Intelligent control method and device of numerical control grinding machine combined with real-time feedback

Abstract

The application discloses a numerical control grinding machine intelligent control method and device combined with real-time feedback, relates to the technical field of grinding machine control, and comprises the following steps: collecting historical processing records of a plurality of numerical control grinding machines in a target factory, dividing the historical processing records, determining a plurality of grinding machine domains of the same kind, performing processing deviation analysis, and obtaining a plurality of static compensation coefficient sets; performing one-time compensation on initial processing parameters, controlling the plurality of numerical control grinding machines to perform grinding processing; performing fixed-point compensation on the one-time compensation parameter set, obtaining a secondary compensation parameter set; and controlling the plurality of numerical control grinding machines to perform grinding processing until the equipment stops continuous operation or the grinding wheel accessory is replaced. The application solves the technical problem that, in the prior art, the processing consistency of batch products is poor due to inconsistent equipment states and changes in the state of the grinding wheel with the processing time in the same workshop, achieves all-round accurate control of the processing process of the numerical control grinding machine, and solves the technical effect of poor processing consistency of batch products.

Description

Technical Field

[0001] This invention relates to the field of grinding machine control technology, and more specifically to a CNC grinding machine intelligent control method and device that incorporates real-time feedback. Background Technology

[0002] In modern manufacturing, CNC grinding machines are key equipment for precision machining, widely used in machining, aerospace, automotive manufacturing, and many other fields. With the rapid development of industrial automation and intelligence, increasingly higher demands are being placed on the precision, efficiency, and product consistency of CNC grinding machines. In the same workshop, multiple CNC grinding machines are often used simultaneously. However, due to factors such as varying degrees of equipment aging, subtle differences in manufacturing processes, and different wear and tear during use, even when machining workpieces with the same product characteristics under the same operating parameters, the consistency of products produced by different CNC grinding machines still varies significantly. Furthermore, during the machining process of a single CNC grinding machine, as machining time increases, the grinding wheel inevitably wears down, and its grinding performance gradually declines. In addition, the cooling and lubrication performance of the coolant also decreases after prolonged use. If fixed control parameters are used throughout the machining process, and the changes in the performance of the grinding wheel and coolant cannot be adapted to in a timely manner, the machining quality will inevitably be further affected, leading to poor product consistency.

[0003] Existing technologies suffer from poor processing consistency in batches of products due to inconsistent equipment conditions within the same workshop and changes in grinding wheel condition over processing time. Summary of the Invention

[0004] This application provides a CNC grinding machine intelligent control method and device that incorporates real-time feedback, which addresses the technical problem of poor processing consistency of batch products in the same workshop due to inconsistent equipment status and changes in grinding wheel status over processing time.

[0005] In view of the above problems, this application provides a method and device for intelligent control of CNC grinding machines that incorporates real-time feedback.

[0006] The first aspect of this application provides an intelligent control method for a CNC grinding machine that incorporates real-time feedback, the method comprising:

[0007] Historical machining records of several CNC grinding machines within the target factory are collected. Based on these records, the CNC grinding machines are divided into several similar grinding machine domains. Machining deviation analysis is performed on each of these domains to obtain multiple sets of static compensation coefficients. Initial machining parameters are determined based on product characteristics. These initial parameters are then compensated using the multiple sets of static compensation coefficients. Several CNC grinding machines are controlled to perform grinding operations according to the primary compensation parameter set. At predetermined adjustment time points, the primary compensation parameter set is compensated at fixed points according to a secondary compensation scheme to obtain a secondary compensation parameter set. This secondary compensation is used to compensate for the effects of grinding wheel wear and reduced cooling performance. Several CNC grinding machines are controlled to perform grinding operations according to the secondary compensation parameter set until the equipment stops continuous operation or the grinding wheel accessories are replaced.

[0008] A second aspect of this application provides an intelligent control device for a CNC grinding machine that incorporates real-time feedback, the device comprising:

[0009] The system comprises the following modules: a static compensation coefficient set acquisition module, which collects historical processing records from several CNC grinding machines within the target factory, divides the CNC grinding machines into multiple similar grinding machine domains based on these records, and performs processing deviation analysis on each of these domains to acquire multiple static compensation coefficient sets; a grinding processing control module, which determines initial processing parameters based on product characteristics, performs primary compensation on these initial processing parameters based on the multiple static compensation coefficient sets, and controls several CNC grinding machines to perform grinding processing according to the primary compensation parameter set; a secondary compensation parameter set acquisition module, which performs fixed-point compensation on the primary compensation parameter set according to a secondary compensation scheme at predetermined adjustment time nodes to acquire a secondary compensation parameter set, wherein the secondary compensation is used to compensate for the effects of grinding wheel wear and reduced cooling performance; and a secondary compensation parameter set control module, which controls several CNC grinding machines to perform grinding processing according to the secondary compensation parameter set until the equipment stops continuous operation or the grinding wheel accessories are replaced.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] Historical machining records of several CNC grinding machines within the target factory are collected. Based on these records, the CNC grinding machines are divided into several similar grinding machine domains. Machining deviation analysis is performed on each of these domains to obtain multiple sets of static compensation coefficients. Initial machining parameters are determined, and primary compensation is applied to these parameters based on the static compensation coefficient sets, controlling the CNC grinding machines to perform grinding operations. At predetermined adjustment time points, secondary compensation is applied to these primary compensation parameter sets according to a secondary compensation scheme to obtain secondary compensation parameter sets. Based on these secondary compensation parameter sets, the CNC grinding machines are controlled to perform grinding operations until the equipment stops continuous operation or grinding wheel accessories are replaced. This achieves comprehensive and precise control of the CNC grinding machine machining process, solving the technical problem of poor consistency in batch product processing. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A schematic diagram of the intelligent control method for CNC grinding machines incorporating real-time feedback provided in this application embodiment;

[0014] Figure 2 A schematic diagram of the structure of an intelligent control device for a CNC grinding machine that incorporates real-time feedback, provided in an embodiment of this application.

[0015] Explanation of reference numerals in the attached diagram: 10 for static compensation coefficient set acquisition module, 20 for grinding process control module, 30 for secondary compensation parameter set acquisition module, and 40 for secondary compensation parameter set control module. Detailed Implementation

[0016] This application provides a CNC grinding machine intelligent control method and device that incorporates real-time feedback, which addresses the technical problem of poor processing consistency of batch products in the same workshop due to inconsistent equipment status and changes in grinding wheel status over processing time.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, this application provides an intelligent control method for CNC grinding machines that incorporates real-time feedback, the method comprising:

[0019] Step S100: Collect historical machining records of several CNC grinding machines in the target factory, divide the several CNC grinding machines according to the historical machining records, determine multiple grinding machine domains of the same type, and perform machining deviation analysis on the multiple grinding machine domains of the same type to obtain multiple sets of static compensation coefficients.

