Numerical control machining equipment state monitoring system based on big data analysis

Abstract

This invention discloses a CNC machining equipment status monitoring system based on big data analysis, belonging to the field of intelligent manufacturing technology. It addresses the problem of analyzing and providing real-time feedback on the correlation between tool wear and machining quality during the machining process. By extracting time-series correlated feature vectors from vibration signals, acoustic emission signals, and spindle axial micro-displacement data, and combining this with a nonlinear regression model to calculate the machining quality degradation index, machining quality is monitored in real time. Adaptive feature weight adjustment based on the type of machining material enhances feature fusion during the machining of different materials, improving the accuracy of quality assessment. A gradient boosting tree model is used to establish a nonlinear mapping relationship between tool wear and surface roughness, enabling accurate prediction of tool wear and machining quality. Finally, a closed-loop dynamic adjustment mechanism automatically adjusts the cutting speed and feed rate based on the machining quality degradation index and fault location results, ensuring the stability of the machining process and product quality.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, specifically to a status monitoring system for CNC machining equipment based on big data analysis. Background Technology

[0002] With the continuous development of the manufacturing industry, CNC machining equipment plays an increasingly important role in modern production. As manufacturing demands become more complex and precision requirements increase, the stability, efficiency, and machining quality of CNC machining equipment directly impact production efficiency and product quality. Therefore, real-time monitoring and predictive maintenance of CNC machining equipment can not only improve operational reliability but also reduce equipment failure rates, minimize downtime, and enhance overall production efficiency. By combining big data analytics and intelligent monitoring technologies, real-time monitoring and analysis of CNC equipment's operating status can provide a scientific basis for equipment maintenance decisions, further promoting the development of intelligent manufacturing.

[0003] Currently, most CNC machining equipment monitoring systems still rely on traditional monitoring methods, typically using single sensor data acquisition and static threshold judgment for fault diagnosis. However, these methods have significant shortcomings. First, traditional methods usually rely on only a single type of data, such as vibration or temperature signals, failing to comprehensively reflect the complexity of the equipment's operating status. Second, these systems often fail to consider the changes in various operating conditions during machining, resulting in poor data fusion and analysis. Furthermore, due to limitations in sensor installation location and type, existing systems have low accuracy in predicting equipment health status and struggle to adapt to changing machining environments. Moreover, existing methods lack in-depth analysis of the relationship between tool wear and machining quality, failing to effectively construct a precise mapping relationship between tool wear and quality degradation, thus affecting the accuracy of equipment maintenance and machining quality optimization. Therefore, there is an urgent need for a technical solution integrating multimodal sensing, big data analysis, and deep learning to improve the monitoring effect of CNC machining equipment and overcome the limitations of existing methods. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a CNC machining equipment status monitoring system based on big data analysis, which solves the problems mentioned in the background.

[0005] To achieve the above objectives, this invention provides the following technical solution: a CNC machining equipment status monitoring system based on big data analysis, comprising the following modules: a multimodal perception fusion module, an implicit machining quality correlation analysis module, a small-sample federated residual graph network module, and a machining parameter execution module; the multimodal perception fusion module is used to collect vibration, temperature, acoustic emission signals, and spindle axial micro-displacement data of the CNC equipment in real time, combine the working condition labels obtained by G-code parsing, construct a dynamic weight matrix based on the machining stage identification results, perform adaptive fusion processing on multi-source heterogeneous data, and output a multi-dimensional time-series feature vector representing the equipment's operating status; the implicit machining quality correlation analysis module is used to receive the multi-dimensional time-series feature vector and online detected workpiece surface roughness and size data, and extract the roughness from the acoustic emission signal through a time-frequency domain attention mechanism. Highly correlated feature subbands, combined with a nonlinear fusion strategy, perform dual-channel time-series feature alignment on the non-stationary characteristics of vibration signals, high-frequency pulse characteristics of acoustic emission, and thermal deformation trends in spindle axial micro-displacement. This constructs a time-series mapping model between machining status and quality degradation, thereby generating a machining quality degradation index and establishing a nonlinear mapping relationship between tool wear and machining quality. The small-sample federated residual graph network module is used to construct a graph structure based on the physical connection relationships of equipment components. It is trained using the federated learning framework in collaboration with residual parameters from multiple factories, initializes graph edge weights using the spindle-bearing stiffness matrix, and outputs predictions of remaining equipment life and results for locating latent faults. The machining parameter execution module receives the machining quality degradation index and fault location results, dynamically adjusts cutting speed and feed rate parameters through a feedback closed loop, and generates preventative maintenance work orders.

[0006] Furthermore, the specific process of collecting vibration, temperature, acoustic emission signals, and spindle axial micro-displacement data of the CNC equipment, combined with the working condition labels obtained from G-code parsing, and constructing a dynamic weight matrix based on the machining stage identification results is as follows: Parse the G-code of the CNC equipment, extract process parameters such as cutting speed and feed rate, and identify whether the current machining stage is roughing, semi-finishing, or finishing; and adjust the weights of vibration signals, acoustic emission signals, and spindle axial micro-displacement data according to the different machining stages; in the finishing stage, the sensitivity to minute vibrations is enhanced by increasing the weight of vibration signals; at the end of the tool change cycle, the ability to capture high-frequency pulses is improved by increasing the weight of acoustic emission signals; in continuous machining mode, the weight of spindle axial micro-displacement data is gradually increased to capture subtle displacement changes caused by thermal deformation.

[0007] Furthermore, the specific process of adaptively fusing multi-source heterogeneous data and outputting a multi-dimensional time-series feature vector characterizing the equipment's operating status is as follows: wavelet packet transform is performed on the vibration signal to extract non-stationary features; peak counting is performed on the acoustic emission signal to capture high-frequency pulse features; and integral calculation is performed on the axial micro-displacement data of the main shaft to quantify the thermal deformation trend. The dynamic weight matrix is ​​then weighted and fused with the non-stationary features of the vibration signal, the high-frequency pulse features of the acoustic emission, and the thermal deformation trend in the axial micro-displacement of the main shaft to generate a multi-dimensional time-series feature vector containing time-domain statistics, frequency-domain energy distribution, and trend accumulation.

[0008] Furthermore, the following is the specific process of aligning the non-stationary features of the vibration signal, the high-frequency pulse features of the acoustic emission signal, and the thermal deformation trend in the axial micro-displacement of the main shaft using a time-frequency domain attention mechanism to extract the feature sub-bands highly correlated with roughness in the acoustic emission signal, combined with a nonlinear fusion strategy: A short-time Fourier transform is performed on the acoustic emission signal to generate a time-frequency graph, dividing it into multiple frequency bands. An attention weight calculation module is used to select sub-bands significantly correlated with changes in surface roughness detected online, focusing on frequency bands with dense burst pulses in the high-frequency band. Wavelet packet decomposition is used to extract the energy distribution of non-stationary frequency bands related to bearing wear in the vibration signal, and the energy entropy is calculated as an indicator of vibration feature degradation. Based on the time-series integral operation of the micro-displacement data, the cumulative effect of the main shaft's thermal deformation is quantified, and the thermal deformation trend is dynamically corrected in conjunction with the ambient temperature change rate. A dual-channel LSTM network is constructed, with the first channel inputting the non-stationary vibration features and the thermal deformation trend, and the second channel inputting the energy distribution of the acoustic emission feature sub-bands. A sliding window algorithm is used to align the high-frequency equipment status data and the low-frequency quality detection data, and cross-rate feature fusion is achieved through timestamp matching, outputting a time-series correlation feature vector between processing status and quality degradation.

