Machine tool automatic control optimization method based on sensor monitoring data

Abstract

The invention relates to the field of automation control, in particular to a machine tool automation control optimization method based on sensor monitoring data, and the method comprises the steps: obtaining a first distance optimization factor according to data feature information in a data window; distance measurement between the windows is optimized according to the first distance optimization factor to obtain a first optimized distance; performing secondary division on the windows, and obtaining a local autocorrelation function of the time sequence data windows according to all the secondary division windows; obtaining a second distance optimization factor according to the local autocorrelation function; optimizing the first optimization distance between the time sequence data windows according to a second distance optimization factor to obtain a second optimization distance; distance measurement is carried out according to the second optimization distance to complete the clustering process, a clustering result is obtained, cutter cutting force control parameters are optimized according to the clustering result, cutter cutting force control is carried out according to the optimized control parameters, and therefore the influence of the micro-vibration condition on cutter cutting force control is eliminated.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic control, and particularly to an optimization method for automatic control of machine tools based on sensor data. Background Art

[0002] Precise control of automatic machine tools is crucial for modern industrial manufacturing, and real-time monitoring of tool cutting force is a key link. The change of tool cutting force directly reflects the interaction between the tool and the workpiece during the machining process. Its monitoring data can not only provide an important basis for optimizing the machine tool control strategy, but also diagnose tool wear, machining quality and other aspects. In order to improve the stability and response speed of the control system, the PID control strategy is usually used to dynamically regulate the cutting force.

[0003] However, the noise caused by tool micro-vibrations in the sensor monitoring data will lead to misjudgment of key features such as cutting force fluctuations and acceleration changes, thus affecting the control effect of the PID control method. Therefore, it has become an urgent problem to effectively analyze and process the micro-vibration situation existing in the monitoring data of the cutting force sensor. Summary of the Invention

[0004] In view of this, the embodiments of the present invention provide an optimization method for automatic control of machine tools based on sensor data to solve the problem of poor PID control effect caused by noise caused by tool micro-vibrations in sensor monitoring data.

[0005] The embodiments of the present invention provide an optimization method for automatic control of machine tools based on sensor data. An optimization method for automatic control of machine tools based on sensor data includes the following steps:

[0006] Collect data from the tool cutting force sensor for machine tool automation to obtain the time-series data of tool cutting force monitoring; perform the first data window division on the time-series data of tool cutting force monitoring to obtain the first time-series data window of all the time-series data of tool cutting force monitoring; obtain the first distance optimization factor by performing feature analysis on the first time-series data window; optimize the distance between the first time-series data windows through the first distance optimization factor to obtain the first optimized distance between the first time-series data windows; perform the second data window division on the first time-series data window of the tool cutting force monitoring time-series data to obtain the second time-series data window of all the tool cutting force monitoring time-series data; obtain the local autocorrelation function of the first time-series data window through fitting analysis of the second time-series data window; obtain the second distance optimization factor between the first time-series data windows through the distribution distance analysis between the local autocorrelation functions; optimize the first optimized distance between the first time-series data windows of the tool cutting force monitoring time-series data through the second distance optimization factor to obtain the second optimized distance; perform clustering analysis through the second optimized distance between the first time-series data windows of the tool cutting force monitoring time-series data to obtain the clustering result of the first time-series data window of the tool cutting force monitoring time-series data; optimize the tool cutting force control parameters through the clustering result to obtain the optimized tool cutting force control parameters, and perform automatic machine tool tool cutting force control through the optimized tool cutting force control parameters.

[0007] The step of collecting data from the tool cutting force sensor for machine tool automation to obtain the time-series data of tool cutting force monitoring and performing the first data window division on the time-series data of tool cutting force monitoring to obtain the first time-series data window of all the time-series data of tool cutting force monitoring includes:

[0008] Set the sampling frequency of the tool cutting force sensor and collect the data of the tool cutting force value in the automatic machine tool according to the set sampling frequency to obtain the time-series data of tool cutting force monitoring; set the length of the time-series data window of tool cutting force monitoring, and perform window division on the time-series data of tool cutting force monitoring according to the length of the time-series data window of tool cutting force monitoring to obtain the first time-series data window of all the time-series data of tool cutting force monitoring.

[0009] Preferably, the step of obtaining the first distance optimization factor by performing feature analysis on the first time-series data window includes:

[0010] Obtain the first time series data window of all the tool cutting force monitoring time series data, and obtain the mean of the numerical means within all the first time series data windows, the mean of the numerical standard deviations within all the first time series data windows, the mean of the numerical skewnesses within all the first time series data windows, and the mean of the numerical kurtoses within all the time series data windows according to all the first time series data windows in the tool cutting force monitoring time series data; form the overall mean vector of all the time series data windows with the mean of the numerical means within all the first time series data windows, the mean of the numerical standard deviations within all the first time series data windows, the mean of the numerical skewnesses within all the first time series data windows, and the mean of the numerical kurtoses within all the time series data windows; obtain the Mahalanobis distance between the first time series data window of the tool cutting force monitoring time series data and the overall mean vector of all the first time series data windows according to the overall mean vector of all the first time series data windows; take the reciprocal of the sum of the constant 1 and the Mahalanobis distance between the first time series data window of the tool cutting force monitoring time series data and the overall mean vector of all the first time series data windows as the first distance optimization factor of the first time series data window of the tool cutting force monitoring time series data.

