculating the absolute value of the difference between the cutting force predicted by the twin and the real-time value of the cutting force of the tool holder on the physical machine tool; the vibration deviation in machine tool machining is represented by calculating the absolute value of the difference between the vibration amplitude predicted by the twin and the real-time amplitude of the vibration of the guide rail on the physical machine tool. All types of deviations together constitute the multi-dimensional deviations in machine tool machining, comprehensively reflecting the consistency between the virtual and real states.

[0042] Specifically, the construction process of the digital twin is as follows: First, based on the structural parameters of the physical CNC machine tool (including the geometric parameters and mechanical properties of the spindle, guide rails, worktable, and tool holder), a digital twin geometric model matching the physical CNC machine tool is constructed using 3D modeling technology (such as SolidWorks), ensuring that the geometric shape and structural relationships are consistent with the physical machine tool. The digital twin geometric model represents a 3D model that accurately digitally replicates the structure, dimensions, and mechanical properties of the physical CNC machine tool. It can completely map the hardware composition, assembly relationships, and motion constraints of the physical machine tool and is the physical foundation for realizing virtual-real mapping. Second, using the multi-source heterogeneous characteristics of qualified machine tools as core input parameters, a data-driven virtual-real mapping correlation model is built to establish the mapping relationship between feature parameters and machine tool operating states (speed, cutting force, vibration, displacement). The virtual-real mapping correlation model represents a model based on data statistical analysis and feature correlation rules, used to accurately predict the machine tool's multi-source heterogeneous characteristics through input machine tool processing. A digital model of the operating status is constructed. The mapping relationship represents the deterministic correspondence between the multi-source heterogeneous characteristics of the machine tool and the operating status parameters of the machine tool. That is, when a specific multi-source heterogeneous characteristic of the machine tool is input (such as a certain vibration kurtosis or cutting force fluctuation amplitude), the model can output the theoretical value of the machine tool's operating status that matches it (such as the corresponding vibration amplitude or cutting force value). This relationship is established based on the statistical regularity of historical data and can accurately reflect the intrinsic connection between the characteristics and the operating status, ensuring the accurate mapping of the twin to the physical machine tool. Next, based on the historical qualified operating data of the physical machine tool (including historical data of multi-process machining and high-precision thin-walled part machining), the twin mapping association model is trained and calibrated so that the deviation between the theoretical prediction value output by the twin and the historical qualified data of the physical machine tool is within the preset calibration threshold range. Finally, a real-time virtual-physical data interaction channel is built based on industrial communication protocols and data caching modules to realize the synchronous transmission and dynamic updating of real-time operating data of the physical machine tool and data of the digital twin, thus completing the construction of the digital twin.

[0043] As described above, the evaluation of the conformity of multi-source heterogeneous features of machine tools and the virtual-physical mapping of digital twins help to identify the effectiveness and accuracy of multi-source heterogeneous features of machine tools, eliminate unqualified features with implicit deviations and parameter distortions, provide a high-quality input foundation for digital twins, and achieve precise synchronization and dynamic calibration of the physical machine tool and digital twin based on qualified features. This enhances the twin model's ability to replicate the machining process and its predictive effectiveness, effectively reducing problems such as insufficient twin mapping accuracy and misjudgment of machining anomaly detection caused by feature distortion. It ensures the accuracy of subsequent anomaly detection and the pertinence of anomaly correction, and improves the stability, controllability, workpiece machining quality, and pass rate of the machining process.