[0020] Specifically, collecting historical machining records from several CNC grinding machines within the target factory is fundamental. These records contain machining process data for different products at different times, such as product dimensions, surface roughness data, and machining parameters used at the time. Based on these rich historical records, the CNC grinding machines are categorized. First, the average dimensional characteristics and average surface roughness of the products machined by each CNC grinding machine within a historical time window are obtained from the historical records. Then, using the same processed products as constraints, multiple clustering operations are performed on the CNC grinding machines based on these averages, resulting in multiple clustering results. Next, the co-occurrence frequency of any two CNC grinding machines in the clustering results is statistically analyzed. Two CNC grinding machines with a co-occurrence frequency greater than a predetermined frequency threshold are considered to have high similarity and are added to the same type of grinding machine domain, ultimately determining multiple similar grinding machine domains. After determining the similar grinding machine domains, machining deviation analysis is performed for each domain. Randomly select a grinding machine domain of the same type (such as the first grinding machine domain of the same type), and obtain the average dimensional deviation and average surface roughness deviation of the domain within the historical time window from the historical processing records. Input these average deviation values ​​into the parameter static compensation model. The model calculates and outputs a set of static compensation coefficients. Repeat this process until multiple sets of static compensation coefficients corresponding to multiple grinding machine domains of the same type are obtained.

[0021] Step S200: Determine the initial processing parameters based on product feature matching, perform a first compensation on the initial processing parameters based on the multiple static compensation coefficient sets, and control several CNC grinding machines to perform grinding processing according to the first compensation parameter set.

[0022] Specifically, initial processing parameters are determined based on product characteristics. These characteristics include the product's material, shape, size requirements, and surface quality standards. By matching these characteristics with preset processing parameter matching rules, suitable initial processing parameters are determined. Then, the initial processing parameters are compensated based on multiple sets of static compensation coefficients obtained in the previous steps. Each set of static compensation coefficients corrects for common deviations of equipment within the same grinding machine domain. These coefficients are applied to the initial processing parameters to obtain a primary compensation parameter set. Finally, several CNC grinding machines are controlled to perform grinding operations according to the primary compensation parameter set. This allows for the compensation of processing deviations caused by differences between machines during initial processing, improving product consistency.

[0023] Step S300: At the predetermined adjustment time node, perform fixed-point compensation on the primary compensation parameter set according to the secondary compensation scheme to obtain the secondary compensation parameter set, wherein the secondary compensation is used to compensate for the impact of grinding wheel wear and cooling performance degradation.

[0024] Specifically, a secondary compensation operation is performed at predetermined adjustment time points. These predetermined adjustment time points are set in advance based on experience or research into the patterns of grinding wheel wear and coolant performance degradation, for example, adjustments are made after a certain processing time or after processing a certain number of products. When constructing the secondary compensation scheme, firstly, product characteristics are used as constraints. Historical processing records are queried to obtain the average grinding wheel wear characteristics and the average cooling performance degradation ratio at multiple continuous operation nodes during continuous CNC grinding. These averages are combined according to the operation nodes to obtain multiple sets of input data, each set corresponding to one operation node. Then, these multiple sets of input data are input into a secondary compensation analysis plugin for analysis. This plugin outputs multiple predetermined secondary compensation coefficient sets after complex internal calculations. Finally, a mapping relationship is established between the operation nodes and the predetermined secondary compensation coefficient sets. Based on this mapping relationship, a secondary compensation scheme is constructed. The primary compensation parameter set is then compensated at fixed points according to the secondary compensation scheme to obtain the secondary compensation parameter set, effectively compensating for the impact of grinding wheel wear and cooling performance degradation on processing consistency.

[0025] Step S400: Based on the set of secondary compensation parameters, control several CNC grinding machines to perform grinding operations until the equipment stops continuous operation or the grinding wheel accessories are replaced.

[0026] Specifically, several CNC grinding machines are controlled to perform grinding operations based on a secondary compensation parameter set. Throughout the machining process, the secondary compensation parameter set is continuously used until the equipment stops continuous operation (e.g., due to completion of production tasks, equipment failure, etc.) or the grinding wheel accessories are replaced (when grinding wheel wear reaches a certain level and affects machining quality). This step ensures that during equipment operation, through reasonable parameter compensation, a high degree of machining consistency is maintained, guaranteeing the quality stability of batch products, improving production efficiency and product qualification rate, reducing scrap and rework rates caused by changes in equipment performance, and enhancing the economic benefits and product competitiveness of the entire production process.

[0027] In one possible implementation, step S300 further includes:

[0028] Step S310: At a predetermined monitoring time point, collect parameter fluctuation ranges and processed product deviations of several CNC grinding machines, and perform parameter optimization analysis based on the parameter fluctuation ranges and processed product deviations to determine several sets of cubic compensation coefficients.

[0029] Step S320: Dynamically fine-tune the set of secondary compensation parameters according to the plurality of third-order compensation coefficients to obtain a plurality of optimized processing parameters, and control a plurality of CNC grinding machines to perform grinding operations.

[0030] Specifically, setting predetermined monitoring time points is crucial for ensuring timely capture of changes in the machining process. These time points can be determined based on experience, equipment operating patterns, or production process requirements, such as at regular machining intervals or after each batch of products is completed. At these specific moments, comprehensive data acquisition is conducted on several CNC grinding machines. The collected parameter fluctuation ranges cover the fluctuations of key machining parameters such as grinding speed, workpiece rotation speed, feed rate, depth of cut, and feed rate. Sensors and data acquisition systems accurately record the maximum, minimum, and average values ​​of these parameters over a certain period to reflect dynamic changes during machining. Simultaneously, the deviations of the machined products are precisely measured, including dimensional and surface roughness deviations, and compared with design requirements. Next, parameter optimization analysis is performed based on the collected parameter fluctuation ranges and machined product deviations. This requires querying historical machining records to collect a set of sample parameter fluctuation ranges, a set of sample machined product deviations, and multiple sets of sample parameter compensation coefficients. Each processing parameter here corresponds to a parameter compensation coefficient, which is used to quantify the impact of parameter adjustment on the processing result. The sample parameter fluctuation range and sample processed product deviation are used as input data, and the sample parameter compensation coefficient is used as supervision information. The machine learning model is trained using this rich sample data. The machine learning model learns from a large amount of historical data to find the intrinsic relationship between parameter fluctuation and product deviation, and then obtains a parameter optimization analyzer. This analyzer can accurately perform parameter optimization analysis based on the currently collected actual parameter fluctuation range and processed product deviation, and determine several sets of cubic compensation coefficients. Each set of coefficients corresponds to a CNC grinding machine to specifically correct deviations in its processing process.

