Intelligent adjustment system and method for tool processing machine tool operation status based on cloud computing

Abstract

This invention relates to the technical field of operating status control systems, and discloses an intelligent adjustment system and method for the operating status of a tool-making machine tool based on cloud computing. The invention first uses wavelet analysis to extract features from the vibration signal of the tool-making machine tool, followed by dimensionality reduction processing. A BP neural network model is then used to identify whether the spindle vibration signal of the tool-making machine tool is abnormal, denoted as the first adjustment factor. Next, the digital instrument image of the tool-making machine tool is processed using the mean method and digital segmentation based on image recognition to identify the numbers on the digital instrument image, denoted as the second adjustment factor. Then, a Hidden Markov Model is constructed to detect the wear of the machining tools on the tool-making machine tool, denoted as the third adjustment factor. Finally, a support vector machine model is used to predict the first, second, and third adjustment factors, thereby achieving intelligent adjustment of the operating status of the tool-making machine tool.

Description

Technical Field

[0001] This invention relates to the technical field of operating status control systems, specifically to a cloud computing-based intelligent adjustment system and method for the operating status of cutting tool machine tools. Background Technology

[0002] Chinese invention patent application CN106292535A discloses a CNC machine tool machining system and a cloud-based CNC machine tool machining system. Specifically, it includes: interconnecting the data of CNC machine tools to achieve data communication on a machine tool simulation platform; sending the workpiece to the CNC machine tool via a machining program; the CNC machine tool machining the workpiece and displaying various information parameters during the machining process on the machine tool simulation platform; simulating the workpiece on the machine tool simulation platform and simultaneously uploading various information parameters during the CNC machine tool machining process to the cloud platform; the cloud platform monitors in real time whether any abnormalities occur during the CNC machine tool machining process; if the CNC machine tool successfully completes the machining without any abnormalities; if the CNC machine tool fails to complete the machining and an abnormality occurs, the cloud platform generates a machining error report and stops the machining task. This system uses simulation to reduce losses caused by machining errors, connects all CNC machine tools, and uploads operating data to the cloud platform, enabling real-time monitoring of the CNC machine tool machining status.

[0003] During the machining process of cutting tools, tool breakage can easily cause damage to the machining equipment and lead to machining safety accidents. Furthermore, traditional tool cutting machine tool condition monitoring mainly relies on experienced machining personnel to subjectively estimate the machine tool's operating status based on factors such as machining noise, cutting vibration, and machining time. This method cannot achieve real-time monitoring of the machine tool's operating status or adjust the relevant operating parameters in real time. Summary of the Invention

[0004] To address the problems in related technologies, this invention provides a cloud computing-based intelligent adjustment system and method for the operating status of cutting tool processing machine tools, thereby overcoming the aforementioned technical problems existing in the prior art.

[0005] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] This invention relates to a cloud computing-based intelligent adjustment method for the operating status of machine tools used in cutting tool processing, comprising the following steps:

[0007] S1. Obtain the vibration signal of the spindle of the tool processing machine tool, use wavelet analysis to extract features from the vibration signal of the tool processing machine tool, then use principal component analysis to reduce the dimension of the extracted vibration signal, and use BP neural network to complete the identification of features, which is denoted as the first adjustment factor;

[0008] S2. Perform image recognition processing on the digital instrument image of the cutting tool machine tool. Preprocess the digital instrument image of the cutting tool machine tool. The pixels of the digital instrument image of the cutting tool machine tool are processed by the mean method and digital segmentation to identify the numbers on the digital instrument image and record them as the second adjustment factor.

[0009] S3. Obtain the wear value of the machining tools of the cutting tool processing machine, construct a hidden Markov model to detect the wear of the machining tools of the cutting tool processing machine, realize the identification of the wear of the machining tools of the cutting tool processing machine, and record it as the third adjustment factor;

[0010] S4. Combining the first, second, and third adjustment factors, construct a sample set for the adjustment of the tool processing machine tool, and use a support vector machine model to predict the first, second, and third adjustment factors to achieve intelligent adjustment of the operating status of the tool processing machine tool.

[0011] This invention collects vibration signals from the spindle of a cutting tool machining center, extracts features from these signals using wavelet analysis, introduces decentralization processing, and then uses principal component analysis to reduce the dimensionality of the extracted vibration signals. A BP neural network model is used to identify whether the vibration signals from the cutting tool spindle are abnormal. Next, the digital instrument images of the cutting tool machining center are preprocessed, including locating the image positions and performing edge detection. The images are then binarized to facilitate digit recognition, and character segmentation is used to extract and identify the digits. Finally, a wear matrix is ​​set up to be processed, and a hidden matrix is ​​established for the wear values ​​in the matrix. The Markov model makes wear value recognition more flexible, enabling the detection of tool wear on cutting machine tools. It is relatively simple to use and has good recognition results. Finally, combining the first adjustment factor (machine tool spindle vibration signal value), the second adjustment factor (machine tool digital instrument value), and the third adjustment factor (machine tool wear value), a support vector machine model is used to combine these three adjustment factors and achieve predictive results. The kernel function transforms the data into high-dimensional features, making it suitable for classification and recognition of small samples, avoiding many limitations of neural networks.