[0044] Preferably, the specific process for machine tool machining anomaly detection is as follows: The multi-dimensional deviation sequence of machine tool machining is input into a preset anomaly identification model (such as a random forest model), and the resulting CNC machine tool machining anomaly (such as tool wear, thermal deformation, geometric errors, vibration anomalies, etc.) is output. Simultaneously, the location of the anomaly is located. The CNC machine tool machining anomaly result represents the specific types of faults or anomalies identified by the preset anomaly identification model after analyzing the multi-dimensional deviation sequence of machine tool machining. These anomalies indicate that the CNC machine tool deviates from its normal operating state during machining and may negatively impact machining accuracy, workpiece quality, or equipment lifespan. Examples include tool wear, thermal deformation, geometric errors, and vibration anomalies. Based on the CNC machine tool machining anomaly result, CNC machine tool machining anomaly discrimination is performed. CNC machine tool machining anomaly discrimination involves determining whether the output CNC machine tool machining anomaly result is non-empty. If it is, the CNC machine tool is determined to have a machining anomaly, and targeted machining anomaly correction is performed based on the anomaly location and anomaly type. Otherwise, the CNC machine tool machining state is determined to be normal, and machining continues. The process involves operating the machine tool and synchronizing its real-time operating data to the digital twin to achieve dynamic matching between the virtual and physical operating states. The specific training process for the anomaly recognition model is as follows: First, based on multi-dimensional deviation data of machine tool processing (including deviation data under normal operating conditions and various abnormal operating conditions such as tool wear and thermal deformation), each set of deviation data is labeled with the corresponding processing anomaly type and anomaly location label to construct a model training dataset. Second, the training dataset is divided into a training set and a test set according to a preset ratio. The training set is used for model training, and the test set is used for performance verification. Next, the training set is input into the anomaly recognition model (such as a random forest model), using multi-dimensional deviations as input features and labeled anomaly information as output targets to train the model and learn the correlation between deviation features and processing anomalies. Finally, the model's recognition accuracy is verified through the test set, and the model's hyperparameters (such as the number and depth of decision trees in the random forest) are iteratively adjusted until the model's anomaly recognition accuracy and anomaly location accuracy both meet the preset accuracy thresholds. The trained model is then saved as the preset anomaly recognition model.

[0045] Specifically, machine tool machining anomaly correction is used to correct CNC machine tool machining anomalies (such as tool wear, thermal deformation, geometric errors, vibration anomalies, etc.), restore the machine tool to normal machining state, ensure workpiece machining accuracy, and improve production stability. The specific process is as follows: A matching relationship between anomaly type and root cause is established based on the results of CNC machine tool machining anomalies. For example, if the anomaly type is tool wear, the root cause is determined to be the exhaustion of tool life or excessive cutting load; if it is thermal deformation, the root cause is determined to be insufficient spindle heat dissipation or excessive machining time leading to temperature accumulation; if it is geometric error, the root cause is determined to be the deviation of guide rail and worktable positioning accuracy. Targeted corrections and improvements are then made based on the anomaly type. For example, the correction prompt for tool wear is: stop machine tool machining operation, disassemble the worn tool, and replace it with a qualified tool of the same model; the correction prompt for thermal deformation is: pause the machining process and activate the enhanced heat dissipation system of key components such as the spindle and guide rail (e.g., increase the cooling fan speed, start the cooling oil circuit); the correction prompt for geometric error is: use a laser interferometer to detect the geometric position of the guide rail and worktable, and tighten the connecting bolts of the guide rail and worktable.