[0009] Furthermore, generating a machining quality degradation index and establishing a nonlinear mapping relationship between tool wear and machining quality includes the following steps: inputting time-series correlated feature vectors into a nonlinear regression model, fusing acoustic emission high-frequency pulse density, vibration energy entropy, and thermal deformation accumulation to calculate the machining quality degradation index; adaptively adjusting feature weights according to the type of machining material, strengthening vibration and thermal deformation features during metal machining, and strengthening acoustic emission features during composite material machining; establishing a correspondence between different tool wear stages and acoustic emission high-frequency pulse features based on tool wear experimental data; fitting the nonlinear influence of tool wear degree on surface roughness using a gradient boosting tree, quantifying the wear amount-roughness transfer coefficient, and establishing a nonlinear mapping relationship between tool wear and machining quality.

[0010] Furthermore, the specific process of constructing the graph structure based on the physical connection relationship of the equipment components is as follows: the spindle, bearing, guide rail, and tool are abstracted as graph nodes, and the mechanical transmission path between nodes is defined as graph edges according to the mechanical assembly relationship; the graph edge weights are initialized using the spindle-bearing stiffness matrix to characterize the propagation intensity of force and thermal load between components.

[0011] Furthermore, based on the federated learning framework, the residual parameters of multiple factories are trained collaboratively. The graph edge weights are initialized using the spindle-bearing stiffness matrix, and the specific process for outputting equipment remaining life prediction and latent fault location results is as follows: Each factory trains a residual graph convolutional network locally, calculates the parameter difference between the local model and the global model as the residual, and only uploads the residual parameters to the cloud for weighted aggregation to generate a global federated model, which is then distributed to each edge node. During federated training, the adjustment range of graph edge weights is constrained according to the spindle-bearing stiffness matrix to avoid physical logic violations caused by purely data-driven approaches. Fault propagation paths are defined for insufficient bearing preload and guide rail lubrication failure, and abnormal nodes are located using a graph attention mechanism. Based on the vibration energy entropy and thermal deformation trend of the graph nodes, the health of equipment components is assessed, the remaining life is inferred using a time series prediction model, and the fault location results are output.

[0012] Furthermore, the specific process of receiving the machining quality degradation index and fault location results, and dynamically adjusting the cutting speed and feed rate parameters through feedback closed loop is as follows: When the machining quality degradation index exceeds the primary threshold, the cutting speed is linearly reduced according to the index increase ratio, and the feed rate is reduced to suppress vibration; if the latent fault location results show that the bearing preload is insufficient, the spindle speed is dynamically adjusted according to the spindle load current to avoid the resonance frequency range.

[0013] Furthermore, the specific process for generating preventive maintenance work orders is as follows: the warning interval is divided according to the quality degradation index, and the index rise rate threshold and absolute value threshold are defined; when the rate of change of the quality degradation index exceeds the rise rate threshold or the quality degradation index exceeds the absolute value threshold, a stop command is triggered and a tool replacement work order is pushed.

[0014] The present invention has the following beneficial effects:

[0015] (1) The CNC machining equipment status monitoring system based on big data analysis, through real-time data acquisition by the multimodal perception fusion module and precise analysis by the implicit machining quality correlation analysis module, can comprehensively monitor the operating status of CNC equipment. The adaptive fusion processing of multi-source heterogeneous data makes the representation of equipment status more accurate, and can timely capture minor fault signs of equipment, thereby effectively improving the fault early warning capability of equipment. By extracting features highly correlated with machining quality and combining them with the time-frequency domain attention mechanism, the system can accurately assess the degradation of machining quality, providing real-time quality control and decision support for the production process.

[0016] (2) A CNC machining equipment status monitoring system based on big data analysis utilizes a small-sample federated residual graph network module. This module, combined with the physical connections of the equipment, achieves cross-factory collaboration through a federated learning framework, accurately predicting the remaining lifespan of the equipment and locating potential hidden faults. The advantage of this module lies in its ability to perform efficient model training and optimization without exposing local equipment data, effectively protecting industrial data privacy. Furthermore, the machining parameter execution module, based on a feedback closed-loop control mechanism, can dynamically adjust machining parameters (such as cutting speed and feed rate) according to real-time monitoring results, thereby optimizing machining accuracy and extending equipment lifespan. By generating preventative maintenance work orders, potential risks can be eliminated in advance, reducing unplanned equipment downtime and improving production efficiency.

[0017] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0018] Figure 1 This is a flowchart of the CNC machining equipment status monitoring system based on big data analysis according to the present invention.

[0019] Figure 2 This is a flowchart of the processing quality degradation index of the present invention. Detailed Implementation

[0020] This application's embodiments address the issues of equipment fault prediction, tool wear monitoring, and machining quality control during CNC machining through a CNC machining equipment status monitoring system based on big data analytics. The system collects real-time data from various sensors on the machining equipment, including vibration signals, acoustic emission signals, spindle speed, temperature, and load current, and utilizes big data analytics to process and analyze this data, thereby achieving precise monitoring of equipment status and fault early warning. Therefore, this application's embodiments provide a scientific and effective solution for equipment monitoring, fault prediction, and machining quality control during CNC machining, significantly improving production efficiency and equipment reliability.

[0021] The overall concept of the solution in this application embodiment is as follows:

[0022] The system collects vibration, temperature, acoustic emission signals, and spindle axial micro-displacement data of CNC equipment in real time. Combined with the working condition labels obtained by G-code parsing, a dynamic weight matrix is ​​constructed based on the identification results of the machining stage. The system performs adaptive fusion processing on multi-source heterogeneous data and outputs a multi-dimensional time-series feature vector representing the operating status of the equipment.

[0023] The system receives multidimensional time-series feature vectors and online detection of workpiece surface roughness and dimensional data. It extracts feature subbands highly correlated with roughness from acoustic emission signals through a time-frequency domain attention mechanism. Combining a nonlinear fusion strategy, it aligns the non-stationary features of vibration signals, high-frequency pulse features of acoustic emission, and thermal deformation trend in spindle axial micro-displacement with dual-channel time-series features. It then constructs a time-series mapping model between machining state and quality degradation, generates a machining quality degradation index, and establishes a nonlinear mapping relationship between tool wear and machining quality.

[0024] A graph structure is constructed based on the physical connection relationships of equipment components. The residual parameters of multiple factories are trained in collaboration with the federated learning framework. The graph edge weights are initialized by combining the spindle-bearing stiffness matrix, and the results of equipment remaining life prediction and latent fault location are output.

