Numerical control machining intelligent monitoring and self-adaptive control method and related device

Abstract

The invention discloses a numerical control machining intelligent monitoring and self-adaptive control method and a related device, and the method comprises the steps: collecting a multi-mode original data set in a high-speed rotation state of a main shaft; performing time-frequency domain conjoint analysis on the data set to generate a time-varying feature matrix; reconstructing an inner hole state digital twinborn model based on the time-varying feature matrix; performing multi-scale analysis on the model, identifying macroscopic geometric deviation and microscopic material change, and generating a layering abnormal index; historical data and abnormal indexes are combined, a deep learning algorithm is applied to analyze the performance evolution trend of the spindle inner hole, and a multi-time-scale state prediction result is output; and according to the prediction result and the current processing task requirement, dynamically generating a decision scheme including processing parameter optimization and a customized maintenance plan, and feeding back the decision scheme to a processing control system in real time. According to the invention, real-time dynamic monitoring of the inner hole of the main shaft can be realized, accurate performance analysis and prediction are provided, and the stability and the machining precision in the machining process are ensured.

Description

Technical Field

[0001] The present invention relates to the field of mechanical processing technology, and in particular to a numerical control processing intelligent monitoring and adaptive control method and related devices. Background Art

[0002] In the field of modern intelligent manufacturing, the intelligent monitoring and control system of CNC machining equipment is the key to ensuring machining accuracy and production efficiency. Among them, the state monitoring and intelligent control of the core components of the equipment during the machining process are particularly important, which is directly related to the performance and machining quality of the entire machining system. With the development of machining technology, especially in the field of high-speed machining and precision machining, even small changes in the state of the equipment during operation may have a significant impact on the machining quality. If the control system cannot detect and adjust these changes in time, it may lead to a decrease in machining accuracy, deterioration of the machining surface quality, and even cause system failures. In severe cases, it may cause equipment damage, production stagnation and economic losses. Therefore, state monitoring and performance evaluation based on intelligent control are particularly important. It can provide early warning for abnormal changes in the machining process, and then realize intelligent control and adjustment, avoiding downtime, losses and unnecessary maintenance caused by equipment abnormalities.

[0003] At present, the state monitoring methods of industrial sites mainly rely on static measurement or offline inspection. These traditional control methods are usually carried out after the equipment stops running, and cannot reflect the dynamic working state in the actual processing process. The shutdown inspection requires interrupting production, which seriously affects production efficiency, and cannot achieve continuous online monitoring and intelligent control. These methods have great limitations. They cannot monitor the equipment status and performance changes in real time and dynamically, and it is difficult to detect small state deviations in time, resulting in the inability to perform effective predictive maintenance and intelligent control. Especially under high-speed processing conditions, traditional control systems are difficult to cope with complex and changeable working conditions, and cannot achieve accurate state perception and adaptive adjustment. Therefore, the existing technology urgently needs a real-time monitoring method based on intelligent control, especially under high-speed processing conditions, which can comprehensively evaluate the equipment status and perform intelligent control to achieve accurate state monitoring and preventive maintenance. Summary of the invention

[0004] The main purpose of the present invention is to solve the problems existing in the prior art of being unable to monitor the dynamic state of the inner hole of the spindle of the machining center in real time and lacking effective preventive maintenance means.

[0005] A first aspect of the present invention provides a method for intelligent monitoring and adaptive control of numerical control machining, the method comprising: When the spindle is rotating at high speed, the vibration signal, acoustic emission signal and current signal of the spindle are collected simultaneously to obtain a multi-modal raw data set; Performing a joint analysis of the multimodal original data set in the time and frequency domains, extracting dynamic characteristic parameters reflecting the state of the spindle inner hole, and generating a time-varying characteristic matrix; Based on the time-varying characteristic matrix, the geometric morphology and material properties of the spindle inner hole are reconstructed using a preset nonlinear mapping model to obtain a digital twin model of the inner hole state; Performing multi-scale analysis on the digital twin model of the inner hole state, identifying macroscopic geometric deviations and microscopic material changes, and generating layered anomaly indicators; Combining historical data and the above-mentioned stratified abnormality indicators, a deep learning algorithm is applied to analyze the performance evolution trend of the spindle inner hole, and a multi-time scale state prediction result is output; According to the state prediction results and the current processing task requirements, decision plans including processing parameter optimization and customized maintenance plans are dynamically generated and fed back to the processing control system in real time.

[0006] Optionally, when the spindle is rotating at a high speed, the vibration signal, acoustic emission signal and current signal of the spindle are collected simultaneously to obtain a multi-modal original data set, including: Obtain the current spindle speed, cutting force and processing material information to obtain the spindle working condition parameters; Determining a time window for signal acquisition according to the spindle operating parameters; In the time window, synchronously collect vibration signals, acoustic emission signals and current signals of the bearing seat, tool interface and motor winding of the main shaft to obtain original multi-modal signals; Synchronously collecting spindle temperature and coolant flow data, and performing time alignment with the original multimodal signal to obtain correlated signal data; The associated signal data are collected and integrated during a complete working cycle of the spindle, and are associated with the processing parameters and environmental factors of the spindle to obtain a multimodal original data set containing spatiotemporal information.

[0007] Optionally, performing a time-frequency domain joint analysis on the multimodal original data set, extracting dynamic characteristic parameters reflecting the state of the spindle inner hole, and generating a time-varying characteristic matrix includes: Performing wavelet time-frequency transformation on the multimodal original data set to obtain a time-frequency representation spectrum; The time-frequency representation spectrum is combined with the spindle operation stage information to divide it into time-frequency sub-spectra of no-load, acceleration, stable cutting and deceleration stages; Performing adaptive threshold processing on the time-frequency sub-spectrum, extracting characteristic frequencies and energy distributions of each stage, and obtaining dynamic frequency characteristics; Combining the dynamic frequency characteristics with the cutting parameters of the spindle, a characteristic parameter set reflecting the geometric state of the inner hole is constructed; Performing a thermal-mechanical coupling analysis on the characteristic parameter set and the spindle temperature data to obtain a comprehensive characteristic vector; The comprehensive characteristic vectors are arranged in time series and normalized to generate a time-varying characteristic matrix reflecting the evolution of the spindle inner hole state.

[0008] Optionally, the preset nonlinear mapping model includes a feature encoding layer, a multimodal fusion layer, a dynamic response layer, a geometric reconstruction layer and a material property estimation layer; Based on the time-varying characteristic matrix, the preset nonlinear mapping model is used to reconstruct the geometric morphology and material properties of the spindle inner hole to obtain the inner hole state digital twin model, including: The time-varying feature matrix is ​​input into the feature encoding layer, and the vibration, acoustic emission and current signals are subjected to feature extraction through a multi-scale convolutional neural network to obtain a multi-modal feature map; In the multimodal fusion layer, the attention mechanism is used to adaptively weight and fuse the multimodal feature maps to obtain a fused feature vector; The fused feature vector is input into the dynamic response layer, and the dynamic response layer adopts a long short-term memory network structure to model the dynamic response of the spindle under different cutting conditions and output a time series feature sequence; In the geometric reconstruction layer, the temporal feature sequence is processed by using a graph convolutional network to obtain a three-dimensional geometric morphology of the spindle inner hole, wherein the three-dimensional geometric morphology includes roundness, straightness and surface roughness; Inputting the time series feature sequence into the material property estimation layer, the material property estimation layer adopts a fully connected neural network structure to obtain the hardness distribution and residual stress state of the inner hole material; Combined with the three-dimensional geometric morphology, the hardness distribution and residual stress state of the inner hole material, and the real-time collected spindle temperature field data, thermal-mechanical coupling analysis and dynamic optimization are performed to obtain a digital twin model of the inner hole state.

[0009] Optionally, the performing of multi-scale analysis on the inner hole state digital twin model to identify macroscopic geometric deviations and microscopic material changes and generate layered anomaly indicators includes: Performing multi-scale wavelet decomposition on the inner hole state digital twin model to obtain a multi-scale feature set reflecting the geometric morphology and material properties of the spindle inner hole at different processing stages; Based on the multi-scale feature set, the macro-geometric deviation features of the spindle inner hole are extracted using the principal component analysis method to obtain the macro-geometric anomaly index; Applying a local binary pattern algorithm to the multi-scale feature set to identify a microscopic material change pattern on the surface of the spindle inner hole and obtain a microscopic material anomaly index; By combining the macroscopic geometric anomaly index and the microscopic material anomaly index, a hierarchical anomaly evaluation system is constructed to generate hierarchical anomaly index containing geometric deviation and material change information.