[0046] like Figure 5 The diagram shows a schematic of the CNC machine tool machining anomaly detection and correction system based on digital twins provided in this application embodiment. The system includes: a machine tool multi-source data noise monitoring module, a machine tool multi-source feature qualification monitoring module, and a machine tool machining anomaly monitoring module. The machine tool multi-source data noise monitoring module is used to monitor the machine tool machining coupling noise adaptability by quantifying parameters of transient impact noise and low-frequency chatter noise interference intensity, in order to decide whether to enable transient impact noise suppression processing and baseline drift correction processing. By monitoring machine tool multi-source data noise, it helps to distinguish the interference characteristics and intensity levels of the two types of coupled noise, achieving directional decision-making for noise processing and separating spike disturbances and baseline drift interference from the data source. The machine tool multi-source feature qualification monitoring module is used to monitor the machine tool machining coupling noise adaptability... After monitoring, a multi-source heterogeneous feature qualification assessment of the machine tool is initiated, and the assessment results determine whether to initiate the digital twin virtual-real mapping operation. Monitoring the qualification of the machine tool's multi-source features helps to screen out high-reliability features suitable for digital twin mapping from the feature dimension, and eliminate invalid features with implicit biases or excessive time-series fluctuations, providing accurate and reliable feature input for virtual-real mapping. The machine tool machining anomaly monitoring module is used to enable machine tool machining anomaly detection after the digital twin virtual-real mapping operation is completed, and decide whether to take corrective measures based on the detection results. By monitoring machine tool machining anomalies, it helps to leverage the advantages of virtual-real collaboration to achieve early detection and early location of machining anomalies, accurately identify the core causes of various anomalies such as tool wear and thermal deformation, and provide clear targeting basis for anomaly correction to ensure the continuity and stability of the machining process.

[0047] As described above, the machine tool multi-source data noise monitoring module, machine tool multi-source feature qualification monitoring module, and machine tool machining anomaly monitoring module help to build a collaborative monitoring mechanism covering the entire process from machine tool machining multi-source data purification to machine tool machining multi-source heterogeneous feature screening, and finally to machining anomaly monitoring and correction. This achieves seamless connection and efficiency superposition of monitoring at each stage, improving the overall operational efficiency of the system. Specifically, the machine tool multi-source data noise monitoring module provides real basic data for the machine tool multi-source feature qualification monitoring module, improving the accuracy of feature qualification judgment. The machine tool multi-source feature qualification monitoring module outputs a high-precision, high-reliability digital twin model for the machine tool machining anomaly monitoring module, enabling anomaly detection to achieve more sensitive anomaly identification and root cause location based on virtual-real comparison. The detection results and correction feedback from the machine tool machining anomaly monitoring module can specifically correct anomalies in the machine tool machining process in a timely manner, forming a dynamic cycle and improving the adaptability of the entire system to complex machining conditions.

[0048] Example 2, as an alternative to the machine tool multi-source heterogeneous feature qualification assessment in Example 1, addresses the issue of micro-deformation and slight fluctuations in cutting load during the machining process of high-precision thin-walled irregular parts (such as thin-walled frames for aerospace and thin-walled cavities for precision instruments) on CNC machine tools. Simultaneously, the workpiece surface quality requirements are extremely high, potentially leading to hidden deviations in some machine tool multi-source heterogeneous features (such as slight deviations in feature values ​​from theoretical values ​​or minor fluctuations in timing features). The single machine tool feature compliance rate cannot identify these hidden anomalies, easily causing misjudgments and affecting subsequent digital twin mapping accuracy. Therefore, an alternative to the machine tool multi-source heterogeneous feature qualification assessment is required. The specific process is as follows: Obtain the relative deviation rate of machine tool features, which characterizes the overall deviation between the actual and preset values ​​of the machine tool multi-source heterogeneous features, and the timing waveform of machine tool features, which characterizes the overall level of timing stability of the machine tool multi-source heterogeneous features. The accuracy quantification parameter of machine tool features is the dynamic coefficient. The result of weighted coupling of the machine tool feature accuracy quantification parameter and the corresponding machine tool feature accuracy influence parameter is used as the machine tool feature accuracy index to quantify the overall accuracy of multi-source heterogeneous features of machine tools. The weighted coupling means multiplying the relative deviation rate of machine tool features with the corresponding relative deviation rate influence value of machine tool features, multiplying the time-series fluctuation coefficient of machine tool features with the corresponding time-series fluctuation influence value of machine tool features, and then summing the two product results. It is determined whether the machine tool feature accuracy index is greater than the preset accuracy threshold. If it is, the corresponding multi-source heterogeneous features of machine tools are marked as qualified multi-source heterogeneous features of machine tools, and a digital twin virtual-real mapping operation is performed based on the qualified multi-source heterogeneous features of machine tools. Otherwise, a machine tool multi-source heterogeneous feature distortion prompt is sent. The preset accuracy threshold is represented by the average value of the machine tool feature accuracy index over a historical time period.

