CN118567294A - Cooperative control method and system of numerical control machine tool - Google Patents

Abstract

The invention relates to the technical field of data processing, and discloses a cooperative control method and a cooperative control system of a numerical control machine tool, wherein the cooperative control method comprises the following steps: carrying out multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data; performing adaptive parameter optimization calculation on the machine tool working state prediction data to obtain a target machining parameter combination; and carrying out dynamic allocation processing on the target machining parameter combination to obtain a cooperative machining strategy and control a plurality of numerical control machines to carry out real-time machining, carrying out control analysis on real-time measurement data and the cooperative machining strategy to obtain a quality compensation instruction and carrying out fault mode identification to obtain a fault prediction result and carrying out control instruction analysis to generate a cooperative control instruction. According to the invention, the accuracy of the working state prediction of the machine tool is improved through multi-model fusion analysis; the self-adaptive parameter optimization calculation realizes the dynamic optimization of the processing parameters, improves the efficiency of multi-machine cooperation, and the dynamic allocation processing mechanism realizes the intelligent cooperation operation of a plurality of machine tools.

Description

Cooperative control method and system for CNC machine tools

Technical Field

The present invention relates to the technical field of data processing, and in particular to a collaborative control method and system for a numerically controlled machine tool.

Background Art

Existing CNC machine tool control methods mainly focus on single-machine control, which collects various machine tool parameters such as position, feed speed, spindle acceleration, etc. for real-time monitoring and adjustment. These methods usually use algorithms such as PID control, fuzzy control or adaptive control to achieve precise positioning and processing quality control of machine tools. At the same time, some advanced control systems have begun to introduce machine learning and artificial intelligence technologies, such as neural networks and genetic algorithms, to improve the intelligence level and processing efficiency of machine tools.

However, these traditional methods have obvious limitations when faced with complex scenarios where multiple CNC machine tools work together. First, single-machine control cannot fully utilize the resource complementarity and information sharing between multiple machine tools, resulting in low overall processing efficiency. Second, the lack of overall optimization and coordination mechanism for the machine tool group makes it difficult to cope with emergencies and dynamic task adjustments. In addition, the computational efficiency and prediction accuracy of existing methods need to be improved when dealing with massive multidimensional data and complex nonlinear relationships.

Summary of the invention

In view of this, an embodiment of the present invention provides a collaborative control method and system for a CNC machine tool, which are used to improve the efficiency and accuracy of the collaborative control of the CNC machine tool.

The present invention provides a collaborative control method for CNC machine tools, comprising: performing multi-dimensional parameter collection and preprocessing on a plurality of CNC machine tools to obtain a standardized parameter data set; performing multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data; performing adaptive parameter optimization calculation on the machine tool working state prediction data to obtain a target processing parameter combination; dynamically allocating and processing the state information of the plurality of CNC machine tools collected in real time and the target processing parameter combination to obtain a collaborative processing strategy; controlling the plurality of CNC machine tools to perform real-time processing through the collaborative processing strategy, collecting real-time measurement data during the real-time processing, and performing control analysis on the real-time measurement data and the collaborative processing strategy to obtain quality compensation instructions; performing fault mode recognition processing on the standardized parameter data set and the quality compensation instructions to obtain a fault prediction result, and performing control instruction analysis on the fault prediction result through a decision tree algorithm to generate a collaborative control instruction.

In the present invention, the step of performing multi-dimensional parameter acquisition and preprocessing on multiple CNC machine tools to obtain a standardized parameter data set includes: performing multi-channel parallel acquisition on the position, feed speed, spindle acceleration, temperature and vibration data of multiple CNC machine tools to obtain original multi-dimensional parameter data; performing wavelet transform filtering on the original multi-dimensional parameter data to obtain noise reduction parameter data; performing feature extraction on the noise reduction parameter data through a principal component analysis algorithm to obtain dimension reduction feature data; performing Z-score standardization on the dimension reduction feature data to obtain preliminary standardized data; discretizing the preliminary standardized data through an adaptive quantization algorithm to obtain discretized parameter data; performing time window segmentation on the discretized parameter data to obtain time series parameter segments; performing frequency domain conversion on the time series parameter segments through Fourier transform to obtain frequency domain feature data; performing peak detection and spectral line extraction on the frequency domain feature data to obtain optimized spectrum features; performing cluster analysis on the optimized spectrum features through a self-organizing map algorithm to obtain parameter clustering results; performing multi-dimensional cross-validation on the parameter clustering results to obtain the standardized parameter data set.

In the present invention, the step of performing multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data includes: performing time series decomposition on the standardized parameter data set to obtain trend items, seasonal items and random items; performing time series prediction on the trend items by autoregressive integral sliding average algorithm to obtain trend prediction results; performing periodic analysis on the seasonal items by Fourier series expansion to obtain seasonal characteristics; performing prediction analysis on the random items by long short-term memory network to obtain random item prediction data; and performing ensemble learning algorithm on the trend prediction results, the seasonal characteristics and the random item prediction. The data are merged to obtain a preliminary state prediction result; the standardized parameter data set is segmented into a sliding window to obtain a plurality of time window samples; the plurality of time window samples are calculated in parallel by a support vector regression algorithm to obtain a plurality of groups of local prediction data; the plurality of groups of local prediction data are dynamically weighted averaged to obtain a comprehensive prediction result; the preliminary state prediction result and the comprehensive prediction result are optimized by a particle swarm optimization algorithm to obtain optimized prediction data; the optimized prediction data are subjected to Monte Carlo simulation to generate a plurality of groups of prediction data, and the machine tool working state prediction data is obtained by confidence interval analysis.

In the present invention, the step of performing adaptive parameter optimization calculation on the machine tool working state prediction data to obtain a target processing parameter combination includes: performing multi-dimensional decomposition on the machine tool working state prediction data to obtain a working state feature vector; performing dimensionality reduction processing on the working state feature vector by a principal component analysis algorithm to obtain an optimized feature set; performing fuzzy clustering analysis on the optimized feature set to obtain a working state mode classification result; performing initial parameter generation on the working state mode classification result by a genetic algorithm to obtain an initial processing parameter set; performing orthogonal experimental design on the initial processing parameter set to obtain a parameter optimization test plan; performing multi-objective optimization calculation on the parameter optimization test plan by a response surface method to obtain an optimized parameter response surface; performing gradient descent search on the optimized parameter response surface to obtain a local optimal parameter combination; performing global optimization on the local optimal parameter combination by a simulated annealing algorithm to obtain a global optimized parameter set; performing sensitivity analysis on the global optimized parameter set to obtain a parameter sensitivity ranking; performing weighted combination on the parameter sensitivity ranking by an adaptive weight adjustment algorithm to obtain the target processing parameter combination.

In the present invention, the step of dynamically allocating and processing the state information of multiple CNC machine tools collected in real time and the target processing parameter combination to obtain a collaborative processing strategy includes: performing real-time data stream processing on the state information of the multiple CNC machine tools to obtain a machine tool state feature matrix; performing trend prediction on the machine tool state feature matrix through multivariate time series analysis to obtain a short-term state prediction result; performing task decomposition on the target processing parameter combination to obtain a subtask parameter set; performing initial matching on the short-term state prediction result and the subtask parameter set to obtain an initial task allocation plan; performing load balancing calculation on the initial task allocation plan to obtain a load balancing coefficient; optimizing and reorganizing the initial task allocation plan and the load balancing coefficient through a genetic algorithm to obtain an optimized task allocation plan; performing conflict detection on the optimized task allocation plan to obtain a task conflict matrix; performing conflict resolution on the task conflict matrix through a graph coloring algorithm to obtain a conflict-free task sequence; performing time window division on the conflict-free task sequence to obtain a dynamic scheduling schedule; and performing strategy optimization on the dynamic scheduling schedule through a reinforcement learning algorithm to obtain the collaborative processing strategy.