[0010] Optionally, the historical data and the layered abnormality index are combined to apply a deep learning algorithm to analyze the performance evolution trend of the spindle inner hole, and output state prediction results at multiple time scales, including: Performing time series alignment on the historical processed data and the hierarchical anomaly indicators to obtain a comprehensive time series data set; Performing time series decomposition on the comprehensive time series data set to obtain multiple time scale components reflecting the change of spindle inner hole performance; Using a deep learning algorithm to extract features from the multiple time scale components, and obtain a multi-scale feature representation of the spindle inner hole performance; Based on the multi-scale feature representation, calculating the predicted values ​​of the spindle inner hole geometric accuracy change, material performance degradation and remaining service life; Performing statistical analysis on the predicted values ​​to obtain a performance prediction index including a prediction result and a confidence interval; According to the performance prediction index, a prediction of the performance change of the spindle inner hole within 24 hours, 7 days and 30 days is generated to obtain a multi-time scale state prediction result.

[0011] Optionally, dynamically generating a decision plan including machining parameter optimization and customized maintenance plan based on the state prediction result and current machining task requirements, and feeding back to the machining control system in real time, including: Extract key parameters such as workpiece material, machining accuracy and surface quality required by the current machining task to obtain task requirement indicators; Matching and analyzing the state prediction result with the task requirement index to obtain an adaptability evaluation result of the spindle performance and the processing requirement; According to the adaptability evaluation results, the adjustment amounts of key machining parameters such as spindle speed, feed speed and cutting depth are calculated to generate an optimized machining parameter solution; Based on the performance degradation trend in the state prediction result and the mission requirement indicator, formulate a customized maintenance plan including maintenance time, replacement parts and adjustment frequency; The optimized processing parameter plan and the customized maintenance plan are integrated into a decision-making plan, and the decision-making plan is transmitted to the processing control system in real time through a data interface.

[0012] A second aspect of the present invention provides a numerical control machining intelligent monitoring and adaptive control device, comprising: The data acquisition module is used to simultaneously collect the vibration signal, acoustic emission signal and current signal of the spindle when the spindle is rotating at a high speed, so as to obtain a multi-modal original data set; A time-frequency analysis module is used to perform a time-frequency domain joint analysis on the multi-modal original data set, extract dynamic characteristic parameters reflecting the state of the spindle inner hole, and generate a time-varying characteristic matrix; A digital twin module is used to reconstruct the geometric morphology and material properties of the spindle inner hole based on the time-varying characteristic matrix and use a preset nonlinear mapping model to obtain a digital twin model of the inner hole state; An anomaly identification module, used to perform multi-scale analysis on the inner hole state digital twin model, identify macroscopic geometric deviations and microscopic material changes, and generate hierarchical anomaly indicators; A state prediction module is used to combine historical data and the layered abnormality index, apply a deep learning algorithm to analyze the performance evolution trend of the spindle inner hole, and output a state prediction result at multiple time scales; The decision generation module is used to dynamically generate decision plans including processing parameter optimization and customized maintenance plans according to the state prediction results and the current processing task requirements, and provide real-time feedback to the processing control system.

[0013] The third aspect of the present invention provides a CNC machining intelligent monitoring and adaptive control device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected through lines; the at least one processor calls the instructions in the memory so that the CNC machining intelligent monitoring and adaptive control device executes the steps of the above-mentioned CNC machining intelligent monitoring and adaptive control method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the above-mentioned CNC machining intelligent monitoring and adaptive control method.

[0015] The CNC machining intelligent monitoring and adaptive control method of the present invention has the following advantages: First, this solution forms a multimodal raw data set by synchronously collecting vibration signals, acoustic emission signals, and current signals while the spindle is rotating at high speed. This approach overcomes the limitations of traditional methods and can monitor the health of the inner hole in real time under the actual working state of the spindle. Compared with static measurement and shutdown inspection, the multimodal signals collected in real time can fully reflect the dynamic state of the inner hole of the spindle, avoiding traditional detection methods that can only reflect static problems. Under high-speed rotating working conditions, the inner hole of the spindle may experience slight deformation, wear, thermal expansion, and other problems, and these changes are often difficult to detect through traditional static detection methods. By collecting multimodal signals in real time, this solution can capture these subtle changes more accurately.

[0016] Next, the scheme processes the multimodal original data set through joint analysis in the time-frequency domain, extracts dynamic characteristic parameters reflecting the state of the spindle inner hole, and generates a time-varying characteristic matrix. Most traditional methods use a single measurement signal, which makes it difficult to comprehensively evaluate the multi-dimensional state changes of the spindle inner hole. By combining vibration, acoustic emission and current signals for time-frequency analysis, this scheme can perform in-depth analysis of the signal at different time and frequency scales, and extract characteristic parameters that can accurately reflect the changes in the inner hole state. Joint analysis in the time-frequency domain can not only capture signal changes in the two dimensions of time and frequency, but also effectively extract key dynamic information related to the geometry and material properties of the spindle inner hole. This processing process greatly enhances the accuracy and reliability of monitoring, and can effectively capture subtle changes that are difficult to detect with traditional methods.

[0017] Subsequently, based on the time-varying characteristic matrix, the solution reconstructs the geometric shape and material properties of the spindle inner hole through a preset nonlinear mapping model, and then obtains a digital twin model of the inner hole state. The digital twin model digitally reproduces the geometric shape and material properties of the spindle inner hole, allowing engineers to observe and analyze the state of the spindle inner hole in real time in a virtual environment. This digital method can make up for the lag and limitations of traditional detection methods, help users deeply understand the changing laws of the inner hole and predict potential failure risks in advance. Through the digital twin model, users can not only obtain the geometric shape data of the spindle inner hole (such as roundness, straightness, etc.), but also understand the hardness distribution and residual stress state of the material in real time, providing data support for subsequent processing and maintenance.

[0018] Based on the digital twin model of the inner hole state, the solution further conducts multi-scale analysis to identify macroscopic geometric deviations and microscopic material changes, and generate layered abnormality indicators. Traditional monitoring methods can often only reflect some general changes in the inner hole of the spindle, and lack accurate monitoring of microscopic details. However, through multi-scale analysis, this solution can identify various abnormal conditions from macro to micro and generate layered abnormality indicators. This layered abnormality indicator can not only reveal the geometric shape deviation of the inner hole of the spindle, but also reflect the material degradation at the micro level, forming a more comprehensive and accurate abnormality assessment system. Through this multi-dimensional and multi-level abnormality identification, this solution can detect potential processing problems or equipment failures at an early stage, greatly improving the effectiveness of monitoring and maintenance.

[0019] Finally, this solution combines historical data and hierarchical abnormal indicators, applies deep learning algorithms to analyze the performance evolution trend of the spindle inner hole, and outputs multi-time scale state prediction results. This process realizes the dynamic prediction of the spindle inner hole state, and can predict the performance changes of the spindle in the future period of time based on historical data and real-time monitoring information. By analyzing the rules in historical data, the deep learning algorithm can accurately capture the changing trend of the spindle inner hole performance and provide a basis for predictive maintenance. By outputting multi-time scale state prediction results, this solution can provide real-time feedback to the machining control system, help the system dynamically adjust the machining parameters according to the health of the spindle, and formulate customized maintenance plans to ensure that the spindle is always in the best working condition.

[0020] In summary, this solution overcomes the shortcomings of traditional detection methods that cannot monitor in real time, cannot fully evaluate, and cannot perform predictive maintenance by comprehensively applying advanced technologies such as multimodal signal acquisition, time-frequency domain joint analysis, digital twin modeling, multi-scale analysis, and deep learning algorithms. Through these technical means, this solution can achieve real-time monitoring, accurate evaluation, and dynamic prediction of the spindle inner hole status, thereby improving machining accuracy, reducing equipment failure rate, and reducing downtime and maintenance costs, effectively improving the operating efficiency and economic benefits of the machining center. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0022] Figure 1 A schematic diagram of an embodiment of a method for intelligent monitoring and adaptive control of numerical control machining in an embodiment of the present invention; Figure 2 A schematic diagram of an embodiment of a numerical control machining intelligent monitoring and adaptive control device in an embodiment of the present invention; Figure 3 It is a schematic diagram of an embodiment of the intelligent monitoring and adaptive control device for numerical control machining in an embodiment of the present invention.