[0049] Specifically, the formula for the relative deviation rate of machine tool features is as follows:

[0050] Where K represents the relative deviation rate of machine tool features, n represents the total number of extracted multi-source heterogeneous features of the machine tool (such as the total number of features like vibration signal kurtosis, cutting force fluctuation amplitude, displacement positioning deviation, etc.), j represents the sequence number of the multi-source heterogeneous features of the machine tool (j=1, 2, 3, ..., n), and A j T represents the actual extracted value of the j-th multi-source heterogeneous feature of the machine tool. j This represents the preset feature calibration value of the j-th multi-source heterogeneous feature of the machine tool.

[0051] Specifically, the formula for the machine tool characteristic timing fluctuation coefficient is as follows:

[0052] Where P represents the machine tool characteristic timing fluctuation coefficient, S j μ represents the standard deviation of the time series value of the j-th multi-source heterogeneous characteristic of the machine tool. j The j-th multi-source heterogeneous feature time series value represents the average value of the j-th machine tool feature. It is obtained by performing an arithmetic mean operation on the time series feature sequence. Its function is to characterize the overall level of the single feature time series value and serve as a benchmark for calculating the degree of fluctuation. The standard deviation of the j-th multi-source heterogeneous feature time series value is obtained by sorting the actual extracted values ​​of the j-th feature according to the processing time sequence to obtain the time series feature sequence. The dispersion of the time series feature sequence is calculated using the standard deviation formula. Its function is to characterize the dispersion level of the single feature time series value.

[0053] Specifically, the formula for the accuracy index of machine tool features is as follows:

[0054] Where Q represents the machine tool feature accuracy index, W1 represents the influence value of the relative deviation rate of machine tool features, and W2 represents the influence value of the time series fluctuation of machine tool features. The influence value of the relative deviation rate of machine tool features is used to reflect the degree of influence of the relative deviation rate of machine tool features on the machine tool feature accuracy index, and the influence value of the time series fluctuation of machine tool features is used to reflect the degree of influence of the time series fluctuation coefficient of machine tool features on the machine tool feature accuracy index.

[0055] It should be noted that the digital twin-based CNC machine tool machining anomaly detection and correction method provided in this application involves a set of weighted parameter systems for matching the relative deviation rate influence value of machine tool features, the time-series fluctuation influence value of machine tool features, and the corresponding degree of influence when evaluating the conformity of multi-source heterogeneous features of this machine tool. This system is pre-set by machine tool process professionals and stored in the digital twin data support system, providing the core basis for the weight matching of the two types of feature quantification parameters and their corresponding influence values, as well as the accurate calculation of accuracy indicators. Specifically, a large amount of historical machine tool machining data is first extracted, covering the relative deviation rate of machine tool features, the time-series fluctuation coefficient of machine tool features, and their corresponding preset benchmark values ​​(historical statistical optimal range) under different workpiece types (such as high-precision thin-walled irregular parts and large structural parts) and different machining processes (such as cutting, drilling, and milling). At the same time, it includes complete data samples of the combination of two types of feature quantification parameters and the actual values ​​of the corresponding accuracy indicators, forming a complete data sample for each parameter group. The system assigns a weighted quantitative value based on the degree of influence of each feature on the accuracy index of the machine tool. For example, when the relative deviation rate of the machine tool features exceeds the preset benchmark value, which increases the risk of feature value distortion, a higher relative deviation rate influence value is matched to strengthen its proportion in the accuracy index calculation. The actual effective values ​​of the relative deviation rate influence value and the temporal fluctuation influence value of the machine tool features under each historical machining scenario are recorded simultaneously. Then, through correlation analysis (such as Pearson correlation coefficient), abnormal correlation data caused by temporary sensor deviations (such as zero drift of displacement sensor, calibration error of cutting force sensor) and sudden machining interference (such as short-term power supply fluctuation, temporary loosening of fixture) are eliminated, and the parameter correspondence with statistical stability is retained. Finally, all effective data are integrated to form this weighted parameter system to ensure the accuracy and quantifiability of weight matching. When the system conducts a qualification assessment of the multi-source heterogeneous features of the machine tool, the influence value matching the current machining scenario can be quickly retrieved from this system.