[0012] Preferably, step S1 includes the following steps:

[0013] S11. Acquire the spindle vibration signal of the tool processing machine tool, and set the spindle vibration signal set as follows: ,in Indicates the first There are one spindle vibration signal, and the characteristic dimension of the spindle vibration signal is... The main shaft vibration signal matrix is ​​obtained. as follows,

[0014] ,

[0015] in, The first element in the spindle vibration signal matrix represents the... Each feature dimension is The main shaft vibration signal; the wavelet packet decomposition level is set to... The number of sub-bands is Then the node is The wavelet packet coefficients of the nodes are The characteristics of the spindle vibration signal are replaced with vibration energy characteristics. The calculation formula is as follows:

[0016]

[0017] Select one row from the spindle vibration signal matrix as the set of vibration signals to be processed. ,in The first element in the spindle vibration signal matrix represents the... Each feature dimension is The vibration energy of the spindle vibration signal in the set of vibration signals to be processed is calculated to obtain the vibration energy set. ,in The first element in the spindle vibration signal matrix represents the... Each feature dimension is The vibration energy of the spindle vibration signal; the relative entropy of the vibration energy is set as... Calculate the dissimilarity of the relative entropy of vibrational energy, denoted as . If the vibration energy in the vibration energy set is greater than the dissimilarity of the relative entropy of the vibration energy, the vibration energy in the vibration energy set is retained; otherwise, the vibration energy in the vibration energy set is replaced by the dissimilarity of the relative entropy of the vibration energy, so as to obtain the set of main shaft vibration signals after feature extraction and generate the matrix of main shaft vibration signals after feature extraction.

[0018] S12. The set of spindle vibration signals after feature extraction is decentered so that the mean value of the spindle vibration signals in the set is 0, resulting in a decentered set of spindle vibration signals. A decentered spindle vibration signal matrix is ​​generated, and the linear transformation matrix is ​​set as follows. The linear transformation matrix maximizes the covariance between the centered principal shaft vibration signal matrix and the linear transformation matrix. The eigenvalues ​​and eigenvectors of the linear transformation matrix are calculated and arranged in descending order to obtain the eigenvalue set and eigenvector set, respectively. The top eigenvalues ​​are then selected from the eigenvalue set. _i ... 1 eigenvector; select the first eigenvalues ​​from the eigenvalue set. The set of eigenvalues ​​and eigenvectors, with the first eigenvectors forming a dimensionality reduction matrix, is then multiplied by a linear transformation matrix to obtain the dimensionality-reduced eigenvectors, denoted as . ,in Indicates the first Each feature vector element;

[0019] S13. Set the input layer nodes of the BP neural network as... The learning rate is 0.01, and the expected error is... The activation function is set to the Sigmoid function; the dimensionality-reduced feature vector is standardized to obtain a standardized feature vector, and the standardized feature vector is selected from the... One element is used as the standardized training set, and other elements in the standardized feature vector are selected as the standardized test set; a maximum number of iterations is set. and the current iteration number The error threshold is The standardized training set is input into the BP neural network to train it. The training continues when the current iteration number is greater than the maximum iteration number or the expected error is less than... If the iteration stops, the trained BP neural network model is obtained; otherwise, the weights are adjusted until the current iteration number is greater than the maximum iteration number or the expected error is less than the maximum. ;

[0020] The standardized test set is input into the trained BP neural network model, and the results of the standardized test set are output. Based on the results of the standardized test set, it is determined whether the vibration signal of the machine tool spindle is abnormal. If the vibration signal of the machine tool spindle is abnormal, the vibration signal value of the machine tool spindle is used as the first adjustment factor; otherwise, a new vibration signal of the machine tool spindle is obtained and S11, S12 and S13 are repeated.

[0021] This invention collects vibration signals from the spindle of a cutting tool machining center, sets a spindle vibration signal matrix, uses wavelet analysis to extract features from the vibration signals, and replaces the spindle vibration signal features with vibration energy features, greatly reducing computational costs. After introducing decentralization processing, principal component analysis is used to reduce the dimensionality of the extracted vibration signals. A BP neural network model is used to identify whether the spindle vibration signals of the cutting tool machining center are abnormal. The processing method has high accuracy.

[0022] Preferably, step S2 includes the following steps:

[0023] S21. Acquire the digital instrument image of the tool processing machine tool, and preprocess the digital instrument image of the tool processing machine tool. The specific process is as follows:

[0024] S211. Obtain the first from the surveillance video. The digital instrument image of the frame tool machining machine tool is selected from the first frame. Corner points of the digital instrument image of the frame tool machining machine; for the first... The digital instrument image of the frame-by-frame cutting tool machining tool is rasterized to obtain a raster, with the first frame being the first raster. Using the lower left corner of the digital instrument image of the frame tool machining machine as the origin, a rectangular coordinate system is established to obtain the grid coordinates, thus obtaining the first... The corner grid coordinates of the digital instrument image of the frame-by-frame cutting tool machine tool; obtained from the surveillance video. The digital instrument images of the frame tool machining machine are compared with the first frame. The corner grid coordinates and the first frame of the digital instrument image of the tool machining machine. The corner grid coordinates of the digital instrument image of the frame tool machining machine are used. If the corner grid coordinates match, then the first [frame] is selected. Get a frame of digital instrumentation images of the tool-cutting machine tool; otherwise, get a new digital instrumentation image of the tool-cutting machine tool.

[0025] S212. Set the horizontal dimension of the digital instrument image of the new tool processing machine tool to... Vertical dimension is The edge detection matrix is ​​obtained. as follows,

[0026] ,

[0027] in, The horizontal dimension of the edge detection matrix is... Vertical dimension is The pixels;

[0028] Calculate the partial derivatives of the pixel values ​​of the pixels in the edge detection matrix in the horizontal and vertical directions, denoted as . and Length of edge detection vector The formula is as follows:

[0029]

[0030] One column of the edge detection matrix is ​​selected as the vertical detection set. The number of pixels in the vertical detection set whose pixel values ​​are greater than the length of the edge detection vector is counted. Other columns in the edge detection matrix are then selected as vertical detection sets. The two vertical detection sets corresponding to the columns with the highest number of pixels whose pixel values ​​are greater than the length of the edge detection vector are denoted as the vertical edges. One row of the edge detection matrix is ​​selected as the horizontal detection set. The number of pixels in the horizontal detection set whose pixel values ​​are greater than the length of the edge detection vector is counted. Other rows in the edge detection matrix are then selected as horizontal detection sets. The two horizontal detection sets corresponding to the rows with the highest number of pixels whose pixel values ​​are greater than the length of the edge detection vector are denoted as the horizontal edges. The digital instrument image enclosed by the vertical and horizontal edges is the preprocessed digital instrument image, thus completing the digital instrument image preprocessing for the cutting tool machine tool.

