Grinding wheel wear degree identification method and device based on adaptive neural fuzzy reasoning

Abstract

The invention discloses a grinding wheel wear degree identification method and device based on adaptive neural fuzzy reasoning. The method comprises the steps that a grinding temperature signal, grinding force and the surface quality of a workpiece are obtained; determining the abrasion stage of the grinding wheel according to the surface quality, the grinding force and the grinding temperature; extracting the total number of thermal pulses in the grinding temperature signal as the number of dynamic effective abrasive particles; a grinding wheel surface topography graph is obtained, abrasive dust and abrasive particle abrasion in the grinding wheel surface topography graph are recognized and distinguished through an HSV segmentation method, the area rate is calculated, and the abrasive dust proportion and the abrasive particle abrasion proportion are obtained; and the abrasive dust proportion, the abrasive particle abrasion proportion and the dynamic effective abrasive particle number serve as input, the abrasion stage serves as output, the self-adaptive neural fuzzy inference system is trained, and the trained self-adaptive neural fuzzy inference system is obtained for abrasion degree recognition. According to the method, the abrasion condition of the grinding wheel can be accurately and conveniently judged by utilizing the self-adaptive neural fuzzy reasoning technology and synthesizing the multi-dimensional characteristic signals in the grinding process.

Description

Technical Field

[0001] The present invention relates to the field of precision machining, and particularly to a method and device for identifying the wear degree of a grinding wheel based on adaptive neuro-fuzzy inference. Background Art

[0002] In modern manufacturing, the requirements for surface quality, geometric accuracy, and machining efficiency in precision machining processes are getting higher and higher, especially in the fields of aerospace, automotive, and mold.

[0003] Diamond grinding wheels are widely used in machining carbide, ceramic materials, and other difficult-to-machine materials. Grinding wheel grinding has the characteristics of high speed and high efficiency, but it is also accompanied by continuous wear of the grinding wheel itself. This wear is manifested in the form of grain shedding, blunting, and chip clogging, etc., which causes the grinding wheel to gradually lose its original machining ability. If the wear degree of the grinding wheel cannot be monitored in time, it will not only affect the machining quality, but even affect the dimensional accuracy and surface integrity of the workpiece. Therefore, realizing the identification of the wear state of the grinding wheel to determine when it needs to be dressed or replaced is of great significance for ensuring production efficiency and product quality.

[0004] Currently, grinding wheel wear detection technologies can generally be divided into two categories: direct detection and indirect detection. Direct detection technologies mainly directly judge the wear state by measuring the wear amount or topography change on the surface of the grinding wheel (such as the stylus method, the imprint method, the optical section method, and the electron microscope observation method). Although these methods are intuitive, they require high-precision detection equipment and usually need to stop the machine for detection, so it is difficult to meet the requirements of on-line real-time detection. Indirect monitoring technologies indirectly infer the wear state of the grinding wheel by analyzing various physical quantities (such as acoustic emission signals, vibration signals, current, voltage, temperature, etc.) during the grinding process. This method can realize on-line monitoring, but due to the complexity of the grinding process, the measurement of a single physical quantity often cannot comprehensively and accurately reflect the wear condition of the grinding wheel. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method and device for identifying the wear degree of a grinding wheel based on adaptive neuro-fuzzy inference to improve the above problems.

[0006] The present invention provides a method for identifying the wear degree of a grinding wheel based on adaptive neuro-fuzzy inference, which includes:

[0007] S101, during the continuous grinding of the workpiece by the grinding wheel, obtain the grinding temperature signal, grinding force, and surface quality of the workpiece in each grinding stage;

[0008] S102, determine the wear stage of the grinding wheel according to the surface quality, grinding force, and grinding temperature of the workpiece;

[0009] S103. Extract the total number of thermal pulses in the grinding temperature signal as the number of dynamically effective abrasive grains.

[0010] S104. Obtain the surface topography map of the grinding wheel after each grinding. Identify and distinguish the grinding debris and abrasive grain wear in the surface topography map of the grinding wheel through the HSV segmentation method and calculate the area ratio to obtain the proportion of grinding debris and the proportion of abrasive grain wear.