[0056] As described above, the multi-source heterogeneous feature qualification assessment of machine tools helps to solve the technical bottleneck of the single machine tool feature compliance rate being unable to identify subtle deviations such as slight deviations in feature values ​​and minor fluctuations in timing features in high-precision thin-walled irregular parts processing scenarios. It achieves comprehensive coverage of the degree of deviation of static feature values ​​and dynamic timing stability, reduces the interference of distorted features on the accuracy of virtual-real mapping from the source, strengthens the ability of digital twins to accurately replicate the physical machine tool state, ensures the accuracy of subsequent processing anomaly detection and the pertinence of anomaly correction, fully adapts to the processing quality and data reliability requirements of high-precision thin-walled irregular parts, reduces the risk of workpiece scrapping and equipment wear caused by subtle deviations, and improves the processing stability and product qualification rate of CNC machine tools under complex working conditions.

Claims

1. A method for detecting and correcting machining anomalies in CNC machine tools based on digital twins, characterized in that, Includes the following steps: The parameters for quantifying the interference intensity of transient impact noise and low-frequency chatter noise are monitored for machine tool machining coupling noise compatibility to determine whether to enable transient impact noise suppression processing and baseline drift correction processing. After the machine tool processing coupling noise adaptability monitoring is completed, the machine tool multi-source heterogeneous feature qualification assessment is initiated, and the digital twin virtual-real mapping operation is determined based on the obtained assessment results. After the digital twin virtual-real mapping operation is completed, machine tool processing anomaly detection is enabled, and a decision is made on whether to take corrective action based on the detection results.