[0025] It receives the machining quality degradation index and fault location results, dynamically adjusts the cutting speed and feed rate parameters through feedback closed loop, and generates preventive maintenance work orders.

[0026] Please see Figure 1 This invention provides a technical solution: a CNC machining equipment status monitoring system based on big data analysis, comprising the following modules: a multimodal perception fusion module, an implicit machining quality correlation analysis module, a small-sample federated residual graph network module, and a machining parameter execution module. The multimodal perception fusion module is used to collect vibration, temperature, acoustic emission signals, and spindle axial micro-displacement data of the CNC equipment in real time. Combined with the working condition labels obtained from G-code parsing, a dynamic weight matrix is ​​constructed based on the machining stage identification results. Adaptive fusion processing is performed on the multi-source heterogeneous data, and a multi-dimensional temporal feature vector representing the equipment's operating status is output. The implicit machining quality correlation analysis module receives the multi-dimensional temporal feature vector and online detected workpiece surface roughness and dimensional data. It extracts data highly correlated with roughness from the acoustic emission signal using a time-frequency domain attention mechanism. The system utilizes a feature subband and a nonlinear fusion strategy to align the non-stationary characteristics of vibration signals, the high-frequency pulse characteristics of acoustic emission, and the thermal deformation trend in the spindle's axial micro-displacement with dual-channel time-series features. This constructs a time-series mapping model between machining status and quality degradation, generating a machining quality degradation index and establishing a nonlinear mapping relationship between tool wear and machining quality. The small-sample federated residual graph network module constructs a graph structure based on the physical connection relationships of equipment components. It trains residual parameters from multiple factories using a federated learning framework, initializes graph edge weights using the spindle-bearing stiffness matrix, and outputs predictions of remaining equipment life and results for locating latent faults. The machining parameter execution module receives the machining quality degradation index and fault location results, dynamically adjusts cutting speed and feed parameters through a feedback closed loop, and generates preventative maintenance work orders.

[0027] In this implementation scheme, the multimodal perception fusion module performs the following functions: This module acquires multiple signals from the CNC machining equipment in real time, including vibration, temperature, acoustic emission (AE) signals, and spindle axial micro-displacement data. Vibration and temperature reflect the mechanical state and temperature changes of the equipment, acoustic emission signals provide clues about material or tool wear, and spindle axial micro-displacement data helps monitor the accuracy and stability of the equipment. G-code parsing: G-codes are the instruction codes of CNC machine tools, providing information about machining conditions (e.g., cutting speed, feed rate, etc.). Combining this information, the system can identify different machining stages and construct a dynamic weight matrix based on the stage identification results to adjust the data fusion strategy. Adaptive fusion: The key to this module is the adaptive fusion of heterogeneous data (such as vibration, temperature, etc.) from different sources and types to output a comprehensive, multi-dimensional time-series feature vector. These feature vectors accurately reflect the operating status of the equipment and provide basic data for subsequent analysis. Implicit machining quality correlation analysis module function: This module is used to analyze the machining quality of the CNC equipment. Input data includes a multi-dimensional time-series feature vector (output by the first module) and online detected workpiece surface roughness and dimensional data. Time-frequency domain attention mechanism: This technique allows the system to dynamically focus on important feature subbands within the time-frequency domain of the signal. For example, this mechanism can identify the acoustic emission signal features most relevant to changes in workpiece surface roughness. Nonlinear fusion strategy: This strategy combines the complex relationships between different signals (such as the non-stationary characteristics of vibration signals, the high-frequency pulse characteristics of acoustic emission, and the thermal deformation trend of spindle micro-displacement) to ensure effective alignment and fusion of features from multiple signals. These fused features are used to construct a time-series mapping model reflecting the relationship between machining status and quality degradation, ultimately generating a machining quality degradation index. Tool wear and quality mapping relationship: This analysis identifies the impact of tool wear on machining quality and establishes a nonlinear mapping relationship between the two, which is crucial for predicting and managing machining quality degradation in advance. Small-sample federated residual graph network module function: The core of this module lies in constructing a graph structure using the physical connections of equipment components. Equipment components (such as spindles and bearings) are connected by graph edges, forming a complex network that reflects the interactions between components. Federated Learning Framework: Federated learning is a distributed machine learning technique that allows multiple factories or devices to share models without exchanging data and jointly train a global model. This framework is particularly suitable for data privacy protection, ensuring that local data within equipment and factories is not leaked. Residual Parameter Training: By training residual parameters, the system can capture abnormal behavior or hidden faults in the equipment. These parameters reflect the "residual" part of the equipment during operation, i.e., the deviation from the expected normal state.Spindle-Bearing Stiffness Matrix Initialization: The spindle and bearings are critical moving parts in CNC machine tools, and their stiffness relationship affects the accuracy and lifespan of the equipment. Initializing the stiffness matrix helps assign appropriate weights to the edges of the graph structure, thereby improving the accuracy of fault location. Output: This module outputs the remaining life prediction and latent fault location results of the equipment, which helps predict the future performance of the equipment and locate potential faults, reducing unexpected downtime. Machining Parameter Execution Module Function: This module receives the machining quality degradation index and fault location results from the latent machining quality correlation analysis module and feeds them back to the machining process in real time. The system dynamically adjusts machining parameters, such as cutting speed and feed rate, based on this feedback to optimize the machining process. Closed-Loop Control: Through a feedback closed-loop mechanism, the system can adjust parameters in real time during machining, ensuring machining quality while reducing quality fluctuations caused by equipment failure or improper operation. Preventive Maintenance: Through fault prediction and quality degradation analysis, the system can generate preventive maintenance work orders to perform equipment inspections and maintenance in advance, thereby reducing unplanned downtime and extending equipment lifespan. Multimodal sensing fusion: Combines data from different types of sensors (such as vibration, temperature, acoustic emission, etc.) to provide more comprehensive and accurate equipment status monitoring. Time-frequency domain attention mechanism: During signal processing, the system can focus on the most important frequencies or time-domain features of the signal, improving the accuracy of the analysis. Nonlinear fusion strategy: When fusing different signal features, it does not use a simple linear weighting method, but considers the complex nonlinear relationships between signals. Federated learning framework: A distributed machine learning method that supports multi-party collaboration without exchanging local data, protecting data privacy. Residual graph network: Captures abnormal features between equipment components through a graph-structured network model, helping to identify potential faults.

[0028] Specifically, the process of collecting vibration, temperature, acoustic emission signals, and spindle axial micro-displacement data from CNC equipment, combined with the working condition labels obtained from G-code parsing, and constructing a dynamic weight matrix based on the machining stage identification results is as follows: Parse the CNC equipment's G-code to extract process parameters such as cutting speed and feed rate, and identify the current machining stage as roughing, semi-finishing, or finishing; adjust the weights of vibration signals, acoustic emission signals, and spindle axial micro-displacement data according to the different machining stages; in the finishing stage, increase the weight of vibration signals to enhance sensitivity to minute vibrations; at the end of the tool change cycle, increase the weight of acoustic emission signals to improve the ability to capture high-frequency pulses; in continuous machining mode, gradually increase the weight of spindle axial micro-displacement data to capture subtle displacement changes caused by thermal deformation.