[0011] S105. Use the proportion of grinding debris, the proportion of abrasive grain wear, and the number of dynamically effective abrasive grains as inputs and the wear stage as the output to train the constructed adaptive neuro-fuzzy inference system to obtain the trained adaptive neuro-fuzzy inference system.

[0012] S106. Identify the wear degree of the grinding wheel according to the trained adaptive neuro-fuzzy inference system.

[0013] Preferably, in step S101, a k-type thermocouple is used to measure the grinding temperature signal and the grinding force, and the k-type thermocouple is polished to 4 - 10 microns.

[0014] Preferably, in step S102, the wear stage of the grinding wheel is divided into the initial stage, the middle stage, the severe stage, and the blunt stage.

[0015] Preferably, step S103 includes:

[0016] Obtain the range of the noise signal in the grinding temperature signal to obtain the noise threshold.

[0017] Use the thermal pulse counting program to analyze the temperature signals of multiple groups in the same wear stage, and statistically calculate the average value of the total number of thermal pulses greater than the noise threshold in the grinding temperature signal as the number of dynamically effective abrasive grains.

[0018] Preferably, step S104 includes:

[0019] Uniformly divide multiple shooting areas in the circumferential direction of the grinding wheel, divide three shooting areas in the front, middle, and rear in the axial direction, and take multiple surface topography maps of the grinding wheel by shooting in the shooting areas through a camera.

[0020] Convert the surface topography map of the grinding wheel from the RGB color space to the HSV color space to obtain an HSV image.

[0021] Adopt color feature extraction technology to obtain the H, S, and V channel information of the HSV image and perform histogram analysis on each channel.

[0022] Determine the threshold ranges of grinding debris blockage and abrasive grain wear in the H, S, and V channels according to the results of the histogram analysis, and create a mask according to the threshold ranges to mark the pixel points within the specified HSV value range on the HSV image.

[0023] Filter pixels of the HSV image using the said mask to obtain the total number of pixels N1 and N2 of the chip blockage and abrasive wear areas;

[0024] Calculate the chip proportion based on the total number of pixels N of the HSV image and the abrasive wear proportion as

[0025] Preferably, step S105 includes:

[0026] Generate a sample data set based on the chip proportion, abrasive wear proportion and dynamic effective abrasive number, and randomly divide the sample data set into independent training data set and test data set;

[0027] Define a fuzzy set: Set 2 to 8 types of initial membership function types to be determined and 2 to 5 numbers of fuzzy subsets to be determined;

[0028] Define the fuzzy system structure: Generate a fuzzy inference system using the grid segmentation method;

[0029] Model training: Set the number of iterations, and use a hybrid neural network algorithm of backpropagation and least squares algorithm to train the fuzzy inference system, so that the output of the fuzzy inference system continuously approaches the training result of the training data set. When the preset conditions are met, the training stops to obtain an adaptive neural network fuzzy inference system model.

[0030] Preferably, it further includes:

[0031] Model optimization: Determine the membership function and the number of fuzzy subsets according to the root mean square error and the number of iterations in training and testing, where whether the training is overfitting is determined by cross-validation;

[0032] The calculation formulas for the root mean square error and cross-validation are as follows:

[0033] Suppose there are k subsets, each subset has a size of n i , and there are a total of n samples, where

[0034]

[0035] For the i-th subset, let its prediction result be Then the root mean square error:

[0036]

[0037] where y ij is the actual value of the j-th sample in the i-th subset, is the predicted value of the j-th sample in the i-th subset;

[0038] Finally, calculate the average RMSE:

[0039]

[0040] In this way, the training error of the model and the average root mean square error of cross - validation are obtained. When the training error is less than 5%, the training is stopped. If the difference between the average root mean square error of cross - validation and the training error is within 2.5%, it indicates that the model fits well; otherwise, it is over - fitting.