2. The method for detecting and correcting machining anomalies in CNC machine tools based on digital twins as described in claim 1, characterized in that, The parameters for quantifying the interference intensity of transient impact noise and low-frequency flutter noise include transient impact characteristic parameters and low-frequency flutter characteristic parameters. The transient impact characteristic parameters include the rate of change of cutting parameters, the peak value of vibration acceleration, and the amplitude of sudden changes in cutting force. The specific acquisition process is as follows: By monitoring the values ​​of cutting parameters before and after switching during multi-process alternating machining, recording the start time and interval of cutting parameter switching, and calculating the ratio of the difference in cutting parameters before and after switching to the switching interval, this value is used as the rate of change of cutting parameters. The cutting parameters refer to the key parameters used to describe the core process state during CNC machine tool cutting. By collecting vibration acceleration data of the machine tool guide rail within a preset time period before and after the start time of cutting parameter switching, time-domain analysis is performed on the collected vibration acceleration data, and the maximum value of the vibration acceleration data within this time period is extracted and taken as the vibration acceleration peak value. By collecting cutting force data within a preset time period before and after the start time of cutting parameter switching, the difference between the maximum cutting force after switching and the steady-state cutting force before switching is calculated and used as the amplitude of the sudden change in cutting force. The low-frequency flutter characteristic parameters include the amplitude of the low-frequency component of vibration acceleration and the amplitude of displacement fluctuation. The specific acquisition process is as follows: By collecting vibration acceleration data of machine tool guide rail within a preset time period, frequency domain decomposition processing is performed on the collected vibration acceleration data to extract the maximum value of vibration acceleration signal amplitude within a preset low frequency band, which is then used as the amplitude of the low frequency component of vibration acceleration. By collecting the machining displacement data of the machine tool table relative to the workpiece within a preset time period, the difference between the maximum and minimum values ​​of the machining displacement data within the preset time period is calculated and used as the displacement fluctuation range. Transient impact noise discrimination is performed based on transient impact characteristic parameters, wherein the transient impact noise discrimination means verifying whether the transient impact characteristic parameters meet the transient impact judgment conditions; If the verification result meets the transient impact judgment criteria, transient impact noise suppression processing shall be adopted; The transient impact determination condition specifically refers to simultaneously satisfying the following conditions: the rate of change of cutting parameters is greater than a preset threshold for the rate of change of cutting parameters, the peak value of vibration acceleration is greater than a preset threshold for the peak value of vibration acceleration, and the amplitude of the sudden change in cutting force is greater than a preset threshold for the amplitude of the sudden change in cutting force. Low-frequency flutter noise discrimination based on low-frequency flutter characteristic parameters; The low-frequency flutter noise discrimination method checks whether the low-frequency flutter characteristic parameters meet the low-frequency flutter judgment conditions. If the verification result meets the low-frequency chatter judgment criteria, baseline drift correction is performed. If the verification result does not meet the low-frequency chatter judgment criteria, the corresponding machine tool processing multi-source data is marked as qualified machine tool processing multi-source data, and machine tool processing features are extracted based on the qualified machine tool processing multi-source data. The machine tool machining multi-source data refers to a collection of multiple types of monitoring data collected by various sensors during the CNC machine tool machining process, reflecting different dimensions of machining status; The low-frequency flutter determination condition specifically refers to the fact that the amplitude of the low-frequency component of the vibration acceleration is greater than the preset low-frequency acceleration component threshold and the displacement fluctuation amplitude is greater than the preset displacement fluctuation amplitude threshold.

3. The method for detecting and correcting machining anomalies in CNC machine tools based on digital twins as described in claim 2, characterized in that, The specific process for suppressing transient impulse noise is as follows: Accurately locate instantaneous peak data, which means extracting multi-source machine tool processing data within a preset time period before and after the start time of cutting parameter switching, and determining whether the peak value of vibration acceleration is greater than the preset vibration peak rejection threshold, and whether the amplitude of the sudden change in cutting force is greater than the preset cutting force peak rejection threshold. If so, the corresponding machine tool processing multi-source data occurrence time is obtained, and peak data time synchronization judgment is performed based on the occurrence time of the machine tool processing multi-source data. Otherwise, the vibration acceleration data and cutting force data corresponding to the machine tool processing multi-source data are judged as qualified machine tool processing multi-source data, and machine tool processing features are extracted based on the qualified machine tool processing multi-source data. The peak data time synchronization judgment means judging whether the peak-parameter switching time deviation is less than a preset time deviation threshold. If it is, the vibration acceleration data and cutting force data corresponding to the machine tool processing multi-source data are judged as instantaneous peak data caused by transient impact noise. Otherwise, the corresponding vibration acceleration data and cutting force data are judged as qualified machine tool processing multi-source data, and machine tool processing features are extracted based on the qualified machine tool processing multi-source data. Peak perturbation suppression based on instantaneous peak data is performed as follows: Remove instantaneous spike data from multi-source machine tool processing data; Select a data segment with a preset interpolation compensation window duration before and after the peak data, and compensate and fill the gaps in the multi-source machine tool processing data after removing the peaks; The data segment represents the vibration acceleration data of the machine tool guideway and the cutting force data of the machine tool spindle within a preset interpolation compensation window before and after the occurrence of the peak data, without being affected by significant transient impact noise. After the transient impact noise suppression process is completed, the signal-to-noise ratio of the machine tool machining data is obtained; Determine whether the signal-to-noise ratio of the machine tool processing data is greater than the preset signal-to-noise ratio threshold. If so, continue to perform low-frequency chatter noise discrimination based on low-frequency chatter characteristic parameters; otherwise, send a transient impact noise suppression processing failure prompt.