[0011] Preferably, the distance optimization between the first time series data windows by the first distance optimization factor to obtain the first optimized distance between the first time series data windows includes:

[0012] Obtain any two first time series data windows of the tool cutting force monitoring time series data, and multiply the respective first distance optimization factors of the any two first time series data windows as the optimization weight for measuring the distance between the any two first time series data windows; take the calculation result of multiplying the optimization weight for measuring the distance between the any two first time series data windows and the Euclidean distance between the any two first time series data windows as the first optimized distance between the first time series data windows.

[0013] Preferably, the second data window division of the first time series data window of the tool cutting force monitoring time series data to obtain all the second time series data windows of the tool cutting force monitoring time series data, and the local autocorrelation function of the first time series data window is obtained through the fitting analysis of the second time series data windows, includes:

[0014] Obtain the set second data window length, divide the first time series data window of the tool cutting force monitoring time series data according to the second data window length for the second time, and obtain the second time series data window in the first time series data window of all the tool cutting force monitoring time series data; Obtain the set autocorrelation lag parameter, and perform local autocorrelation function evaluation corresponding to the first time series data window of the tool cutting force monitoring time series data on all the second time series data windows according to the set autocorrelation lag parameter to obtain the local autocorrelation function of the first time series data window.

[0015] Preferably, the obtaining of the second distance optimization factor between the first time series data windows through the analysis of the distribution distance between the local autocorrelation functions includes:

[0016] Obtain any two first time series data windows of the tool cutting force monitoring time series data, obtain the distribution distance between the local autocorrelation functions corresponding to the any two first time series data windows respectively, and use the reciprocal of the sum of the constant 1 and the distribution distance between the local autocorrelation functions corresponding to the any two first time series data windows respectively as the second distance optimization factor between the first time series data windows.

[0017] Preferably, the optimizing of the first optimization distance between the first time series data windows of the tool cutting force monitoring time series data through the second distance optimization factor to obtain the second optimization distance includes:

[0018] Obtain the first optimization distance between the first time series data windows and the second distance optimization factor between the first time series data windows, and use the calculation result of multiplying the second distance optimization factor between the first time series data windows by the first optimization distance between the first time series data windows as the second optimization distance between the first time series data windows of the tool cutting force monitoring time series data.

[0019] Preferably, the performing of clustering analysis through the second optimization distance between the first time series data windows of the tool cutting force monitoring time series data to obtain the clustering result of the first time series data windows of the tool cutting force monitoring time series data includes:

[0020] Obtain the set number of clusters, and perform clustering according to the second optimization distance between the first time series data windows of the tool cutting force monitoring time series data as the distance metric to obtain the clustering result of the first time series data windows of the tool cutting force monitoring time series data.

[0021] Preferably, the optimizing of the tool cutting force control parameters through the clustering result to obtain the optimized tool cutting force control parameters, and the performing of automatic machine tool tool cutting force control through the optimized tool cutting force control parameters includes:

[0022] Obtain the clustering result and the real-time tool cutting force monitoring time series data of the automated machine tool, perform a first time series data window division on the real-time tool cutting force monitoring time series data of the automated machine tool to obtain the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool, and perform cluster evaluation on the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool according to the clustering result to obtain the cluster category to which the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool belongs; optimize the tool cutting force control parameters of the automated machine tool according to the cluster category information of the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool, and output a control signal of the optimized tool cutting force control parameters to the machine tool actuator to realize the control optimization of the cutting force.

[0023] Preferably, the optimization of the tool cutting force control parameters of the automated machine tool according to the cluster category information of the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool includes:

[0024] Obtain the average cutting force, standard deviation mean, and average zero-crossing rate of the cluster center corresponding to each cluster category in the clustering result of the first time series data window of the tool cutting force monitoring time series data; obtain the cluster category information of the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool, and use the calculation result of dividing the cutting force mean corresponding to the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool by the cutting force mean of the cluster category to which the first time series data window of the real-time tool cutting force monitoring time series data belongs as the proportional gain adjustment factor; use the calculation result of dividing the cutting force standard deviation corresponding to the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool by the cutting force standard deviation of the cluster category to which the first time series data window of the real-time tool cutting force monitoring time series data belongs as the integral gain adjustment factor; use the calculation result of dividing the average zero-crossing rate of the cutting force corresponding to the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool by the average zero-crossing rate of the cutting force of the cluster category to which the first time series data window of the real-time tool cutting force monitoring time series data belongs as the differential gain adjustment factor; obtain the real-time PID control parameters of the automated machine tool; use the product of the proportional gain adjustment factor and the proportional gain parameter of the real-time PID control of the automated machine tool as the proportional gain optimization parameter; use the integral gain parameter of the real-time PID control of the automated machine tool divided by the integral gain adjustment factor as the integral gain optimization parameter; use the product of the differential gain adjustment factor and the differential gain parameter of the real-time PID control of the automated machine tool as the differential gain optimization parameter.