[0023] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0026] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, which must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0027] An embodiment of the present application provides a method for intelligent monitoring and adaptive control of CNC machining. Figure 1 A flow chart of a method for intelligent monitoring and adaptive control of CNC machining provided in one embodiment of the present application.

[0028] refer to Figure 1 , when the spindle is rotating at high speed, the vibration signal, acoustic emission signal and current signal of the spindle are collected simultaneously to obtain a multi-modal raw data set; In one embodiment of the present application, when the spindle is rotating at a high speed, the vibration signal, acoustic emission signal and current signal of the spindle are collected simultaneously to obtain a multimodal original data set, including: obtaining the current speed, cutting force and processing material information of the spindle to obtain the spindle operating parameters; determining the time window for signal collection according to the spindle operating parameters; within the time window, synchronously collecting the vibration signal, acoustic emission signal and current signal of the spindle bearing seat, tool interface and motor winding to obtain the original multimodal signal; synchronously collecting the spindle temperature and coolant flow data, time-aligning them with the original multimodal signal to obtain correlated signal data; collecting and integrating the correlated signal data during the complete working cycle of the spindle, and correlating them with the spindle processing parameters and environmental factors to obtain a multimodal original data set containing spatiotemporal information.

[0029] Specifically, it is first necessary to obtain the current spindle speed, cutting force and processing material information in order to obtain the spindle working parameters. The core goal of this process is to provide accurate basic data support for subsequent signal acquisition and data analysis. The process of obtaining spindle speed, cutting force and processing material information needs to be implemented through a series of sensors and data acquisition systems installed on the spindle and processing equipment. The spindle speed can be obtained through a speed sensor, which can use a photoelectric encoder or a magnetic sensor, which can accurately measure the rotation speed of the spindle and transmit data to the central control system in real time. The speed sensor can accurately record the spindle speed changes in different processing stages to ensure that speed fluctuations or abnormal conditions can be captured.

[0030] Similarly, cutting force monitoring needs to be achieved through force sensors. Cutting force reflects the load on the tool when it contacts the workpiece, which is critical to machining accuracy and the health of the spindle bore. In precision machining, changes in cutting force can indicate problems such as tool wear, uneven hardness of the workpiece material, or improper cutting parameters. Force sensors such as strain gauges and piezoelectric sensors installed on the tool holder, tool interface, or workpiece table can measure the forces generated in different directions during the cutting process in real time, and feed their data back to the monitoring system in real time. These data are very important for evaluating the working status of the spindle and detecting potential problems in a timely manner.

[0031] In addition, processing material information is an important factor affecting the state of the spindle inner hole. The hardness, surface roughness and cutting characteristics of the processing material will have different load effects on the spindle, so it is crucial to obtain this information. Through the CNC system settings and the management of the material parameter library, the relevant data of the workpiece material, such as hardness value, material category, etc., can be dynamically read. Combined with the cutting force and temperature data during processing, the impact of the material on the spindle load can be evaluated. In actual operation, the CNC system usually automatically obtains material data through the control panel or sensor, and matches it with the processing task parameters to provide comprehensive working condition information for spindle monitoring.

[0032] Once the spindle operating parameters are obtained, the time window for signal acquisition can be determined based on these data. The time window refers to the range and duration of signal data acquisition within a specific time period, which is usually closely related to the spindle operating parameters. For example, when the spindle speed is high or the cutting force fluctuates greatly, the signal acquisition frequency needs to be increased accordingly in order to capture subtle changes in the spindle during high-speed machining. Under stable working conditions, the time window can be appropriately extended to obtain more signal data for comprehensive analysis. By dynamically adjusting the acquisition window according to the operating parameters, the accuracy and timeliness of data acquisition can be effectively improved to avoid missing any critical state changes.

[0033] After determining the time window for signal acquisition, the next step is to collect signals from the key components of the spindle. These components include the spindle's bearing seat, tool interface, and motor winding. The purpose of collecting signals from these components is to monitor vibration, thermal expansion, wear, and other problems that may occur during the operation of the spindle. Vibration sensors, acoustic emission sensors, and current sensors can be used to collect vibration signals, acoustic emission signals, and current fluctuation data from these components during high-speed rotation. Vibration signals can reflect mechanical anomalies in the spindle during rotation, especially wear or failure of bearings; acoustic emission signals are used to capture early warnings of minor faults such as cracks and friction inside the spindle; and current signals reflect changes in the spindle motor load, which is very helpful for analyzing whether the spindle is overloaded or whether abnormal loads occur. By synchronously collecting these signal data, a comprehensive dynamic monitoring system can be formed.

[0034] In addition to these mechanical signals, temperature data and coolant flow data are also crucial. The spindle generates a lot of heat during high-speed rotation. The increase in temperature may cause thermal expansion of the material, which in turn affects the geometry of the spindle inner hole and leads to a decrease in machining accuracy. Therefore, the operating temperature of the spindle must be monitored in real time through a temperature sensor. At the same time, the coolant flow plays a vital role in temperature control. Too low a coolant flow may cause the spindle to overheat, thus affecting the machining quality. Therefore, it is necessary to obtain real-time flow data through a coolant flow sensor. These temperature and coolant flow data need to be time-aligned with vibration, acoustic emission, and current signals. Time alignment refers to synchronizing data from different sources according to timestamps to ensure that the changes of various signals in the same time period can be accurately compared and analyzed. For example, in a certain processing stage of the spindle, the vibration signal and current signal may fluctuate abnormally, and the temperature may also rise. After time alignment, it is possible to accurately determine which factor caused the change in spindle performance.

[0035] Through these synchronously collected data, combined with the working parameters and environmental factors of the spindle, the data can be fully integrated to obtain a multimodal original data set containing spatiotemporal information. The integration of spatiotemporal information refers to the comprehensive combination and analysis of data in the time dimension and the space dimension. In the time dimension, it is ensured that the signal data can reflect the state changes of the spindle in different working stages (such as no-load, acceleration, stable cutting, deceleration, etc.); in the space dimension, by uniformly managing the data of different components of the spindle (such as bearing seats, tool interfaces, motor windings, etc.), a multimodal data set is formed. For example, in the stable cutting stage, the vibration signal and cutting force data of the spindle may be relatively stable, while in the acceleration stage, the volatility of these data may increase. Through the integration of spatiotemporal information, it is possible to clearly see the performance of the spindle under different working conditions, and then judge whether its working state is normal.

[0036] By constructing this multimodal data set, this solution can more accurately monitor the spindle inner hole status in real time and predict performance. The spatiotemporal information of the data not only helps the system better understand the working status of the spindle, but also provides a reliable basis for subsequent dynamic modeling, fault diagnosis and performance optimization. The advantage of the whole process is that by integrating data from different signal sources and integrating spatiotemporal information, comprehensive and accurate status monitoring of the spindle can be achieved, avoiding the singleness and limitations of traditional methods, and ensuring efficient and real-time maintenance and decision support.

[0037] Please continue to refer to Figure 1 , performing a time-frequency domain joint analysis on the multimodal original data set, extracting dynamic characteristic parameters reflecting the state of the spindle inner hole, and generating a time-varying characteristic matrix; In one embodiment of the present application, the multimodal original data set is subjected to a joint analysis in the time-frequency domain to extract dynamic characteristic parameters reflecting the state of the spindle inner hole and generate a time-varying characteristic matrix, including: performing a wavelet time-frequency transform on the multimodal original data set to obtain a time-frequency representation spectrum; combining the time-frequency representation spectrum with the spindle operation stage information to divide it into time-frequency sub-spectra of no-load, acceleration, stable cutting and deceleration stages; performing adaptive threshold processing on the time-frequency sub-spectra to extract the characteristic frequency and energy distribution of each stage to obtain dynamic frequency characteristics; combining the dynamic frequency characteristics with the cutting parameters of the spindle to construct a characteristic parameter set reflecting the geometric state of the inner hole; performing a thermal-mechanical coupling analysis on the characteristic parameter set and the spindle temperature data to obtain a comprehensive characteristic vector; arranging the comprehensive characteristic vector in time series and normalizing it to generate a time-varying characteristic matrix reflecting the evolution of the spindle inner hole state.