[0029] In this implementation plan, G-code parsing: G-code (or CNC code) is a set of instructions that controls the machining process of a CNC machine tool. Each instruction corresponds to a specific operation in the machining process, such as cutting speed, feed rate, tool path, etc. In this step, the G-code is first parsed to extract process parameters such as cutting speed, feed rate, and tool diameter. These parameters determine the specific requirements of the machining process and provide basic data for the subsequent construction of the dynamic weight matrix. Machining stage identification: Based on the parameters extracted from the G-code (such as cutting speed, feed rate, etc.), the system determines whether the current CNC machining stage is roughing, semi-finishing, or finishing. Each machining stage has different requirements and characteristics: Roughing stage: Usually large cutting amount, high cutting speed, mainly for material removal. Semi-finishing stage: Moderate cutting amount, mainly for contour trimming, with high precision requirements. Finishing stage: Small cutting amount, low cutting speed, mainly for fine machining, requiring high precision and low surface roughness. Different machining stages have different signal characteristics, therefore, data weights need to be adjusted according to the characteristics of different stages during signal fusion. Weight Adjustment: Based on different machining stages, the system needs to adjust the weights of each signal source to ensure sensitive capture of minute changes in the machining state. Finishing Stage: Due to the high machining accuracy required in finishing, the system needs to increase the weight of vibration signals at this stage. This is because vibration signals can help monitor minute mechanical vibrations, which are highly sensitive to during finishing; enhancing the vibration signal helps the system detect fine machining errors. End of Tool Change Cycle: At the end of the tool change cycle, tool wear can lead to workpiece surface quality issues. Therefore, increasing the weight of acoustic emission signals is important. Acoustic emission signals reflect high-frequency pulses generated during tool-workpiece contact; increasing their weight helps to better capture anomalies caused by tool wear or breakage. Continuous Machining Mode: In continuous machining mode, the spindle temperature and mechanical vibration can cause thermal deformation due to the ongoing machining process. To monitor displacement changes caused by thermal deformation, it is necessary to gradually increase the weight of the spindle axial micro-displacement data. These minute axial displacement changes can directly affect machining accuracy; therefore, weighting this signal in this mode helps to capture the effects of thermal deformation. Dynamic Weighting Matrix: This matrix dynamically adjusts the weights of each signal based on different processing stages and real-time acquired signal data. Using the operating condition labels and processing stage information obtained through G-code parsing, the system can generate a dynamic weighting matrix that reflects the importance of each signal (vibration, acoustic emission, micro-displacement) in the current processing stage in real time. The construction of this weighting matrix enables the system to intelligently adjust its data fusion strategy according to the processing stage and signal characteristics, thereby improving the accuracy and sensitivity of equipment status monitoring.

[0030] Specifically, the process of adaptively fusing multi-source heterogeneous data and outputting a multi-dimensional time-series feature vector characterizing the equipment's operating status is as follows: wavelet packet transform is performed on the vibration signal to extract non-stationary features; peak counting is performed on the acoustic emission signal to capture high-frequency pulse features; and integral calculation is performed on the axial micro-displacement data of the main shaft to quantify the thermal deformation trend. The dynamic weight matrix is ​​then weighted and fused with the non-stationary features of the vibration signal, the high-frequency pulse features of the acoustic emission, and the thermal deformation trend in the axial micro-displacement of the main shaft to generate a multi-dimensional time-series feature vector containing time-domain statistics, frequency-domain energy distribution, and trend accumulation.

[0031] In this implementation scheme, wavelet packet transform is used to analyze non-stationary signals, effectively extracting local time-frequency features from vibration signals. Signal decomposition: The vibration signal is decomposed into multiple frequency bands using wavelet packet transform, each representing the signal characteristics of a different frequency band. Extraction of non-stationary features: Time-domain analysis is performed on each decomposed frequency band to calculate features such as root mean square (RMS), peak value, and mean value, used to capture the non-stationarity of the vibration signal. Formula representation: Where: x(t) is the vibration signal, and T is the total signal time. RMS is the root mean square value of the signal, used to describe the signal energy. Acoustic emission signals typically reflect high-frequency vibrations and pulse activity during machining, especially when tools are worn or malfunctioning, manifesting as high-frequency pulses. Peak counting: By analyzing the high-frequency portion of the acoustic emission signal, the number of peaks exceeding a certain threshold is counted. These peaks correspond to key events in the machining process, such as tool contact and wear. High-frequency pulse characteristics: By peak counting, rapidly changing characteristics in the signal are captured, thereby understanding anomalies and problems in the machining process. Formula expression: Where: x(t) is the acoustic emission signal, and N is the number of signal sampling points. θ is a set threshold, defined as the minimum value of the peak value; exceeding this value is considered a valid pulse. 1 is an indicator function; when |x i |> θ is 1, otherwise it is 0. The spindle axial micro-displacement data reflects the axial displacement caused by thermal deformation during the cutting process. This data usually needs to be integrated to quantify the thermal deformation trend. Integration: Integrating the spindle axial micro-displacement yields the cumulative effect of thermal deformation. The integration operation reflects the trend of minute displacement changes caused by temperature rise and thermal expansion. Formula expression: Where: v axial (t) is the time series data of the micro-displacement of the main shaft axial direction. D thermalThis is the thermal deformation trend, representing the cumulative change in the axial micro-displacement of the main shaft. The non-stationary characteristics of the vibration signal, the high-frequency pulse characteristics of the acoustic emission signal, and the thermal deformation trend of the axial micro-displacement of the main shaft are weighted and fused. Weighted fusion uses a dynamic weight matrix to adjust the contribution of different features to the final fusion result. Dynamic weight matrix: generated from the G-code parsing and processing stage identification results mentioned earlier, dynamically adjusting the weights of each signal feature to adapt to different processing stages. Weighted fusion: the features of each signal are weighted according to their weights, ultimately resulting in a multi-dimensional time-series feature vector containing time-domain statistics, frequency-domain energy distribution, and trend accumulation. Multi-dimensional time-series feature vector: through the weighted fused features, the system ultimately generates a multi-dimensional time-series feature vector containing multiple features (such as time-domain statistics, frequency-domain energy distribution, and trend accumulation), characterizing the operating status of the equipment. Time-domain statistics: including mean, variance, peak value, root mean square, etc., reflecting the basic characteristics of the signal. Frequency-domain energy distribution: reflecting the energy distribution of the signal in the frequency domain, used to identify signal features of different frequency components. Cumulative trend: such as thermal deformation trend, it represents the cumulative effect of the signal over time. This feature vector can comprehensively reflect the operating status of CNC equipment under different working conditions, providing a basis for equipment fault prediction and maintenance decisions.