In the present invention, the collaborative processing strategy is used to control the multiple CNC machine tools to perform real-time processing, and real-time measurement data is collected during the real-time processing, and the real-time measurement data and the collaborative processing strategy are controlled and analyzed to obtain quality compensation instruction steps, including: parsing the collaborative processing strategy to obtain a control instruction sequence for each machine tool; executing the control instruction sequence for each machine tool through a real-time control interface to control multiple CNC machine tools to perform real-time processing to obtain real-time processing status data; performing multi-source data fusion on the real-time processing status data to obtain a comprehensive state feature vector; and performing Kalman filtering algorithm on the real-time processing status data. The comprehensive state feature vector is subjected to noise suppression to obtain an optimized state estimate; the optimized state estimate is subjected to statistical process control analysis to obtain a process capability index; the process capability index is evaluated through a fuzzy inference system to obtain a quality state grade; an association analysis is performed on the quality state grade and the collaborative processing strategy to obtain a strategy influencing factor; the strategy influencing factor is modeled through a support vector regression algorithm to obtain a quality prediction model; a multi-objective optimization calculation is performed on the quality prediction model to obtain a compensation parameter set; the compensation parameter set is dynamically adjusted through an adaptive control algorithm to obtain the quality compensation instruction.

In the present invention, the fault pattern recognition processing is performed on the standardized parameter data set and the quality compensation instruction to obtain a fault prediction result, and the control instruction analysis is performed on the fault prediction result through a decision tree algorithm to generate a collaborative control instruction step, including: time series alignment of the standardized parameter data set and the quality compensation instruction to obtain a synchronous data matrix; multi-scale decomposition of the synchronous data matrix through wavelet packet transform to obtain a characteristic coefficient set; principal component analysis of the characteristic coefficient set to obtain a reduced-dimensional feature vector; fault pattern classification of the reduced-dimensional feature vector through a support vector machine algorithm to obtain a preliminary fault category; fuzzy set operation is performed on the preliminary fault category to obtain a fault membership matrix; nonlinear mapping of the fault membership matrix is performed through an artificial neural network to obtain a fault severity assessment result; time series prediction is performed on the fault severity assessment result to obtain the fault prediction result; rule extraction is performed on the fault prediction result through the decision tree algorithm to obtain a fault handling rule set; priority sorting of the fault handling rule set is performed to obtain a hierarchical control strategy; and dynamic optimization of the hierarchical control strategy is performed through a model predictive control algorithm to obtain the collaborative control instruction.

The present invention also provides a collaborative control system for a numerically controlled machine tool, comprising:

The acquisition module is used to collect and preprocess multi-dimensional parameters of multiple CNC machine tools to obtain standardized parameter data sets;

An analysis module, used for performing multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data;

A calculation module, used for performing adaptive parameter optimization calculation on the machine tool working state prediction data to obtain a target processing parameter combination;

An allocation module, used for dynamically allocating and processing the state information of the plurality of CNC machine tools and the target processing parameter combination collected in real time to obtain a collaborative processing strategy;

A processing module, used to control the multiple CNC machine tools to perform real-time processing through the collaborative processing strategy, collect real-time measurement data during the real-time processing, and control and analyze the real-time measurement data and the collaborative processing strategy to obtain quality compensation instructions;

The identification module is used to perform fault mode recognition processing on the standardized parameter data set and the quality compensation instruction to obtain a fault prediction result, and perform control instruction analysis on the fault prediction result through a decision tree algorithm to generate a collaborative control instruction.

In the technical solution provided by the present invention, through multi-dimensional parameter collection and preprocessing technology, the operating status of multiple machine tools is fully and accurately captured, laying a solid data foundation for efficient collaborative control; the multi-model fusion analysis method effectively improves the accuracy of machine tool working status prediction, making collaborative decision-making more accurate; the adaptive parameter optimization calculation realizes the dynamic optimization of processing parameters, significantly improving the efficiency of multi-machine collaboration; the dynamic allocation processing mechanism realizes the intelligent collaborative operation of multiple machine tools, fully utilizes equipment resources, and greatly improves the overall production efficiency; the real-time quality control and compensation mechanism ensures the quality consistency during the collaborative processing process and reduces the errors caused by differences between machine tools; the fault mode recognition and prediction function improves the reliability of the collaborative system, reduces downtime, and ensures continuous and efficient operation; the application of the decision tree algorithm makes collaborative control The generation of instructions is more intelligent, which improves the response speed and adaptability; by integrating multi-source data and multiple advanced algorithms, a comprehensive machine tool collaborative control system is constructed, which not only improves the overall processing accuracy, but also reduces the need for manual intervention, greatly improving the level of automation; the system's self-learning and adaptive capabilities enable the collaborative control strategy to be continuously optimized to cope with complex and changing production environments; the combination of real-time and predictive features enables the system to foresee and prevent possible problems while ensuring the current collaborative processing quality, thereby improving the continuity and stability of collaboration; the implementation of the collaborative control strategy enables the entire processing system to show higher flexibility and adaptability, and can quickly respond to changes in production needs, thereby improving overall production efficiency; finally, this collaborative control method based on data-driven and intelligent algorithms significantly improves the efficiency and accuracy of multi-machine collaboration.

BRIEF DESCRIPTION OF THE DRAWINGS

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

FIG1 is a flow chart of a collaborative control method for a CNC machine tool according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a collaborative control system for a CNC machine tool according to an embodiment of the present invention.

DETAILED DESCRIPTION

The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. 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.

In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

For ease of understanding, the specific process of the embodiment of the present invention is described below. Please refer to FIG. 1, which is a flow chart of a collaborative control method for a CNC machine tool according to an embodiment of the present invention. As shown in FIG. 1, the method includes the following steps:

S101, performing multi-dimensional parameter collection and preprocessing on multiple CNC machine tools to obtain a standardized parameter data set;

S102, performing multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data;

S103, performing adaptive parameter optimization calculation on the machine tool working state prediction data to obtain a target processing parameter combination;

S104, dynamically allocate and process the state information and target processing parameter combinations of multiple CNC machine tools collected in real time to obtain a collaborative processing strategy;

S105, controlling multiple CNC machine tools to perform real-time processing through a collaborative processing strategy, collecting real-time measurement data during the real-time processing, and controlling and analyzing the real-time measurement data and the collaborative processing strategy to obtain quality compensation instructions;

S106, performing fault pattern recognition processing on the standardized parameter data set and the quality compensation instruction to obtain a fault prediction result, and performing control instruction analysis on the fault prediction result through a decision tree algorithm to generate a coordinated control instruction.

It should be noted that the position, feed speed, spindle acceleration, temperature, vibration and other data of the machine tool are collected through high-precision sensors, and wavelet transform is used for noise reduction. Then, principal component analysis (PCA) is used for feature extraction and dimension reduction. Finally, the parameters of different dimensions are unified to the same scale through Z-score standardization. Multi-model fusion analysis is performed on the standardized parameter data set to obtain the machine tool working status prediction data. This step comprehensively uses time series decomposition, autoregressive integrated moving average (ARIMA) model, long short-term memory (LSTM) network and other methods to predict the working status of the machine tool from multiple angles, and then the model is fused through integrated learning algorithms such as random forest or gradient boosting tree to improve the accuracy and robustness of the prediction.

Then, the machine tool working state prediction data is adaptively optimized to obtain the target processing parameter combination. This step uses genetic algorithms or particle swarm optimization algorithms to dynamically adjust the cutting speed, feed rate, cutting depth and other processing parameters according to the predicted working state to achieve the best balance between processing efficiency and quality. Based on the target processing parameter combination, the state information of multiple CNC machine tools collected in real time is dynamically allocated and processed to obtain a collaborative processing strategy. This step uses reinforcement learning algorithms, such as Q-learning or deep Q network (DQN), to dynamically allocate and schedule processing tasks according to the real-time status of each machine tool and the characteristics of the processing task to achieve multi-machine collaborative optimization. Multiple CNC machine tools are controlled by collaborative processing strategies for real-time processing, and real-time measurement data during real-time processing is collected. These data and collaborative processing strategies are controlled and analyzed to obtain quality compensation instructions. This step uses the Kalman filter algorithm to suppress noise in real-time data, and then analyzes the fluctuation of processing quality through the statistical process control (SPC) method. Finally, fuzzy control or adaptive control algorithms are used to generate compensation instructions to adjust processing parameters in real time to maintain processing quality.