[0041] An embodiment of the present invention also provides a device for identifying the wear degree of a grinding wheel based on adaptive neuro - fuzzy inference, which includes:

[0042] An acquisition unit, configured to acquire the grinding temperature signal, grinding force, and surface quality of the workpiece in each grinding stage during the continuous grinding of the workpiece by the grinding wheel;

[0043] A wear stage determination unit, configured to determine the wear stage of the grinding wheel according to the surface quality, grinding force, and grinding temperature of the workpiece;

[0044] An extraction unit, configured to extract the total number of thermal pulses in the grinding temperature signal as the number of dynamically effective abrasive grains;

[0045] A proportion acquisition unit, configured to acquire the surface topography map of the grinding wheel after each grinding, identify and distinguish the grinding chips and abrasive grain wear in the surface topography map of the grinding wheel by the HSV segmentation method, calculate the area ratio, and obtain the proportion of grinding chips and the proportion of abrasive grain wear;

[0046] A training unit, configured to use the proportion of grinding chips, the proportion of abrasive grain wear, and the number of dynamically effective abrasive grains as inputs, and the wear stage as the output, to train the constructed adaptive neuro - fuzzy inference system, and obtain the trained adaptive neuro - fuzzy inference system;

[0047] An identification unit, configured to identify the wear degree of the grinding wheel according to the trained adaptive neuro - fuzzy inference system.

[0048] In summary, the present invention uses computer image processing and adaptive neuro - fuzzy inference technology, comprehensively analyzes the multi - dimensional characteristic signals (grinding wheel surface state, number of dynamically effective abrasive grains) in the grinding process to analyze the wear of the grinding wheel, and can accurately and conveniently judge the wear condition of the grinding wheel without disassembling the grinding wheel. Brief Description of the Drawings

[0049] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0050] Figure 1Schematic flow chart of the grinding wheel wear degree identification method based on adaptive neuro-fuzzy inference provided by the first embodiment of the present invention;

[0051] Figure 2 Flow chart of the thermal pulse counting program provided by the embodiment of the present invention;

[0052] Figure 3 Schematic diagram of the grinding wheel surface image acquisition area provided by the embodiment;

[0053] Figure 4 Fixed-point grinding wheel wear diagram under different material removal volumes V provided by the embodiment;

[0054] Figure 5 Schematic flow chart of the adaptive neuro-fuzzy inference system provided by the embodiment;

[0055] Figure 6(a) and Figure 6(b) are the structural framework diagrams of the adaptive neuro-fuzzy inference system of the present invention;

[0056] Figure 7 Schematic structural diagram of the grinding wheel wear degree identification device based on adaptive neuro-fuzzy inference provided by the second embodiment of the present invention. Detailed implementation manners

[0057] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0058] Please refer to Figure 1 , the embodiment of the present invention provides a grinding wheel wear degree identification method based on adaptive neuro-fuzzy inference, which includes:

[0059] S101, during the continuous grinding of the workpiece by the grinding wheel, obtain the grinding temperature signal, grinding force, and surface quality of the workpiece in each grinding stage.

[0060] Specifically, in this embodiment, a wear experiment of the grinding wheel grinding the workpiece can be carried out, and the grinding temperature, grinding force, and surface quality of the workpiece in each stage during the wear process of the diamond grinding wheel are recorded.

[0061] Among them, a k-type thermocouple can be used to measure the grinding temperature and grinding force. To reduce the measurement error, the thickness of the thermocouple needs to be polished to 4 - 10 microns. The recorded grinding temperature signal and grinding force signal are filtered in dewesoft software and then the data is exported.

[0062] S102. Determine the wear stage of the grinding wheel based on the surface quality, grinding force, and grinding temperature of the workpiece.

[0063] Among them, the wear stage of the grinding wheel in different grinding stages can be divided according to the change of the grinding force. For example, it can be divided into the initial stage, the middle stage, the severe stage, and the dull stage. Of course, the wear stage can also be set according to actual needs, and the present invention does not make specific limitations.

[0064] S103. Extract the total number of thermal pulses in the grinding temperature signal as the dynamic effective grain number.