[0038] Specifically, in this scheme, in order to accurately monitor the changes in the state of the inner hole of the spindle, it is necessary to perform time-frequency analysis on the multimodal original data set. First, the data set collects signal data in multiple dimensions such as vibration signals, acoustic emission signals, current signals, etc. of the spindle under different working conditions through different sensors. In order to deeply analyze the changes in these signals, it is necessary to use the wavelet time-frequency transform method to convert the signal into a time-frequency representation spectrum. Wavelet time-frequency transform is a method that can analyze signals in both time and frequency dimensions at the same time. It is suitable for processing non-stationary signals, such as the state changes of the spindle during acceleration, deceleration, and cutting. Compared with the traditional Fourier transform, the wavelet transform can provide more accurate local frequency information because it can capture the frequency fluctuations of the signal with high resolution in a short time, so that it can accurately identify when the spindle has sudden vibrations, temperature changes or load fluctuations.

[0039] For example, if the spindle enters the high-speed cutting stage, the speed increases rapidly, and the vibration signal and current signal may also undergo instantaneous changes. These changes are very rapid and local, and traditional time domain analysis methods may not be able to effectively capture such instantaneous changes. However, wavelet time-frequency transform can accurately capture such rapid changes and avoid information loss by retaining frequency information at the same time with high time resolution.

[0040] Next, the signal obtained through the time-frequency representation spectrum is further divided into the signal in combination with the spindle's operating stage information. The working process of the spindle is usually divided into several different operating stages, including no-load, acceleration, stable cutting and deceleration. In different working stages, the vibration mode, load change and temperature fluctuation of the spindle are different. For example, in the no-load state, the vibration frequency of the spindle is usually low, while in the acceleration stage, due to the rapid change in speed, the vibration frequency and cutting force will increase significantly. Therefore, by combining the information of the operating stage, the time-frequency representation spectrum is divided into four sub-spectra: no-load, acceleration, stable cutting and deceleration, making the analysis more refined.

[0041] The key to this process is that the vibration characteristics of each stage are different. For example, in the no-load stage, the energy distribution of the vibration may be concentrated in the lower frequency range; while in the cutting process, especially in the stable cutting stage, the frequency and energy distribution of the vibration signal will change due to the increase of cutting force and spindle load. By dividing the staged time-frequency sub-spectrum, the dynamic characteristics of each stage can be analyzed in a targeted manner to avoid confusing the signal changes under different working conditions.

[0042] After obtaining the time-frequency sub-spectra of each stage, the next step is to perform adaptive threshold processing on these sub-spectra. The purpose of this processing method is to automatically set the threshold according to the signal characteristics of each stage to filter out the most representative frequency components and remove noise interference. Specifically, the working state of the spindle at different stages may cause changes in different frequency ranges. During the acceleration stage, the spindle may experience higher-frequency vibrations, and these frequency signals may be mixed with some irrelevant noise signals. Therefore, through adaptive threshold processing, the threshold can be dynamically adjusted according to the vibration characteristics of each stage, thereby extracting the frequency characteristics that truly represent the health status of the spindle.

[0043] For example, during the deceleration phase, the spindle speed decreases, the high-frequency components in the vibration signal may gradually weaken, while the low-frequency components become more prominent. Adaptive threshold processing can help us accurately identify these changes and analyze the health status of each stage through the distribution diagram of frequency and energy, especially identifying potential anomalies or faults.

[0044] Subsequently, by combining the dynamic frequency features extracted from the time-frequency sub-spectrum and the cutting parameters of the spindle (such as cutting depth, feed speed, etc.), a feature parameter set reflecting the geometric state of the spindle inner hole can be constructed. These cutting parameters are important factors affecting the geometry of the spindle inner hole and the workload. For example, excessive cutting depth will increase the load on the spindle, causing problems such as vibration and thermal expansion. By combining these cutting parameters with the dynamic frequency features for analysis, a more accurate geometric feature reflecting the state of the spindle inner hole can be constructed. In this way, not only can the geometric changes of the spindle inner hole be monitored, but also the relationship between these changes and the machining process can be understood, providing data support for subsequent fault prediction and optimization.

[0045] Next, the solution performs a thermo-mechanical coupling analysis on the spindle temperature data and the characteristic parameter set to obtain a comprehensive characteristic vector. Thermo-mechanical coupling analysis refers to analyzing the influence of spindle temperature on mechanical properties while combining mechanical characteristics to comprehensively evaluate the working state of the spindle. The influence of temperature on spindle materials is particularly significant, especially during high-speed cutting, when the increase in temperature can cause material expansion and thermal deformation, thereby causing geometric deviations. Therefore, the thermo-mechanical coupling analysis combines the temperature change of the spindle with mechanical parameters such as vibration and stress, and obtains a more accurate spindle inner hole state through a coupling model. In this way, temperature and mechanical effects can be considered simultaneously, avoiding the errors caused by analyzing only temperature or only mechanical state.

[0046] Finally, these comprehensive feature vectors are arranged in time series and normalized to obtain a time-varying feature matrix that reflects the evolution of the spindle inner hole state. The time-varying feature matrix is ​​a dynamic matrix containing the spindle inner hole state change information. It can not only reflect the state of the spindle at different time periods, but also provide a basis for subsequent state prediction and anomaly detection. Through normalization, it can ensure that each feature is within the same scale range, thereby avoiding the influence of different parameter scales on the analysis results.

[0047] Overall, this solution effectively improves the ability to accurately identify and predict changes in the spindle bore state through wavelet time-frequency transform, time-frequency sub-spectrum division, dynamic frequency feature extraction, and thermal-mechanical coupling analysis. Each step of the design takes into account the dynamic performance of the spindle under high-speed cutting and complex working conditions, ensuring that the monitoring and fault warning of the spindle bore are more accurate and timely. Through the combination of these technologies, the level of spindle health management can be improved, precise maintenance and optimization can be achieved, processing efficiency can be improved, and the failure rate can be reduced.

[0048] Please continue to refer to Figure 1 , based on the time-varying characteristic matrix, using a preset nonlinear mapping model, reconstructing the geometric morphology and material properties of the spindle inner hole to obtain a digital twin model of the inner hole state; In one embodiment of the present application, the preset nonlinear mapping model includes a feature encoding layer, a multimodal fusion layer, a dynamic response layer, a geometric reconstruction layer and a material property estimation layer; The method is based on the time-varying feature matrix and utilizes a preset nonlinear mapping model to reconstruct the geometric morphology and material properties of the inner hole of the spindle to obtain a digital twin model of the inner hole state, including: inputting the time-varying feature matrix into the feature encoding layer, extracting features of vibration, acoustic emission and current signals through a multi-scale convolutional neural network to obtain a multimodal feature map; in the multimodal fusion layer, utilizing an attention mechanism to adaptively weight and fuse the multimodal feature map to obtain a fused feature vector; inputting the fused feature vector into the dynamic response layer, wherein the dynamic response layer adopts a long short-term memory network structure to dynamically analyze the spindle under different cutting conditions. The response is modeled and a time series feature sequence is output; in the geometric reconstruction layer, the time series feature sequence is processed by a graph convolutional network to obtain the three-dimensional geometric morphology of the inner hole of the spindle, and the three-dimensional geometric morphology includes roundness, straightness and surface roughness; the time series feature sequence is input into the material property estimation layer, and the material property estimation layer adopts a fully connected neural network structure to obtain the hardness distribution and residual stress state of the inner hole material; combined with the three-dimensional geometric morphology, the hardness distribution and residual stress state of the inner hole material, and the real-time collected spindle temperature field data, thermal-mechanical coupling analysis and dynamic optimization are carried out to obtain a digital twin model of the inner hole state.

[0049] Specifically, in the feature encoding layer, the vibration signal, acoustic emission signal and current signal are first processed. Since these signals have obvious changes in time and frequency, the use of multi-scale convolutional neural network (CNN) can extract important features from the signals at different time scales. Convolutional neural network can automatically learn local and global features in different signals, avoiding the complexity of manual feature selection. Specifically, the vibration signal can help understand the vibration mode and potential mechanical anomalies of the spindle, the acoustic emission signal can detect cracks or wear inside the spindle, and the current signal reflects the load and working status of the motor. For example, during the cutting process, if the current signal suddenly increases, it may be due to an increase in the spindle load or an overload phenomenon. Through the feature extraction of CNN, this change can be efficiently captured, thereby providing support for subsequent fault diagnosis.

