Numerically-controlled machine tool cutter anomaly detection method and system and storage medium

Abstract

The invention relates to a numerically-controlled machine tool cutter anomaly detection method and system and a storage medium, and relates to the technical field of broken cutter detection, and the method comprises the steps: obtaining cutter image data collected by an image collection device; obtaining a direct detection data set of the cutter according to the cutter image data, wherein the direct detection data set comprises offset data, damage data and jumping data; acquiring cutter signal data acquired by signal acquisition equipment; obtaining an indirect detection data set of the cutter according to the cutter signal data, wherein the indirect detection data set comprises power data, load data, cutting force data and signal characteristic data; and constructing a tool anomaly detection model, and inputting the direct detection data set and the indirect detection data set of the tool into the tool anomaly detection model to obtain a tool anomaly detection result. The method has the effect of improving the tool anomaly detection accuracy and reliability.

Description

Technical Field

[0001] The present application relates to the technical field of broken tool detection, and in particular to a method, system and storage medium for detecting tool anomalies in CNC machine tools. Background Art

[0002] When processing pneumatic tool or electric tool accessories, it is usually necessary to use a CNC machine tool. A CNC machine tool is an automated machine tool equipped with a program control system. It can input information into the CNC device through an information carrier, and after calculation and processing, the CNC device sends various control signals to control the movement of the machine tool, and automatically processes the accessories according to the shape and size required by the drawing.

[0003] Since the accessories of pneumatic tools and electric tools are small in size and have relatively precise structures, they require high precision in CNC machine tool processing. When the CNC machine tool tool has abnormal conditions such as broken tools, bending or position shifting, it will have a serious impact on the processing precision. In the prior art, a broken tool detection device is usually used to sample and detect the control signal of the tool, but the tool abnormality information obtained in this way has defects such as single type and limited information, resulting in poor accuracy and reliability of tool abnormality detection. Summary of the invention

[0004] The purpose of this application is to provide a method, system and storage medium for detecting tool anomalies in CNC machine tools to improve the accuracy and reliability of tool anomaly detection.

[0005] In the first aspect, the present application provides a method for detecting abnormality of a CNC machine tool tool using the following technical solution: A method for detecting abnormality of a tool of a numerical control machine tool, comprising: Acquire tool image data acquired by an image acquisition device; Obtaining a direct detection data set of the tool according to the tool image data, wherein the direct detection data set includes offset data, breakage data and runout data; Acquire tool signal data collected by a signal acquisition device; Obtaining an indirect detection data set of the tool according to the tool signal data, wherein the indirect detection data set includes power data, load data, cutting force data and signal characteristic data; A tool anomaly detection model is constructed, and the direct detection data set and indirect detection data set of the tool are input into the tool anomaly detection model to obtain the tool anomaly detection result.

[0006] By adopting the above-mentioned technical scheme, the present application directly detects the tool through image data, and at the same time indirectly detects the tool through signal data, and combines the direct detection data set and the indirect detection data set to predict the abnormal detection results, making the detection range wider and more comprehensive. Even if an unexpected situation causes a certain data collection failure, it will not affect the final detection result, thereby improving the accuracy and reliability of tool abnormality detection.

[0007] Further, the direct detection data set of the tool is obtained according to the tool image data, and the direct detection data set includes offset data, breakage data and runout data, including: Obtain the actual position of the tool according to the tool image data; Obtaining tool standard data, and obtaining the tool standard position according to the tool standard data; The tool length difference and radius difference are obtained according to the standard position of the tool and the actual position of the tool, and the length difference and radius difference are constructed as offset data; Add bias data to the direct detection dataset.

[0008] By adopting the above technical solution, length and radius offset is the basis of tool detection, which is used to ensure the accuracy of tool position, thereby reducing processing errors caused by tool size deviation. Therefore, this application calculates offset data through image recognition and adds direct detection data set to improve the accuracy of tool abnormality detection.

[0009] Further, the direct detection data set of the tool is obtained according to the tool image data, and the direct detection data set includes offset data, breakage data and runout data, and also includes: Identify the actual appearance of the tool based on the tool image data; Acquire tool standard data, and obtain tool standard appearance according to the tool standard data; The damage data is obtained by comparing the actual appearance of the tool with the standard appearance of the tool; Add the damaged data to the direct detection dataset.

[0010] By adopting the above technical solution, the tool damage may include cracks, notches or blunting, which will directly affect the processing quality and tool life. Therefore, this application obtains damage data through image recognition and adds it to the direct detection data set to improve the accuracy of tool abnormality detection.

[0011] Further, the direct detection data set of the tool is obtained according to the tool image data, and the direct detection data set includes offset data, breakage data and runout data, and also includes: Obtain tool image data collected at different times; The runout data is obtained according to the tool image data collected at different times; Add the beating data to the direct detection dataset.