[0065] Specifically, as Figure 2 shown, the dynamic effective grain number can be statistically calculated through a thermal pulse counting program:

[0066] First, import the grinding temperature signal into the program in the form of data. Each row is a data point, with a total of i rows, and each column is the number of grinding temperature signal samples, with a total of j columns. Traverse each column of data through nested for loops, subtract adjacent data points to obtain the data fluctuation value Td, then save the values of Td greater than the noise threshold into the Wt array and count. Finally, store the total number of thermal pulses of c samples in the form of 1 row and c columns into the Pt array and output. The average value of the total number of thermal pulses of c samples will be used as the dynamic effective grain number of the grinding arc area passing through the thermocouple.

[0067] S104. Obtain the surface topography map of the grinding wheel after each grinding, identify and distinguish the grinding chips and abrasive wear in the surface topography map of the grinding wheel through the HSV segmentation method, and calculate the area ratio to obtain the proportion of grinding chips and the proportion of abrasive wear.

[0068] Specifically, S104 includes:

[0069] First, evenly divide multiple shooting areas in the circumferential direction of the grinding wheel, divide the front, middle, and rear shooting areas in the axial direction, and obtain multiple surface topography maps of the grinding wheel by shooting in the shooting areas through a camera.

[0070] As Figure 3 shown, evenly divide 6 - 20 shooting areas in the circumferential direction of the grinding wheel, divide it into the front, middle, and rear shooting areas in the axial direction, and make marks at 6 acquisition points with a marker pen for fixed-point shooting. Figure 4 It is the fixed-point grinding wheel wear degree effect diagram of the grinding wheel under different material removal amounts V.

[0071] Then, convert the surface topography map of the grinding wheel from the RGB color space to the HSV color space to obtain the HSV image.

[0072] Among them, the photographed surface topography map of the grinding wheel is imported into the computer, and then in an image processing software, such as matlab, the imread function is used to read the image, and the rgb2hsv function is used to convert the original image from RGB to the HSV color space to obtain the HSV image.

[0073] Next, according to the results of the histogram analysis, the threshold ranges of chip clogging and abrasive grain wear in the H, S, and V channels are determined, and a mask is created according to the threshold ranges to mark the pixel points within the specified HSV value range on the HSV image.

[0074] Among them, specifically, the hsvImage function is used to obtain the H, S, and V channel information of the original image through the indexes (:,:,1), (:,:,2), and (:,:,3) respectively, and then the imshow function is used to display the images of each channel to perform histogram analysis on each channel. Through these analyses, the threshold ranges of chips and abrasive grains can be determined, so as to create a mask to mark the qualified pixels and perform pixel screening on the original HSV image.

[0075] Then, the mask is applied to perform pixel screening on the HSV image to obtain the total number of pixels N1 and N2 of the chip clogging and abrasive grain wear areas;

[0076] Finally, according to the total number of pixels N of the HSV image, the chip proportion is calculated and the abrasive grain wear proportion is

[0077]

[0078] S105, taking the chip proportion, the abrasive grain wear proportion, and the dynamic effective abrasive grain number as inputs and the wear stage as the output, training the constructed adaptive neuro-fuzzy inference system to obtain the trained adaptive neuro-fuzzy inference system.

[0079] As Figure 5 shown, in this embodiment, after the data preparation is complete, the construction of the adaptive neuro-fuzzy inference system can be carried out. Among them, the adaptive neuro-fuzzy inference system (ANFIS) combines the self-learning ability of the neural network and the inference ability of fuzzy logic, can establish a relatively accurate model in a complex non-linear system, and has strong fault tolerance and robustness. The construction process specifically includes:

[0080] First, the previously obtained data set (chip clogging proportion, abrasive grain wear proportion, as inputs, and the grinding wheel wear degree as the output) is divided into a training set and a test set;

[0081] Select the type of membership function (MF) under the grid segmentation method, such as trimf (MF1), trapmf (MF2), gbellmf1 (MF3), gaussmf (MF4), etc., and input the number of MFs (i.e., the number of fuzzy subsets corresponding to the input) to establish the FIS system and perform fuzzy processing on the input data.