[0025] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: The present invention collects data from the tool cutting force sensor for machine tool automation to obtain the tool cutting force monitoring time series data; performs the first data window division on the tool cutting force monitoring time series data to obtain the first time series data windows of all the tool cutting force monitoring time series data; obtains the first distance optimization factor through feature analysis of the first time series data windows; optimizes the distance between the first time series data windows through the first distance optimization factor to obtain the first optimized distance between the first time series data windows; performs the second data window division on the first time series data windows of the tool cutting force monitoring time series data to obtain the second time series data windows of all the tool cutting force monitoring time series data; obtains the local autocorrelation function of the first time series data windows through fitting analysis of the second time series data windows; obtains the second distance optimization factor between the first time series data windows through the distribution distance analysis between the local autocorrelation functions; optimizes the first optimized distance between the first time series data windows of the tool cutting force monitoring time series data through the second distance optimization factor to obtain the second optimized distance; performs clustering analysis through the second optimized distance between the first time series data windows of the tool cutting force monitoring time series data to obtain the clustering result of the first time series data windows of the tool cutting force monitoring time series data; optimizes the tool cutting force control parameters through the clustering result to obtain the optimized tool cutting force control parameters, and controls the tool cutting force of the automated machine tool through the optimized tool cutting force control parameters. Among them, by extracting the statistical features of all the tool cutting force monitoring time series data windows, the overall feature vector of all the time series data windows is obtained, and the first optimization factor is obtained according to the difference between the feature vector of each time series data window and the overall feature vector, so as to eliminate the influence of the slightly fluttering data windows with large standard deviations, abnormal skewness or kurtosis in the distance measurement process, and further obtain the second optimization factor through the local autocorrelation information of the second window in the time series data window and perform further distance optimization, so as to accurately evaluate the distance between the data windows with different fluctuation modes, obtain accurate clustering results, and improve the optimization accuracy of the PID control parameters through the clustering results. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0027] Figure 1 is the method flow chart of an optimization method for machine tool automation control based on sensor data according to the embodiments of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] Embodiments of the present disclosure will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where like or similar reference numerals denote like or similar elements or elements having like or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0029] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above accompanying drawings are used to distinguish similar objects and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used for this toothpaste can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0030] In order to illustrate the technical solution of the present invention, specific embodiments will be used for illustration below.

[0031] The specific scenario targeted by the present invention is: optimizing the cutting force control of the cutting tool of an automated machine tool in the automated machine tool.

[0032] See Figure 1 , which is a method flow chart of an automated machine tool control optimization method based on sensor data provided by Embodiment 1 of the present invention. As Figure 1 shown, an automated machine tool control optimization method based on sensor data may include:

[0033] Step S101, collect the tool cutting force monitoring time series data through the tool cutting force sensor of the automated machine tool and divide the tool cutting force monitoring time series data into time series data windows.

[0034] For an automated machine tool equipped with a tool cutting force sensor, the cutting force of the tool in the automated machine tool can be data-collected through the tool cutting force sensor. In the embodiments of the present invention, all known tool cutting force monitoring time series data of the tool cutting force sensor are collected. Preferably, the sampling frequency of the tool cutting force sensor is set to 0.5 in the embodiments of the present invention. One tool cutting force value is collected once, and the tool cutting force monitoring time series data of the automated machine tool is obtained according to the set sampling frequency of the tool cutting force sensor.

[0035] Based on the above data collection method, the tool cutting force monitoring time series data of the tool in the automated machine tool is obtained. After obtaining the tool cutting force monitoring time series data of the tool in the automated machine tool, the cutting force monitoring time series data needs to be window-divided according to the set window length. The window length is set to , divide the time series data of the cutting force of the tool in the automated machine tool into windows to obtain all the time series data windows of the cutting force monitoring of the tools in the automated machine tool.

[0036] Step S102: Obtain the first distance optimization factor of the time series data window of the cutting force monitoring of the tool according to the data characteristic information in the time series data window of the cutting force monitoring of the tool, and optimize the distance measurement between the windows according to the first distance optimization factor of the time series data window of the cutting force monitoring of the tool, so as to obtain the first optimized distance between the time series data windows of the cutting force monitoring of the tool.

[0037] After obtaining the time series data of the cutting force sensor, it is necessary to perform clustering analysis through the time series data window of the cutting force monitoring of the tool, so as to divide the time series data of the cutting force monitoring of the tool into different clusters, and dynamically adjust the PID control parameters of the cutting force of the tool through the change information of the time series data window of the cutting force in the cluster. Therefore, in the process of controlling the cutting force of the tool in the automated machine tool by the PID control method, the PID control parameters are adjusted according to different modes corresponding to the cutting force, so that the PID control process can more accurately control the cutting force of the tool in the automated machine tool.

[0038] However, in the process of clustering the time series data windows of the cutting force monitoring of the tool in the automated machine tool, it divides the time series data windows of the cutting force monitoring of the automated machine tool into different clusters by measuring the distance between the time series data windows of the cutting force. However, in the process of measuring the distance between the time series monitoring data windows of the cutting force, due to the micro-vibration of the time series data of the cutting force monitoring, an error will occur in the distance measurement between two time series monitoring data windows during the distance measurement process between the time series data windows. Therefore, in the distance measurement of the clustering process, it is necessary to optimize the distance measurement between the time series data windows of the cutting force monitoring to eliminate the influence of the micro-vibration during the distance measurement process between the time series data windows of the cutting force monitoring, so as to ensure the accurate cluster division of the time series monitoring data windows of the cutting force.