Finally, the standardized parameter data set and quality compensation instructions are processed for fault pattern recognition to obtain fault prediction results, and the fault prediction results are analyzed for control instructions through the decision tree algorithm to generate collaborative control instructions. This step uses support vector machines (SVM) or convolutional neural networks (CNN) to classify fault patterns, and then uses the decision tree algorithm to extract fault handling rules, and finally generates collaborative control instructions to achieve intelligent collaborative control and fault prevention of multiple machine tools.

For example, in a processing unit containing 5 CNC machine tools, each machine tool is equipped with 10 sensors, collecting 100 data points per second. After wavelet transform filtering, the noise level was reduced by 80%. PCA analysis reduced the original 50-dimensional features (5 machine tools × 10 sensors) to 15 dimensions, retaining 95% of the information. Multi-model fusion analysis combines the prediction results of three models, ARIMA, LSTM and random forest, to predict the machine tool status in the next hour. Adaptive parameter optimization uses genetic algorithms to find the optimal combination of processing parameters, including spindle speed, feed rate and cutting depth after 100 generations of iterations. The collaborative processing strategy was trained for 10,000 rounds through a deep Q network to learn the optimal task allocation scheme under different machine tool load conditions. During real-time processing, Kalman filtering reduced measurement noise by 70%, and SPC analysis found that the standard deviation of processing accuracy was 0.01mm. Fault mode recognition uses the SVM algorithm, trained on 2,000 historical fault samples, to identify 5 common fault modes. Finally, the decision tree algorithm generated a decision tree with 50 nodes for quickly generating collaborative control instructions. Through this series of processing, efficient collaborative control of 5 CNC machine tools was achieved, significantly improving the overall processing efficiency and quality stability.

By executing the above steps, through multi-dimensional parameter collection and preprocessing technology, the operating status of multiple machine tools is fully and accurately captured, laying a solid data foundation for efficient collaborative control; the multi-model fusion analysis method effectively improves the accuracy of machine tool working status prediction, making collaborative decision-making more accurate; the adaptive parameter optimization calculation realizes the dynamic optimization of processing parameters, significantly improving the efficiency of multi-machine collaboration; the dynamic allocation processing mechanism realizes the intelligent collaborative operation of multiple machine tools, fully utilizes equipment resources, and greatly improves the overall production efficiency; the real-time quality control and compensation mechanism ensures the quality consistency in the collaborative processing process and reduces the errors caused by differences between machine tools; the fault mode recognition and prediction function improves the reliability of the collaborative system, reduces downtime, and ensures continuous and efficient operation; the application of the decision tree algorithm makes the collaborative control index The generation of commands is more intelligent, which improves the response speed and adaptability; by integrating multi-source data and multiple advanced algorithms, a comprehensive machine tool collaborative control system is constructed, which not only improves the overall processing accuracy, but also reduces the need for manual intervention, greatly improving the level of automation; the system's self-learning and adaptive capabilities enable the collaborative control strategy to be continuously optimized to cope with complex and changing production environments; the combination of real-time and predictive features enables the system to foresee and prevent possible problems while ensuring the current collaborative processing quality, thereby improving the continuity and stability of collaboration; the implementation of the collaborative control strategy enables the entire processing system to show higher flexibility and adaptability, and can quickly respond to changes in production needs, thereby improving overall production efficiency; finally, this collaborative control method based on data-driven and intelligent algorithms significantly improves the efficiency and accuracy of multi-machine collaboration.

In a specific embodiment, the process of executing step S101 may specifically include the following steps:

(1) Perform multi-channel parallel acquisition of the position, feed speed, spindle acceleration, temperature and vibration data of multiple CNC machine tools to obtain original multi-dimensional parameter data;

(2) Perform wavelet transform filtering on the original multi-dimensional parameter data to obtain noise reduction parameter data;

(3) Extract features from the noise reduction parameter data using the principal component analysis algorithm to obtain dimension-reduced feature data;

(4) Perform Z-score normalization on the reduced dimension feature data to obtain preliminary normalized data;

(5) Discretize the preliminary standardized data through an adaptive quantization algorithm to obtain discretized parameter data;

(6) Segment the discretized parameter data into time windows to obtain time series parameter segments;

(7) Convert the time series parameter fragments into frequency domain through Fourier transform to obtain frequency domain feature data;

(8) Perform peak detection and spectrum line extraction on the frequency domain feature data to obtain optimized spectrum features;

(9) Cluster analysis is performed on the optimized spectrum features through the self-organizing mapping algorithm to obtain parameter clustering results;

(10) Perform multi-dimensional cross-validation on the parameter clustering results to obtain a standardized parameter data set.

Specifically, the position, feed speed, spindle acceleration, temperature and vibration data of multiple CNC machine tools are first collected in parallel through multiple channels to obtain the original multi-dimensional parameter data. This step uses a high-precision sensor network to simultaneously collect multiple physical quantities to form a multi-dimensional data matrix. Then, the original multi-dimensional parameter data is subjected to wavelet transform filtering to obtain denoised parameter data. Wavelet transform can effectively remove high-frequency noise while retaining the important features of the signal by decomposing the signal into wavelet coefficients of different scales. The denoised parameter data is feature extracted by the principal component analysis algorithm to obtain reduced dimension feature data. Principal component analysis (PCA) calculates the covariance matrix of the data to find the main direction of data change, reduce the data dimension and retain most of the information. The reduced dimension feature data is subjected to Z-score standardization to obtain preliminary standardized data. Z-score standardization converts features of different scales to the same standard scale to eliminate the influence of dimension.

The preliminary standardized data is discretized by the adaptive quantization algorithm to obtain discretized parameter data. Adaptive quantization dynamically adjusts the quantization interval according to the data distribution to improve the accuracy of discretization. The discretized parameter data is segmented into time windows to obtain time series parameter fragments. Time window segmentation divides the continuous data stream into segments of fixed length for subsequent analysis. The time series parameter fragments are converted into frequency domain by Fourier transform to obtain frequency domain feature data. Fourier transform converts the time domain signal into frequency domain representation to reveal the frequency composition of the signal. Peak detection and spectral line extraction are performed on the frequency domain feature data to obtain optimized spectrum features. Peak detection identifies significant peaks in the spectrum, and spectral line extraction retains the most representative frequency components. The optimized spectrum features are clustered by the self-organizing map algorithm to obtain parameter clustering results. Self-organizing map (SOM) is an unsupervised learning algorithm that can map high-dimensional data to low-dimensional space while maintaining the topological structure of the data. Finally, the parameter clustering results are cross-validated in multiple dimensions to obtain a standardized parameter data set. Cross-validation evaluates the stability and reliability of clustering results through repeated training and testing.

For example, in a machining cell containing five CNC machine tools, each machine tool is equipped with five sensors for position, feed speed, spindle acceleration, temperature, and vibration, with a sampling frequency of 1kHz. The raw data forms a 5×5×1000 three-dimensional matrix (5 machine tools, 5 parameters, 1000 sampling points). The high-frequency noise is removed by using a 3-layer decomposition using db4 wavelet. PCA analysis reduces the 25-dimensional features (5 machine tools × 5 parameters) to 10 dimensions, retaining 95% of the information. Z-score normalization normalizes all features to a distribution with a mean of 0 and a standard deviation of 1. Adaptive quantization discretizes continuous data into 100 levels. Segmentation is performed using a 10-second time window (10,000 sampling points). A 1024-point FFT is performed on each time window to obtain frequency domain features. Peak detection identifies the top 5 main peaks in each spectrum. A 10×10 SOM network clusters the spectrum features into 100 prototype vectors. The stability of the clustering results was evaluated through 10-fold cross validation.

In a specific embodiment, the process of executing step S102 may specifically include the following steps:

(1) Perform time series decomposition on the standardized parameter data set to obtain trend terms, seasonal terms, and random terms;

(2) The trend item is predicted in time series by using the autoregressive integrated moving average algorithm to obtain the trend prediction result;

(3) Perform periodic analysis on seasonal terms through Fourier series expansion to obtain seasonal characteristics;

(4) Performing prediction analysis on random items through long short-term memory networks to obtain random item prediction data;

(5) The trend prediction results, seasonal characteristics and random item prediction data are integrated through the ensemble learning algorithm to obtain the preliminary state prediction results;

(6) Perform sliding window segmentation on the standardized parameter data set to obtain multiple time window samples;

(7) Using the support vector regression algorithm, multiple time window samples are calculated in parallel to obtain multiple sets of local prediction data;

(8) Perform dynamic weighted averaging on multiple sets of local prediction data to obtain a comprehensive prediction result;

(9) Optimize the parameters of the preliminary state prediction results and the comprehensive prediction results through the particle swarm optimization algorithm to obtain optimized prediction data;

(10) Monte Carlo simulation is performed on the optimized prediction data to generate multiple sets of prediction data, and the machine tool working status prediction data is obtained through confidence interval analysis.