[0050] The dynamic response layer uses a long short-term memory network (LSTM) to perform time series modeling on the fused feature vectors. LSTM is a neural network structure that is very suitable for processing time series data. It can effectively capture long-term dependencies in signals. In this solution, LSTM is used to model the dynamic response of the spindle under different cutting conditions, and can accurately describe the behavioral changes of the spindle in different stages such as acceleration, deceleration, and stable cutting. For example, during the acceleration stage, the load and vibration of the spindle may increase dramatically, and LSTM can capture these changes by memorizing historical data and predict the performance of the spindle under future conditions. This time series modeling capability provides reliable time-dependent features for subsequent state evaluation, fault prediction, and performance optimization.

[0051] In the geometric reconstruction layer, the time series feature sequence is processed, and the three-dimensional geometric shape of the spindle is reconstructed using a graph convolutional network (GCN). The key role of this layer is to convert the features extracted from the time series data into the geometry of the spindle inner hole, such as important parameters such as roundness, straightness, and surface roughness. Unlike traditional convolutional neural networks, graph convolutional networks can process graph-structured data, making them very suitable for analyzing the geometric changes of the spindle inner hole. For example, the inner hole of the spindle may undergo thermal expansion during high-speed cutting, resulting in slight changes in the geometry. Through GCN, these slight geometric changes can be accurately mapped to three-dimensional space to obtain a more intuitive and high-precision inner hole geometry model, which provides key data for optimizing machining accuracy and predicting potential faults.

[0052] Next, the material property estimation layer uses a fully connected neural network (FCN) to process the time series feature sequence to obtain the hardness distribution and residual stress state of the inner hole material. Hardness and residual stress are important factors affecting the performance and life of the spindle. Uneven hardness of the spindle inner hole material may cause material degradation, thereby affecting the processing accuracy; residual stress may cause spindle deformation or cracks. Therefore, through the fully connected neural network, the material properties of the spindle inner hole can be accurately estimated, further improving the overall picture of the inner hole state.

[0053] Finally, the thermo-mechanical coupling analysis combines the material properties and the working state of the spindle, and further optimizes the digital twin model by analyzing the impact of temperature changes and mechanical loads on the spindle inner hole state. Since the spindle may generate a lot of heat during high-speed cutting, the increase in temperature may cause thermal expansion of the spindle, which will affect the geometry and performance. The thermo-mechanical coupling analysis combines temperature data with mechanical vibration, load and other factors, and considers the joint impact of thermal and mechanical effects on the spindle inner hole state, thereby obtaining a more accurate inner hole state digital twin model.

[0054] This digital twin model can provide real-time status monitoring of the spindle inner hole by integrating multiple information such as geometric morphology, material properties, and temperature field data, and can provide early warning of potential failures and provide a basis for processing optimization and maintenance decisions. Through this comprehensive analysis model, the health status of the spindle can be predicted more accurately, reducing production downtime caused by failures and improving processing efficiency and accuracy.

[0055] Through the combination of the above technologies, this solution can model the spindle inner hole state from multiple dimensions, which not only reflects the dynamic response of the spindle under different cutting conditions, but also takes into account the characteristics of the inner hole material and the temperature influence, providing a scientific basis for the precise monitoring, fault diagnosis and maintenance optimization of the spindle. The advantage of this method is that through a multi-level deep learning architecture and efficient data fusion, it can accurately capture the dynamic changes of the spindle state, thereby realizing intelligent early warning and optimization solution generation, significantly improving the stability and economic benefits of the production line.

[0056] Please continue to refer to Figure 1 , performing multi-scale analysis on the digital twin model of the inner hole state, identifying macroscopic geometric deviations and microscopic material changes, and generating layered anomaly indicators; In one embodiment of the present application, the multi-scale analysis of the inner hole state digital twin model is performed to identify macro-geometric deviations and micro-material changes, and generate hierarchical anomaly indicators, including: performing multi-scale wavelet decomposition on the inner hole state digital twin model to obtain a multi-scale feature set that reflects the geometric morphology and material properties of the spindle inner hole at different processing stages; based on the multi-scale feature set, using the principal component analysis method to extract the macro-geometric deviation characteristics of the spindle inner hole to obtain a macro-geometric anomaly indicator; applying a local binary pattern algorithm to the multi-scale feature set to identify the micro-material change pattern on the surface of the spindle inner hole to obtain a micro-material anomaly indicator; combining the macro-geometric anomaly indicators and the micro-material anomaly indicators to construct a hierarchical anomaly assessment system to generate a hierarchical anomaly indicator containing geometric deviation and material change information.

[0057] Specifically, in order to accurately monitor and evaluate the state of the spindle inner hole, the digital twin model of the spindle inner hole is first processed through multi-scale wavelet decomposition. Wavelet decomposition is a powerful tool that can simultaneously analyze signals in two dimensions: time and frequency. Through wavelet transform, the signal is decomposed into sub-signals of different scales, each of which represents the change of the signal in a different frequency range. For the monitoring of the spindle inner hole state, this means that information such as the slight deformation, temperature change, and wear of the inner hole during the machining process can be extracted from different time scales. Wavelet decomposition can effectively capture local changes in the signal, especially those non-stationary signal fluctuations, such as the rapid changes that may occur when the spindle goes from acceleration to stable cutting stage.

[0058] For example, during high-speed cutting, the spindle bore may expand slightly due to temperature rise. Through wavelet decomposition, these tiny changes in the high-frequency range can be accurately captured. And through this multi-scale decomposition, not only can these changes be identified, but also the different frequency components in each time period can be accurately distinguished. For example, in the high-load cutting stage, the change of vibration frequency may be more obvious than in the stable cutting stage, and these changes will be reflected in different scales in the wavelet decomposition, helping us to identify the working status of the spindle at different times.

[0059] After obtaining the multi-scale feature set, the next step is to extract the macro-geometric deviation features through principal component analysis (PCA). Principal component analysis is a statistical method for dimensionality reduction. It can convert multi-dimensional data in the original feature space into a few principal components through linear transformation, which retain the variance of the data to the greatest extent. In this scheme, PCA is used to extract the main features that best represent the geometric state of the spindle inner hole from the multi-scale feature set, such as macro-geometric parameters such as roundness and straightness. Through PCA, the core information of geometric deviation can be extracted from complex, multi-dimensional signal data.

[0060] For example, if a spindle has a slight geometric deviation after running for a long time, PCA can extract these deviations from vibration signals and other features and convert them into specific geometric indicators, such as roundness error. If the spindle has geometric deviations during stable cutting, these deviations may affect the machining accuracy. PCA can help identify such geometric changes and provide a basis for subsequent maintenance decisions.

[0061] Next, the local binary pattern algorithm (LBP) is used to identify the microscopic material change patterns on the surface of the spindle inner hole. LBP is an algorithm commonly used for texture and surface pattern recognition. It extracts the characteristics of small surface changes by binarizing the local area of ​​the image. In this scheme, LBP is used to analyze the microscopic material changes that may occur on the surface of the inner hole, such as tiny cracks caused by excessive wear or thermal expansion, changes in surface roughness, etc. Through the local binary pattern algorithm, tiny surface damage patterns can be extracted from the vibration signal. Although these microscopic changes are not easy to detect in the macroscopic geometry, they have a crucial impact on the long-term performance and machining accuracy of the spindle.

[0062] For example, after long-term use, the inner surface of the spindle may have microcracks due to friction and temperature changes. These cracks will show specific patterns in the acoustic emission signal and vibration signal. The LBP algorithm can effectively identify these microscopic damages, thereby obtaining microscopic material abnormality indicators, providing a reliable basis for early detection of potential problems.

[0063] After obtaining the macroscopic geometric anomaly indicators and microscopic material anomaly indicators, the next step is to combine them to build a hierarchical anomaly assessment system. The core of this system is to combine the two aspects of abnormal information, geometric deviation and material change, to generate a comprehensive abnormality assessment indicator. In this way, the health status of the spindle inner hole can be evaluated more comprehensively and accurately. Specifically, the geometric deviation reflects the overall shape change of the spindle, while the material change provides more microscopic damage information on the inner hole surface. By combining this information, the condition of the spindle can be fully understood, which can provide decision support for maintenance and optimization.

[0064] For example, if the roundness deviation and surface wear of the spindle change in different working stages, relying solely on any one indicator of geometric deviation or material change cannot fully reflect the true state of the spindle. Through the hierarchical abnormality assessment system, taking these two changes into consideration, it is possible to more accurately identify whether the spindle has potential failures and provide more precise strategies for maintenance. This method effectively avoids the limitations of single-dimensional analysis and enhances the system's early warning capabilities.