[0039] Specifically, after obtaining the time series data window of the cutting force monitoring, due to the influence of micro-vibration on the data points in the window, the data points show different statistical characteristics. For example, in the data window with slight chatter, the standard deviation and peak value of the cutting force value are larger than those of the data window with normal cutting. Traditional clustering algorithms directly use the Euclidean distance to calculate the distance between data windows, treat all data points equally, and ignore the differences in the statistical characteristics of data points, resulting in the clustering results being insensitive to micro-vibration.

[0040] To solve the problem of micro-vibration, first, it is necessary to obtain the first distance optimization factor of the tool cutting force monitoring time-series data window according to the data characteristic information in the tool cutting force monitoring time-series data window, including:

[0041] Obtain all the time-series data windows in the tool cutting force monitoring time-series data, and obtain the mean of the numerical means, the mean of the numerical standard deviations, the mean of the numerical skewnesses, and the mean of the numerical kurtoses within all the time-series data windows according to all the time-series data windows in the tool cutting force monitoring time-series data; form the overall mean vector of all the time-series data windows with the mean of the numerical means, the mean of the numerical standard deviations, the mean of the numerical skewnesses, and the mean of the numerical kurtoses within all the time-series data windows; obtain the Mahalanobis distance between the tool cutting force monitoring time-series data window and the overall mean vector of all the time-series data windows according to the overall mean vector of all the time-series data windows; take the reciprocal of the sum of the constant 1 and the Mahalanobis distance between the tool cutting force monitoring time-series data window and the overall mean vector of all the time-series data windows as the first distance optimization factor of the tool cutting force monitoring time-series data window.

[0042] In one embodiment, assume that the mean feature vector of the th tool cutting force monitoring time-series data window is , the overall mean vector of all the time-series data windows is , and the covariance matrix of the statistical characteristics of all the time-series data windows is . Then, the calculation expression of the first distance optimization factor of the tool cutting force monitoring time-series data window is:

[0043]

[0044] Among them, represents the first distance optimization factor of the tool cutting force monitoring time-series data window; represents the mean feature vector of the th tool cutting force monitoring time-series data window; represents the overall mean feature vector of all the time-series data windows; represents vector transpose; represents the inverse matrix of the covariance matrix of the statistical characteristics of all the time-series data windows; represents the constant .

[0045] It should be noted that The Mahalanobis distance is between the mean feature vector of the tool cutting force monitoring time series data window and the overall mean feature vector of all time series data windows. The Mahalanobis distance weights each feature, taking into account the correlation between different features. For example, if the mean and standard deviation are positively correlated, then in the direction where the mean is large and the standard deviation is also large, the Mahalanobis distance will be relatively small because it considers this change to be "normal". For data points that deviate from this correlation (such as a large mean but a small standard deviation, or a small mean but a large standard deviation), the Mahalanobis distance will be large. Convert the Mahalanobis distance into a coefficient within the range of When the statistical features of a data window deviate far from the center of the data set (large Mahalanobis distance), its value will be small; when the statistical features of a data window are close to the center of the data set (small Mahalanobis distance), its value will be close to .

[0046] After obtaining the first distance optimization factor of the tool cutting force monitoring time series data window, the distance metric between windows can be optimized according to the first distance optimization factor of the tool cutting force monitoring time series data window, and the first optimized distance between the tool cutting force monitoring time series data windows is obtained, including:

[0047] Obtain any two of the tool cutting force monitoring time series data windows, multiply the first distance optimization factors of the any two tool cutting force monitoring time series data windows respectively as the optimization weight of the distance metric between the any two tool cutting force monitoring time series data windows; the calculation result of multiplying the optimization weight of the distance metric between the any two tool cutting force monitoring time series data windows by the Euclidean distance between the any two tool cutting force monitoring time series data windows is used as the first optimized distance between the tool cutting force monitoring time series data windows.

[0048] In an embodiment, assume that the th tool cutting force monitoring time series data window is , and the th tool cutting force monitoring time series data window is , then the calculation expression of the first optimized distance between the tool cutting force monitoring time series data windows is:

[0049]

[0050] where represents the first optimized distance between the th tool cutting force monitoring time series data window and the th tool cutting force monitoring time series data window; represents the The first distance optimization factor of the cutting force monitoring time series data window of a tool; Indicates the first distance optimization factor of the cutting force monitoring time series data window of a tool; Indicates the th feature of the mean feature vector; Indicates the th feature value in the mean feature vector of the cutting force monitoring time series data window of the th tool; Indicates the th feature value in the mean feature vector of the cutting force monitoring time series data window of the th tool; Indicates the Euclidean distance between the th cutting force monitoring time series data window of a tool and the th cutting force monitoring time series data window of a tool.

[0051] It should be noted that the optimization factors and play a weighting role. If the statistical features of two data windows and are both far from the center of the data set ( and are both small), then their distance will be amplified. If the statistical features of two data windows and are both close to the center of the data set ( and are both large, close to 1), then their distance is close to the traditional Euclidean distance. If the statistical feature of one data window is far from the center of the data set ( value is small), while the statistical feature of the other data window is close to the center of the data set ( value is large), then their distance will be affected by the data window that is far from the center, and the distance will also be amplified to a certain extent.