It should be noted that the standardized parameter data set is first decomposed into time series to obtain trend items, seasonal items and random items. The time series decomposition adopts the X-12-ARIMA method, which can effectively separate long-term trends, periodic changes and random fluctuations. Then, the trend items are predicted by the autoregressive integrated moving average (ARIMA) algorithm to obtain the trend prediction results. The ARIMA model combines three components: autoregression (AR), difference (I) and moving average (MA), which can capture the autocorrelation, trend and stationary characteristics of the time series. The seasonal items are analyzed periodically by Fourier series expansion to obtain seasonal characteristics. The Fourier series decomposes the periodic signal into a series of linear combinations of sine and cosine functions, which helps to identify and quantify periodic patterns of different frequencies. At the same time, the random items are predicted and analyzed by the long short-term memory (LSTM) network to obtain the random item prediction data. LSTM is a special recurrent neural network that can effectively capture long-term dependencies and is suitable for processing nonlinear and non-stationary characteristics in time series.

Then, the trend forecast results, seasonal characteristics and random item forecast data are integrated through an ensemble learning algorithm to obtain the preliminary state forecast results. Ensemble learning, such as random forest or gradient boosting tree, improves the accuracy and robustness of the overall forecast by combining the forecast results of multiple base learners. At the same time, the standardized parameter data set is segmented by sliding windows to obtain multiple time window samples. The sliding window technology allows the model to capture local time dependencies and improve the flexibility of forecasting. The support vector regression (SVR) algorithm is used to parallelly calculate multiple time window samples to obtain multiple sets of local forecast data. SVR realizes nonlinear regression by mapping the input to a high-dimensional feature space and constructing an optimal hyperplane in it. Dynamic weighted averaging is performed on multiple sets of local forecast data to obtain a comprehensive forecast result. Dynamic weighted averaging dynamically adjusts the weight according to the performance of each local forecast, thereby improving the accuracy of the overall forecast. The particle swarm optimization (PSO) algorithm is used to optimize the parameters of the preliminary state forecast results and the comprehensive forecast results to obtain optimized forecast data. The PSO algorithm simulates swarm intelligence and finds the global optimal solution by moving particles in the solution space, effectively optimizing the parameters of the forecast model. Finally, the optimized prediction data is simulated by Monte Carlo to generate multiple sets of prediction data, and the prediction data of the machine tool working status is obtained through confidence interval analysis. Monte Carlo simulation evaluates the uncertainty of the prediction results through multiple random sampling, while confidence interval analysis provides the reliability range of the prediction results.

For example, in a machining unit containing five CNC machine tools, the spindle speed is selected as the key parameter for analysis. First, the X-12-ARIMA method is used to decompose the historical data of one year (one data point per hour, a total of 8760 points) to obtain trend items, seasonal items, and random items. The ARIMA (1,1,1) model is applied to the trend item to predict the trend for the next 7 days. The daily and weekly cycles in the seasonal item are identified through Fourier series expansion. The random item is modeled using an LSTM network (a single-layer LSTM with 64 hidden units) to predict random fluctuations in the next 24 hours. The random forest algorithm (containing 100 decision trees) is used to fuse the prediction results of these three parts to obtain a preliminary state prediction.

At the same time, the data was segmented using a 24-hour sliding window (with a step size of 1 hour) to obtain 8736 samples. SVR (using the RBF kernel function) was applied to each sample for local prediction, and 8736 sets of local prediction results were obtained. These local predictions were fused by exponential weighted averaging (with higher weights for recent data) to obtain a comprehensive prediction result. The PSO algorithm (100 particles, 500 iterations) was used to optimize the fusion weights and model parameters. Finally, 1000 Monte Carlo simulations were performed to generate 1000 sets of possible future states, and the final prediction interval of the machine tool working state was obtained by calculating the 95% confidence interval.

In a specific embodiment, the process of executing step S103 may specifically include the following steps:

(1) Perform multi-dimensional decomposition on the machine tool working state prediction data to obtain the working state feature vector;

(2) The working state feature vector is reduced in dimension through the principal component analysis algorithm to obtain the optimized feature set;

(3) Perform fuzzy clustering analysis on the optimized feature set to obtain the classification results of the working status mode;

(4) Generate initial parameters for the classification results of the working state mode through genetic algorithm to obtain the initial processing parameter set;

(5) Conduct orthogonal experimental design on the initial processing parameter set to obtain the parameter optimization test plan;

(6) Using the response surface method, multi-objective optimization calculations are performed on the parameter optimization test plan to obtain the optimized parameter response surface;

(7) Perform gradient descent search on the optimization parameter response surface to obtain the local optimal parameter combination;

(8) Globally optimize the local optimal parameter combination through the simulated annealing algorithm to obtain the global optimal parameter set;

(9) Perform sensitivity analysis on the global optimization parameter set and obtain parameter sensitivity ranking;

(10) The parameter sensitivity rankings are weighted and combined through an adaptive weight adjustment algorithm to obtain the target processing parameter combination.

Specifically, the machine tool working state prediction data is first decomposed in multiple dimensions to obtain the working state feature vector. Multidimensional decomposition includes time domain, frequency domain and time-frequency domain analysis, extracting features such as mean, variance, peak frequency, wavelet energy, etc. to form a high-dimensional feature vector. Then, the working state feature vector is reduced in dimension by the principal component analysis (PCA) algorithm to obtain the optimized feature set. PCA calculates the covariance matrix of the features to find the main direction of change, project the high-dimensional data into a low-dimensional space, and retain the most important information. Fuzzy clustering analysis is performed on the optimized feature set to obtain the working state pattern classification result. Fuzzy clustering allows data points to belong to multiple clusters to different degrees, and better handles the fuzzy boundaries between working states. The initial parameters of the working state pattern classification result are generated by the genetic algorithm to obtain the initial processing parameter set. The genetic algorithm simulates the natural selection and genetic process, and generates highly adaptable initial parameters through operations such as crossover and mutation.

The initial processing parameter set is subjected to orthogonal experimental design to obtain the parameter optimization test plan. Orthogonal experimental design can examine the influence of multiple factors within a small number of experiments and improve the test efficiency. The multi-objective optimization calculation of the parameter optimization test plan is performed by the response surface method to obtain the optimized parameter response surface. The response surface method establishes a mathematical model between the input parameters and the output response to predict the optimal parameter combination. The optimized parameter response surface is searched by gradient descent to obtain the local optimal parameter combination. The gradient descent method iterates along the negative gradient direction of the objective function to quickly find the local optimal solution. The local optimal parameter combination is globally optimized by the simulated annealing algorithm to obtain the global optimal parameter set. The simulated annealing algorithm helps to jump out of the local optimum and find the global optimal solution by introducing random perturbations. The sensitivity analysis of the global optimization parameter set is performed to obtain the parameter sensitivity ranking. The sensitivity analysis evaluates the influence of each parameter on the output and helps to identify the key parameters. Finally, the parameter sensitivity ranking is weighted and combined by the adaptive weight adjustment algorithm to obtain the target processing parameter combination. The adaptive weight adjustment dynamically assigns weights according to the importance of the parameters to obtain the final optimization parameter combination.

For example, in the process of machining optimization of a five-axis CNC milling machine, the predicted working state data was first decomposed and 20 features were extracted, including statistical features and spectral features of spindle speed, feed speed, cutting force, etc. Through PCA analysis, the 20-dimensional features were reduced to 6 dimensions, retaining 95% of the information. Fuzzy C-means clustering was performed on these 6-dimensional features, and the working states were divided into 3 categories: high-speed light cutting, medium-speed conventional, and low-speed heavy cutting.