[0031] The secondary compensation parameter set is dynamically fine-tuned based on several determined tertiary compensation coefficients. While the secondary compensation parameter set is determined considering factors such as grinding wheel wear and coolant performance degradation, other dynamic changes can still affect machining consistency during actual processing. The tertiary compensation coefficients further correct for these real-time changes. The tertiary compensation coefficients are then calculated with the corresponding parameters of the secondary compensation parameter set, for example, through a weighted summation or other reasonable calculation method, to obtain optimized machining parameters. These optimized machining parameters fully consider various factors, including the current operating status of the equipment, the degree of grinding wheel wear, coolant performance, and dynamic changes during the machining process. Finally, several CNC grinding machines are controlled to perform grinding operations based on these optimized machining parameters. This ensures that throughout the entire machining process, the CNC grinding machines can continuously adjust the machining parameters based on real-time feedback, maintaining a high-precision machining state. This effectively improves the dimensional and surface roughness consistency of batch products, enhances product quality and production efficiency, and reduces production costs.

[0032] In one possible implementation, step S310 further includes:

[0033] Step S311: Query historical processing records, collect sample parameter fluctuation range set, sample processed product deviation set, and multiple sample parameter compensation coefficient sets. Among them, the processing parameters include grinding speed, workpiece rotation speed, feed rate, cutting depth and feed rate, and each processing parameter corresponds to a parameter compensation coefficient.

[0034] Step S312: Using the sample parameter fluctuation range and sample processed product deviation as inputs, and the sample parameter compensation coefficient as supervision, train the machine learning model using the sample parameter fluctuation range set, sample processed product deviation set, and multiple sample parameter compensation coefficient sets to obtain the parameter optimization analyzer.

[0035] Step S313: Using the parameter optimization analyzer, perform parameter optimization analysis based on the parameter fluctuation range and the deviation of the processed product.

[0036] Specifically, the first step is to delve into the valuable data resource of historical machining records. These records encompass detailed processing information for various products using CNC grinding machines over extended periods. For the grinding speed parameter, careful analysis of speed variation data under different machining tasks is crucial to determine its maximum, minimum, and average values ​​within each machining cycle, thus constructing a sample parameter fluctuation range set for grinding speed. For example, during the machining of a specific part, the initial grinding speed might be set at 2000 rpm. Due to various factors, the speed might fluctuate between 1950 rpm and 2050 rpm during the machining process; this fluctuation range constitutes a sample parameter fluctuation range for grinding speed. Similarly, fluctuation ranges for other machining parameters such as workpiece rotation speed, feed rate, depth of cut, and tool feed rate are collected and summarized to form a complete sample parameter fluctuation range set. Simultaneously, comprehensive quality inspection is conducted on historically machined products. The difference between the actual and design dimensions of each product, as well as the deviation of surface roughness from the standard roughness, are precisely measured. These dimensional and surface roughness deviation data are then compiled to obtain a sample machined product deviation set. For example, a batch of processed products might have a designed size of 100 mm, but the actual measured size fluctuates between 99.8 mm and 100.2 mm. These deviations become part of the deviation set of the sample processed products. Furthermore, to provide effective supervision information for subsequent machine learning model training, multiple sets of sample parameter compensation coefficients need to be obtained from historical processing records. For each processing parameter, the compensation coefficients previously used to adjust that parameter to improve product quality are reviewed. These compensation coefficients are closely related to the fluctuations in processing parameters and product deviations. For instance, when fluctuations in grinding speed are found to cause product size deviations, the compensation coefficients previously used to correct the grinding speed are recorded. Multiple sets of sample parameter compensation coefficients are formed for different processing conditions and deviation types. The coefficients in each set correspond to the corresponding processing parameters, laying a solid data foundation for subsequent model learning and parameter optimization analysis.

[0037] The model's input feature vectors are derived from data on grinding speed fluctuations, workpiece rotation speed fluctuations, feed rate fluctuations, depth of cut fluctuations, and tool feed rate fluctuations within the sample parameter fluctuation range set, as well as dimensional deviations and surface roughness deviations from the sample processed product deviation set. This input data comprehensively reflects the dynamic changes of the CNC grinding machine during past processing and the resulting product quality deviations. A neural network model is then constructed, comprising an input layer, hidden layers, and an output layer. The number of neurons in the input layer corresponds to the dimension of the input feature vector, and it receives the parameter fluctuation range and processed product deviation data. The hidden layer, by setting a certain number of neurons (adjustable according to data complexity and model performance requirements), uses the connection weights between neurons and activation functions (such as the ReLU function) to learn complex nonlinear relationships in the data. The output layer outputs the same number of predicted compensation coefficients as the number of processing parameters, with each coefficient corresponding to one processing parameter. During training, the sample parameter compensation coefficient set serves as supervision information, where the compensation coefficient corresponding to each processing parameter explicitly indicates the direction and magnitude of adjustments to be made for the corresponding parameter fluctuations and product deviations. The model calculates predicted compensation coefficients based on the input data and compares them with the actual sample parameter compensation coefficients. By calculating the error between the two (e.g., mean squared error), the backpropagation algorithm is used to propagate the error from the output layer to the hidden and input layers, adjusting the connection weights between neurons so that the model's predicted compensation coefficients gradually approach the actual sample parameter compensation coefficients. As training continues, the model learns more deeply from the sample data, gradually mastering the complex nonlinear relationship between parameter fluctuation ranges, processed product deviations, and parameter compensation coefficients. When the model's performance metrics (e.g., mean squared error, accuracy) on the training and validation sets reach the expected standards, it indicates that the model has the ability to accurately predict compensation coefficients based on the input parameter fluctuation ranges and processed product deviations. At this point, training is complete, and the resulting parameter optimization analyzer can receive real-time collected parameter fluctuation range and processed product deviation data, and quickly output the corresponding optimized compensation coefficients. This provides a key basis for the dynamic fine-tuning of subsequent CNC grinding machine machining parameters, thereby achieving precise control of the machining process and effectively improving the processing consistency and quality stability of batch products. In addition to neural network algorithms, support vector machines, decision tree regression, and other algorithms can also be applied to the model training process according to actual conditions to meet the needs of different scenarios.