[0052] Step S103, perform a second data window division on the cutting force monitoring time series data window of the tool, and obtain the local autocorrelation function of the cutting force monitoring time series data window according to all the second data windows in the cutting force monitoring time series data window.

[0053] After obtaining the first optimized distance between the cutting force monitoring time series data windows of the tool, although the first distance optimization factor considers the statistical features (mean, standard deviation, skewness, kurtosis) of the cutting force data within the data window and enhances The sensitivity of the algorithm to the overall distribution difference of data. However, in actual cutting processes, different cutting states often correspond to different cutting force fluctuation patterns. For example, during normal cutting, the cutting force signal shows relatively random fluctuations; while during slight chatter, the cutting force signal usually exhibits a certain degree of periodicity or intermittency. This is because chatter, as a self-excited vibration, has a frequency related to the natural frequency of the machine tool-tool-workpiece system. In addition, tool wear, uneven workpiece materials, etc. may also cause the cutting force to present specific fluctuation patterns. Although the statistical characteristics of the first distance optimization factor can reflect the fluctuation amplitude to a certain extent, they are essentially "static" and cannot describe the dynamic characteristics of the cutting force signal changing with time, nor can they distinguish data windows with different fluctuation patterns. Since different cutting states (especially slight chatter) often correspond to different fluctuation patterns, and these patterns contain important information about the machining process, accurately identifying these fluctuation patterns is crucial for cutting state monitoring. The early detection of slight chatter is particularly important because timely detection and taking measures can avoid the further development of chatter and ensure machining quality and tool life. At the same time, different cutting states require different PID control parameters. If these states cannot be accurately distinguished, the adaptive adjustment of PID parameters cannot be achieved. Therefore, after obtaining the first optimization distance between the tool cutting force monitoring time series data windows, it is necessary to further divide the tool cutting force monitoring time series data windows into second data windows, and obtain the local autocorrelation function of the tool cutting force monitoring time series data windows according to all the second data windows in the tool cutting force monitoring time series data windows, including:

[0054] Obtain the set length of the second data window, divide the tool cutting force monitoring time series data window according to the length of the second data window to obtain all the second data windows in the tool cutting force monitoring time series data window; obtain the set autocorrelation lag parameter, and evaluate the local autocorrelation function corresponding to the tool cutting force monitoring time series data window for all the second data windows in the tool cutting force monitoring time series data window according to the set autocorrelation lag parameter to obtain the local autocorrelation function corresponding to the tool cutting force monitoring time series data window.

[0055] In one embodiment, the length of the second data window is set to , and each tool cutting force time series data window is divided into second data windows. The length of the second data window can be adjusted according to the actual scenario and there is no requirement. It should be noted that the central data point in the time series data window is removed and divided during the secondary division process to obtain all the second data windows.

[0056] In one embodiment, the autocorrelation lag parameter is set to , and the autocorrelation lag parameter can be adjusted according to the actual scenario without requirements, so as to obtain the autocorrelation function of each second data window corresponding to the autocorrelation lag parameter.

[0057] After obtaining the autocorrelation function of each second data window in the tool cutting force monitoring time series data window, average the autocorrelation functions of all second data windows in the tool cutting force monitoring time series data window, and use the averaged result as the local autocorrelation function of the tool cutting force monitoring time series data window.

[0058] Step S104, obtain the second distance optimization factor between the tool cutting force monitoring time series data windows according to the local autocorrelation function of the tool cutting force monitoring time series data window, and optimize the first optimization distance between the tool cutting force monitoring time series data windows according to the second distance optimization factor to obtain the second optimization distance between the tool cutting force monitoring time series data windows.

[0059] After obtaining the local autocorrelation function of the tool cutting force monitoring time series data window, the second distance optimization factor between the tool cutting force monitoring time series data windows can be obtained according to the local autocorrelation function of the tool cutting force monitoring time series data window, including:

[0060] Obtain any two of the tool cutting force monitoring time series data windows, obtain the distribution distance between the local autocorrelation functions corresponding to the any two tool cutting force monitoring time series data windows, and use the reciprocal of the sum of the constant 1 and the distribution distance between the local autocorrelation functions corresponding to the any two tool cutting force monitoring time series data windows as the second distance optimization factor between the tool cutting force monitoring time series data windows.

[0061] In one embodiment, assume that the local autocorrelation function of the th tool cutting force monitoring time series data window is , and the local autocorrelation function of the th tool cutting force monitoring time series data window is , then the calculation expression of the second distance optimization factor between the th tool cutting force monitoring time series data window and the th tool cutting force monitoring time series data window is:

[0062]

[0063] Wherein, represents the second distance optimization factor between the th tool cutting force monitoring time series data window and the th tool cutting force monitoring time series data window; represents the local autocorrelation function of the th tool cutting force monitoring time series data window; represents the local autocorrelation function of the th tool cutting force monitoring time series data window; represents the distance between the local autocorrelation function of the th tool cutting force monitoring time series data window and the local autocorrelation function of the th tool cutting force monitoring time series data window; represents a constant .

[0064] After obtaining the second distance optimization factor between the tool cutting force monitoring time series data windows, the first optimization distance between the tool cutting force monitoring time series data windows can be optimized according to the second distance optimization factor to obtain the second optimization distance between the tool cutting force monitoring time series data windows, including:

[0065] Obtain the first optimization distance between the tool cutting force monitoring time series data windows and the second distance optimization factor between the tool cutting force monitoring time series data windows, and use the calculation result of multiplying the second distance optimization factor between the tool cutting force monitoring time series data windows by the first optimization distance between the tool cutting force monitoring time series data windows as the second optimization distance between the tool cutting force monitoring time series data windows.