[0082] After that, select the hybrid neural network algorithm of backpropagation and least squares to train the model, and select 40 for the number of iterations. In this process, the membership function parameters and rule parameters in the system will be automatically optimized. When the model output continuously approaches and reaches the preset condition (in this embodiment, the error is less than 5%), the training stops. Since the optimal number of iterations is different due to the different types of membership functions and the number of fuzzy subsets, and the number of iterations is related to the running time of the program, it can be used as a reference quantity for selecting the optimal model.

[0083] As shown in FIGS. 6(a) and 6(b), the ANFIS structure used in this embodiment consists of five layers: a fuzzification layer, a basic rule layer, a normalization layer, a defuzzification layer, and an output layer. The number of fuzzy rules is obtained by the product of the number of fuzzy subsets of each input. For example, in FIG. 6(a), the fuzzy rules are 2×2×2 = 8 (3 input variables, each variable has 2 fuzzy subsets); while in FIG. 6(b), the fuzzy rules are 3×3×3 = 27 (3 input variables, each variable has 3 fuzzy subsets).

[0084] The form of the fuzzy rule is: if x is A and y is B, then z is C, where x and y are input variables, z is the output variable, and A, B, and C are different fuzzy sets;

[0085] During the modeling process, the number of fuzzy subsets of each input variable can be freely set according to needs. However, considering the running speed and modeling requirements, the number of MFs corresponding to each input parameter in this case is selected in the range of 2 to 5.

[0086] Since the parameters of the input variables are fixed, the quality of the model mainly depends on the type of common MFs (4 types) and the number of MFs corresponding to each input parameter (4 cases of 2 to 5 are selected in this article). For the convenience of model screening, in this embodiment, the root mean square error (RMSE) obtained after the operation of the ANFIS system is used for the initial screening of the model to measure the accuracy of the prediction. The smaller the RMSE value, the higher the prediction accuracy. To evaluate the stability and generalization ability of the model and avoid the situation of overfitting of the model on a specific training set, k-fold cross-validation is introduced. Through multiple trainings and evaluations, the performance of the model can be more comprehensively understood.

[0087] Among them, the calculation formulas for the root mean square error and cross-validation are as follows:

[0088] Suppose there are k subsets, and the size of each subset is n i , and there are a total of n samples, among which

[0089]

[0090] For the i-th subset, let its prediction result be Then the root mean square error:

[0091]

[0092] where y ij is the actual value of the j-th sample in the i-th subset, is the predicted value of the j-th sample in the i-th subset;

[0093] Finally, calculate the average RMSE:

[0094]

[0095] By training the model with different MF types and MF quantities, the data in Table 1 is obtained in this embodiment. First, control the root mean square errors of training, testing, and validation within 10%. If it is greater than 10%, this parameter is excluded. Then, compare the average root mean square error of cross-validation. When the average root mean square error is more than 2.5% greater than the training error, the model is considered overfitted. Finally, compare the validation error and the number of iterations. The smaller the validation error and the number of iterations, the higher the prediction accuracy of the model and the shorter the prediction time. Through a series of judgments, the optimal parameters of the model in this case are: the membership function selects the trapezoidal membership function, and the number of fuzzy subsets is 4.

[0096] Table 1 Comparison of RMSE, cross-validation, and number of iterations under different MF types and MF quantities

[0097]

[0098] S106. Identify the degree of grinding wheel wear according to the trained adaptive neuro-fuzzy inference system.

[0099] In this embodiment, according to the obtained adaptive neuro-fuzzy inference system, the input predicts the dynamically effective abrasive grain number collected from the grinding wheel, the proportion of grinding chips, and the proportion of abrasive grain wear after image processing. The degree of wear of the grinding wheel can be obtained through the output of the system.

[0100] In summary, the present invention utilizes computer image processing and adaptive neuro-fuzzy inference technology to comprehensively analyze the multi-dimensional characteristic signals (grinding wheel surface state, dynamic effective abrasive number) during the grinding process to determine the wear of the grinding wheel, and can accurately and conveniently judge the wear condition of the grinding wheel without disassembling the grinding wheel.