[0031] S22. Convert the preprocessed digital instrument image into a binary matrix to be processed, where each element in the binary matrix represents a pixel; calculate the average grayscale value of each pixel in the binary matrix to be processed, denoted as . The gray values ​​of the pixels in the binary matrix to be processed are greater than... Add to the first partitioning matrix, and make sure the gray value of the pixel in the binary matrix to be processed is less than or equal to the gray value of the pixel. Add to the second partitioning matrix; repeat S22, setting the first threshold as... ,when Less than When As a threshold for binarized grayscale values;

[0032] The preprocessed digital instrument image is divided into a foreground digital instrument image and a background digital instrument image to obtain a binarized digital instrument image. Set the grayscale value of the preprocessed digital instrument image to... Binarized digital instrument image The calculation formula is as follows:

[0033]

[0034] S23. Establish a rectangular coordinate system with the lower left corner of the binarized digital instrument image as the origin, and project the binarized digital instrument image onto the coordinate system. Axial projection Axis projection set; from the origin along Axis search, from The grayscale value of the current pixel in the axis projection set is 0 and Starting with a grayscale value of 1 for the next pixel in the axis projection set, until... The grayscale value of the current pixel in the axis projection set is 1 and The process ends when the grayscale value of the next pixel in the axis projection set is 0, resulting in the digital instrument image; repeat step S23 until the process continues along the axis. After the axis search is completed, a digital instrument image set is obtained, and the numbers in the digital instrument image set are identified. When the identified numbers in the digital instrument image set are abnormal, the digital instrument value of the cutting tool processing machine tool is used as a second adjustment factor.

[0035] This invention preprocesses the digital instrument image of the cutting tool processing machine tool, rasterizes the digital instrument image to locate its position, and uses edge detection to detect vertical and horizontal edges, which helps to suppress image noise. Then, the digital instrument image is binarized to grayscale, and the characters and background of the digital instrument image are segmented, while other grayscale details are discarded for easy recognition. The digital grayscale value is extracted using character segmentation, thereby recognizing the numbers.

[0036] Preferably, step S3 includes the following steps:

[0037] S31. Set the wear state of the machining tools on the cutting machine tool, forming a set of wear states for the machining tools on the cutting machine tool, denoted as... ,in Indicates the first The wear states are defined as follows: The characteristic dimension of the wear states in the set of wear states of the machining tools of the cutting tool processing machine is... Generate the wear matrix to be processed. as follows,

[0038] ,

[0039] in, This represents the first element in the wear matrix to be processed. Types of wear states, feature dimensions are The wear value;

[0040] Select one row from the wear matrix to be processed and denote it as the wear set to be processed. ,in Indicates the first Types of wear states, feature dimensions are Find the minimum wear value in the wear set to be processed. The wear set to be processed is set as the hidden state of the hidden Markov model, and the wear matrix to be processed is extracted. Extracting the feature vectors of the dimensional dimension yields the extracted vector sequence matrix. The Hidden Markov Model is trained using the maximum likelihood estimation method to obtain the Hidden Markov Model.

[0041] S32. Calculate the time required to extract vectors from the extracted vector sequence matrix, and set... At any given time, the wear values ​​of the extracted vectors are used to form a wear value set. Then, the probability of the wear state of the wear value set is calculated, and the wear value of the machining tool of the cutting tool is calculated using the probability of the wear state of the wear value set.

[0042] Over time The wear value of the machining tool on the cutting machine is continuously increased and calculated at the current moment, and a second threshold is set. When the wear value of the machining tool on the cutting machine at the current moment is greater than If the wear value of the machining tool on the cutting tool machine tool is abnormal at the current moment, the wear value of the machining tool machine tool will be used as the third adjustment factor.

[0043] This invention establishes a hidden Markov model for the wear values ​​of cutting tools and machining tools in a set wear matrix, making the identification of wear values ​​more flexible. Continuous wear values ​​are obtained through probability calculation, enabling the detection of wear on cutting tools and machining tools. The method is relatively simple and the identification effect is good.

[0044] Preferably, step S4 includes the following steps:

[0045] S41, the first adjustment factor is the spindle vibration signal value of the tool processing machine tool, the second adjustment factor is the digital instrument value of the tool processing machine tool, and the third adjustment factor is the tool wear value of the tool processing machine tool; select the first, second, and third adjustment factors... A data set comprising the tool processing machine tool adjustment sample set ,in Indicates the first Vibration signal value of a machine tool spindle. Indicates the first Digital instrument readings of a single cutting tool machining center. Indicates the first Wear value of machining tools on individual cutting machine tools;

[0046] set up The VC dimension represents the adjustment sample set of the cutting tool machining machine. Let represent the empirical risk of the tool processing machine tool adjustment sample set, and the probability distribution of the tool processing machine tool adjustment sample set is: , Represents random numbers and Then the minimum empirical risk of adjusting the sample set of the tool processing machine is... The formula is as follows:

[0047]