4. The method for detecting and correcting machining anomalies in CNC machine tools based on digital twins as described in claim 2, characterized in that, The specific process of baseline drift correction is as follows: Extract the time-domain baseline of vibration acceleration data within a preset low-frequency band, and the time-domain baseline of processing displacement data within a preset time period; The time-domain baseline extraction method for the vibration acceleration data is as follows: the vibration acceleration data in the preset low-frequency band is sorted according to the acquisition time sequence, and an overall trend fitting line corresponding to the vibration acceleration data is generated; The time-domain baseline extraction method for the processing displacement data is as follows: sort the processing displacement data within a preset time period according to the processing time sequence, and generate an overall trend fitting line corresponding to the processing displacement data; Obtain the baseline drift trend slope and drift accumulation; The low-frequency flutter correlation coefficient between the baseline drift trend slope, drift accumulation, and low-frequency flutter characteristic parameters was calculated based on multivariate correlation analysis. If the correlation coefficient of low-frequency flutter is greater than the preset correlation coefficient threshold, it is determined to be a baseline drift feature caused by low-frequency flutter noise, and low-frequency flutter adaptive baseline correction is performed. Otherwise, the corresponding machine tool processing multi-source data is marked as qualified machine tool processing multi-source data, and machine tool processing features are extracted based on the qualified machine tool processing multi-source data. The specific process for performing low-frequency flutter-adaptive baseline correction is as follows: Input the baseline drift trend slope and drift accumulation into the preset baseline compensation value mapping table, and query to obtain the vibration acceleration baseline compensation value and displacement baseline compensation value; According to the acquisition sequence, the original amplitude of each vibration acceleration data in the preset low frequency band and the corresponding vibration acceleration baseline compensation value are calculated by difference to obtain the corrected vibration acceleration data. According to the processing sequence, the original amplitude of each processing displacement data within the preset time period and the displacement baseline compensation value under the corresponding time period are calculated by difference to obtain the corrected processing displacement data. Smoothing is performed on multi-source machine tool machining data after baseline correction for low-frequency chatter adaptation. Verify the results of low-frequency flutter noise correction.

5. The method for detecting and correcting machining anomalies in CNC machine tools based on digital twins as described in claim 4, characterized in that, The specific process for verifying the low-frequency flutter noise correction results is as follows: Obtain the low-frequency interference suppression rate and baseline stability coefficient; The baseline stability coefficient includes the acceleration baseline stability coefficient and the machining displacement baseline stability coefficient; The acceleration baseline stability coefficient is represented by the fluctuation variance of the corrected vibration acceleration data baseline; The stability coefficient of the machining displacement baseline is represented by the variance of the fluctuation of the corrected machining displacement data baseline; Determine whether the low-frequency interference suppression rate and baseline stability coefficient meet the low-frequency correction qualification conditions. If they do, mark the corresponding machine tool processing multi-source data as qualified machine tool processing multi-source data and perform machine tool processing feature extraction based on the qualified machine tool processing multi-source data. Otherwise, send a low-frequency chatter noise correction failure prompt. The low-frequency correction qualification condition means that the low-frequency interference suppression rate is less than the preset suppression rate threshold, the acceleration baseline stability coefficient is less than the preset vibration baseline stability threshold, and the processing displacement baseline stability coefficient is less than the preset displacement baseline stability threshold. The qualification assessment of multi-source heterogeneous features of machine tools is initiated based on the extracted multi-source heterogeneous features of machine tools. The machine tool multi-source heterogeneous features refer to a set of non-homogeneous features extracted from multi-source data of qualified machine tool processing, which can characterize the processing state of CNC machine tools from different dimensions.