[0032] Specifically, the process of extracting feature sub-bands highly correlated with roughness from the acoustic emission signal using a time-frequency domain attention mechanism, and aligning the non-stationary features of the vibration signal, the high-frequency pulse features of the acoustic emission, and the thermal deformation trend in the axial micro-displacement of the spindle using a nonlinear fusion strategy is as follows: A short-time Fourier transform is performed on the acoustic emission signal to generate a time-frequency graph, which is then divided into multiple frequency bands. An attention weight calculation module is used to select sub-bands significantly correlated with changes in surface roughness detected online, focusing on frequency bands with dense burst pulses in the high-frequency band. Wavelet packet decomposition is used to extract the energy distribution of non-stationary frequency bands related to bearing wear in the vibration signal, and energy entropy is calculated as an indicator of vibration feature degradation. Based on the time-series integral operation of the micro-displacement data, the cumulative effect of spindle thermal deformation is quantified, and the thermal deformation trend is dynamically corrected in conjunction with the ambient temperature change rate. A dual-channel LSTM network is constructed, with the first channel inputting the non-stationary vibration features and the thermal deformation trend, and the second channel inputting the energy distribution of the acoustic emission feature sub-bands. A sliding window algorithm is used to align high-frequency equipment status data with low-frequency quality detection data, and cross-rate feature fusion is achieved through timestamp matching, outputting a time-series correlation feature vector between processing status and quality degradation.

[0033] In this implementation scheme, the Short-Time Fourier Transform (STFT) is first performed on the acoustic emission signal to generate a time-frequency diagram. The STFT combines information from the time and frequency domains of the signal, allowing us to observe the energy distribution of the signal across different frequency bands at different time points. The formula is as follows: Where: x(τ) is the original acoustic emission signal. h(t-τ) is a window function that controls the locality in the time domain. X(t,f) is the time-frequency plot, representing the signal distribution in the time-frequency domain. Subband division and attention weight calculation: The time-frequency plot is divided into multiple frequency bands. Through the attention weight calculation module, frequency bands that are significantly related to changes in surface roughness are selected. Particular focus is placed on frequency bands with burst pulses in the high-frequency range, as these bands are often closely related to changes in processing quality (such as roughness changes). Through the attention mechanism, the attention weight 'a' can be calculated. i To focus on the relevant frequency bands: Where: f i It is the characteristic value of the i-th frequency band. i The attention weight of frequency band i determines its importance in the model. Non-stationary feature extraction of vibration signals: Wavelet packet decomposition is performed on the vibration signal to extract the energy distribution of non-stationary frequency bands related to bearing wear. Wavelet packet transform decomposes the signal into different frequency bands and analyzes the energy distribution of each band by selecting an appropriate mother wavelet and frequency bandwidth. The formula for wavelet packet transform is: Where: ψ(t) is the mother wavelet, α is the scaling parameter, and β is the translation parameter. W ψ (α,β) are wavelet packet coefficients, representing the signal distribution at different scales and locations. Energy entropy: Calculated by analyzing the energy distribution across different frequency bands, energy entropy serves as a degradation index of vibration characteristics. Energy entropy reflects the complexity and uncertainty of energy distribution in a vibration signal and effectively represents the signal's degradation trend. The formula for calculating energy entropy is: Where: p i is the probability distribution of the i-th energy frequency band, representing the proportion of that frequency band in the overall signal energy. N is the total number of frequency bands. Time-series integration: Integrating the spindle axial micro-displacement data quantifies the cumulative effect of thermal deformation. The thermal deformation trend represents the cumulative effect of micro-displacement caused by temperature changes during processing. Through integration, the cumulative change in spindle axial micro-displacement can be obtained. The formula is as follows: Where: v axial (t) is the instantaneous value of the micro-displacement of the main shaft axial direction. D thermal (t) represents the thermal deformation trend, indicating the cumulative effect of spindle thermal deformation. Ambient temperature correction: The thermal deformation trend is dynamically corrected based on the rate of change of ambient temperature. Changes in ambient temperature affect the thermal expansion and contraction of the spindle, therefore correction is needed based on actual ambient temperature variations. The correction formula is: D thermal (t)=D thermal(t)+λ·ΔT(t); where: ΔT(t) is the change in ambient temperature. λ is the coefficient of temperature's influence on thermal deformation. LSTM Network: Construct a dual-channel Long Short-Time Memory (LSTM) network to process time-series data. The first channel inputs the non-stationary characteristics of the vibration signal and the trend of principal axis thermal deformation, and the second channel inputs the energy distribution of the acoustic emission characteristic subband. The basic formula of the LSTM network is as follows: h t =f(W h h t-1 +W x x t +b); where: h t It is the hidden state at the current moment. W h and W x This is the weight matrix, used to connect the hidden state from the previous time step and the current input. b is the bias term, and x... t This is the input at the current moment. Dual-channel design: The first channel processes vibration and thermal deformation data, and the second channel processes acoustic emission features. A sliding window algorithm aligns high-frequency equipment status data with low-frequency quality inspection data. Feature vector output: A time-series correlation feature vector between processing status and quality degradation is output through a dual-channel LSTM network. This feature vector integrates signal information from different sources to reflect the equipment's operating status and processing quality. The final output time-series feature vector is a high-dimensional vector containing correlation information of vibration, acoustic emission, and spindle micro-displacement signals at different time points, effectively reflecting the equipment's health status and quality degradation during processing.

[0034] Please see Figure 2 Specifically, generating a machining quality degradation index and establishing a nonlinear mapping relationship between tool wear and machining quality includes the following steps: inputting time-series correlated feature vectors into a nonlinear regression model, fusing acoustic emission high-frequency pulse density, vibration energy entropy, and thermal deformation accumulation to calculate the machining quality degradation index; adaptively adjusting feature weights according to the type of machining material, strengthening vibration and thermal deformation features during metal machining, and strengthening acoustic emission features during composite material machining; establishing a correspondence between different tool wear stages and acoustic emission high-frequency pulse features based on tool wear experimental data; fitting the nonlinear influence of tool wear degree on surface roughness using a gradient boosting tree, quantifying the wear amount-roughness transfer coefficient, and establishing a nonlinear mapping relationship between tool wear and machining quality.