[0065] Through the combination of multi-scale wavelet decomposition, principal component analysis, local binary pattern algorithm and hierarchical anomaly assessment system, the solution can comprehensively evaluate the state of the spindle inner hole from multiple dimensions. This comprehensive analysis method can not only accurately capture the changes in the geometry and material state of the spindle inner hole, but also provide accurate fault prediction and maintenance suggestions through hierarchical anomaly indicators, thereby improving processing accuracy and equipment stability. The design of each step takes into account the needs of actual applications, ensuring accurate monitoring and prediction of the spindle state, greatly improving the efficiency and safety of the production process.

[0066] Please continue to refer to Figure 1 , combining historical data and the above-mentioned stratified abnormality indicators, applying deep learning algorithms to analyze the performance evolution trend of the spindle inner hole, and outputting multi-time scale state prediction results; In one embodiment of the present application, the historical data and the stratified abnormality indicators are combined, and a deep learning algorithm is applied to analyze the performance evolution trend of the spindle inner hole, and a multi-time scale state prediction result is output, including: aligning the historical processing data and the stratified abnormality indicators in time series to obtain a comprehensive time series data set; performing time series decomposition on the comprehensive time series data set to obtain multiple time scale components reflecting the performance changes of the spindle inner hole; using a deep learning algorithm to extract features of the multiple time scale components to obtain a multi-scale feature representation of the spindle inner hole performance; based on the multi-scale feature representation, calculating the predicted values ​​of the spindle inner hole geometric accuracy change, material performance degradation and remaining service life; performing statistical analysis on the predicted values ​​to obtain a performance prediction indicator including a prediction result and a confidence interval; based on the performance prediction indicator, generating a prediction reflecting the performance change of the spindle inner hole within 24 hours, 7 days and 30 days to obtain a multi-time scale state prediction result.

[0067] Specifically, one of the core goals of this solution is to accurately predict the performance changes of the spindle inner hole through historical processing data and real-time monitoring of layered abnormal indicators, combined with time series analysis and deep learning technology. To achieve this goal, it is first necessary to align the historical processing data with the layered abnormal indicators in time series to obtain a comprehensive time series data set. The key to this process is that the historical processing data reflects the performance of the spindle under past working conditions, while the layered abnormal indicators provide real-time abnormal warnings by monitoring the inner hole status. By aligning the two, the past status can be combined with the current status to construct a data set with timeliness and historical continuity, providing a solid foundation for subsequent analysis.

[0068] The process of time series alignment involves synchronizing historical data and real-time data in the time dimension. Because the working status of the spindle changes over time, data at different time points have different representativeness. By aligning historical processing data and real-time layered abnormal indicators, each data point can be accurately matched to the corresponding time point, thus forming a continuous time series data set. This data set can not only reflect the dynamic changes of the spindle inner hole, but also compare the past operating status with the current status, helping to identify the trend of spindle performance changes and potential failure risks.

[0069] For example, if the historical data records the changes in the vibration signal of the spindle under different loads and speeds, and the real-time data provides the geometric and material change indicators of the spindle inner hole, then through time series alignment, these data can be integrated into a unified time series data set. In this way, the relationship between the vibration, geometric deviation and material state of the spindle can be observed at the time of analysis, and the health of the spindle can be identified more accurately.

[0070] After obtaining the comprehensive time series data set, the next step is to perform time series decomposition on the data set. The purpose of time series decomposition is to separate the components of different time scales in the comprehensive time series data to help analyze the changes in spindle performance from multiple levels. Specifically, time series decomposition can decompose the data into multiple time scale components, each of which reflects the different frequency characteristics of the spindle performance changes. Through this decomposition, we can see the performance differences of the spindle in different time periods and under different working conditions, and then identify the long-term trend and short-term fluctuations of the spindle bore performance.

[0071] For example, after the spindle has been working under high load for a long time, the vibration signal and temperature change may fluctuate greatly in the short term, which may reflect the rapid wear or temperature rise of the spindle. However, the temperature and vibration changes of the spindle during the normal working stage are relatively stable. Through time series decomposition, these short-term fluctuations and long-term trends can be clearly distinguished, providing more accurate data for subsequent analysis.

[0072] After completing the time series decomposition, deep learning algorithms are used to extract features from multiple time scale components. Deep learning, especially neural network structures such as convolutional neural networks (CNN) and long short-term memory networks (LSTM), have shown strong capabilities in feature extraction of time series data. Through these algorithms, key features of spindle bore performance can be extracted from multi-scale components. These features include geometric changes of the spindle under different working conditions, material performance degradation, and vibration modes. By learning features from large amounts of data, deep learning algorithms can automatically discover potential patterns in the data, avoiding the complexity of manual feature selection.

[0073] For example, after the spindle has been cutting for a long time, the temperature and vibration signals may show obvious trend changes. The deep learning algorithm can extract the performance change pattern of the spindle inner hole by learning the time series characteristics of these signals, and generate a multi-scale feature representation that can accurately reflect the health status of the spindle. These features can be further used to predict the geometric accuracy change, material performance degradation and remaining service life of the spindle.

[0074] Based on the obtained multi-scale feature representation, the system can further calculate the predicted values ​​of the spindle inner hole geometric accuracy change, material performance degradation and remaining service life. The core of this process is to establish a comprehensive prediction model by combining multi-scale features with the historical and real-time working conditions of the spindle. Specifically, the geometric accuracy change can be analyzed by measuring the geometric characteristics of the inner hole such as roundness and straightness; material performance degradation is evaluated by monitoring material hardness, residual stress and wear; the remaining service life is calculated by analyzing the spindle's working history, current status and future working environment. The remaining effective use time of the spindle. These predicted values ​​provide maintenance personnel with detailed performance predictions to help make maintenance decisions in advance and prevent spindle failures or failures.

[0075] After obtaining these predicted values, statistical analysis is used to analyze the predicted results and generate a performance prediction index that includes the predicted results and confidence intervals. The purpose of statistical analysis is to give the reliability and accuracy of each predicted value by combining multiple predicted results. The confidence interval provides a range for the predicted results, indicating the possible fluctuation range of the predicted values ​​under a certain confidence level. For example, if the predicted value of the geometric accuracy change of the spindle is 0.1mm and the confidence interval is [0.08mm, 0.12mm], it means that the geometric accuracy change of the spindle is likely to fluctuate within this range in future working cycles. Through the confidence interval, users can understand the credibility of the predicted results and make more reasonable maintenance decisions based on this information.

[0076] Finally, based on the generated performance prediction indicators, the system can generate predictions that reflect the performance changes of the spindle inner hole within 24 hours, 7 days, and 30 days, thereby obtaining multi-time scale state prediction results. These prediction results provide the health status of the spindle in different time periods, helping users understand the performance evolution process of the spindle and provide decision support for future work. For example, if the prediction shows that the geometric accuracy of the spindle will change significantly in the next 7 days, the system can notify maintenance personnel in advance to make adjustments, thereby avoiding production downtime caused by spindle failure.

[0077] Through these steps, the entire prediction process can not only accurately capture the performance changes of the spindle, but also provide maintenance personnel with accurate early warnings based on the analysis results at different time scales. The advantage of this method is that it combines historical data, real-time monitoring information and deep learning technology, and can provide accurate spindle health prediction through precise time series analysis, avoiding the lag and limitations of traditional methods, thereby improving the use efficiency of the spindle and the stability of production.

[0078] Please continue to refer to Figure 1, according to the state prediction results and the current processing task requirements, a decision plan including processing parameter optimization and customized maintenance plan is dynamically generated and fed back to the processing control system in real time.

[0079] In one embodiment of the present application, a decision plan including processing parameter optimization and a customized maintenance plan is dynamically generated based on the state prediction results and the current processing task requirements, and is fed back to the processing control system in real time, including: extracting key parameters such as workpiece material, processing accuracy and surface quality in the current processing task requirements to obtain task requirement indicators; matching and analyzing the state prediction results with the task requirement indicators to obtain an adaptability evaluation result between the spindle performance and the processing requirements; calculating the adjustment amount of key processing parameters such as spindle speed, feed speed and cutting depth based on the adaptability evaluation results, and generating an optimized processing parameter plan; based on the performance degradation trend in the state prediction results and the task requirement indicators, formulating a customized maintenance plan including maintenance time, replacement parts and adjustment frequency; integrating the optimized processing parameter plan and the customized maintenance plan into a decision plan, and transmitting the decision plan to the processing control system in real time through a data interface.