[0038] Using a pre-trained parameter optimization analyzer, key parameter optimization analysis is performed based on real-time collected parameter fluctuation ranges and processed product deviations. First, the fluctuation range data of machining parameters such as grinding speed, workpiece rotation speed, feed rate, depth of cut, and tool feed rate, acquired from the CNC grinding machine's sensors at the current moment, are input into the parameter optimization analyzer. This data reflects the real-time dynamic changes in the equipment's operation during machining. Simultaneously, the dimensional deviations and surface roughness deviations of the product being machined, obtained through precise measurements, are also input. Leveraging the complex mapping relationship between parameter fluctuation ranges, processed product deviations, and parameter compensation coefficients learned during the training phase, the parameter optimization analyzer quickly processes the input real-time data. Through its complex internal calculation logic, it comprehensively considers the degree of influence of each machining parameter fluctuation on product deviations and their interactions. For example, it analyzes the correlation between grinding speed fluctuations and product dimensional deviations, and the impact of workpiece rotation speed fluctuations on surface roughness deviations. Based on these analyses, the parameter optimization analyzer calculates the most suitable compensation coefficient for each machining parameter, forming a set of three compensation coefficients for the current machining state. This set of coefficients can accurately guide the subsequent dynamic fine-tuning of the secondary compensation parameter set, enabling the CNC grinding machine to adapt to various changes in the machining process in a timely manner, ensuring that it maintains a high-precision machining state throughout the entire machining process, effectively improving the dimensional consistency and surface roughness consistency of batch products, enhancing product quality and production efficiency, and reducing scrap rate and production costs caused by machining deviations.

[0039] In one possible implementation, step S100 further includes:

[0040] Step S110: Based on the historical processing records, obtain the average values ​​of multiple dimensional features and multiple surface roughness values ​​of multiple processed products from several CNC grinding machines within the historical time window.

[0041] Step S120: Using the same processed product as a constraint, cluster several CNC grinding machines multiple times based on the average values ​​of multiple dimensional features and multiple average values ​​of surface roughness to obtain multiple clustering results.

[0042] Step S130: Calculate the co-occurrence frequency of any two CNC grinding machines in the multiple clustering results, and add any two CNC grinding machines with a frequency greater than a predetermined threshold to the same type of grinding machine domain to obtain the multiple types of grinding machine domains.

[0043] Specifically, this involves in-depth analysis of historical processing records, a crucial source of information. These records meticulously document rich data on various products processed by numerous CNC grinding machines over a past period. For each CNC grinding machine and multiple processed products, precise dimensional characteristic information is extracted, such as the product's diameter, length, and hole diameter. The average value of these dimensional data within the historical time window is calculated, resulting in multiple dimensional characteristic averages. Simultaneously, high-precision measuring equipment is used to acquire surface roughness data for each processed product, and its average value is calculated, yielding multiple surface roughness averages. These average data provide a macroscopic reflection of the average quality level and characteristics of products processed by each CNC grinding machine within a specific historical period.

[0044] Using the same processed product as the core constraint, this study delves into and utilizes multiple dimensional feature averages and multiple surface roughness averages to perform multiple clustering operations on several CNC grinding machines, thereby obtaining multiple clustering results. First, for each processed product, the average dimensional feature and average surface roughness values ​​corresponding to the relevant CNC grinding machine processing data are precisely extracted. For example, when machining a certain precision gear, the average dimensional feature values ​​such as tooth thickness and addendum circle diameter, as well as the average surface roughness, obtained by each CNC grinding machine in machining this gear, are considered. These product-specific average data are used as the basis for clustering, and advanced clustering algorithms (such as hierarchical clustering and DBSCAN algorithm) are employed for clustering operations. During the clustering process, the algorithm classifies the CNC grinding machines into different categories based on the similarity of their dimensional and surface roughness features when machining the product. The similarity criterion comprehensively considers the differences between the dimensional feature averages and the closeness of the surface roughness averages; the smaller the difference and the higher the closeness, the stronger the similarity is considered, and the more likely they are to be clustered into one class. Because there are various processed products, the above clustering process is repeated for each product, resulting in multiple clustering results. Each clustering result clearly reflects the similarity grouping of CNC grinding machines in terms of processing quality and performance when processing a specific product. This provides key classification information for accurately identifying equipment groups with common characteristics and implementing targeted control strategies, which helps improve the overall accuracy and stability of CNC grinding in the workshop and achieve high-quality consistency in mass production.

[0045] Based on the previously obtained clustering results, a comprehensive analysis of the combinations of any two CNC grinding machines was conducted. For each pair of CNC grinding machines, the number of times they appeared in the same group in different clustering results was carefully counted, serving as the basis for calculating the co-occurrence frequency. For example, in a scenario with five clustering results, if CNC grinding machine A and CNC grinding machine B are grouped in the same group in three of the clustering results, then their co-occurrence frequency is 3. Next, a predetermined frequency threshold was introduced as a key criterion. This threshold was set based on a deep understanding of the overall performance distribution and similarity requirements of CNC grinding machines in the workshop. The co-occurrence frequency of each pair of CNC grinding machines was compared with the predetermined frequency threshold one by one. When the co-occurrence frequency of a pair of CNC grinding machines was greater than the predetermined frequency threshold, it meant that these two machines showed high similarity and correlation in multiple product processing clusters, and they had relatively consistent performance characteristics and quality performance when processing different products. At this time, these two CNC grinding machines were added to the same type of grinding machine domain. By statistically analyzing and filtering the co-occurrence frequencies of all CNC grinding machine combinations, multiple same type of grinding machine domains were gradually constructed. The CNC grinding machines within each grinding machine domain share stronger commonalities, which provides a solid foundation for subsequent intelligent control measures such as determining unified compensation coefficients and optimizing machining parameters for equipment within the domain. This helps improve the overall efficiency of CNC grinding machine processing in the workshop and the consistency of product quality, effectively reducing machining errors and production costs caused by equipment differences.

[0046] In one possible implementation, step S100 further includes:

[0047] Step S140: Randomly select a first type of grinding machine domain, and according to the historical processing records, obtain the average dimensional deviation and the average surface roughness deviation of the first type of grinding machine domain within the historical time window.

[0048] Step S150: Input the mean value of the size deviation and the mean value of the surface roughness deviation into the parameter static compensation model, output the first static compensation coefficient set, and add it to the plurality of static compensation coefficient sets.

[0049] Specifically, a first grinding machine domain was randomly selected from several established similar grinding machine domains. This random selection ensured the representativeness and universality of subsequent analyses. Then, historical machining records were thoroughly analyzed, focusing on the machining data of CNC grinding machines within this first similar grinding machine domain over a historical time window. For the dimensional information of the machined products, the deviation from the design standard dimensions was precisely calculated. The average dimensional deviation was obtained by statistically averaging a large number of machined product dimensional deviation data. Simultaneously, the surface roughness of the products was also measured, and its deviation from the ideal surface roughness was measured, and the average surface roughness deviation was calculated. These average deviation data comprehensively reflect the average deviation level and quality stability of this similar grinding machine domain during historical machining processes.