[0066] In an embodiment, the calculation expression of the second optimization distance between the tool cutting force monitoring time series data windows is:

[0067]

[0068] wherein, represents the second optimization distance between the th tool cutting force monitoring time series data window and the th tool cutting force monitoring time series data window; represents the second distance optimization factor between the th tool cutting force monitoring time series data window and the th tool cutting force monitoring time series data window; represents the first optimization distance between the th tool cutting force monitoring time series data window and the th tool cutting force monitoring time series data window.

[0069] It should be noted that when the fluctuation patterns of two data windows are similar, is smaller, Larger; when the difference in fluctuation patterns is relatively large, relatively large, relatively small. Specifically, if the fluctuation patterns of two data windows and are similar ( relatively large, close to ), then the influence on the first optimized distance is relatively small, and the second optimized distance is close to . If the fluctuation patterns of two data windows and are quite different ( relatively small), then their distance will be reduced, such that even if their statistical characteristics do not differ much, they may be classified into different clusters.

[0070] Step S105: Perform distance measurement based on the second optimized distance between the tool cutting force monitoring time series data windows and complete the clustering process to obtain the clustering result of the tool cutting force monitoring time series data windows. Optimize the tool cutting force control parameters of the automated machine tool according to the clustering result, and perform tool cutting force control of the automated machine tool according to the optimized tool cutting force control parameters.

[0071] After obtaining the second optimized distance between the tool cutting force monitoring time series data windows, distance measurement can be performed based on the second optimized distance between the tool cutting force monitoring time series data windows and the clustering process can be completed to obtain the clustering result of the tool cutting force monitoring time series data windows, including:

[0072] Obtain the set number of clusters, and perform clustering with the second optimized distance between the tool cutting force monitoring time series data windows as the distance metric to obtain the clustering result of the tool cutting force monitoring time series data windows. In the embodiments of the present invention, the set number of clusters is obtained by the elbow method. Determining the number of clusters by the elbow method is a well-known technique and will not be elaborated here.

[0073] After obtaining the clustering result of the tool cutting force monitoring time series data windows, the tool cutting force control parameters of the automated machine tool can be optimized according to the clustering result, and tool cutting force control of the automated machine tool can be performed according to the optimized tool cutting force control parameters, including:

[0074] ​Obtain the clustering result of the tool cutting force monitoring time series data window and the real-time tool cutting force monitoring time series data of the automated machine tool. Divide the real-time tool cutting force monitoring time series data of the automated machine tool into windows to obtain the real-time tool cutting force monitoring time series data window of the automated machine tool. Evaluate the cluster of the real-time tool cutting force monitoring time series data window of the automated machine tool according to the clustering result of the tool cutting force monitoring time series data window to obtain the cluster category to which the real-time tool cutting force monitoring time series data window of the automated machine tool belongs; optimize the tool cutting force control parameters of the automated machine tool according to the cluster category information of the real-time tool cutting force monitoring time series data window of the automated machine tool. Specifically, optimizing the tool cutting force control parameters of the automated machine tool according to the cluster category information of the real-time tool cutting force monitoring time series data window of the automated machine tool includes:

[0075] Obtain the average cutting force, average standard deviation, and average zero-crossing rate of the cluster center corresponding to each cluster category in the clustering result of the tool cutting force monitoring time series data window; obtain the cluster category information of the real-time tool cutting force monitoring time series data window of the automated machine tool, and use the calculation result of dividing the average cutting force of the real-time tool cutting force monitoring time series data window of the automated machine tool by the average cutting force of the cluster category of the real-time tool cutting force monitoring time series data window as the proportional gain adjustment factor; use the calculation result of dividing the cutting force standard deviation of the real-time tool cutting force monitoring time series data window of the automated machine tool by the cutting force standard deviation of the cluster category of the real-time tool cutting force monitoring time series data window as the integral gain adjustment factor; use the calculation result of dividing the average zero-crossing rate of the cutting force of the real-time tool cutting force monitoring time series data window of the automated machine tool by the average zero-crossing rate of the cutting force of the cluster category of the real-time tool cutting force monitoring time series data window as the derivative gain adjustment factor; obtain the real-time PID control parameters of the automated machine tool; use the product of the proportional gain adjustment factor and the proportional gain parameter of the real-time PID control of the automated machine tool as the proportional gain optimization parameter; use the integral gain parameter of the real-time PID control of the automated machine tool divided by the integral gain adjustment factor as the integral gain optimization parameter; use the product of the derivative gain adjustment factor and the derivative gain parameter of the real-time PID control of the automated machine tool as the derivative gain optimization parameter. Output the optimized tool cutting force PID control parameters, including the proportional gain optimization parameter, the integral gain optimization parameter, and the derivative gain optimization parameter, as control signals to the machine tool actuator to achieve the control optimization of the cutting force.