[0080] Specifically, the goal of this solution is to closely integrate the machining task requirements with the actual working status of the spindle. Through precise analysis and calculation, the machining parameters are adjusted in real time and a customized maintenance plan is formulated to ensure the machining quality and long-term stability of the equipment. First of all, the task requirement indicators are extracted to convert the core requirements of the machining task, such as workpiece material, machining accuracy, surface quality, etc., into specific parameters. These parameters have a direct impact on the working status of the spindle, such as the cutting force of different materials, the error range of machining accuracy requirements, and the requirements of surface quality on spindle performance, etc. These all need to be accurately extracted and converted into task requirement indicators.

[0081] For example, when performing high-precision aluminum alloy processing, the material hardness and surface finish requirements will affect the cutting force and spindle load. Therefore, the system will extract task demand indicators by collecting workpiece material hardness information, processing accuracy and surface quality standards. These indicators provide a basis for subsequent evaluation of spindle performance and adjustment of processing parameters.

[0082] Once these task demand indicators are obtained, the next step is to match the state prediction results with the task demand indicators. The purpose of this process is to evaluate the adaptability between the current performance of the spindle and the processing requirements. If the health status of the spindle does not meet the processing accuracy and surface quality requirements, the system will automatically identify the mismatch through matching analysis. For example, suppose the prediction shows that the spindle has a slight geometric deviation under the current working conditions, but the processing task requires higher accuracy. The system will automatically identify this gap and make subsequent adjustments based on this gap.

[0083] Specifically, matching analysis is to calculate the difference between each performance parameter (such as vibration, temperature, cutting force, etc.) and the actual demand by comparing the task demand indicators and the spindle performance prediction results. If the difference exceeds a certain preset threshold, the system will identify potential problems and provide reference for subsequent machining parameter adjustments. This matching analysis can not only help identify existing problems, but also provide early warning of potential performance degradation, thereby helping to optimize the machining process.

[0084] Next, based on the results of the adaptability assessment, the system will calculate the adjustments to key machining parameters such as spindle speed, feed rate and cutting depth. The core of this process is to dynamically adjust the machining parameters according to the task requirements and the current performance status of the spindle. For example, if the vibration signal of the spindle indicates that there is a slight deviation in its geometric accuracy, the system may recommend reducing the cutting depth, reducing the speed or adjusting the feed rate to reduce the spindle load and improve the machining accuracy. Calculating these adjustments is not just a simple increase or decrease operation, but relies on a comprehensive analysis of multiple factors such as spindle status, machining accuracy and material properties to ensure that the adjustment of each parameter can effectively improve the machining quality and reduce spindle wear.

[0085] For example, when the spindle vibration increases, the system may calculate the need to reduce the cutting depth and speed based on the existing vibration signal and temperature changes, thereby reducing the load during the processing and reducing the wear of the spindle. This adjustment can not only avoid damage to the spindle due to excessive load, but also improve the quality of the processed surface and ensure the stability of the processing accuracy.

[0086] Based on the performance degradation trend and task demand indicators in the status prediction results, the system will develop a customized maintenance plan, which includes maintenance time, replacement parts, and adjustment frequency. The design basis of the customized maintenance plan is the historical data of spindle performance degradation and the current status prediction results. For example, if the system predicts that the spindle may fail due to high temperature and vibration in the next month, the system will arrange the maintenance window of the spindle in advance to avoid sudden failures. Through this prediction, users can inspect and maintain the equipment in advance according to the plan, reducing downtime caused by sudden failures.

[0087] This customized maintenance plan has obvious advantages over the traditional scheduled maintenance plan. Traditional maintenance plans usually perform equipment inspections and repairs based on fixed cycles, while this customized maintenance plan based on real-time spindle performance evaluation is more flexible and can dynamically adjust the maintenance cycle based on the actual use of the spindle, greatly improving equipment utilization and production efficiency.

[0088] Finally, by integrating the optimized processing parameter plan and customized maintenance plan into a decision plan, and transmitting the decision plan to the processing control system in real time through the data interface, it is ensured that every link in the processing process can be fine-tuned according to real-time data. This step realizes the automation and intelligence of processing parameters and maintenance plans, enabling the system to respond to the state changes of the spindle in real time and quickly adjust the processing process. Through this real-time feedback mechanism, the system can ensure that the processing process is always kept in the best state, avoid spindle failure due to excessive load or untimely maintenance, and further improve processing accuracy and production efficiency.

[0089] For example, if the system detects that the temperature of the spindle exceeds the set threshold, the real-time transmission decision plan will automatically adjust the cutting depth and speed, and notify the operator to perform necessary inspections and maintenance. This not only ensures machining accuracy, but also avoids production interruptions through real-time adjustment and early warning mechanisms.

[0090] In summary, this solution has established a flexible, intelligent and efficient processing optimization and equipment maintenance system through the steps of extracting task demand indicators, matching analysis, processing parameter optimization, customized maintenance plan and real-time decision transmission. Each step is closely combined with the actual state of the spindle and the requirements of the processing task, which not only ensures the processing accuracy and surface quality, but also improves production efficiency and reduces downtime through real-time feedback and dynamic adjustment, thereby significantly improving the stability and economic benefits of the overall processing process.

[0091] The above describes the CNC machining intelligent monitoring and adaptive control method in the embodiment of the present invention. The following describes the CNC machining intelligent monitoring and adaptive control device in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an intelligent monitoring and adaptive control device for numerical control machining includes: The data acquisition module 201 is used to simultaneously acquire the vibration signal, acoustic emission signal and current signal of the spindle when the spindle is rotating at a high speed, so as to obtain a multi-modal original data set; The time-frequency analysis module 202 is used to perform a time-frequency domain joint analysis on the multi-modal original data set, extract dynamic characteristic parameters reflecting the state of the spindle inner hole, and generate a time-varying characteristic matrix; The digital twin module 203 is used to reconstruct the geometric morphology and material properties of the spindle inner hole based on the time-varying characteristic matrix and use a preset nonlinear mapping model to obtain a digital twin model of the inner hole state; An anomaly identification module 204 is used to perform multi-scale analysis on the inner hole state digital twin model, identify macroscopic geometric deviations and microscopic material changes, and generate hierarchical anomaly indicators; A state prediction module 205 is used to combine historical data and the layered abnormality index, apply a deep learning algorithm to analyze the performance evolution trend of the spindle inner hole, and output a state prediction result at multiple time scales; The decision generation module 206 is used to dynamically generate a decision plan including processing parameter optimization and customized maintenance plan according to the state prediction results and the current processing task requirements, and provide real-time feedback to the processing control system.

[0092] above Figure 2 The intelligent monitoring and adaptive control device for CNC machining in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The intelligent monitoring and adaptive control device for CNC machining in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0093] Figure 3 3 is a structural diagram of a CNC machining intelligent monitoring and adaptive control device provided by an embodiment of the present invention. The CNC machining intelligent monitoring and adaptive control device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the CNC machining intelligent monitoring and adaptive control device 300. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the CNC machining intelligent monitoring and adaptive control device 300 to implement the steps of the above-mentioned CNC machining intelligent monitoring and adaptive control method.

[0094] The CNC machining intelligent monitoring and adaptive control device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 3 The structure of the CNC machining intelligent monitoring and adaptive control device shown does not constitute a limitation on the CNC machining intelligent monitoring and adaptive control device provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0095] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the CNC machining intelligent monitoring and adaptive control method.

[0096] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0097] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0098] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A method for intelligent monitoring and adaptive control of numerical control machining, characterized in that: include: When the spindle is rotating at high speed, the vibration signal, acoustic emission signal and current signal of the spindle are collected simultaneously to obtain a multi-modal raw data set; Performing a joint analysis of the multimodal original data set in the time and frequency domains, extracting dynamic characteristic parameters reflecting the state of the spindle inner hole, and generating a time-varying characteristic matrix; Based on the time-varying characteristic matrix, the geometric morphology and material properties of the spindle inner hole are reconstructed using a preset nonlinear mapping model to obtain a digital twin model of the inner hole state; Performing multi-scale analysis on the digital twin model of the inner hole state, identifying macroscopic geometric deviations and microscopic material changes, and generating layered anomaly indicators; Combining historical data and the above-mentioned stratified abnormality indicators, a deep learning algorithm is applied to analyze the performance evolution trend of the spindle inner hole, and a multi-time scale state prediction result is output; According to the state prediction results and the current processing task requirements, decision plans including processing parameter optimization and customized maintenance plans are dynamically generated and fed back to the processing control system in real time.