[0050] Once the average dimensional deviation and average surface roughness deviation of the first type of grinding machine domain within the historical time window are obtained, they are input into the parametric static compensation model. This parametric static compensation model can be an intelligent model built based on machine learning algorithms, or a mapping table summarized from a large amount of experimental data and engineering experience. If it is a machine learning model, it has already learned the complex relationships between many different combinations of average dimensional deviation and average surface roughness deviation and the corresponding static compensation coefficient set during the training phase. When the current average dimensional deviation and average surface roughness deviation are input, the model will quickly perform calculations and inferences based on its internal neural network structure algorithm logic. Through feature extraction and pattern recognition of the input data, the model predicts the most suitable set of static compensation coefficients for the current deviation, which includes precise adjustment amounts of compensation coefficients for various processing parameters (such as grinding speed, workpiece rotation speed, feed rate, depth of cut, and feed rate) to correct processing deviations caused by inherent differences in the equipment. If a mapping table is used, a table of correspondences between the average dimensional deviation, average surface roughness deviation, and static compensation coefficient set is established in advance based on rich historical data and practical experience. After inputting the average dimensional deviation and average surface roughness deviation, the corresponding first static compensation coefficient set is directly obtained by looking up a table. This method is simple and direct, and can quickly provide a compensation scheme based on past mature experience. Regardless of the method used to obtain the first static compensation coefficient set, it is added to multiple static compensation coefficient sets, providing a key basis for subsequent unified compensation and machining parameter optimization for different types of grinding machines. This effectively improves the consistency of CNC grinding machine-processed products, ensures the quality stability of mass production, increases production efficiency, and reduces production costs.

[0051] In one possible implementation, step S300 further includes:

[0052] Step S330: Using the product characteristics as constraints, query historical processing records to obtain the average values ​​of multiple grinding wheel wear characteristics under multiple continuous operation nodes during continuous operation of several CNC grinding machines, as well as the average values ​​of multiple cooling performance reduction ratios under multiple continuous operation nodes.

[0053] Step S340: Combine the average wear characteristics of the multiple grinding wheels and the average cooling performance reduction ratio according to the operation node to obtain multiple sets of input data, wherein each set of input data corresponds to one operation node.

[0054] Step S350: Input the multiple sets of input data into the secondary compensation analysis plugin for analysis, and output multiple predetermined secondary compensation coefficient sets.

[0055] Step S360: Establish a mapping relationship between multiple job nodes and multiple predetermined secondary compensation coefficient sets, and construct the secondary compensation scheme according to the mapping relationship.

[0056] Specifically, using product characteristics as key constraints, historical processing records were thoroughly investigated. For several CNC grinding machines operating continuously, the average wear characteristics of the grinding wheels were accurately obtained at multiple continuous operation nodes. Grinding wheel wear characteristics can be measured in various ways, such as the reduction in grinding wheel radius and the degree of change in grinding wheel surface morphology. The average values ​​of these characteristics at each operation node were calculated, thus obtaining multiple average grinding wheel wear characteristics. Simultaneously, the cooling performance of the coolant during continuous operation was measured, such as changes in the coolant's heat dissipation efficiency, and the average percentage decrease in performance relative to the initial performance at each continuous operation node was calculated. The acquisition of this data provides a solid foundation for subsequent analysis.

[0057] The acquired average values ​​of multiple grinding wheel wear characteristics and multiple average values ​​of cooling performance degradation rates are carefully combined according to the corresponding work nodes. For example, the average grinding wheel wear characteristics at a certain work node are combined with the average cooling performance degradation rate at that node to form a set of input data, and each set of input data uniquely corresponds to a work node. This combination method can fully present the status of the two key factors, grinding wheel wear and coolant performance degradation, at each work node, providing comprehensive and accurate input information for the secondary compensation analysis plugin.

[0058] After carefully combined sets of input data are fed into the secondary compensation analysis plugin, a series of complex and orderly analytical operations begin within the plugin. First, the plugin extracts features from each set of input data, deeply analyzing the information contained in the average grinding wheel wear characteristics and the average cooling performance reduction ratio. For the average grinding wheel wear characteristics, the plugin further analyzes the degree of grinding wheel wear, wear patterns, and potential impact on machining accuracy and surface quality. For the average cooling performance reduction ratio, it assesses in detail the effect of reduced coolant heat dissipation capacity on thermal deformation and tool wear during machining. Next, based on pre-defined algorithm models—which are trained on extensive historical data and optimized through engineering experience—the plugin performs comprehensive analysis based on the extracted features. It utilizes machine learning algorithms (such as neural networks) to calculate the compensation coefficients that need to be adjusted for different machining parameters (grinding speed, workpiece rotation speed, feed rate, depth of cut, and feed rate, etc.). For example, when severe grinding wheel wear and a significant drop in coolant performance are detected, the plugin may calculate compensation coefficients for parameters such as reducing grinding speed, appropriately adjusting feed rate, and depth of cut to maintain machining accuracy and surface quality. This reduces the impact of accelerated grinding wheel wear on the workpiece, while also considering heat dissipation issues caused by decreased coolant performance to prevent thermal deformation from affecting machining accuracy. After a complex calculation and optimization process, the plugin outputs a predetermined set of secondary compensation coefficients for each set of input data. These coefficient sets precisely define the compensation adjustment amount for each machining parameter at the corresponding work node, providing a crucial basis for subsequently constructing secondary compensation schemes. This ensures that the CNC grinding machine maintains stable machining performance even with grinding wheel wear and decreased coolant performance, improving product quality consistency, reducing scrap rate, and increasing production efficiency.

[0059] A one-to-one mapping relationship is established between multiple operation nodes and multiple predetermined sets of secondary compensation coefficients. Based on this mapping relationship, a complete secondary compensation scheme is constructed. During actual machining, when the CNC grinding machine reaches a specific operation node, it quickly and accurately obtains the corresponding predetermined set of secondary compensation coefficients according to this mapping relationship, and adjusts the machining parameters in real time. This ensures that throughout the continuous operation, the CNC grinding machine can dynamically adjust the machining parameters according to changes in grinding wheel wear and coolant performance, effectively maintaining the consistency of machined products, improving product quality and production efficiency, extending grinding wheel life, and reducing production costs.

[0060] In one possible implementation, step S350 further includes:

[0061] Step S351: Collect the sample grinding wheel wear feature set and the sample cooling performance reduction ratio set, and obtain the secondary compensation coefficient set under different sample grinding wheel wear features and different sample cooling performance reduction ratio conditions, and construct multiple sample secondary compensation coefficient sets.