[0076] In summary, in the embodiments of the present invention, data of the tool cutting force sensor for machine tool automation is collected to obtain the time series data of tool cutting force monitoring; the time series data of tool cutting force monitoring is subjected to the first data window division to obtain the first time series data window of all the time series data of tool cutting force monitoring; through feature analysis of the first time series data window, the first distance optimization factor is obtained; the distance between the first time series data windows is optimized by the first distance optimization factor to obtain the first optimized distance between the first time series data windows; the first time series data window of the time series data of the tool cutting force monitoring is subjected to the second data window division to obtain the second time series data window of all the time series data of tool cutting force monitoring; fitting analysis is performed through the second time series data window to obtain the local autocorrelation function of the first time series data window; through the distribution distance analysis between the local autocorrelation functions, the second distance optimization factor between the first time series data windows is obtained; the first optimized distance between the first time series data windows of the time series data of the tool cutting force monitoring is optimized by the second distance optimization factor to obtain the second optimized distance; clustering analysis is performed through the second optimized distance between the first time series data windows of the time series data of the tool cutting force monitoring to obtain the clustering result of the first time series data window of the time series data of tool cutting force monitoring; the tool cutting force control parameters are optimized through the clustering result to obtain the optimized tool cutting force control parameters, and the tool cutting force of the automated machine tool is controlled through the optimized tool cutting force control parameters. Among them, by extracting the statistical features of all the time series data windows of the tool cutting force monitoring, the overall feature vector of all the time series data windows is obtained, and the first optimization factor is obtained according to the difference between the feature vector of each time series data window and the overall feature vector, so as to eliminate the influence of the slightly fluttering data windows with large standard deviation, abnormal skewness or kurtosis in the distance measurement process, and on this basis, the second optimization factor is further obtained through the local autocorrelation information of the second window in the time series data window and further distance optimization is performed, so as to accurately evaluate the distance between the data windows of different fluctuation modes, obtain accurate clustering results, and improve the optimization accuracy of the PID control parameters through the clustering results.

[0077] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A machine tool automation control optimization method based on sensor monitoring data, characterized in that: include: Collect data from the cutting force sensor of the machine tool automation tool to obtain the cutting force monitoring time series data; Performing a first data window division on the tool cutting force monitoring time series data to obtain a first time series data window of all the tool cutting force monitoring time series data; obtaining a first distance optimization factor by performing feature analysis on the first time series data window; Optimizing the distance between the first time series data windows by using a first distance optimization factor to obtain a first optimized distance between the first time series data windows; Perform a second data window division on the first time series data window of the tool cutting force monitoring time series data to obtain a second time series data window of all the tool cutting force monitoring time series data; perform fitting analysis on the second time series data window to obtain a local autocorrelation function of the first time series data window; obtain a second distance optimization factor between the first time series data windows through distribution distance analysis between the local autocorrelation functions; optimize the first optimized distance between the first time series data windows of the tool cutting force monitoring time series data through the second distance optimization factor to obtain a second optimized distance; perform cluster analysis on the second optimized distance between the first time series data windows of the tool cutting force monitoring time series data to obtain a clustering result of the first time series data window of the tool cutting force monitoring time series data; optimize the tool cutting force control parameters through the clustering results to obtain the optimized tool cutting force control parameters, and perform automated machine tool tool cutting force control through the optimized tool cutting force control parameters.

2. The machine tool automation control optimization method based on sensor data according to claim 1 is characterized in that: The data collection of the tool cutting force sensor of the machine tool automation is performed to obtain the tool cutting force monitoring time series data, and the tool cutting force monitoring time series data is divided into a first data window to obtain the first time series data window of all the tool cutting force monitoring time series data, including: The sampling frequency of the tool cutting force sensor is set and data collection of the tool cutting force values ​​in the automated machine tool is performed according to the set sampling frequency to obtain the tool cutting force monitoring time series data; the window length of the tool cutting force monitoring time series data is set and the tool cutting force monitoring time series data is divided into windows according to the window length of the tool cutting force monitoring time series data to obtain the first time series data window of all the tool cutting force monitoring time series data.

3. The machine tool automation control optimization method based on sensor data according to claim 1 is characterized in that: The step of obtaining a first distance optimization factor by performing feature analysis on the first time series data window includes: Obtain the first time series data window of all tool cutting force monitoring time series data, and obtain the mean of the numerical means in all first time series data windows, the mean of the numerical standard deviations in all first time series data windows, the mean of the numerical skewness in all first time series data windows, and the mean of the numerical kurtosis in all time series data windows according to all first time series data windows in the tool cutting force monitoring time series data; and add the mean of the numerical means in all first time series data windows, the mean of the numerical standard deviations in all first time series data windows, the mean of the numerical skewness in all first time series data windows, and the mean of the numerical kurtosis in all time series data windows. The overall mean vector of all the time series data windows is formed according to the mean of the numerical kurtosis in the window; the Mahalanobis distance between the first time series data window of the tool cutting force monitoring time series data and the overall mean vector of all the first time series data windows is obtained according to the overall mean vector of all the first time series data windows; the constant 1 and the reciprocal of the sum of the Mahalanobis distances between the first time series data window of the tool cutting force monitoring time series data and the overall mean vector of all the first time series data windows are used as the first distance optimization factor of the first time series data window of the tool cutting force monitoring time series data.