[0035] In this implementation scheme, the feature vector includes: Acoustic emission high-frequency pulse density: reflects the instantaneous pulse information during the material cutting process, representing the high-frequency noise characteristics generated during the contact between the tool and the workpiece, and is usually closely related to tool wear. Vibration energy entropy: can characterize the complexity of the vibration signal, and the trend of vibration energy change is closely related to tool wear and the stability of the machining process. Cumulative thermal deformation: quantifies the thermal deformation trend of the spindle or tool through the integral calculation of micro-displacement data. Tool wear is usually accompanied by thermal deformation, which in turn affects the machining quality. Nonlinear regression model input: inputs the time-series correlated feature vector generated in the previous step into the nonlinear regression model. The goal of this regression model is to calculate the dynamic degradation index of machining quality based on multiple features (such as acoustic emission high-frequency pulse density, vibration energy amplitude, and cumulative thermal deformation). Nonlinear regression can handle complex nonlinear relationships between features, thereby more accurately describing the degradation process of machining quality. Formula expression: D=σ(W2·ReLU(W1·[E AE E Vib E Thermal ]+b1)+b2); [E AE E Vib E Thermal ] is the input feature vector, representing the high-frequency pulse density E in the acoustic emission signal. AE Energy entropy E of vibration signal Vib and cumulative thermal deformation E Thermal W1 and W2 are the model's weight matrices, controlling the nonlinear relationship between features and the degradation index. b1 and b2 are bias terms. ReLU is the activation function, used to increase nonlinear features. σ is the output activation function (sigmoid function), ultimately outputting the processing quality degradation index D. Feature weights are adjusted according to the processing material type: To enable the regression model to adapt to different types of processing materials, in metal processing and composite material processing, the influence of different signals is amplified by adaptively adjusting the feature weights. For example, in metal processing, vibration and thermal deformation features contribute more to quality degradation, while in composite material processing, the high-frequency pulse characteristics of acoustic emission signals may be more representative. The adaptive adjustment formula is: w feature (M)=α M ·w vibration +β M ·w AE +γ M ·w thermal ; where:w feature (M) represents the feature weights for the processed material M. α M ,β M ,γ M These are adjustment coefficients related to the material type M, corresponding to vibration, acoustic emission, and thermal deformation characteristics, respectively. vibration ,w AE,w thermal These are the initial weights for vibration, acoustic emission, and thermal deformation characteristics. During metal processing, the weights for vibration and thermal deformation (α) are... M and γ M The weight of acoustic emission characteristics (β) is relatively large during composite material processing. M The correlation between tool wear stages and acoustic emission characteristics is relatively large. This involves establishing the correspondence between different tool wear stages (e.g., light wear, moderate wear, heavy wear) and the high-frequency pulse characteristics of acoustic emission signals (e.g., pulse density, frequency) using experimental data. Typically, as tool wear intensifies, the pulse density and frequency of the acoustic emission signal change, reflecting changes in the contact state between the tool and the working surface and the cutting force. Modeling this stage usually employs empirical formulas or interpolation methods based on experimental data, as shown in the following formula: Where: P AE (W) represents the high-frequency acoustic emission pulse density under the tool wear level W. φ a (W) Basis functions (such as interpolation functions) related to the wear stage W. ξ a The weights of each basis function represent the contribution of acoustic emission features. This formula establishes a nonlinear relationship between wear level and acoustic emission signal features. Gradient Boosting Tree Fitting: Gradient Boosting Tree (GBDT) is a powerful regression method that effectively captures the nonlinear influence of tool wear on surface roughness. By using tool wear data (e.g., wear amount) and surface roughness data as training sets, a gradient boosting tree model is used to fit the relationship between them. The formula is: R0 surface =GBT(W tool ); where: R surface It refers to surface roughness. W tool This refers to the degree of tool wear. GBT is a gradient boosting tree model, representing the nonlinear mapping of tool wear to surface roughness. Wear-roughness transfer coefficient: Based on the output of the gradient boosting tree, this coefficient quantifies the transfer between tool wear and surface roughness. This coefficient reflects the degree of influence of wear on surface roughness. The formula for the transfer coefficient is: Where: K wear It is the transfer coefficient between wear and surface roughness. dR surface It refers to a minute change in surface roughness. dW tool This represents a small change in tool wear. This coefficient is used to quantify the impact of wear on machining quality and can serve as an important parameter in subsequent quality prediction. Nonlinear mapping relationship establishment: Combining the wear roughness transfer coefficient fitted by the gradient lifting tree and the machining quality degradation index D, a nonlinear mapping relationship between tool wear and machining quality was established. This mapping relationship can reflect the comprehensive impact of tool wear on machining quality degradation. The nonlinear mapping formula is: Q quality =g(W tool,K wear ,D); where: Q quality It is the final processing quality indicator. g(·) is a non-linear mapping function, which is usually modeled using neural networks or support vector regression (SVR).

[0036] Specifically, the process of constructing a graph structure based on the physical connection relationship of equipment components is as follows: the spindle, bearing, guide rail, and tool are abstracted as graph nodes, and the mechanical transmission path between nodes is defined as graph edges according to the mechanical assembly relationship; the graph edge weights are initialized using the spindle-bearing stiffness matrix to characterize the propagation intensity of force and thermal load between components.

[0037] In this implementation scheme, the nodes are defined as follows: The core components of the equipment (such as spindles, bearings, guide rails, and cutting tools) are abstracted as nodes in the graph. Each node represents a component, and the physical connections between them are represented by edges. Edges are defined as follows: The mechanical transmission paths between nodes are defined as edges in the graph based on the mechanical assembly relationships. For example, the connection between the spindle and the bearing is represented by an edge, which reflects the mechanical transmission between them. This transmission path includes not only force transmission but may also include the influence of thermal loads. Graph edge weight initialization: The weight of each edge is used to describe the intensity of force and thermal load propagation between components. The spindle-bearing stiffness matrix is ​​a crucial basis for initializing edge weights; it reflects the stiffness characteristics between the bearing and the spindle, thus providing a quantitative basis for the intensity of the mechanical transmission process. The stiffness matrix typically contains information about stiffness coefficients, representing the displacement response between components under specific loads.

[0038] Specifically, the process of training residual parameters across multiple factories using a federated learning framework, initializing graph edge weights using the spindle-bearing stiffness matrix, and outputting equipment remaining life prediction and latent fault location results is as follows: Each factory trains a residual graph convolutional network locally, calculates the parameter difference between the local model and the global model as the residual, uploads only the residual parameters to the cloud for weighted aggregation, generates a global federated model, and distributes it to each edge node; during federated training, the adjustment range of graph edge weights is constrained according to the spindle-bearing stiffness matrix to avoid physical logic violations caused by purely data-driven approaches; fault propagation paths are defined for insufficient bearing preload and guide rail lubrication failure, and abnormal nodes are located using a graph attention mechanism; based on the vibration energy entropy and thermal deformation trend of graph nodes, the health of equipment components is assessed, the remaining life is inferred using a time series prediction model, and fault location results are output.