[0048] To minimize the minimum empirical risk of the tool processing machine tool adjustment sample set, an error threshold is set. When the minimum empirical risk of adjusting the sample set of the tool processing machine is less than The minimum empirical risk of adjusting the sample set of the tool processing machine is taken as the insensitive loss parameter, denoted as... Otherwise, adjust and This minimizes the minimum empirical risk of adjusting the sample set of the cutting tool machine tool;

[0049] S42. Set the linear regression function of the support vector machine as follows: , and ,in , , , , Since is a constant, the formula for fitting the linear regression function of the support vector machine is: , and ,in Indicates the first Vibration signal value of a machine tool spindle. Indicates the first Digital instrument readings of a single cutting tool machining center. Indicates the first Wear value of machining tools on individual cutting machine tools;

[0050] Optimization conditions and constraints are set, and the linear regression function of the support vector machine is processed to obtain the final regression function. The data in the tool processing machine tool adjustment sample set is normalized to obtain a normalized sample set, which is then divided into a normalized training set and a normalized test set. The kernel function in the support vector machine is set to the radial basis function, and the loss error is... The error threshold is The normalized sample set is input into the support vector machine and iterated continuously until the loss error is less than 1. The trained support vector machine model is obtained; then the normalized test set is input into the trained support vector machine model, and the third threshold is set to... When the difference between the output result and the actual result is less than the actual result... If the trained support vector machine model is correct, then it is the final support vector machine model; otherwise, repeat steps S41 and S42 until the difference between the output result and the actual result is less than the actual result. ;

[0051] S43. Build a tool processing machine tool operation platform, transmit the current first adjustment factor, second adjustment factor and third adjustment factor to the tool processing machine tool operation platform, predict the first adjustment factor, second adjustment factor and third adjustment factor to obtain prediction data, and realize intelligent adjustment of the tool processing machine tool operation status and repair of abnormalities based on the prediction data.

[0052] This invention combines the vibration signal value of the spindle of the cutting tool machine tool as the first adjustment factor, the digital instrument value of the cutting tool machine tool as the second adjustment factor, and the wear value of the cutting tool machine tool as the third adjustment factor. It uses a support vector machine model to combine the first, second, and third adjustment factors and achieve the prediction effect. The data is transformed into high-dimensional features through kernel functions, which is suitable for classification and recognition of small samples and avoids many limitations of neural networks.

[0053] This embodiment also discloses a cloud computing-based intelligent adjustment system for the operating status of a cutting tool machine tool, specifically including: a spindle vibration signal extraction module for the cutting tool machine tool, a digital instrument image recognition module, a tool wear value detection module for the cutting tool machine tool, and an operating status adjustment module for the cutting tool machine tool;

[0054] The spindle vibration signal extraction module of the cutting tool processing machine tool is used to extract and reduce the dimensionality of the spindle vibration signal, and to identify the spindle vibration signal through a neural network;

[0055] The digital instrument image recognition module is used to perform image recognition processing on digital instrument images;

[0056] The tool wear detection module for cutting tools is used to construct a hidden Markov model to detect the wear of cutting tools on the cutting tool processing machine.

[0057] The tool processing machine tool operation status adjustment module is used to train a support vector machine model by combining the first adjustment factor, the second adjustment factor and the third adjustment factor to realize intelligent adjustment of the tool processing machine tool operation status.

[0058] The present invention has the following beneficial effects:

[0059] 1. This invention collects vibration signals from the spindle of a tool-making machine tool, uses wavelet analysis to extract features from the vibration signals, greatly reducing computational costs. After introducing decentralization processing, principal component analysis is used to reduce the dimensionality of the extracted vibration signals. A BP neural network model is used to identify whether the vibration signals from the spindle of the tool-making machine tool are abnormal. The processing method has high accuracy.

[0060] 2. This invention preprocesses the digital instrument image of the tool processing machine tool, locates the position of the digital instrument image and performs edge detection, and then binarizes the digital instrument image to make the numbers easier to identify; then it sets the wear matrix to be processed, establishes a hidden Markov model for the wear values ​​in the matrix, and realizes the detection of tool wear of the tool processing machine tool. The method is relatively simple and the recognition effect is good.

[0061] 3. This invention combines the first adjustment factor (spindle vibration signal value of the tool processing machine), the second adjustment factor (digital instrument value of the tool processing machine), and the third adjustment factor (tool wear value of the tool processing machine) with a support vector machine model to achieve prediction results. This method is suitable for classification and recognition of small samples and avoids many limitations of neural networks.

[0062] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0063] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, the drawings can be obtained from these drawings without creative effort.

[0064] Figure 1 This invention provides a flowchart illustrating the intelligent adjustment process of the operating status of a cloud-based intelligent adjustment system for the operating status of a cutting tool machine tool. Detailed Implementation

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] In the description of this invention, it should be understood that the terms "opening", "upper", "lower", "top", "middle", "inner", etc., which indicate orientation or positional relationship, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the invention.