Using genetic algorithm, 50 groups of initial machining parameters were generated with machining efficiency and surface quality as the objective function, including spindle speed, feed rate, cutting depth, etc. 25 groups of experimental schemes were designed using L25 (56) orthogonal table. A quadratic polynomial model was established by response surface methodology to describe the relationship between parameters and machining efficiency and surface quality. Using gradient descent method on the response surface, the local optimal solution was found: spindle speed 8000rpm, feed rate 2000mm/min, cutting depth 0.5mm.

The simulated annealing algorithm was applied, with the initial temperature set to 100, the cooling coefficient to 0.95, and 1000 iterations to obtain the global optimization parameters: spindle speed 8200rpm, feed rate 2100mm/min, and cutting depth 0.48mm. Sensitivity analysis showed that the sensitivity ratio of spindle speed, feed rate, and cutting depth was 3:2:1. Finally, the adaptive weight algorithm was used to dynamically adjust the parameter weights according to the sensitivity to obtain the final target processing parameter combination: spindle speed 8180rpm, feed rate 2080mm/min, and cutting depth 0.49mm.

Specifically, the parameter sensitivity rankings are weighted and combined by an adaptive weight adjustment algorithm to obtain the target processing parameter combination, specifically including: according to the parameter sensitivity ratio of 3:2:1, the weights of the spindle speed, feed speed and cutting depth are initialized to 0.5, 0.33 and 0.17 respectively;

Define the objective function f = α × (1/processing time) + β × (1/surface roughness), where α and β are the balance coefficients of processing efficiency and surface quality respectively;

The weighted average of the global optimization parameters (8200rpm, 2100mm/min, 0.48mm) and the local optimal parameters (8000rpm, 2000mm/min, 0.5mm) was used to obtain the initial target processing parameter combination: spindle speed = 8200×0.5+8000×0.5=8100rpm, feed speed = 2100×0.33+2000×0.67≈2033mm/min, cutting depth = 0.48×0.17+0.5×0.83≈0.497mm;

Calculate the objective function value f1 of the initial parameter combination;

Fine-tune the weight value, for example, adjust the weight of the spindle speed to 0.52, recalculate the parameter combination and the objective function value f2; compare f2 and f1, if f2>f1, accept the new weight and parameter combination, otherwise keep the original value; repeat the fine-tuning, recalculation and comparison steps until the objective function value changes less than the preset threshold value of 0.001 or reaches the maximum number of iterations of 100;

Finally, the target processing parameter combination was obtained: spindle speed 8180rpm, feed speed 2080mm/min, and cutting depth 0.49mm.

In a specific embodiment, the process of executing step S104 may specifically include the following steps:

(1) Perform real-time data stream processing on the status information of multiple CNC machine tools to obtain the machine tool status feature matrix;

(2) The trend of the machine tool state feature matrix is predicted through multivariate time series analysis to obtain the short-term state prediction results;

(3) Decompose the target processing parameter combination into tasks to obtain subtask parameter sets;

(4) Initially match the short-term state prediction results and the subtask parameter set to obtain the initial task allocation plan;

(5) Perform load balancing calculation on the initial task allocation scheme to obtain the load balancing coefficient;

(6) The initial task allocation scheme and load balancing coefficient are optimized and reorganized through genetic algorithm to obtain the optimized task allocation scheme;

(7) Perform conflict detection on the optimized task allocation scheme to obtain the task conflict matrix;

(8) Resolve the conflicts in the task conflict matrix through the graph coloring algorithm to obtain a conflict-free task sequence;

(9) Divide the conflict-free task sequence into time windows to obtain a dynamic scheduling schedule;

(10) The dynamic scheduling schedule is optimized through reinforcement learning algorithm to obtain the collaborative processing strategy.

Specifically, the state information of multiple CNC machine tools is processed in real time data streams to obtain the machine tool state feature matrix. The real-time data stream processing uses sliding window technology to batch process the continuously arriving sensor data, extract statistical features such as mean, variance, and peak value, and form a multidimensional feature matrix. Then, the trend of the machine tool state feature matrix is predicted by multivariate time series analysis to obtain the short-term state prediction result. The multivariate time series analysis uses the vector autoregression (VAR) model to consider the mutual influence between different variables and predict the state changes of the machine tool in the short term in the future. The target processing parameter combination is task-decomposed to obtain the subtask parameter set. Task decomposition is based on the processing technology requirements, and the overall task is divided into multiple subtasks, each of which contains specific processing parameters. The short-term state prediction results and the subtask parameter set are initially matched to obtain the initial task allocation plan. The initial matching uses a greedy algorithm to preliminarily allocate subtasks according to the current state and predicted state of the machine tool.

It should be noted that when the overall task is divided into multiple subtasks, the task decomposition is performed based on the given target processing parameter combination (spindle speed 8180rpm, feed speed 2080mm/min, cutting depth 0.49mm). Assuming that the milling task can be decomposed into the following subtasks:

Roughing subtask 1: Spindle speed: 8000rpm (slightly lower to improve stability); feed rate: 2200mm/min (slightly higher to improve material removal rate); cutting depth: 0.7mm (increased to improve efficiency).

Semi-finishing subtask 2: Spindle speed: 8180rpm (maintain target value); feed speed: 2080mm/min (maintain target value); cutting depth: 0.49mm (maintain target value).

Finishing subtask 3: Spindle speed: 8300rpm (slightly higher, to improve surface quality); feed rate: 1900mm/min (slightly lower, to improve accuracy); cutting depth: 0.3mm (reduced, to improve surface finish).

Contour processing subtask 4: Spindle speed: 8200rpm (slightly higher, suitable for complex contours); feed speed: 1800mm/min (lower, improved accuracy); cutting depth: 0.4mm (slightly lower, suitable for contour processing).

Groove processing subtask 5: Spindle speed: 7800rpm (slightly lower, to improve stability); feed rate: 1600mm/min (lower, to adapt to groove processing); cutting depth: 0.55mm (slightly higher, to adapt to groove depth requirements).

When machining a precision valve body made of aluminum alloy, including planes, contours and internal grooves, the machining process is as follows based on the previously decomposed subtasks:

Rough machining subtask 1: Parameters: 8000rpm, 2200mm/min, 0.7mm; Purpose: To quickly remove most of the excess material and form the basic outline of the part; Duration: 20 minutes.

Semi-finishing subtask 2: Parameters: 8180rpm, 2080mm/min, 0.49mm; Purpose: To further process the main surface of the part and improve dimensional accuracy; Duration: 15 minutes.

Finishing subtask 3: Parameters: 8300rpm, 1900mm/min, 0.3mm; Purpose: To process key planes and mating surfaces to achieve high surface finish; Duration: 10 minutes.

Contour machining subtask 4: Parameters: 8200rpm, 1800mm/min, 0.4mm; Purpose: Accurately machine the external contour and transition surface of the valve body; Duration: 12 minutes.

Groove processing subtask 5: Parameters: 7800rpm, 1600mm/min, 0.55mm; Purpose: Process the flow channel and sealing groove inside the valve body; Duration: 8 minutes.

The entire machining process: The machine tool first performs roughing tasks to quickly remove a large amount of material; then switches to semi-finishing parameters to improve overall accuracy; then performs finishing to process key planes and mating surfaces; then performs contouring to accurately form the external contour; and finally performs groove machining to complete the internal structure. Total machining time: 65 minutes

The above example specifically shows how to combine the decomposed subtasks in an orderly manner to form a complete processing flow. Each subtask optimizes the parameters for a specific processing target, which together ensures the overall processing quality and efficiency of the part. In this way, the reasonable decomposition and application of the original target processing parameters are achieved, making the entire processing process more flexible and efficient. This method can better adapt to the different characteristics and precision requirements of complex workpieces, and ultimately obtain high-quality finished parts.

The initial task allocation scheme is subjected to load balancing calculation to obtain the load balancing coefficient. The load balancing calculation takes into account the processing capacity, current load and task complexity of the machine tools, and generates coefficients that reflect the load conditions of each machine tool. The initial task allocation scheme and the load balancing coefficient are optimized and reorganized by the genetic algorithm to obtain the optimized task allocation scheme. The genetic algorithm uses load balancing and processing efficiency as fitness functions, and generates a better allocation scheme through crossover and mutation operations. Conflict detection is performed on the optimized task allocation scheme to obtain the task conflict matrix. Conflict detection identifies potential conflicts in time and resources, and generates a matrix representing the conflict relationship between tasks. The task conflict matrix is conflict-resolved by the graph coloring algorithm to obtain a conflict-free task sequence. The graph coloring algorithm regards conflicting tasks as vertices in the graph, uses different colors to represent different execution time periods, and minimizes the number of colors required to resolve conflicts.