[0039] In this implementation scheme, the federated learning framework trains residual parameters locally: each factory deploys and trains a Graph Convolutional Network (GCN) model locally. Each factory uses its own data for residual training, calculating the parameter differences (i.e., residuals) between the local model and the global model. This residual training method effectively preserves the characteristics of the local model while ensuring the consistency of the global model. Uploading residual parameters: each factory only uploads the residual parameters generated during training to the cloud, instead of directly uploading the complete data. This protects the data privacy of each factory and avoids the bandwidth and storage burden caused by large-scale data transmission. Global model aggregation: the cloud performs weighted aggregation of all uploaded residual parameters to generate a global federated model. This global model contains fused information from the data of each factory and reflects the production status of all factories. A weighted average method is typically used during aggregation to ensure that the contribution of each factory is reasonably adjusted according to its data volume. Model deployment to edge nodes: the aggregated global model is deployed to each edge node (i.e., each factory) for local prediction and fault detection. Stiffness Matrix Constraint: During federated training, to avoid violations of physical logic caused by purely data-driven approaches (i.e., relying solely on data for training while ignoring the physical characteristics of the equipment), the spindle-bearing stiffness matrix is ​​used as a constraint to limit the adjustment range of graph edge weights. This ensures that the physical characteristics of the equipment are not exceeded when adjusting the model. Significance of Physical Constraints: The transmission of mechanical and thermal loads is not only affected by data characteristics but must also conform to actual physical laws. Therefore, by using the stiffness matrix for constraints, the model training can be made more consistent with physical reality, avoiding overfitting of the data and loss of predictive ability for the true state of the equipment. Fault Propagation Path Definition: For common faults in equipment components (such as insufficient bearing preload or guide rail lubrication failure), it is necessary to define the fault propagation path. By analyzing the mechanical connections and thermal conduction relationships between nodes, the propagation path of the fault from one node to other nodes is determined. Insufficient bearing preload may lead to increased vibration, thereby affecting the stability of the entire spindle system; guide rail lubrication failure may lead to increased friction, thereby affecting the machining accuracy of the tool. Graph Attention Mechanism: The graph attention mechanism (GAT) is used to identify and locate abnormal nodes in the graph. In a graph structure, certain nodes (such as faulty components) may have a significant impact on the overall system. Graph attention mechanisms highlight important nodes by calculating the weights and correlations between them, thus effectively locating the source of the fault. For example, if the vibration energy or thermal deformation trend of a node is abnormal, it may affect other nodes through the graph's propagation mechanism, indicating that the node may be the source of the fault. Equipment health assessment: Based on the vibration energy entropy and thermal deformation trend of the graph nodes, the health status of each component can be assessed. Vibration energy entropy reflects the complexity of the vibration signal, thus reflecting the health status of the equipment; thermal deformation trend reflects the cumulative damage caused by temperature changes.These two indicators allow for a comprehensive assessment of equipment health. Remaining life prediction: By combining time series forecasting models (e.g., LSTM) with the aforementioned health assessment indicators, the remaining lifespan of the equipment is predicted. These indicators, as input, enable dynamic prediction of potential failures or performance degradation within a future period through time-series models, providing a reference for equipment maintenance and replacement. Fault localization: Through graph attention mechanisms and the definition of fault propagation paths, combined with equipment health assessment and remaining life prediction, the model can output fault localization results, indicating which components may have latent faults or are about to fail. These results provide guidance for maintenance personnel, helping them intervene before equipment problems occur, avoiding downtime and production interruptions.

[0040] Specifically, the process of receiving the machining quality degradation index and fault location results, and dynamically adjusting the cutting speed and feed rate parameters through feedback closed loop is as follows: When the machining quality degradation index exceeds the primary threshold, the cutting speed is linearly reduced according to the index increase ratio, and the feed rate is reduced to suppress vibration; if the latent fault location results show that the bearing preload is insufficient, the spindle speed is dynamically adjusted according to the spindle load current to avoid the resonance frequency range.

[0041] This implementation plan describes the specific process of dynamically adjusting the cutting speed and feed rate parameters using a feedback closed-loop system.

[0042] Upon receiving the machining quality degradation index and fault location results, the system dynamically adjusts the cutting speed and feed rate based on this information to optimize the machining process and reduce fault risks. The specific steps are as follows: Quality degradation index exceeding the initial threshold: When the machining quality degradation index (such as surface roughness, vibration energy, etc.) reaches the set initial threshold, it indicates a significant downward trend in machining quality. At this point, the system linearly reduces the cutting speed according to the ratio of the index increase. Simultaneously, to reduce vibration, the system reduces the feed rate. This process helps slow tool wear, improve machining accuracy, and prevent further quality degradation. Hidden fault location results indicating insufficient bearing preload: When the fault location results indicate insufficient bearing preload, the system dynamically adjusts the spindle speed based on the spindle load current information. By adjusting the speed, it avoids entering the resonant frequency range, thereby preventing equipment vibration deterioration and protecting equipment safety.

[0043] Specifically, the process of generating preventive maintenance work orders is as follows: the warning interval is divided according to the quality degradation index, and the index rise rate threshold and absolute value threshold are defined; when the rate of change of the quality degradation index exceeds the rise rate threshold or the quality degradation index exceeds the absolute value threshold, a stop command is triggered and a tool replacement work order is pushed.

[0044] In this implementation plan, based on changes in the quality degradation index, the system can automatically generate preventative maintenance work orders to avoid further equipment damage and ensure production stability. The specific steps are as follows: **Quality Degradation Index Warning Range Division:** First, the system divides the quality degradation index into different warning ranges. Specifically, a rate of increase threshold and an absolute value threshold are defined to trigger warnings. **Rate of Increase Threshold:** When the rate of change of the quality degradation index (i.e., the time-varying rate of change of the degradation index) exceeds the set rate of increase threshold, the system will issue a warning signal. **Absolute Value Threshold:** When the quality degradation index itself exceeds the set absolute value threshold, the system will also trigger a warning. **Triggering a Shutdown Command and Generating a Tool Replacement Work Order:** Once the quality degradation index exceeds the warning threshold, the system will automatically trigger a shutdown command and simultaneously generate a tool replacement work order. This preventative maintenance helps ensure that equipment does not suffer irreversible damage due to excessive wear.

[0045] In summary, this application has at least the following effects:

[0046] This CNC machining equipment status monitoring system, based on big data analytics, extracts time-series correlated feature vectors and calculates the machining quality degradation index. It can monitor tool wear and machining quality changes in real time, providing timely warnings and adjusting machining parameters to ensure machining accuracy and stability. Based on the machining quality degradation index and fault location results, the system can dynamically adjust cutting speed and feed rate to optimize the machining process, reduce vibration, and improve machining stability, thereby enhancing machining efficiency and product quality. By generating a quality degradation index and comparing it with warning intervals, the system can promptly trigger preventative maintenance measures, such as stop orders and tool replacement work orders, avoiding equipment failures and production stoppages, and improving equipment utilization. By establishing a nonlinear mapping relationship between tool wear and machining quality, the system can accurately quantify the impact of tool wear on surface roughness, providing scientific wear assessment and control schemes, thus improving the controllability and accuracy of the production process. Employing a federated learning framework, the system combines data from multiple factories for collaborative training, improving the model's generalization ability and accuracy. This ensures that while each factory makes individual adjustments, it can share and optimize the global model, improving the overall machining quality and efficiency of the production line.