[0067] This implementation discloses a cloud computing-based intelligent adjustment method for the operating status of machine tools used in cutting tool processing, specifically including the following:

[0068] S1. Obtain the vibration signal of the spindle of the tool processing machine tool, use wavelet analysis to extract features from the vibration signal of the tool processing machine tool, then use principal component analysis to reduce the dimension of the extracted vibration signal, and use BP neural network to complete the identification of features, which is denoted as the first adjustment factor;

[0069] S1 includes the following steps:

[0070] S11. Acquire the spindle vibration signal of the tool processing machine tool, and set the spindle vibration signal set as follows: ,in Indicates the first There are one spindle vibration signal, and the characteristic dimension of the spindle vibration signal is... The main shaft vibration signal matrix is ​​obtained. as follows,

[0071] ,

[0072] in, The first element in the spindle vibration signal matrix represents the... Each feature dimension is The main shaft vibration signal; the wavelet packet decomposition level is set to... The number of sub-bands is Then the node is The node wavelet packet coefficients are The characteristics of the spindle vibration signal are replaced with vibration energy characteristics. The calculation formula is as follows:

[0073]

[0074] Select one row from the spindle vibration signal matrix as the set of vibration signals to be processed. ,in The first element in the spindle vibration signal matrix represents the... Each feature dimension is The vibration energy of the spindle vibration signal in the set of vibration signals to be processed is calculated to obtain the vibration energy set. ,in The first element in the spindle vibration signal matrix represents the... Each feature dimension is The vibration energy of the spindle vibration signal; the relative entropy of the vibration energy is set as... Calculate the dissimilarity of the relative entropy of vibrational energy, denoted as . If the vibration energy in the vibration energy set is greater than the dissimilarity of the relative entropy of the vibration energy, the vibration energy in the vibration energy set is retained; otherwise, the vibration energy in the vibration energy set is replaced by the dissimilarity of the relative entropy of the vibration energy, so as to obtain the set of main shaft vibration signals after feature extraction and generate the matrix of main shaft vibration signals after feature extraction.

[0075] S12. The set of spindle vibration signals after feature extraction is de-centered, so that...

[0076] The spindle vibration signal set after feature extraction has a mean of 0, resulting in a decentralized spindle vibration signal set and generating a decentralized spindle vibration signal matrix.

[0077] Let the linear transformation matrix be The linear transformation matrix maximizes the covariance between the centered principal shaft vibration signal matrix and the linear transformation matrix. The eigenvalues ​​and eigenvectors of the linear transformation matrix are calculated and arranged in descending order to obtain the eigenvalue set and eigenvector set, respectively. The top eigenvalues ​​are then selected from the eigenvalue set. 1 eigenvalue, before selecting the eigenvector set. 1 eigenvector; select the first eigenvalues ​​from the eigenvalue set. Before the set of eigenvalues ​​and eigenvectors The eigenvectors form a dimension reduction matrix. Multiplying this dimension reduction matrix by the linear transformation matrix yields the dimension-reduced eigenvectors, denoted as . ,in Indicates the first 1 eigenvector element;

[0078] S13. Set the input layer nodes of the BP neural network as... The learning rate is 0.01, and the expected error is... The activation function is set to the Sigmoid function; the dimensionality-reduced feature vector is standardized to obtain a standardized feature vector, and the standardized feature vector is selected from the... One element is used as the standardized training set, and other elements in the standardized feature vector are selected as the standardized test set; a maximum number of iterations is set. And the current iteration number, the error threshold is The standardized training set is input into the BP neural network to train it. The training continues when the current iteration number is greater than the maximum iteration number or the expected error is less than... If the iteration stops, the trained BP neural network model is obtained; otherwise, the weights are adjusted until the current iteration number is greater than the maximum iteration number or the expected error is less than the maximum. ;

[0079] The standardized test set is input into the trained BP neural network model, and the results of the standardized test set are output. Based on the results of the standardized test set, it is determined whether the vibration signal of the machine tool spindle is abnormal. If the vibration signal of the machine tool spindle is abnormal, the vibration signal value of the machine tool spindle is used as the first adjustment factor; otherwise, a new vibration signal of the machine tool spindle is obtained and S11, S12 and S13 are repeated.

[0080] S2. Perform image recognition processing on the digital instrument image of the cutting tool machine tool. Preprocess the digital instrument image of the cutting tool machine tool. The pixels of the digital instrument image of the cutting tool machine tool are processed by the mean method and digital segmentation to identify the numbers on the digital instrument image and record them as the second adjustment factor.

[0081] S2 includes the following steps:

[0082] S21. Acquire the digital instrument image of the tool processing machine tool, and preprocess the digital instrument image of the tool processing machine tool. The specific process is as follows:

[0083] S211. Obtain the first from the surveillance video. The digital instrument image of the frame tool machining machine tool is selected from the first frame. Corner points of the digital instrument image of the frame tool machining machine; for the first... The digital instrument image of the frame-by-frame cutting tool machining tool is rasterized to obtain a raster, with the first frame being the first raster. Using the lower left corner of the digital instrument image of the frame tool machining machine as the origin, a rectangular coordinate system is established to obtain the grid coordinates, thus obtaining the first... The corner grid coordinates of the digital instrument image of the frame-by-frame cutting tool machine tool; obtained from the surveillance video. The digital instrument images of the frame tool machining machine are compared with the first frame. The corner grid coordinates and the first frame of the digital instrument image of the tool machining machine. The corner grid coordinates of the digital instrument image of the frame tool machining machine are used. If the corner grid coordinates match, then the first [frame] is selected. Get a frame of digital instrumentation images of the tool-cutting machine tool; otherwise, get a new digital instrumentation image of the tool-cutting machine tool.

[0084] S212. Set the horizontal dimension of the digital instrument image of the new tool processing machine tool to... Vertical dimension is The edge detection matrix is ​​obtained. as follows,

[0085] ,

[0086] in, The horizontal dimension of the edge detection matrix is... Vertical dimension is The pixels;

[0087] Calculate the partial derivatives of the pixel values ​​of the pixels in the edge detection matrix in the horizontal and vertical directions, denoted as . and Length of edge detection vector The calculation formula is as follows:

[0088]

[0089] One column of the edge detection matrix is ​​selected as the vertical detection set. The number of pixels in the vertical detection set whose pixel values ​​are greater than the length of the edge detection vector is counted. Other columns in the edge detection matrix are then selected as vertical detection sets. The two vertical detection sets corresponding to the columns with the highest number of pixels whose pixel values ​​are greater than the length of the edge detection vector are denoted as the vertical edges. One row of the edge detection matrix is ​​selected as the horizontal detection set. The number of pixels in the horizontal detection set whose pixel values ​​are greater than the length of the edge detection vector is counted. Other rows in the edge detection matrix are then selected as horizontal detection sets. The two horizontal detection sets corresponding to the rows with the highest number of pixels whose pixel values ​​are greater than the length of the edge detection vector are denoted as the horizontal edges. The digital instrument image enclosed by the vertical and horizontal edges is the preprocessed digital instrument image, thus completing the digital instrument image preprocessing for the cutting tool machine tool.