The conflict-free task sequence is divided into time windows to obtain a dynamic scheduling schedule. The time window division takes into account the execution time of the task and the available time of the machine tool to generate a detailed execution plan. Finally, the dynamic scheduling schedule is optimized through the reinforcement learning algorithm to obtain the collaborative processing strategy. Reinforcement learning algorithms, such as deep Q networks (DQN), learn the optimal scheduling strategy by simulating the long-term benefits of different scheduling decisions.

For example, in a machining unit containing 5 CNC machine tools, each machine tool generates 100 data points per second, including spindle speed, feed rate, cutting force, etc. Using a 10-second sliding window, 20 statistical features are extracted to form a 5x20 state feature matrix. The state changes in the next hour are predicted by the VAR model. A complex aviation parts machining task is decomposed into 50 subtasks, each of which contains specific machining parameters. After the initial matching, the load balancing coefficient is calculated, ranging from 0.6 to 1.2.

The genetic algorithm used 100 individuals and evolved for 500 generations to obtain the optimal allocation solution. Conflict detection found that 15 pairs of tasks had time conflicts and generated a 50x50 conflict matrix. The graph coloring algorithm successfully resolved all conflicts using 6 colors. The conflict-free sequence was divided into 12 2-hour time windows to form a preliminary schedule. Finally, the DQN reinforcement learning model learned the optimal dynamic adjustment strategy by simulating 10,000 rounds. The final collaborative processing strategy can dynamically adjust the task allocation and execution order according to the real-time machine tool status and task progress, realizing efficient collaborative work of multiple machine tools.

In a specific embodiment, the process of executing step S105 may specifically include the following steps:

(1) Analyze the collaborative machining strategy to obtain the control instruction sequence of each machine tool;

(2) Execute the control instruction sequence of each machine tool through the real-time control interface to control multiple CNC machine tools to perform real-time processing and obtain real-time processing status data;

(3) Perform multi-source data fusion on real-time processing status data to obtain a comprehensive state feature vector;

(4) The Kalman filter algorithm is used to suppress the noise of the comprehensive state feature vector to obtain the optimized state estimate;

(5) Perform statistical process control analysis on the optimized state estimate to obtain the process capability index;

(6) Evaluate the process capability index through the fuzzy reasoning system to obtain the quality status level;

(7) Conduct correlation analysis between quality status level and collaborative processing strategy to obtain strategy influencing factors;

(8) Model the strategy influencing factors through the support vector regression algorithm to obtain a quality prediction model;

(9) Perform multi-objective optimization calculation on the quality prediction model to obtain a compensation parameter set;

(10) The compensation parameter set is dynamically adjusted through the adaptive control algorithm to obtain the quality compensation instruction.

Specifically, the collaborative processing strategy is parsed to obtain the control instruction sequence of each machine tool. The parsing process converts the high-level strategy into specific machine tool operation instructions, such as spindle speed, feed speed and other parameter settings. The control instruction sequence of each machine tool is executed through the real-time control interface to control multiple CNC machine tools for real-time processing and obtain real-time processing status data. The real-time control interface uses industrial Ethernet or fieldbus technology to achieve high-speed and reliable communication between the control system and the machine tool. Multi-source data fusion is performed on the real-time processing status data to obtain a comprehensive state feature vector. Multi-source data fusion uses the Dempster-Shafer evidence theory to integrate information from different sensors to form a more comprehensive and accurate state description. The noise of the comprehensive state feature vector is suppressed by the Kalman filter algorithm to obtain the optimized state estimate. The Kalman filter effectively removes measurement noise and provides the optimal state estimate through the prediction-correction iterative process.

Statistical process control analysis is performed on the optimized state estimate to obtain the process capability index. Statistical process control (SPC) uses control chart technology to calculate process capability indices such as Cp and Cpk to evaluate the stability and capability of the processing process. The process capability index is evaluated through the fuzzy inference system to obtain the quality status level. The fuzzy inference system builds a fuzzy rule base based on expert knowledge to convert the quantitative process capability index into a qualitative quality status level. The quality status level and the collaborative processing strategy are analyzed to obtain the strategy influencing factors. The association analysis uses the Apriori algorithm to explore the potential relationship between the processing strategy and the quality status and identify the key influencing factors. The strategy influencing factors are modeled through the support vector regression (SVR) algorithm to obtain the quality prediction model. SVR uses kernel functions to map data to high-dimensional space, build a nonlinear regression model, and predict processing quality.

The quality prediction model is optimized by multi-objective calculation to obtain the compensation parameter set. The multi-objective optimization adopts the NSGA-II algorithm, which considers multiple objectives such as processing quality, efficiency and cost at the same time to generate the Pareto optimal solution set. Finally, the compensation parameter set is dynamically adjusted by the adaptive control algorithm to obtain the quality compensation instruction. The adaptive control algorithm is based on the model predictive control (MPC) framework and dynamically adjusts the compensation parameters according to the real-time processing status and quality prediction.

For example, in a scenario where a five-axis CNC milling machine processes aircraft engine blades, the collaborative processing strategy is parsed into a sequence of 500 G-code instructions. The real-time control interface executes these instructions at a frequency of 100 Hz, while collecting real-time data of 10 parameters such as spindle speed, feed rate, and cutting force. Multi-source data fusion integrates these 10 parameters into a 30-dimensional state feature vector. The Kalman filter assumes that both process noise and observation noise are Gaussian distributed. Through iterative calculation, the standard deviation of the measurement noise is reduced from the original 0.1 mm to 0.01 mm.

Statistical process control analysis calculated Cp=1.5 and Cpk=1.3, where Cp is the process capability index and Cpk is the process capability index correction value. This indicates that the machining process is basically stable but there is an offset. The fuzzy inference system converts these indices into a "good" quality status level based on five fuzzy rules. Correlation analysis found that cutting speed and coolant flow are the two main factors affecting quality. The SVR model uses the RBF kernel function and is trained on 200 sets of historical data to establish a mapping relationship between influencing factors and machining quality.

The NSGA-II algorithm uses a population size of 100 and evolves for 500 generations to obtain a set of compensation solutions containing 10 parameters. The adaptive control algorithm calculates the optimal control input every 0.1 seconds within the 5-second prediction time domain and dynamically adjusts the cutting parameters. The quality compensation instructions finally generated include specific operations such as increasing the cutting speed by 5% and increasing the coolant flow by 10%, which effectively improves the machining accuracy and surface quality of the blade.

In a specific embodiment, the process of executing step S106 may specifically include the following steps:

(1) Time series alignment of the standardized parameter data set and mass compensation instructions to obtain a synchronized data matrix;

(2) Perform multi-scale decomposition of the synchronization data matrix through wavelet packet transform to obtain a set of characteristic coefficients;

(3) Perform principal component analysis on the feature coefficient set to obtain the reduced-dimensional feature vector;

(4) Using the support vector machine algorithm to classify the fault modes of the reduced-dimensional feature vectors, we can obtain preliminary fault categories.

(5) Perform fuzzy set operations on the preliminary fault categories to obtain the fault membership matrix;

(6) Perform nonlinear mapping of the fault membership matrix through an artificial neural network to obtain the fault severity assessment result;

(7) Perform time series prediction on the fault severity assessment results to obtain fault prediction results;

(8) Extract rules from fault prediction results through decision tree algorithm to obtain a set of fault handling rules;

(9) Prioritize the fault handling rule set to obtain a hierarchical control strategy;

(10) The hierarchical control strategy is dynamically optimized through the model predictive control algorithm to obtain the collaborative control instructions.

It should be noted that the standardized parameter data set and the quality compensation instructions are aligned in time series to obtain a synchronized data matrix. The time series alignment uses the dynamic time warping (DTW) algorithm to solve the problem of different sampling frequencies and time delays and ensure the time consistency of the data. Then, the synchronized data matrix is decomposed at multiple scales by wavelet packet transform to obtain a set of characteristic coefficients. Wavelet packet transform decomposes the signal into subbands of different frequencies and time scales, providing the time-frequency local characteristics of the signal.