[0047] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0049] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0050] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0051] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0052] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A CNC machining equipment status monitoring system based on big data analysis, characterized in that, It includes the following modules: multimodal perception fusion module, implicit processing quality correlation analysis module, small sample federated residual graph network module, and processing parameter execution module; The multimodal perception fusion module is used to collect vibration, temperature, acoustic emission signals and spindle axial micro-displacement data of CNC equipment in real time. Combined with the working condition labels obtained by G-code parsing, a dynamic weight matrix is ​​constructed based on the identification results of the processing stage. The module performs adaptive fusion processing on multi-source heterogeneous data and outputs a multi-dimensional time-series feature vector representing the operating status of the equipment. The implicit machining quality correlation analysis module is used to receive multi-dimensional time-series feature vectors and online detection of workpiece surface roughness and size data. It extracts feature subbands highly correlated with roughness from the acoustic emission signal through a time-frequency domain attention mechanism. Combined with a nonlinear fusion strategy, it performs dual-channel time-series feature alignment on the non-stationary features of vibration signals, high-frequency pulse features of acoustic emission, and thermal deformation trend in spindle axial micro-displacement. It constructs a time-series mapping model between machining state and quality degradation, and then generates a machining quality degradation index and establishes a nonlinear mapping relationship between tool wear and machining quality. The small-sample federated residual graph network module is used to construct a graph structure based on the physical connection relationship of equipment components, train the residual parameters of multiple factories in coordination according to the federated learning framework, initialize the graph edge weights in combination with the spindle-bearing stiffness matrix, and output the equipment remaining life prediction and latent fault location results. The machining parameter execution module is used to receive the machining quality degradation index and fault location results, dynamically adjust the cutting speed and feed parameters through feedback closed loop, and generate preventive maintenance work orders. Generating a machining quality degradation index and establishing a nonlinear mapping relationship between tool wear and machining quality includes the following steps: The time-series correlated feature vectors are input into the nonlinear regression model, and the high-frequency pulse density of acoustic emission, vibration energy entropy and thermal deformation accumulation are fused to calculate the processing quality degradation index. The feature weights are adaptively adjusted according to the type of material being processed, with vibration and thermal deformation features being enhanced during metal processing and acoustic emission features being enhanced during composite material processing. Based on the experimental data of tool wear, a correspondence between different wear stages of the tool and the characteristics of high-frequency acoustic emission pulses was established; By fitting the nonlinear effect of tool wear on surface roughness using gradient boosting tree fitting, the wear amount-roughness transfer coefficient is quantified, and a nonlinear mapping relationship between tool wear and machining quality is established. The specific process of receiving the machining quality degradation index and fault location results, and dynamically adjusting the cutting speed and feed rate parameters through a feedback closed loop, is as follows: When the machining quality degradation index exceeds the primary threshold, the cutting speed is reduced linearly according to the exponential increase ratio, and the feed rate is reduced to suppress vibration. If the latent fault location results show that the bearing preload is insufficient, the speed should be dynamically adjusted according to the spindle load current to avoid the resonant frequency range. The specific process for generating preventative maintenance work orders is as follows: The warning intervals are divided according to the quality degradation index, and the threshold for the rate of increase of the index and the threshold for the absolute value are defined. When the rate of change of the quality degradation index exceeds the rate of increase threshold or the absolute value threshold, a stop command is triggered and a tool change order is sent.

2. The CNC machining equipment status monitoring system based on big data analysis according to claim 1, characterized in that: The specific process of collecting vibration, temperature, acoustic emission signals, and spindle axial micro-displacement data of CNC equipment, combining them with the working condition labels obtained through G-code parsing, and constructing a dynamic weight matrix based on the machining stage identification results is as follows: Analyze the G-code of CNC equipment, extract process parameters such as cutting speed and feed rate, and identify whether the current machining stage is roughing, semi-finishing or finishing. The weights of vibration signals, acoustic emission signals, and spindle axial micro-displacement data are adjusted according to different processing stages. During the finishing stage, the sensitivity to minute vibrations is enhanced by increasing the weight of the vibration signal. At the end of the tool change cycle, the ability to capture high-frequency pulses is improved by enhancing the weight of the acoustic emission signal. In continuous machining mode, the weight of the spindle axial micro-displacement data is gradually increased to capture subtle displacement changes caused by thermal deformation.

3. The CNC machining equipment status monitoring system based on big data analysis according to claim 2, characterized in that: The specific process of adaptively fusing multi-source heterogeneous data and outputting a multi-dimensional time-series feature vector representing the operating status of the equipment is as follows: Wavelet packet transform is used to extract non-stationary features from vibration signals, peak counting is used to capture high-frequency pulse features from acoustic emission signals, and integral calculation is used to quantify the thermal deformation trend from the axial micro-displacement data of the main shaft. The dynamic weight matrix is ​​weighted and fused with the non-stationary characteristics of the vibration signal, the high-frequency pulse characteristics of acoustic emission, and the thermal deformation trend in the micro-displacement of the main shaft to generate a multi-dimensional time-series feature vector containing time-domain statistics, frequency-domain energy distribution, and trend accumulation.

4. The CNC machining equipment status monitoring system based on big data analysis according to claim 1, characterized in that: The specific process of extracting feature subbands highly correlated with roughness from the acoustic emission signal using a time-frequency domain attention mechanism, and then aligning the non-stationary characteristics of the vibration signal, the high-frequency pulse characteristics of the acoustic emission, and the thermal deformation trend in the axial micro-displacement of the main shaft using a nonlinear fusion strategy is as follows: The acoustic emission signal is subjected to short-time Fourier transform to generate a time-frequency diagram, which is divided into multiple frequency bands. Through the attention weight calculation module, the sub-bands that are significantly related to the changes in surface roughness detected online are selected, and the frequency band with dense burst pulses in the high-frequency band is focused. Wavelet packet decomposition is used to extract the energy distribution of non-stationary frequency bands related to bearing wear in the vibration signal, and the energy entropy is calculated as an index of vibration characteristic degradation. Based on time-series integral calculation of micro-displacement data, the cumulative effect of spindle thermal deformation is quantified, and the thermal deformation trend is dynamically corrected by combining the ambient temperature change rate. A dual-channel LSTM network is constructed. The first channel is input to the vibration non-stationary characteristics and thermal deformation trend, and the second channel is input to the energy distribution of the acoustic emission characteristic subband. By aligning high-frequency equipment status data with low-frequency quality detection data using a sliding window algorithm, and achieving cross-rate feature fusion through timestamp matching, a time-series correlation feature vector of processing status and quality degradation is output.

5. The CNC machining equipment status monitoring system based on big data analysis according to claim 1, characterized in that: The specific process of constructing the graph structure based on the physical connection relationships of the device components is as follows: The spindle, bearings, guide rails, and cutting tools are abstracted as graph nodes, and the mechanical transmission paths between nodes are defined as graph edges based on the mechanical assembly relationships. The graph edge weights are initialized using the spindle-bearing stiffness matrix to characterize the propagation intensity of force and thermal load between components.

6. The CNC machining equipment status monitoring system based on big data analysis according to claim 5, characterized in that: The specific process of training residual parameters across multiple factories using a federated learning framework, initializing graph edge weights using the spindle-bearing stiffness matrix, and outputting equipment remaining life prediction and latent fault location results is as follows: Each factory trains a residual graph convolutional network locally, calculates the parameter difference between the local model and the global model as the residual, and only uploads the residual parameters to the cloud for weighted aggregation to generate a global federated model and distribute it to each edge node; During federated training, the adjustment range of edge weights in the constraint graph is determined based on the spindle-bearing stiffness matrix. Define the fault propagation path for insufficient bearing preload and guide rail lubrication failure, and locate abnormal nodes through graph attention mechanism; Based on the vibration energy entropy and thermal deformation trend of the graph nodes, the health of equipment components is assessed, the remaining life is inferred through a time series prediction model, and the fault location results are output.

Images