[0090] S22. Convert the preprocessed digital instrument image into a binary matrix to be processed, where each element in the binary matrix represents a pixel; calculate the average grayscale value of each pixel in the binary matrix to be processed, denoted as . The gray values ​​of the pixels in the binary matrix to be processed are greater than... Add to the first partitioning matrix, and make sure the gray value of the pixel in the binary matrix to be processed is less than or equal to the gray value of the pixel. Add to the second partitioning matrix; repeat S22, setting the first threshold as... ,when Less than ,Will As a threshold for binarized grayscale values;

[0091] The preprocessed digital instrument image is divided into a foreground digital instrument image and a background digital instrument image to obtain a binarized digital instrument image. Set the grayscale value of the preprocessed digital instrument image to... Binarized digital instrument image The calculation formula is as follows:

[0092]

[0093] S23. Establish a rectangular coordinate system with the lower left corner of the binarized digital instrument image as the origin, and project the binarized digital instrument image onto the coordinate system. Axial projection Axis projection set; from the origin along Axis search, from The grayscale value of the current pixel in the axis projection set is 0 and Starting with a grayscale value of 1 for the next pixel in the axis projection set, until... The grayscale value of the current pixel in the axis projection set is 1 and The process ends when the grayscale value of the next pixel in the axis projection set is 0, thus obtaining the digital instrument image number; S23 is repeated until the search along the axis is completed, thus obtaining the digital instrument image number set, and the numbers in the digital instrument image number set are identified; when the identified numbers in the digital instrument image number set are abnormal, the digital instrument value of the cutting tool processing machine tool is used as the second adjustment factor.

[0094] S3. Obtain the wear value of the machining tools of the cutting tool processing machine, construct a hidden Markov model to detect the wear of the machining tools of the cutting tool processing machine, realize the identification of the wear of the machining tools of the cutting tool processing machine, and record it as the third adjustment factor;

[0095] S3 includes the following steps:

[0096] S31. Set the wear state of the machining tools on the cutting machine tool, forming a set of wear states for the machining tools on the cutting machine tool, denoted as... ,in Indicates the first The wear states are defined as follows: The characteristic dimension of the wear states in the set of wear states of the machining tools of the cutting tool processing machine is... Generate the wear matrix to be processed. as follows,

[0097] ,

[0098] in, This represents the first element in the wear matrix to be processed. Types of wear states, feature dimensions are The wear value;

[0099] Select one row from the wear matrix to be processed and denote it as the wear set to be processed. ,in Indicates the first Types of wear states, feature dimensions are Find the minimum wear value in the wear set to be processed. The wear set to be processed is set as the hidden state of the hidden Markov model, and the wear matrix to be processed is extracted. Extracting the feature vectors of the dimensional dimension yields the extracted vector sequence matrix. The Hidden Markov Model is trained using the maximum likelihood estimation method to obtain the Hidden Markov Model.

[0100] S32. Calculate the time required to extract vectors from the extracted vector sequence matrix, and set... At any given time, the wear values ​​of the extracted vectors are used to form a wear value set. Then, the probability of the wear state of the wear value set is calculated, and the wear value of the machining tool of the cutting tool is calculated using the probability of the wear state of the wear value set.

[0101] As time goes by The wear value of the machining tool on the cutting tool machine is continuously increased and calculated at the current moment. A second threshold is set. When the wear value of the machining tool on the cutting tool machine at the current moment is greater than... If the wear value of the machining tool on the cutting tool processing machine is abnormal at the current moment, the wear value of the machining tool on the cutting tool processing machine will be used as the third adjustment factor.

[0102] S4. Combining the first, second, and third adjustment factors, construct a sample set for the adjustment of the tool processing machine tool, and use a support vector machine model to predict the first, second, and third adjustment factors to achieve intelligent adjustment of the operating status of the tool processing machine tool;

[0103] S4 includes the following steps:

[0104] S41, the first adjustment factor is the spindle vibration signal value of the tool processing machine tool, the second adjustment factor is the digital instrument value of the tool processing machine tool, and the third adjustment factor is the tool wear value of the tool processing machine tool; select the first, second, and third adjustment factors... A data set comprising the tool processing machine tool adjustment sample set ,in Indicates the first Vibration signal value of a machine tool spindle. Indicates the first Digital instrument readings of a single cutting tool machining center. Indicates the first Wear value of machining tools on individual cutting machine tools;

[0105] set up The VC dimension represents the adjustment sample set of the cutting tool machining machine. Let represent the empirical risk of the tool processing machine tool adjustment sample set, and the probability distribution of the tool processing machine tool adjustment sample set is: , Represents random numbers and Then the minimum empirical risk of adjusting the sample set of the tool processing machine is... The formula is as follows:

[0106]

[0107] To minimize the minimum empirical risk of the tool processing machine tool adjustment sample set, an error threshold is set. When the minimum empirical risk of adjusting the sample set of the tool processing machine is less than The minimum empirical risk of adjusting the sample set of the tool processing machine is taken as the insensitive loss parameter, denoted as... Otherwise, adjust and This minimizes the minimum empirical risk of adjusting the sample set of the cutting tool machine tool;