Perform principal component analysis on the characteristic coefficient set to obtain the reduced-dimensional feature vector. Principal component analysis (PCA) projects high-dimensional data into a low-dimensional space by calculating the eigenvalues and eigenvectors of the covariance matrix, retaining the main information. Then, the support vector machine (SVM) algorithm is used to classify the fault mode of the reduced-dimensional feature vector to obtain the preliminary fault category. SVM constructs the optimal separation hyperplane in the high-dimensional feature space to achieve multi-category classification of fault modes.

Perform fuzzy set operations on the preliminary fault categories to obtain the fault membership matrix. Fuzzy set operations convert the deterministic classification results into fuzzy membership, reflecting the uncertainty of fault classification. Perform nonlinear mapping on the fault membership matrix through artificial neural networks to obtain the fault severity assessment results. Artificial neural networks, such as multilayer perceptrons (MLPs), use their nonlinear fitting capabilities to convert fuzzy membership into quantitative severity scores.

Perform time series prediction on the fault severity assessment results to obtain fault prediction results. Time series prediction uses a long short-term memory (LSTM) network to capture the long-term dependencies of fault development. Use a decision tree algorithm to extract rules from the fault prediction results to obtain a set of fault handling rules. Decision tree algorithms, such as C4.5, build a tree structure through the information gain criterion to convert complex prediction results into explainable rules.

The fault handling rule set is prioritized to obtain a hierarchical control strategy. Priority sorting takes into account factors such as the severity of the fault, the scope of impact, and the difficulty of handling. The analytic hierarchy process (AHP) is used to determine the weight of the rule. Finally, the hierarchical control strategy is dynamically optimized through the model predictive control (MPC) algorithm to obtain collaborative control instructions. MPC solves the optimal control sequence in the prediction time domain to achieve collaborative control of multiple machine tool systems.

For example, in a processing unit containing 5 CNC machine tools, each machine tool collects 10 parameters with a sampling frequency of 100Hz. The DTW algorithm is used to align these parameters with the quality compensation instructions once per second to form a 5×11×100 synchronous data matrix. A 4-layer wavelet packet decomposition is applied to this matrix to obtain the characteristic coefficients of 16 subbands. PCA analysis reduces the original 1600-dimensional features (5×11×16×2) to 50 dimensions, retaining 95% of the information.

The SVM uses the RBF kernel function to train a classifier that can identify 10 fault modes on historical data. Fuzzy set operations convert the output of the SVM into a 10×10 membership matrix. A 3-layer (50-20-1) MLP network maps the membership matrix to a fault severity score from 0 to 10. The LSTM network (containing 64 hidden units) predicts the fault development trend in the next 24 hours. The C4.5 decision tree algorithm extracts 20 IF-THEN rules from the prediction results.

The AHP method assigns weights to these 20 rules based on expert scores to form a hierarchical control strategy. Finally, the MPC algorithm solves the quadratic programming problem in a 5-minute prediction time domain with a step size of 0.1 seconds to generate the optimal collaborative control instruction sequence. These instructions include specific operations such as adjusting processing parameters, reallocating tasks, and starting backup equipment, realizing intelligent fault prevention and collaborative control of multi-machine tool systems.

The embodiment of the present invention further provides a collaborative control system for a CNC machine tool, as shown in FIG2 , and the collaborative control system for a CNC machine tool specifically includes:

The acquisition module 201 is used to perform multi-dimensional parameter acquisition and preprocessing on multiple CNC machine tools to obtain a standardized parameter data set;

An analysis module 202 is used to perform multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data;

A calculation module 203 is used to perform adaptive parameter optimization calculation on the machine tool working state prediction data to obtain a target processing parameter combination;

An allocation module 204 is used to dynamically allocate the state information of the plurality of CNC machine tools and the target processing parameter combination collected in real time to obtain a collaborative processing strategy;

The processing module 205 is used to control the multiple CNC machine tools to perform real-time processing through the collaborative processing strategy, collect real-time measurement data during the real-time processing, and control and analyze the real-time measurement data and the collaborative processing strategy to obtain quality compensation instructions;

The identification module 206 is used to perform fault mode identification processing on the standardized parameter data set and the quality compensation instruction to obtain a fault prediction result, and perform control instruction analysis on the fault prediction result through a decision tree algorithm to generate a coordinated control instruction.

Through the collaborative work of the above modules, through multi-dimensional parameter collection and preprocessing technology, the operating status of multiple machine tools is fully and accurately captured, laying a solid data foundation for efficient collaborative control; the multi-model fusion analysis method effectively improves the accuracy of machine tool working status prediction, making collaborative decision-making more accurate; adaptive parameter optimization calculation realizes dynamic optimization of processing parameters, significantly improving the efficiency of multi-machine collaboration; dynamic allocation processing mechanism realizes the intelligent collaborative operation of multiple machine tools, fully utilizes equipment resources, and greatly improves overall production efficiency; real-time quality control and compensation mechanism ensures quality consistency during collaborative processing and reduces errors caused by differences between machine tools; fault mode recognition and prediction functions improve the reliability of the collaborative system, reduce downtime, and ensure continuous and efficient operation; the application of decision tree algorithm makes collaborative control The generation of control instructions is more intelligent, which improves the response speed and adaptability; by integrating multi-source data and multiple advanced algorithms, a comprehensive machine tool collaborative control system is constructed, which not only improves the overall processing accuracy, but also reduces the need for manual intervention, greatly improving the level of automation; the system's self-learning and adaptive capabilities enable the collaborative control strategy to be continuously optimized to cope with complex and changing production environments; the combination of real-time and predictive characteristics enables the system to foresee and prevent possible problems while ensuring the current collaborative processing quality, thereby improving the continuity and stability of collaboration; the implementation of the collaborative control strategy enables the entire processing system to show higher flexibility and adaptability, and can quickly respond to changes in production needs, thereby improving overall production efficiency; finally, this collaborative control method based on data-driven and intelligent algorithms significantly improves the efficiency and accuracy of multi-machine collaboration.

Based on the same idea as the method in the above-mentioned embodiment, the system provided by the present application can implement the method in the above-mentioned embodiment. For the convenience of explanation, the structural diagram of the system embodiment only shows the part related to the embodiment of the present application. Those skilled in the art can understand that the illustrated structure does not constitute a limitation on the system, and may include more or fewer modules than shown in the diagram, or a combination of certain modules, or different module arrangements.

The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the embodiments, a person skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims (8)

1. The cooperative control method of the numerical control machine tool is characterized by comprising the following steps of:

carrying out multidimensional parameter acquisition and pretreatment on a plurality of numerical control machine tools to obtain a standardized parameter data set;

Performing multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data;

Performing adaptive parameter optimization calculation on the machine tool working state prediction data to obtain a target machining parameter combination;

dynamically distributing the state information of the plurality of numerical control machine tools acquired in real time and the target machining parameter combination to obtain a cooperative machining strategy;

controlling the numerical control machines to process in real time through the cooperative processing strategy, collecting real-time measurement data in the real-time processing process, and controlling and analyzing the real-time measurement data and the cooperative processing strategy to obtain a quality compensation instruction;

and carrying out fault mode identification processing on the standardized parameter data set and the quality compensation instruction to obtain a fault prediction result, and carrying out control instruction analysis on the fault prediction result through a decision tree algorithm to generate a cooperative control instruction.

2. The cooperative control method of numerically-controlled machine tools according to claim 1, wherein the step of performing multidimensional parameter acquisition and preprocessing on the plurality of numerically-controlled machine tools to obtain the standardized parameter data set comprises:

carrying out multichannel parallel acquisition on the position, feeding speed, main shaft acceleration, temperature and vibration data of a plurality of numerical control machine tools to obtain original multidimensional parameter data;

Performing wavelet transformation filtering processing on the original multidimensional parameter data to obtain noise reduction parameter data;

feature extraction is carried out on the noise reduction parameter data through a principal component analysis algorithm, so that dimension reduction feature data are obtained;

performing Z-score standardization processing on the dimension reduction characteristic data to obtain preliminary standardization data;

Performing discretization processing on the preliminary standardized data through a self-adaptive quantization algorithm to obtain discretization parameter data;

performing time window segmentation on the discretization parameter data to obtain a time sequence parameter segment;

Performing frequency domain conversion on the time sequence parameter fragments through Fourier transformation to obtain frequency domain characteristic data;

performing peak detection and spectral line extraction on the frequency domain feature data to obtain optimized spectrum features;

Performing cluster analysis on the optimized spectrum features through a self-organizing map algorithm to obtain a parameter clustering result;

and carrying out multi-dimensional cross validation on the parameter clustering result to obtain the standardized parameter data set.