[0062] Step S352: Construct a sample dataset based on the sample grinding wheel wear feature set, the sample cooling performance degradation ratio set, and multiple sample secondary compensation coefficient sets, and divide the sample dataset into M equal datasets.

[0063] Step S353: Using the M datasets, supervise the training of the BP network, and when it tends to converge, perform cross-validation on the BP network by interacting with the M datasets to obtain M convergent quadratic compensation analysis branches.

[0064] Step S354: Construct the quadratic compensation analysis plugin based on the M convergent quadratic compensation analysis branches, wherein the output of the quadratic compensation analysis plugin is the average of the output results of the M convergent quadratic compensation analysis branches.

[0065] Specifically, a wide range of sample grinding wheel wear characteristic sets and sample cooling performance degradation ratio sets were collected through various detection methods and data acquisition techniques. For the sample grinding wheel wear characteristics, high-precision measuring instruments were used to monitor key indicators such as surface morphology changes, abrasive grain shedding, and wheel size reduction at different wear stages. These data were then compiled into a sample grinding wheel wear characteristic set. Simultaneously, professional coolant performance testing equipment was used to determine the reduction ratio of coolant heat dissipation efficiency and the degree of lubrication performance degradation under different usage durations or operating conditions, constructing a sample cooling performance degradation ratio set. Then, for different combinations of sample grinding wheel wear characteristics and sample cooling performance degradation ratios, based on extensive experimental data and practical machining experience, corresponding secondary compensation coefficient sets were obtained. These coefficient sets clarified how to adjust machining parameters such as grinding speed, workpiece rotation speed, feed rate, depth of cut, and tool feed rate to ensure machining quality under specific wear and cooling performance conditions, thereby constructing multiple sample secondary compensation coefficient sets.

[0066] The carefully collected sample sets of grinding wheel wear features, sample sets of cooling performance degradation ratios, and corresponding sets of multiple sample secondary compensation coefficients are integrated. The sample grinding wheel wear feature set contains characteristic data of grinding wheels under different wear levels obtained through various advanced detection technologies, such as changes in grinding wheel surface texture, abrasive wear state, and reduction in grinding wheel diameter. The sample cooling performance degradation ratio set covers the degradation ratio data of key indicators such as heat dissipation efficiency and lubrication performance of coolant relative to the initial state under long-term use or different operating conditions. These data, along with the corresponding sample secondary compensation coefficient sets (which include compensation adjustments to machining parameters under different wear and cooling performance conditions), are combined according to certain rules to construct a comprehensive sample dataset reflecting the relationship between grinding wheel wear, cooling performance changes, and compensation coefficients. To ensure the comprehensiveness, accuracy, and generalization ability of the model in subsequent training, this sample dataset needs to be equally partitioned. The sample dataset is divided into M datasets, with each dataset maintaining as much consistency as possible in terms of data size and feature distribution. For example, if there are N sample data points in the sample dataset, then each partitioned dataset contains approximately N / M data points. During the partitioning process, the diversity of sample data must be fully considered to ensure that each dataset covers different degrees of grinding wheel wear characteristics, the proportion of cooling performance degradation, and the corresponding compensation coefficients. This allows each dataset to be used independently for model training while also collaborating to improve the performance of the final model. This partitioning method lays a solid foundation for subsequent supervised training of the BP network using M datasets, helping to train a more accurate and reliable secondary compensation analysis model. This enables effective responses to the problems of grinding wheel wear and coolant performance degradation during CNC grinding, ensuring the stability and consistency of the processed product quality.

[0067] For each dataset, the sample grinding wheel wear characteristics and the sample cooling performance degradation ratio are used as inputs to the BP network, and the corresponding sample quadratic compensation coefficient set is used as the expected output, thus initiating the supervised training process of the BP network. During training, the BP network calculates the transmission results of the input data between neurons in each layer through forward propagation to obtain the predicted quadratic compensation coefficient set. Then, the error between the prediction result and the expected output is calculated, and the error is propagated back from the output layer to the input layer using the backpropagation algorithm. Based on the error, the connection weights and thresholds between neurons are adjusted to continuously optimize the network parameters and reduce the prediction error. As training continues, the error of the BP network model corresponding to each dataset on the validation set gradually decreases and tends to converge, indicating that the model has gradually learned the complex mapping relationship between the input data (grinding wheel wear characteristics and cooling performance degradation ratio) and the output data (quadratic compensation coefficient set). When each BP network model tends to converge, to further improve the stability and generalization ability of the model, cross-validation is performed on M datasets. Each dataset is used as the validation set, and the remaining M-1 datasets are used as the training set to train and evaluate the BP network again. In this process, the network continuously adjusts its parameters to adapt to the characteristics of different datasets, thereby further optimizing the model. Through this cross-validation process, each dataset participates in the training and validation of the model, ultimately resulting in M ​​convergent quadratic compensation analysis branches. Each branch is a fully trained and optimized BP network model, possessing the ability to accurately predict the set of quadratic compensation coefficients based on grinding wheel wear characteristics and the proportion of cooling performance degradation. Furthermore, they exhibit good adaptability and stability when facing different data distributions, providing a solid foundation for the subsequent integration and construction of a high-precision quadratic compensation analysis plugin.

[0068] In the intelligent control system of CNC grinding machines, the key step in achieving accurate compensation is to integrate and construct a secondary compensation analysis plugin based on M convergent secondary compensation analysis branches. After obtaining the M convergent secondary compensation analysis branches, the plugin is constructed. In the plugin's architecture, these M branches are integrated into a cohesive whole. When new average values ​​of grinding wheel wear characteristics and the average proportion of cooling performance degradation are input into the plugin, the M convergent secondary compensation analysis branches process the input data simultaneously. Each branch independently calculates and outputs its corresponding set of secondary compensation coefficients based on the complex mapping relationship between input and output learned during training. The plugin then summarizes the outputs of these M branches. By calculating the average of these M outputs, the final set of secondary compensation coefficients is obtained as the plugin's output. This integration method fully utilizes the advantages of multiple branches, effectively reducing the errors and uncertainties that may exist in a single model. Since each branch uses different data subsets during training, their understanding and processing of the data differ. By calculating the average, the predictive information from multiple branches can be integrated, making the final output set of secondary compensation coefficients more accurate and reliable. This helps to more accurately address wheel wear and coolant performance degradation during CNC grinding, allowing for real-time adjustment of machining parameters based on actual working conditions. This ensures product consistency and quality stability, improves production efficiency, reduces scrap rates, and enhances the overall performance and reliability of the CNC grinding system.