[0108] S42. Set the linear regression function of the support vector machine as follows: , and ,in , , , , Since is a constant, the formula for fitting the linear regression function of the support vector machine is: , and ,in Indicates the first Vibration signal value of a machine tool spindle. Indicates the first Digital instrument readings of a single cutting tool machining center. Indicates the first Wear value of machining tools on individual cutting machine tools;

[0109] Optimization conditions and constraints are set, and the linear regression function of the support vector machine is processed to obtain the final regression function. The data in the tool processing machine tool adjustment sample set is normalized to obtain a normalized sample set, which is then divided into a normalized training set and a normalized test set. The kernel function in the support vector machine is set to the radial basis function, and the loss error is... The error threshold is The normalized sample set is input into the support vector machine and iterated continuously until the loss error is less than 1. The trained support vector machine model is obtained; then the normalized test set is input into the trained support vector machine model, and the third threshold is set to... When the difference between the output result and the actual result is less than the actual result... If the trained support vector machine model is correct, then it is the final support vector machine model; otherwise, repeat steps S41 and S42 until the difference between the output result and the actual result is less than the actual result.

[0110] S43. Build a tool processing machine tool operation platform, transmit the current first adjustment factor, second adjustment factor and third adjustment factor to the tool processing machine tool operation platform, predict the first adjustment factor, second adjustment factor and third adjustment factor to obtain prediction data, and realize intelligent adjustment of the tool processing machine tool operation status and repair of abnormalities based on the prediction data.

[0111] This embodiment also discloses a cloud computing-based intelligent adjustment system for the operating status of a cutting tool machine tool, specifically including: a cutting tool machine tool spindle vibration signal extraction module, a digital instrument image recognition module, a cutting tool machine tool wear value detection module, and a cutting tool machine tool operating status adjustment module;

[0112] The spindle vibration signal extraction module of the cutting tool processing machine tool is used to extract and reduce the dimensionality of the spindle vibration signal, and to identify the spindle vibration signal through a neural network;

[0113] The digital instrument image recognition module is used to perform image recognition processing on digital instrument images;

[0114] The tool wear detection module for cutting tools is used to construct a hidden Markov model to detect the wear of cutting tools on the cutting tool processing machine.

[0115] The tool processing machine tool operation status adjustment module is used to train a support vector machine model by combining the first adjustment factor, the second adjustment factor and the third adjustment factor to realize intelligent adjustment of the tool processing machine tool operation status.

[0116] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0117] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A cloud computing-based intelligent adjustment method for the operating status of machine tools used in cutting tool processing, characterized in that, Includes the following steps: S1. Acquire the vibration signal of the machine tool spindle, use wavelet analysis to extract features from the vibration signal, and use a BP neural network to identify the features. If the vibration signal of the machine tool spindle is abnormal, use the vibration signal value of the machine tool spindle as the first adjustment factor; otherwise, acquire a new vibration signal of the machine tool spindle and repeat step S1. S2. Perform image recognition processing on the digital instrument image of the cutting tool machine tool. Preprocess the digital instrument image of the cutting tool machine tool. The pixels of the digital instrument image of the cutting tool machine tool are processed by the mean method and digital segmentation to identify the numbers on the digital instrument image. When the numbers in the digital instrument image set are abnormal, the digital instrument value of the cutting tool machine tool is used as the second adjustment factor. S3. Obtain the wear value of the machining tools on the cutting machine tool, and construct a hidden Markov model to detect the wear of the machining tools on the cutting machine tool over time. The wear value of the machining tool on the cutting machine is continuously increased and calculated at the current moment, and a second threshold is set. When the wear value of the machining tool on the cutting machine at the current moment is greater than At that time, the wear value of the machining tools of the cutting tool processing machine was used as the third adjustment factor; S4. Combining the first, second, and third adjustment factors, construct a tool processing machine tool adjustment sample set and train a support vector machine model; transmit the current first, second, and third adjustment factors to the tool processing machine tool operation platform, use the support vector machine model to predict the first, second, and third adjustment factors, obtain prediction data, and the tool processing machine tool operation platform realizes intelligent adjustment of the tool processing machine tool operation status and performs maintenance on abnormalities based on the prediction data; S3 includes the following steps: S31. Set the wear state of the machining tools of the cutting tool processing machine to form a set of wear states of the machining tools of the cutting tool processing machine to generate a wear matrix to be processed; use the maximum likelihood estimation method to train the hidden Markov model on the wear matrix to be processed to obtain the hidden Markov model. S32, Settings The wear values ​​of the extracted vectors at any given time are used to form a wear value set. The probability of the wear state of the wear value set is calculated using the hidden Markov model. The wear value of the machining tool of the cutting tool processing machine is obtained using the probability of the wear state of the wear value set. S41, the first adjustment factor is the spindle vibration signal value of the tool processing machine tool, the second adjustment factor is the digital instrument value of the tool processing machine tool, and the third adjustment factor is the tool wear value of the tool processing machine tool; select the first, second, and third adjustment factors... A data set comprising the tool processing machine tool adjustment sample set ,in Indicates the first Vibration signal value of a machine tool spindle. Indicates the first Digital instrument readings of a single cutting tool machining center. Indicates the first Wear value of machining tools on individual cutting machine tools; set up The VC dimension represents the adjustment sample set of the cutting tool machining machine. Let represent the empirical risk of the tool processing machine tool adjustment sample set, and the probability distribution of the tool processing machine tool adjustment sample set is: , Represents random numbers and Then the minimum empirical risk of adjusting the sample set of the tool processing machine is... The formula is as follows: To minimize the minimum empirical risk of the tool processing machine tool adjustment sample set, an error threshold is set. When the minimum empirical risk of adjusting the sample set of the tool processing machine is less than The minimum empirical risk of adjusting the sample set of the tool processing machine is taken as the insensitive loss parameter, denoted as... Otherwise, adjust and This minimizes the minimum empirical risk of adjusting the sample set of the cutting tool machine tool.