2. The CNC machining intelligent monitoring and adaptive control method according to claim 1 is characterized in that: When the spindle is rotating at a high speed, the vibration signal, acoustic emission signal and current signal of the spindle are collected simultaneously to obtain a multi-modal original data set, including: Obtain the current spindle speed, cutting force and processing material information to obtain the spindle working condition parameters; Determining a time window for signal acquisition according to the spindle operating parameters; In the time window, synchronously collect vibration signals, acoustic emission signals and current signals of the bearing seat, tool interface and motor winding of the main shaft to obtain original multi-modal signals; Synchronously collecting spindle temperature and coolant flow data, and performing time alignment with the original multimodal signal to obtain correlated signal data; The associated signal data are collected and integrated during a complete working cycle of the spindle, and are associated with the processing parameters and environmental factors of the spindle to obtain a multimodal original data set containing spatiotemporal information.

3. The CNC machining intelligent monitoring and adaptive control method according to claim 1 is characterized in that: The method of performing a time-frequency domain joint analysis on the multi-modal original data set, extracting dynamic characteristic parameters reflecting the state of the spindle inner hole, and generating a time-varying characteristic matrix includes: Performing wavelet time-frequency transformation on the multimodal original data set to obtain a time-frequency representation spectrum; The time-frequency representation spectrum is combined with the spindle operation stage information to divide it into time-frequency sub-spectra of no-load, acceleration, stable cutting and deceleration stages; Performing adaptive threshold processing on the time-frequency sub-spectrum, extracting characteristic frequencies and energy distributions of each stage, and obtaining dynamic frequency characteristics; Combining the dynamic frequency characteristics with the cutting parameters of the spindle, a characteristic parameter set reflecting the geometric state of the inner hole is constructed; Performing a thermal-mechanical coupling analysis on the characteristic parameter set and the spindle temperature data to obtain a comprehensive characteristic vector; The comprehensive characteristic vectors are arranged in time series and normalized to generate a time-varying characteristic matrix reflecting the evolution of the spindle inner hole state.

4. The CNC machining intelligent monitoring and adaptive control method according to claim 1 is characterized in that: The preset nonlinear mapping model includes a feature encoding layer, a multimodal fusion layer, a dynamic response layer, a geometric reconstruction layer and a material property estimation layer; Based on the time-varying characteristic matrix, the preset nonlinear mapping model is used to reconstruct the geometric morphology and material properties of the spindle inner hole to obtain the inner hole state digital twin model, including: The time-varying feature matrix is ​​input into the feature encoding layer, and the vibration, acoustic emission and current signals are subjected to feature extraction through a multi-scale convolutional neural network to obtain a multi-modal feature map; In the multimodal fusion layer, the attention mechanism is used to adaptively weight and fuse the multimodal feature maps to obtain a fused feature vector; The fused feature vector is input into the dynamic response layer, and the dynamic response layer adopts a long short-term memory network structure to model the dynamic response of the spindle under different cutting conditions and output a time series feature sequence; In the geometric reconstruction layer, the temporal feature sequence is processed by using a graph convolutional network to obtain a three-dimensional geometric morphology of the spindle inner hole, wherein the three-dimensional geometric morphology includes roundness, straightness and surface roughness; Inputting the time series feature sequence into the material property estimation layer, the material property estimation layer adopts a fully connected neural network structure to obtain the hardness distribution and residual stress state of the inner hole material; Combined with the three-dimensional geometric morphology, the hardness distribution and residual stress state of the inner hole material, and the real-time collected spindle temperature field data, thermal-mechanical coupling analysis and dynamic optimization are performed to obtain a digital twin model of the inner hole state.

5. The CNC machining intelligent monitoring and adaptive control method according to claim 1 is characterized in that: The multi-scale analysis of the inner hole state digital twin model is performed to identify macroscopic geometric deviations and microscopic material changes, and generate layered abnormality indicators, including: Performing multi-scale wavelet decomposition on the inner hole state digital twin model to obtain a multi-scale feature set reflecting the geometric morphology and material properties of the spindle inner hole at different processing stages; Based on the multi-scale feature set, the macro-geometric deviation features of the spindle inner hole are extracted using the principal component analysis method to obtain the macro-geometric anomaly index; Applying a local binary pattern algorithm to the multi-scale feature set to identify a microscopic material change pattern on the surface of the spindle inner hole and obtain a microscopic material anomaly index; By combining the macroscopic geometric anomaly index and the microscopic material anomaly index, a hierarchical anomaly evaluation system is constructed to generate hierarchical anomaly index containing geometric deviation and material change information.

6. The CNC machining intelligent monitoring and adaptive control method according to claim 1 is characterized in that: The historical data and the layered abnormality index are combined to analyze the performance evolution trend of the spindle inner hole using a deep learning algorithm, and the state prediction results of multiple time scales are output, including: Performing time series alignment on the historical processed data and the hierarchical anomaly indicators to obtain a comprehensive time series data set; Performing time series decomposition on the comprehensive time series data set to obtain multiple time scale components reflecting the change of spindle inner hole performance; Using a deep learning algorithm to extract features from the multiple time scale components, and obtain a multi-scale feature representation of the spindle inner hole performance; Based on the multi-scale feature representation, calculating the predicted values ​​of the spindle inner hole geometric accuracy change, material performance degradation and remaining service life; Performing statistical analysis on the predicted values ​​to obtain a performance prediction index including a prediction result and a confidence interval; According to the performance prediction index, a prediction of the performance change of the spindle inner hole within 24 hours, 7 days and 30 days is generated to obtain a multi-time scale state prediction result.

7. The CNC machining intelligent monitoring and adaptive control method according to claim 1 is characterized in that: According to the state prediction results and the current processing task requirements, a decision plan including processing parameter optimization and customized maintenance plan is dynamically generated and fed back to the processing control system in real time, including: Extract key parameters such as workpiece material, machining accuracy and surface quality required by the current machining task to obtain task requirement indicators; Matching and analyzing the state prediction result with the task requirement index to obtain an adaptability evaluation result of the spindle performance and the processing requirement; According to the adaptability evaluation results, the adjustment amounts of key machining parameters such as spindle speed, feed speed and cutting depth are calculated to generate an optimized machining parameter solution; Based on the performance degradation trend in the state prediction result and the mission requirement indicator, formulate a customized maintenance plan including maintenance time, replacement parts and adjustment frequency; The optimized processing parameter plan and the customized maintenance plan are integrated into a decision-making plan, and the decision-making plan is transmitted to the processing control system in real time through a data interface.

8. A CNC machining intelligent monitoring and adaptive control device, characterized in that: The CNC machining intelligent monitoring and adaptive control device adopts the CNC machining intelligent monitoring and adaptive control method according to any one of claims 1 to 7, and the CNC machining intelligent monitoring and adaptive control device comprises: The data acquisition module is used to simultaneously collect the vibration signal, acoustic emission signal and current signal of the spindle when the spindle is rotating at a high speed, so as to obtain a multi-modal original data set; A time-frequency analysis module is used to perform a time-frequency domain joint analysis on the multi-modal original data set, extract dynamic characteristic parameters reflecting the state of the spindle inner hole, and generate a time-varying characteristic matrix; A digital twin module is used to reconstruct the geometric morphology and material properties of the spindle inner hole based on the time-varying characteristic matrix and use a preset nonlinear mapping model to obtain a digital twin model of the inner hole state; An anomaly identification module, used to perform multi-scale analysis on the inner hole state digital twin model, identify macroscopic geometric deviations and microscopic material changes, and generate hierarchical anomaly indicators; A state prediction module is used to combine historical data and the layered abnormality index, apply a deep learning algorithm to analyze the performance evolution trend of the spindle inner hole, and output a state prediction result at multiple time scales; The decision generation module is used to dynamically generate decision plans including processing parameter optimization and customized maintenance plans according to the state prediction results and the current processing task requirements, and provide real-time feedback to the processing control system.

9. A CNC machining intelligent monitoring and adaptive control device, characterized in that: The CNC machining intelligent monitoring and adaptive control device comprises: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory so that the CNC machining intelligent monitoring and adaptive control device executes the steps of the CNC machining intelligent monitoring and adaptive control method as described in any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the steps of the CNC machining intelligent monitoring and adaptive control method as described in any one of claims 1 to 7 are implemented.

Images