[0069] Example 2, based on the same inventive concept as the intelligent control method for CNC grinding machines combined with real-time feedback in the foregoing examples, such as... Figure 2 As shown, this application provides an intelligent control device for CNC grinding machines that incorporates real-time feedback. The device and method embodiments in this application are based on the same inventive concept. The device includes:

[0070] The static compensation coefficient set acquisition module 10 is used to collect historical processing records of several CNC grinding machines in the target factory, divide the several CNC grinding machines according to the historical processing records, determine multiple grinding machine domains of the same type, and perform processing deviation analysis on the multiple grinding machine domains of the same type to obtain multiple static compensation coefficient sets.

[0071] The grinding process control module 20 is used to determine the initial processing parameters according to the product characteristics, perform a first compensation on the initial processing parameters based on the multiple static compensation coefficient sets, and control several CNC grinding machines to perform grinding processing according to the first compensation parameter set.

[0072] The secondary compensation parameter set acquisition module 30 is used to perform fixed-point compensation on the primary compensation parameter set according to the secondary compensation scheme at a predetermined adjustment time node to acquire the secondary compensation parameter set. The secondary compensation is used to compensate for the effects of grinding wheel wear and cooling performance degradation.

[0073] The secondary compensation parameter set control module 40 is used to control several CNC grinding machines to perform grinding operations according to the secondary compensation parameter set until the equipment stops continuous operation or the grinding wheel accessories are replaced.

[0074] Furthermore, the secondary compensation parameter set acquisition module 30 also includes:

[0075] The three-dimensional compensation coefficient set determination unit is used to collect parameter fluctuation ranges and processing product deviations of several CNC grinding machines at a predetermined monitoring time node, and perform parameter optimization analysis based on the parameter fluctuation ranges and processing product deviations to determine several three-dimensional compensation coefficient sets.

[0076] An optimized machining parameter acquisition unit is used to dynamically fine-tune the set of secondary compensation parameters according to the plurality of third-order compensation coefficients to obtain a plurality of optimized machining parameters and control a plurality of CNC grinding machines to perform grinding operations.

[0077] Furthermore, the unit for determining the set of third-order compensation coefficients also includes:

[0078] The historical processing record query unit is used to query historical processing records, collect sample parameter fluctuation range set, sample processed product deviation set, and multiple sample parameter compensation coefficient sets. The processing parameters include grinding speed, workpiece rotation speed, feed rate, cutting depth, and tool feed rate, and each processing parameter corresponds to a parameter compensation coefficient.

[0079] The parameter optimization analyzer acquisition unit is used to train a machine learning model by taking the sample parameter fluctuation range and sample processed product deviation as inputs, and using the sample parameter compensation coefficient as supervision, and using the sample parameter fluctuation range set, sample processed product deviation set and multiple sample parameter compensation coefficient sets to obtain the parameter optimization analyzer.

[0080] A parameter optimization analysis unit is used to perform parameter optimization analysis based on the parameter fluctuation range and the deviation of the processed product using the parameter optimization analyzer.

[0081] Furthermore, the static compensation coefficient set acquisition module 10 also includes:

[0082] A surface roughness mean acquisition unit is used to acquire, based on the historical processing records, the mean values ​​of multiple dimensional features and the mean values ​​of multiple surface roughness of multiple processed products from several CNC grinding machines within a historical time window.

[0083] Multiple clustering result acquisition units are used to perform multiple clustering on several CNC grinding machines based on the average values ​​of multiple dimensional features and multiple average values ​​of surface roughness, with the same processed product as a constraint, to obtain multiple clustering results.

[0084] Multiple similar grinding machine domain acquisition units are used to count the co-occurrence frequency of any two CNC grinding machines in the multiple clustering results, and add any two CNC grinding machines with a frequency greater than a predetermined threshold to the similar grinding machine domain to obtain the multiple similar grinding machine domains.

[0085] Furthermore, the static compensation coefficient set acquisition module 10 also includes:

[0086] The first similar grinding machine domain selection unit is used to randomly select a first similar grinding machine domain and obtain the average dimensional deviation and average surface roughness deviation of the first similar grinding machine domain within the historical time window based on the historical processing records.

[0087] The first static compensation coefficient set output unit is used to input the average value of the size deviation and the average value of the surface roughness deviation into the parameter static compensation model, output the first static compensation coefficient set, and add it to the plurality of static compensation coefficient sets.

[0088] Furthermore, the secondary compensation parameter set acquisition module 30 also includes:

[0089] The grinding wheel wear characteristic mean acquisition unit is used to query historical processing records with the product characteristics as constraints, and to obtain the mean values ​​of multiple grinding wheel wear characteristics under multiple continuous operation nodes during continuous operation of several CNC grinding machines, as well as the mean values ​​of multiple cooling performance reduction ratios under multiple continuous operation nodes.

[0090] A multi-set input data acquisition unit is used to combine the average values ​​of multiple grinding wheel wear characteristics and the average values ​​of multiple cooling performance reduction ratios according to the operation nodes to obtain multiple sets of input data, wherein each set of input data corresponds to one operation node.

[0091] A predetermined secondary compensation coefficient set output unit is used to input the multiple sets of input data into the secondary compensation analysis plug-in for analysis and output multiple predetermined secondary compensation coefficient sets.

[0092] A secondary compensation scheme construction unit is used to establish a mapping relationship between multiple work nodes and multiple predetermined secondary compensation coefficient sets, and to construct the secondary compensation scheme according to the mapping relationship.

[0093] Furthermore, the predetermined secondary compensation coefficient set output unit further includes:

[0094] The sample secondary compensation coefficient set construction unit is used to collect the sample grinding wheel wear feature set and the sample cooling performance reduction ratio set, and obtain the secondary compensation coefficient set under different sample grinding wheel wear features and different sample cooling performance reduction ratio conditions, and construct multiple sample secondary compensation coefficient sets.

[0095] The sample dataset construction unit constructs a sample dataset based on the sample grinding wheel wear feature set, the sample cooling performance degradation ratio set, and multiple sample secondary compensation coefficient sets, and equally divides the sample dataset into M datasets.

[0096] The convergent quadratic compensation analysis branch acquisition unit is used to perform supervised training on the BP network using the M datasets respectively, and when it tends to converge, to perform cross-validation on the BP network by interacting with the M datasets to obtain M convergent quadratic compensation analysis branches.

[0097] A secondary compensation analysis plugin construction unit is provided, which integrates and constructs the secondary compensation analysis plugin based on the M convergent secondary compensation analysis branches, wherein the output of the secondary compensation analysis plugin is the average of the output results of the M convergent secondary compensation analysis branches.