3. The cooperative control method of a numerically-controlled machine tool according to claim 1, wherein the step of performing multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data comprises:

Performing time sequence decomposition on the standardized parameter data set to obtain trend items, season items and random items;

Carrying out time sequence prediction on the trend item through an autoregressive integral moving average algorithm to obtain a trend prediction result;

Periodically analyzing the seasonal items through Fourier series expansion to obtain seasonal features;

Carrying out predictive analysis on the random item through a long-short-term memory network to obtain random item predictive data;

Fusing the trend prediction result, the seasonal feature and the random item prediction data through an ensemble learning algorithm to obtain a preliminary state prediction result;

carrying out sliding window segmentation on the standardized parameter data set to obtain a plurality of time window samples;

carrying out parallel calculation on the plurality of time window samples through a support vector regression algorithm to obtain a plurality of groups of local prediction data;

Carrying out dynamic weighted average on the multiple groups of local prediction data to obtain a comprehensive prediction result;

parameter optimization is carried out on the preliminary state prediction result and the comprehensive prediction result through a particle swarm optimization algorithm, so that optimized prediction data are obtained;

and carrying out Monte Carlo simulation on the optimized prediction data to generate a plurality of groups of prediction data, and analyzing the confidence interval to obtain the machine tool working state prediction data.

4. The cooperative control method of a numerically-controlled machine tool according to claim 1, wherein the step of performing adaptive parameter optimization calculation on the machine tool operation state prediction data to obtain a target machining parameter combination includes:

Performing multidimensional decomposition on the machine tool working state prediction data to obtain a working state feature vector;

Performing dimension reduction processing on the working state feature vector through a principal component analysis algorithm to obtain an optimized feature set;

Performing fuzzy cluster analysis on the optimized feature set to obtain a working state mode classification result;

Generating initial parameters of the working state mode classification result through a genetic algorithm to obtain an initial processing parameter set;

carrying out orthogonal test design on the initial processing parameter set to obtain a parameter optimization test scheme;

performing multi-objective optimization calculation on the parameter optimization test scheme by a response surface method to obtain an optimized parameter response surface;

gradient descent search is carried out on the optimized parameter response curved surface, and a local optimal parameter combination is obtained;

Performing global optimization on the local optimal parameter combination through a simulated annealing algorithm to obtain a global optimization parameter set;

performing sensitivity analysis on the global optimization parameter set to obtain parameter sensitivity ordering;

And carrying out weighted combination on the parameter sensitivity ordering by using an adaptive weight adjustment algorithm to obtain the target processing parameter combination.

5. The cooperative control method of a numerically-controlled machine tool according to claim 1, wherein the step of dynamically allocating the combination of the state information of the plurality of numerically-controlled machines acquired in real time and the target machining parameter to obtain the cooperative machining strategy includes:

carrying out real-time data flow processing on the state information of the plurality of numerical control machine tools to obtain a machine tool state characteristic matrix;

Trend prediction is carried out on the machine tool state feature matrix through multivariate time sequence analysis, and a short-term state prediction result is obtained;

performing task decomposition on the target processing parameter combination to obtain a subtask parameter set;

Performing initial matching on the short-term state prediction result and the subtask parameter set to obtain an initial task allocation scheme;

carrying out load balancing calculation on the initial task allocation scheme to obtain a load balancing coefficient;

Optimizing and reorganizing the initial task allocation scheme and the load balancing coefficient through a genetic algorithm to obtain an optimized task allocation scheme;

Performing conflict detection on the optimized task allocation scheme to obtain a task conflict matrix;

carrying out conflict resolution on the task conflict matrix through a graph coloring algorithm to obtain a conflict-free task sequence;

Performing time window division on the conflict-free task sequence to obtain a dynamic scheduling time table;

And carrying out strategy optimization on the dynamic scheduling schedule through a reinforcement learning algorithm to obtain the collaborative processing strategy.

6. The cooperative control method of a numerically-controlled machine tool according to claim 1, wherein the step of controlling the plurality of numerically-controlled machines to perform real-time machining by the cooperative machining strategy, collecting real-time measurement data in a real-time machining process, and performing control analysis on the real-time measurement data and the cooperative machining strategy to obtain the quality compensation instruction includes:

analyzing the cooperative machining strategy to obtain control instruction sequences of all machine tools;

executing the control instruction sequences of the machine tools through a real-time control interface to control the numerical control machine tools to process in real time, so as to obtain real-time processing state data;

carrying out multi-source data fusion on the real-time processing state data to obtain a comprehensive state feature vector;

noise suppression is carried out on the comprehensive state feature vector through a Kalman filtering algorithm, and an optimized state estimation value is obtained;

carrying out statistical process control analysis on the optimized state estimation value to obtain a process capability index;

Evaluating the process capability index through a fuzzy reasoning system to obtain a quality state grade;

performing association analysis on the quality state grade and the collaborative processing strategy to obtain a strategy influence factor;

Modeling the strategy influence factors through a support vector regression algorithm to obtain a quality prediction model;

Performing multi-objective optimization calculation on the quality prediction model to obtain a compensation parameter set;

and dynamically adjusting the compensation parameter set through an adaptive control algorithm to obtain the quality compensation instruction.

7. The cooperative control method of a numerically-controlled machine tool according to claim 6, wherein the step of performing failure mode recognition processing on the standardized parameter data set and the quality compensation instruction to obtain a failure prediction result, and performing control instruction analysis on the failure prediction result through a decision tree algorithm to generate a cooperative control instruction includes:

performing time sequence alignment on the standardized parameter data set and the quality compensation instruction to obtain a synchronous data matrix;

performing multi-scale decomposition on the synchronous data matrix through wavelet packet transformation to obtain a characteristic coefficient set;

Performing principal component analysis on the characteristic coefficient set to obtain a dimension-reducing characteristic vector;

Performing fault mode classification on the dimension reduction feature vector through a support vector machine algorithm to obtain a preliminary fault class;

Performing fuzzy set operation on the preliminary fault category to obtain a fault membership matrix;

Nonlinear mapping is carried out on the fault membership matrix through an artificial neural network, so that a fault severity assessment result is obtained;

performing time sequence prediction on the fault severity evaluation result to obtain a fault prediction result;

performing rule extraction on the fault prediction result through the decision tree algorithm to obtain a fault processing rule set;

the priority ranking is carried out on the fault processing rule set, and a hierarchical control strategy is obtained;

and dynamically optimizing the hierarchical control strategy through a model predictive control algorithm to obtain the cooperative control instruction.

8. A cooperative control system of a numerical control machine tool for performing the cooperative control method of a numerical control machine tool according to any one of claims 1 to 7, comprising:

the acquisition module is used for carrying out multidimensional parameter acquisition and pretreatment on a plurality of numerical control machine tools to obtain a standardized parameter data set;

the analysis module is used for carrying out multi-model fusion analysis on the standardized parameter data set to obtain machine tool working state prediction data;

The calculation module is used for carrying out self-adaptive parameter optimization calculation on the machine tool working state prediction data to obtain a target machining parameter combination;

The distribution module is used for dynamically distributing the state information of the plurality of numerical control machine tools acquired in real time and the target machining parameter combination to obtain a cooperative machining strategy;

The processing module is used for controlling the numerical control machines to process in real time through the cooperative processing strategy, collecting real-time measurement data in the real-time processing process, and controlling and analyzing the real-time measurement data and the cooperative processing strategy to obtain a quality compensation instruction;

The identification module is used for carrying out fault mode identification processing on the standardized parameter data set and the quality compensation instruction to obtain a fault prediction result, and carrying out control instruction analysis on the fault prediction result through a decision tree algorithm to generate a cooperative control instruction.