2. The intelligent adjustment method for the operating status of a tool processing machine tool based on cloud computing according to claim 1, characterized in that, S1 includes the following steps: S11. Obtain the spindle vibration signal of the tool processing machine tool and generate the spindle vibration signal matrix; use wavelet analysis to replace the features of the spindle vibration signal with the vibration energy features to obtain the spindle vibration signal set after feature extraction, and generate the spindle vibration signal matrix after feature extraction. S12. The set of spindle vibration signals after feature extraction is decentered to generate a decentered spindle vibration signal matrix. The decentered spindle vibration signal matrix is ​​then dimensionality-reduced to obtain the dimensionality-reduced feature vector. S13. Set the input layer nodes of the BP neural network as... The learning rate is 0.01, and the expected error is... The activation function is set to the Sigmoid function; the dimensionality-reduced feature vector is standardized to obtain a standardized feature vector, and the standardized feature vector is selected from the... One element is used as the standardized training set, and other elements in the standardized feature vector are selected as the standardized test set; Set the maximum number of iterations and the current iteration number The error threshold is The standardized training set is input into the BP neural network to train it. The training continues when the current iteration number is greater than the maximum iteration number or the expected error is less than... If the iteration stops, the trained BP neural network model is obtained; otherwise, the weights are adjusted until the current iteration number is greater than the maximum iteration number or the expected error is less than the maximum. .

3. The intelligent adjustment method for the operating status of a tool processing machine tool based on cloud computing according to claim 2, characterized in that, The standardized test set is input into the trained BP neural network model, and the results of the standardized test set are output. The abnormality of the spindle vibration signal of the tool processing machine is obtained based on the results of the standardized test set. If the vibration signal of the spindle of the tool processing machine is abnormal, the vibration signal value of the spindle of the tool processing machine is used as the first adjustment factor.

4. The intelligent adjustment method for the operating status of a tool processing machine tool based on cloud computing according to claim 2, characterized in that, S2 includes the following steps: S21. Acquire the digital instrument image of the tool processing machine tool, and preprocess the digital instrument image of the tool processing machine tool. S22. Convert the preprocessed digital instrument image into a binary matrix to be processed, and perform binarization processing on the pixels in the binary matrix to be processed to obtain a binary digital instrument image. S23. Establish a rectangular coordinate system with the lower left corner of the binarized digital instrument image as the origin, and project the binarized digital instrument image onto the coordinate system. Axial projection, along Axis search is performed to obtain a set of digital instrument image numbers, and the numbers in the set of digital instrument image numbers are identified; when the identified numbers in the set of digital instrument image numbers are abnormal, the digital instrument value of the cutting tool processing machine tool is used as a second adjustment factor.

5. The intelligent adjustment method for the operating status of a tool processing machine tool based on cloud computing according to claim 4, characterized in that, S21 includes the following steps: S211. Obtain the first from the surveillance video. The digital instrument images of the frame tool machining machine are compared with the first frame. The corner grid coordinates and the first frame of the digital instrument image of the tool machining machine. The corner grid coordinates of the digital instrument image of the tool-making machine tool are used to locate the digital instrument image of the tool-making machine tool. S212. Perform edge detection processing on the pixels in the edge detection matrix to complete the digital instrument image preprocessing of the cutting tool processing machine tool.

6. The intelligent adjustment method for the operating status of a tool processing machine tool based on cloud computing according to claim 4, characterized in that, S4 further includes the following steps: S42. Define the linear regression function, optimization conditions, and constraint objectives for the support vector machine (SVM). The final regression function is obtained after processing the optimization conditions and constraint objectives. Normalize the data in the tool processing machine tool adjustment sample set to obtain a normalized sample set. Divide the normalized sample set into a normalized training set and a normalized test set. Set the kernel function in the SVM to the radial basis function (RBF) kernel function, and the loss error to... The error threshold is The normalized sample set is input into the support vector machine and iterated continuously until the loss error is less than 1. The trained support vector machine model is obtained; then the normalized test set is input into the trained support vector machine model, and the third threshold is set to... When the difference between the output result and the actual result is less than the actual result... If the trained support vector machine model is correct, then it is the final support vector machine model; otherwise, repeat steps S41 and S42 until the difference between the output result and the actual result is less than the actual result. .

7. A cloud-based intelligent adjustment system for the operating status of a tool-making machine tool, implementing the cloud-based intelligent adjustment method for the operating status of a tool-making machine tool as described in any one of claims 1-6, characterized in that, Specifically include: The machine tool spindle vibration signal extraction module, the digital instrument image recognition module, the machine tool wear value detection module, and the machine tool operating status adjustment module are all included. The spindle vibration signal extraction module of the cutting tool processing machine tool is used to extract and reduce the dimensionality of the spindle vibration signal, and to identify the spindle vibration signal through a neural network; The digital instrument image recognition module is used to perform image recognition processing on digital instrument images; The tool wear detection module for cutting tools is used to construct a hidden Markov model to detect the wear of cutting tools on the cutting tool processing machine. The tool processing machine tool operation status adjustment module is used to train a support vector machine model by combining the first adjustment factor, the second adjustment factor and the third adjustment factor to realize intelligent adjustment of the tool processing machine tool operation status.

Images