[0101] Please refer to Figure 7 , the second embodiment of the present invention further provides a device for identifying the wear degree of a grinding wheel based on adaptive neuro-fuzzy inference, which includes:

[0102] An acquisition unit 210, configured to acquire the grinding temperature signal, grinding force, and surface quality of the workpiece during each grinding stage during the continuous grinding of the workpiece by the grinding wheel;

[0103] A wear stage determination unit 220, configured to determine the wear stage of the grinding wheel according to the surface quality, grinding force, and grinding temperature of the workpiece;

[0104] An extraction unit 230, configured to extract the total number of heat pulses in the grinding temperature signal as the dynamic effective abrasive number;

[0105] A ratio acquisition unit 240, configured to acquire the surface topography map of the grinding wheel after each grinding, identify and distinguish the grinding debris and abrasive wear in the surface topography map of the grinding wheel by the HSV segmentation method, calculate the area ratio, and obtain the ratio of grinding debris and the ratio of abrasive wear;

[0106] A training unit 250, configured to use the ratio of grinding debris, the ratio of abrasive wear, and the dynamic effective abrasive number as inputs and the wear stage as the output to train the constructed adaptive neuro-fuzzy inference system to obtain a trained adaptive neuro-fuzzy inference system;

[0107] An identification unit 260, configured to identify the wear degree of the grinding wheel according to the trained adaptive neuro-fuzzy inference system.

[0108] The third embodiment of the present invention further provides a device for identifying the wear degree of a grinding wheel based on adaptive neuro-fuzzy inference, which includes a memory and a processor, and a computer program is stored in the memory, and the computer program can be executed by the processor to implement the method for identifying the wear degree of a grinding wheel based on adaptive neuro-fuzzy inference as described above.

[0109] In several embodiments provided by the embodiments of the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0110] In addition, each functional module in various embodiments of the present invention may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0111] If the described function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, an electronic device, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes. It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article, or device. Without further limitations, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article, or device including the said element.

[0112] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for identifying the degree of grinding wheel wear based on adaptive neural fuzzy reasoning, characterized in that: include: S101, during the process of continuous grinding of the workpiece by the grinding wheel, obtaining a grinding temperature signal, a grinding force and a surface quality of the workpiece at each grinding stage; S102, determining the wear stage of the grinding wheel according to the surface quality of the workpiece, the grinding force and the grinding temperature signal; S103, extracting the total number of heat pulses in the grinding temperature signal as the number of dynamic effective abrasive particles; S104, obtaining a grinding wheel surface topography after each grinding, identifying and distinguishing the wear debris and abrasive wear in the grinding wheel surface topography by HSV segmentation method, and calculating the area ratio to obtain the wear debris ratio and abrasive wear ratio; S105, taking the wear debris ratio, abrasive wear ratio and dynamic effective abrasive number as inputs and the wear stage as outputs, training the constructed adaptive neural fuzzy inference system to obtain a trained adaptive neural fuzzy inference system; S106, identifying the degree of grinding wheel wear according to the trained adaptive neural fuzzy inference system.

2. The method for identifying the degree of grinding wheel wear based on adaptive neural fuzzy reasoning according to claim 1 is characterized in that: In step S101, a K-type thermocouple is used to measure the grinding temperature signal and the grinding force, and the K-type thermocouple is ground to 4-10 microns.

3. The method for identifying the degree of grinding wheel wear based on adaptive neural fuzzy reasoning according to claim 1 is characterized in that: In step S102 , the wear stages of the grinding wheel are divided into initial, middle, severe and blunt stages.

4. The method for identifying the degree of grinding wheel wear based on adaptive neural fuzzy reasoning according to claim 1 is characterized in that: Step S103 includes: Obtain the range of the noise signal in the grinding temperature signal and obtain the noise threshold; The thermal pulse counting program was used to analyze multiple groups of temperature signals at the same wear stage, and the average value of the total number of thermal pulses greater than the noise threshold in the grinding temperature signal was counted as the dynamic effective abrasive number.