[0098] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0099] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0100] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A CNC grinding machine intelligent control method combining real-time feedback, characterized in that, The methods include: Collect historical machining records of several CNC grinding machines in the target factory, divide the CNC grinding machines according to the historical machining records, determine multiple grinding machine domains of the same type, and perform machining deviation analysis on the multiple grinding machine domains of the same type to obtain multiple sets of static compensation coefficients; The initial processing parameters are determined based on the matching of product characteristics. The initial processing parameters are compensated once based on the multiple sets of static compensation coefficients. Several CNC grinding machines are controlled to perform grinding processing according to the set of compensation parameters. At the predetermined adjustment time point, the primary compensation parameter set is compensated at fixed points according to the secondary compensation scheme to obtain the secondary compensation parameter set, wherein the secondary compensation is used to compensate for the impact of grinding wheel wear and cooling performance degradation. Based on the set of secondary compensation parameters, several CNC grinding machines are controlled to perform grinding operations until the equipment stops continuous operation or the grinding wheel accessories are replaced. The secondary compensation scheme includes: Using the product characteristics as constraints, query historical processing records to obtain the average values ​​of multiple grinding wheel wear characteristics under multiple continuous operation nodes during continuous operation of several CNC grinding machines, as well as the average values ​​of multiple cooling performance reduction ratios under multiple continuous operation nodes. The average values ​​of multiple grinding wheel wear characteristics and the average values ​​of multiple cooling performance reduction ratios are combined according to the operation nodes to obtain multiple sets of input data, wherein each set of input data corresponds to one operation node; The multiple sets of input data are input into the secondary compensation analysis plugin for analysis, and multiple predetermined secondary compensation coefficient sets are output. Establish a mapping relationship between multiple work nodes and multiple predetermined sets of secondary compensation coefficients, and construct the secondary compensation scheme based on the mapping relationship.

2. The intelligent control method for CNC grinding machines combining real-time feedback according to claim 1, characterized in that, The method also includes: At predetermined monitoring time points, parameter fluctuation ranges and product deviations of several CNC grinding machines are collected, and parameter optimization analysis is performed based on the parameter fluctuation ranges and product deviations to determine several sets of triple compensation coefficients. The set of secondary compensation parameters is dynamically fine-tuned based on the aforementioned triple compensation coefficients to obtain several optimized processing parameters, thereby controlling several CNC grinding machines to perform grinding operations.

3. The intelligent control method for CNC grinding machines combined with real-time feedback according to claim 2, characterized in that, Based on the parameter fluctuation range and the deviation of the processed product, parameter optimization analysis is performed, including: Query historical processing records, collect sample parameter fluctuation range set, sample processed product deviation set, and multiple sample parameter compensation coefficient set. Among them, the processing parameters include grinding speed, workpiece rotation speed, feed rate, depth of cut and feed rate, and each processing parameter corresponds to a parameter compensation coefficient. Using the sample parameter fluctuation range and sample processed product deviation as inputs, and supervised by the sample parameter compensation coefficient, a machine learning model is trained using the sample parameter fluctuation range set, the sample processed product deviation set, and multiple sample parameter compensation coefficient sets to obtain a parameter optimization analyzer. Using the parameter optimization analyzer, parameter optimization analysis is performed based on the parameter fluctuation range and the deviation of the processed product.

4. The intelligent control method for CNC grinding machines combined with real-time feedback according to claim 3, characterized in that, Based on the historical machining records, several CNC grinding machines are divided into multiple grinding machine domains of the same type, including: Based on the historical processing records, the average values ​​of multiple dimensional features and multiple average surface roughness of multiple processed products from several CNC grinding machines within the historical time window are obtained; Using the same processed products as a constraint, several CNC grinding machines are clustered multiple times based on the average values ​​of multiple dimensional features and multiple average values ​​of surface roughness, resulting in multiple clustering results; The co-occurrence frequency of any two CNC grinding machines in the multiple clustering results is statistically analyzed, and any two CNC grinding machines with a frequency greater than a predetermined threshold are added to the same type of grinding machine domain to obtain the multiple same type of grinding machine domains.

5. The intelligent control method for CNC grinding machines combined with real-time feedback according to claim 4, characterized in that, Machining deviation analysis was performed on the multiple similar grinding machine domains to obtain multiple sets of static compensation coefficients, including: Randomly select a first type of grinding machine domain, and based on the historical processing records, obtain the average dimensional deviation and average surface roughness deviation of the first type of grinding machine domain within the historical time window; The average dimensional deviation and the average surface roughness deviation are input into the static compensation model, and the first static compensation coefficient set is output and added to the plurality of static compensation coefficient sets.

6. The intelligent control method for CNC grinding machines combining real-time feedback according to claim 1, characterized in that, The secondary compensation analysis plugin is constructed, including: Collect sample grinding wheel wear feature set and sample cooling performance reduction ratio set, and obtain secondary compensation coefficient set under different sample grinding wheel wear features and different sample cooling performance reduction ratio conditions to construct multiple sample secondary compensation coefficient sets; A sample dataset is constructed based on the sample grinding wheel wear feature set, the sample cooling performance degradation ratio set, and multiple sample secondary compensation coefficient sets, and the sample dataset is equally divided into M datasets. Using the M datasets, supervised training of the BP network is performed respectively. When the network tends to converge, the M datasets are interacted to perform cross-validation on the BP network, resulting in M ​​convergent quadratic compensation analysis branches. The quadratic compensation analysis plugin is constructed based on the M convergent quadratic compensation analysis branches, wherein the output of the quadratic compensation analysis plugin is the average of the output results of the M convergent quadratic compensation analysis branches.

7. A CNC grinding machine intelligent control device incorporating real-time feedback, characterized in that: The device is used to execute the intelligent control method for CNC grinding machines combined with real-time feedback as described in any one of claims 1-6, and the device comprises: The static compensation coefficient set acquisition module is used to collect historical processing records of several CNC grinding machines in the target factory, divide the several CNC grinding machines according to the historical processing records, determine multiple grinding machine domains of the same type, and perform processing deviation analysis on the multiple grinding machine domains of the same type to obtain multiple static compensation coefficient sets. A grinding process control module is used to determine initial processing parameters based on product feature matching, perform a first compensation on the initial processing parameters based on the multiple static compensation coefficient sets, and control several CNC grinding machines to perform grinding processes according to the first compensation parameter set. The secondary compensation parameter set acquisition module is used to perform fixed-point compensation on the primary compensation parameter set according to the secondary compensation scheme at a predetermined adjustment time node to obtain the secondary compensation parameter set. The secondary compensation is used to compensate for the effects of grinding wheel wear and cooling performance degradation. The secondary compensation parameter set control module is used to control several CNC grinding machines to perform grinding operations according to the secondary compensation parameter set until the equipment stops continuous operation or the grinding wheel accessories are replaced.

Images