4. The machine tool automation control optimization method based on sensor data according to claim 1, characterized in that: The step of optimizing the distance between the first time series data windows by using the first distance optimization factor to obtain the first optimized distance between the first time series data windows includes: Obtain the first time series data windows of any two of the tool cutting force monitoring time series data, and multiply the first distance optimization factors of each of the any two first time series data windows as the optimization weight of the distance measurement between the any two first time series data windows; and use the calculation result of multiplying the optimization weight of the distance measurement between the any two first time series data windows and the Euclidean distance between the any two first time series data windows as the first optimized distance between the first time series data windows.

5. The machine tool automation control optimization method based on sensor data according to claim 1, characterized in that: The first time series data window of the tool cutting force monitoring time series data is divided into a second data window to obtain a second time series data window of all tool cutting force monitoring time series data, and the local autocorrelation function of the first time series data window is obtained by fitting analysis through the second time series data window, including: Obtain a set second data window length, divide the first time series data window of the tool cutting force monitoring time series data twice according to the second data window length, and obtain the second time series data window in the first time series data window of all the tool cutting force monitoring time series data; obtain a set autocorrelation lag parameter, and evaluate the local autocorrelation function corresponding to the first time series data window of the tool cutting force monitoring time series data for all the second time series data windows according to the set autocorrelation lag parameter to obtain the local autocorrelation function of the first time series data window.

6. The machine tool automation control optimization method based on sensor data according to claim 1, characterized in that: The obtaining of the second distance optimization factor between the first time series data windows through the distribution distance analysis between the local autocorrelation functions includes: Obtain the first time series data windows of any two of the tool cutting force monitoring time series data, obtain the distribution distance between the local autocorrelation functions corresponding to each of the two first time series data windows, and use the constant 1 and the inverse of the sum of the distribution distances between the local autocorrelation functions corresponding to each of the two first time series data windows as the second distance optimization factor between the first time series data windows.

7. The machine tool automation control optimization method based on sensor data according to claim 1, characterized in that: The step of optimizing the first optimized distance between the first time series data windows of the tool cutting force monitoring time series data by using the second distance optimization factor to obtain the second optimized distance includes: Obtain the first optimized distance between the first time series data windows and the second distance optimization factor between the first time series data windows, and use the result of multiplying the second distance optimization factor between the first time series data windows by the first optimized distance between the first time series data windows as the second optimized distance between the first time series data windows of the tool cutting force monitoring time series data.

8. The machine tool automation control optimization method based on sensor data according to claim 1, characterized in that: The clustering analysis is performed by using the second optimized distance between the first time series data windows of the tool cutting force monitoring time series data to obtain the clustering result of the first time series data window of the tool cutting force monitoring time series data, including: Obtain the set number of clusters, and use the second optimized distance between the first time series data windows of the tool cutting force monitoring time series data as a distance metric to perform Clustering is performed to obtain the clustering result of the first time series data window of the tool cutting force monitoring time series data.

9. The machine tool automation control optimization method based on sensor data according to claim 1, characterized in that: The method of optimizing the tool cutting force control parameters by using the clustering results to obtain the optimized tool cutting force control parameters, and performing automatic machine tool tool cutting force control by using the optimized tool cutting force control parameters, comprises: Acquire the clustering result and the real-time tool cutting force monitoring time series data of the automated machine tool, divide the real-time tool cutting force monitoring time series data of the automated machine tool into a first time series data window, acquire the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool, perform cluster evaluation on the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool according to the clustering result, and acquire the cluster to which the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool belongs; optimize the tool cutting force control parameters of the automated machine tool according to the cluster information of the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool, and output the control signal of the optimized tool cutting force control parameters to the machine tool actuator, so as to realize the optimization of cutting force control.

10. The machine tool automation control optimization method based on sensor data according to claim 9, characterized in that: The cluster information of the first time series data window of the real-time tool cutting force monitoring time series data of the automatic machine tool is used to optimize the tool cutting force control parameters of the automatic machine tool, including: Obtain the cutting force mean, standard deviation mean and average zero-crossing rate of the cluster center corresponding to each cluster in the clustering result of the first time series data window of the tool cutting force monitoring time series data; obtain the cluster information to which the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool belongs, and divide the cutting force mean corresponding to the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool by the cutting force mean of the cluster to which the first time series data window of the real-time tool cutting force monitoring time series data belongs as the proportional gain adjustment factor; divide the cutting force standard deviation corresponding to the first time series data window of the real-time tool cutting force monitoring time series data of the automated machine tool by the cutting force standard deviation of the cluster to which the first time series data window of the real-time tool cutting force monitoring time series data belongs as the proportional gain adjustment factor. Integral gain adjustment factor; taking the average zero-crossing rate of the cutting force corresponding to the first time series data window of the real-time tool cutting force monitoring time series data of the automatic machine tool divided by the average zero-crossing rate of the cutting force of the cluster to which the first time series data window of the real-time tool cutting force monitoring time series data belongs as the differential gain adjustment factor; obtaining the real-time PID control parameter of the automatic machine tool; taking the product of the proportional gain adjustment factor and the proportional gain parameter of the real-time PID control of the automatic machine tool as the proportional gain optimization parameter; taking the integral gain parameter of the real-time PID control of the automatic machine tool divided by the integral gain adjustment factor as the integral gain optimization parameter; taking the product of the differential gain adjustment factor and the differential gain parameter of the real-time PID control of the automatic machine tool as the differential gain optimization parameter.

Images