5. The method for identifying the degree of grinding wheel wear based on adaptive neural fuzzy reasoning according to claim 1 is characterized in that: Step S104 includes: The grinding wheel is evenly divided into multiple shooting areas in the circumferential direction, and three shooting areas, namely, front, middle and rear, are divided in the axial direction. A camera is used to shoot in the shooting areas to obtain multiple grinding wheel surface topography images; Convert the grinding wheel surface topography image from RGB color space to HSV color space to obtain an HSV image; The color feature extraction technology is used to obtain the H, S, and V channel information of the HSV image, and the histogram analysis is performed on each channel; Determine the threshold ranges of wear debris blockage and abrasive wear of H, S, and V channels based on the results of histogram analysis, and create masks based on the threshold ranges to mark the pixels on the HSV image that are within the specified HSV value range; Applying the mask to perform pixel screening on the HSV image to obtain the total pixels N1 and N2 of the wear debris clogging and abrasive wear areas; According to the total number of pixels N of the HSV image, the wear debris ratio is calculated and abrasive wear ratio is 6. The method for identifying the degree of grinding wheel wear based on adaptive neural fuzzy reasoning according to claim 5 is characterized in that: Step S105 includes: Generate a sample data set according to the wear debris ratio, abrasive wear ratio and dynamic effective abrasive number, and randomly divide the sample data set into an independent training data set and a test data set; Define fuzzy sets: set 2 to 8 types of initial membership functions to be determined and 2 to 5 numbers of fuzzy subsets to be determined; Define the fuzzy system structure: Use the grid partitioning method to generate the fuzzy inference system; Model training: setting the number of iterations, using a hybrid neural network algorithm of back propagation and least squares algorithm to train the fuzzy inference system, so that the output of the fuzzy inference system is constantly close to the training result of the training data set, and the training stops when the preset conditions are met to obtain an adaptive neural network fuzzy inference system model.

7. The method for identifying the degree of grinding wheel wear based on adaptive neural fuzzy reasoning according to claim 6 is characterized in that: Also includes: Model optimization: The number of membership functions and fuzzy subsets is determined based on the root mean square error and the number of iterations in training and testing. Whether the training is overfitting is determined by cross-validation. The root mean square error and cross validation calculation formula are as follows: Assume there are k subsets, each of size n i , there are n samples in total, among which For the i-th subset, let its prediction result be Then the root mean square error is: where y ij is the actual value of the jth sample in the ith subset, is the predicted value of the jth sample in the i-th subset; Finally, calculate the average RMSE: In this way, we can get the training error of the model and the average root mean square error of cross validation. When the training error is less than 5%, the training is stopped. If the difference between the average root mean square error of cross validation and the training error is within 2.5%, it means that the model fits well, otherwise it is overfitting.

8. A grinding wheel wear degree identification device based on adaptive neural fuzzy reasoning, characterized in that: include: An acquisition unit is used to acquire the grinding temperature signal, grinding force and surface quality of the workpiece at each grinding stage during the continuous grinding of the workpiece by the grinding wheel; A wear stage determination unit, used for determining the wear stage of the grinding wheel according to the surface quality of the workpiece, the grinding force and the grinding temperature signal; An extraction unit, used for extracting the total number of heat pulses in the grinding temperature signal as the number of dynamic effective abrasive particles; The proportion acquisition unit is used to obtain the grinding wheel surface topography after each grinding, identify and distinguish the wear debris and abrasive wear in the grinding wheel surface topography by HSV segmentation method, calculate the area ratio, and obtain the wear debris proportion and abrasive wear proportion; A training unit is used to train the constructed adaptive neural fuzzy inference system by taking the wear debris ratio, abrasive wear ratio and dynamic effective abrasive grain number as input and the wear stage as output, so as to obtain a trained adaptive neural fuzzy inference system; The recognition unit is used to recognize the degree of grinding wheel wear based on the trained adaptive neuro-fuzzy inference system.

Images