6. The method for detecting and correcting machining anomalies in CNC machine tools based on digital twins as described in claim 5, characterized in that, The specific process for the qualification assessment of the multi-source heterogeneous characteristics of the machine tool is as follows: Obtain the machine tool feature compliance rate; If the machine tool feature compliance rate is greater than the preset feature compliance rate threshold, the corresponding machine tool multi-source heterogeneous feature is marked as a qualified machine tool multi-source heterogeneous feature, and a digital twin virtual-real mapping operation is performed based on the qualified machine tool multi-source heterogeneous feature. Otherwise, a machine tool multi-source heterogeneous feature distortion prompt is sent.

7. The method for detecting and correcting machining anomalies in CNC machine tools based on digital twins as described in claim 6, characterized in that, The specific process of the digital twin's virtual-real mapping is as follows: The multi-source heterogeneous characteristics of qualified machine tools are input into the digital twin, and the real-time operation data of the physical machine tool is collected simultaneously. The theoretical prediction value output by the twin is calculated and the multi-dimensional deviation sequence of machine tool processing is compared with the real-time operation data of the physical machine tool. Based on the multi-dimensional deviation of machine tool processing, machine tool processing anomaly detection is enabled.

8. The method for detecting and correcting machining anomalies in CNC machine tools based on digital twins as described in claim 7, characterized in that, The specific process for detecting machine tool processing anomalies is as follows: Input the multi-dimensional deviation sequence of machine tool processing into the preset anomaly identification model, output the abnormal results of CNC machine tool processing, and locate the location of the anomaly; Based on the abnormal results of CNC machine tool processing, identify CNC machine tool processing abnormalities; The CNC machine tool machining anomaly discrimination method determines whether the output CNC machine tool machining anomaly result is non-empty. If it is, it determines that the CNC machine tool has a machining anomaly and performs targeted machine tool machining anomaly correction. Otherwise, it determines that the CNC machine tool machining status is normal, continues to execute machining operations, and synchronously updates the real-time running data of the physical machine tool to the digital twin to complete the dynamic matching of virtual and physical running status. The specific process for correcting machine tool machining anomalies is as follows: Establish a matching relationship between anomaly types and root causes based on the abnormal results of CNC machine tool processing; Targeted improvements and corrections are made based on the type of anomaly.

9. The method for detecting and correcting machining anomalies in CNC machine tools based on digital twins as described in claim 5, characterized in that, The qualification assessment of the multi-source heterogeneous characteristics of the machine tool also includes: Obtain quantitative parameters for the accuracy of machine tool features; The result of weighted coupling of the quantitative parameters of machine tool feature accuracy and the corresponding parameters affecting machine tool feature accuracy is used as the machine tool feature accuracy index. If the accuracy index of the machine tool feature is greater than the preset accuracy threshold, the corresponding multi-source heterogeneous feature of the machine tool is marked as a qualified multi-source heterogeneous feature of the machine tool, and a digital twin virtual-real mapping operation is performed based on the qualified multi-source heterogeneous feature of the machine tool. Otherwise, a machine tool multi-source heterogeneous feature distortion prompt is sent.

10. A CNC machine tool machining anomaly detection and correction system based on digital twins, characterized in that, include: Machine tool multi-source data noise monitoring module, machine tool multi-source characteristic qualification monitoring module, and machine tool processing anomaly monitoring module; The machine tool multi-source data noise monitoring module is used to monitor the machine tool processing coupling noise adaptability of parameters that quantify the interference intensity of transient impact noise and low-frequency chatter noise, so as to decide whether to enable transient impact noise suppression processing and baseline drift correction processing. The machine tool multi-source feature qualification monitoring module is used to start the machine tool multi-source heterogeneous feature qualification assessment after the machine tool processing coupling noise adaptability monitoring is completed, and to determine whether to start the digital twin virtual-real mapping operation based on the obtained assessment results. The machine tool processing anomaly monitoring module is used to activate machine tool processing anomaly detection after the digital twin virtual-real mapping operation is completed, and to decide whether to take corrective measures for machine tool processing anomalies based on the detection results.

Images