[0012] By adopting the above technical solution, tool runout is the radial runout during the rotation of the tool. Excessive runout may affect the processing accuracy and even cause damage to the tool. Therefore, this application obtains runout data through image recognition and adds it to the direct detection data set to improve the accuracy of tool abnormality detection.

[0013] Furthermore, the indirect detection data set of the tool is obtained according to the tool signal data, and the indirect detection data set includes power data, load data, cutting force data and signal characteristic data, including: The power signal, load signal and cutting force signal of the tool are sampled according to the preset sampling frequency and number of sampling points; Obtaining power data according to a sampling result of the power signal; Obtaining load data according to a sampling result of the load signal; Obtaining cutting force data according to the sampling result of the cutting force signal; Signal characteristic data is obtained according to the sampling results of power signal, load signal and cutting force signal.

[0014] By adopting the above technical solution, the present application obtains various data of the tool in the processing control stage by sampling the tool signal, and extracts the signal characteristics during the processing to make a more accurate analysis of the tool status and processing conditions, so as to improve the accuracy of tool abnormality detection.

[0015] Further, obtaining signal characteristic data according to the power signal, the load signal and the cutting force signal includes: Feature extraction is performed on power signal, load signal and cutting force signal respectively; The feature extraction result is used as signal feature data.

[0016] By adopting the above-mentioned technical solution, signal feature extraction can mine valuable information in the signal and distinguish it from noise and interference to improve the accuracy and efficiency of subsequent processing or analysis. Therefore, this application performs feature extraction on power signals, load signals and cutting force signals respectively to improve the accuracy of tool abnormality detection.

[0017] Furthermore, the tool anomaly detection model is constructed, and the direct detection data set and the indirect detection data set of the tool are input into the tool anomaly detection model to obtain the tool anomaly detection result, which specifically includes: Construct an initial model for tool anomaly detection; Acquire a tool abnormality detection result sample in a tool abnormality detection database and a corresponding direct detection data set sample and an indirect detection data set sample, wherein the tool abnormality detection result sample includes a plurality of abnormal samples and a plurality of normal samples; The sample of the abnormal detection result of the filling tool is used as the output feature, and the corresponding direct detection data set sample and indirect detection data set sample are used as the input feature to construct a sample set; The initial model for tool anomaly detection is trained according to the sample set to obtain a trained tool anomaly detection model; The direct detection data set and the indirect detection data set of the tool are input into the trained tool anomaly detection model to obtain the tool anomaly detection result.

[0018] By adopting the above-mentioned technical scheme, the present application constructs an initial model for tool anomaly detection, and uses tool anomaly detection result samples and corresponding direct detection data set samples and indirect detection data set samples to train the initial model, and uses deep learning methods to enable the model to learn the feature relationship between input samples and output samples, so that the tool anomaly detection results can be obtained based on the input direct detection data set and indirect detection data set, thereby improving the accuracy and intelligence of tool anomaly detection.

[0019] Furthermore, the initial model for tool anomaly detection is trained according to the sample set to obtain a trained tool anomaly detection model, which specifically includes: Divide the sample set into a training set and a test set, and preset the number of training rounds; Input the tool anomaly detection result samples in the training set and the corresponding direct detection data set samples and indirect detection data set samples into the tool anomaly detection initial model, and train according to the preset number of training rounds; When the initial model for tool anomaly detection reaches a preset number of training rounds, a trained tool anomaly detection model is obtained; Input the direct detection data set samples and the indirect detection data set samples in the test set into the trained tool anomaly detection model to obtain the tool anomaly detection result prediction samples; The model parameters of the tool anomaly detection model are adjusted according to the tool anomaly detection result samples in the test set and the corresponding tool anomaly detection result prediction samples, and the trained tool anomaly detection model is obtained according to the adjusted model parameters.

[0020] By adopting the above-mentioned technical solution, the present application tests the trained model through a test set, and adjusts the model parameters according to the tool anomaly detection result samples and the corresponding tool anomaly detection result prediction samples, thereby obtaining a trained tool anomaly detection model, improving the accuracy of the tool anomaly detection model, and thus improving the accuracy of tool anomaly detection.

[0021] In the second aspect, the present application provides a CNC machine tool tool anomaly detection system that adopts the following technical solution: A CNC machine tool tool abnormality detection system, comprising: An image data acquisition module, used to acquire tool image data acquired by an image acquisition device; A direct detection data set acquisition module, used to obtain a direct detection data set of the tool according to the tool image data, wherein the direct detection data set includes offset data, breakage data and runout data; A tool signal data acquisition module is used to acquire tool signal data collected by a signal acquisition device; An indirect detection data set acquisition module is used to obtain an indirect detection data set of a tool according to the tool signal data, wherein the indirect detection data set includes power data, load data, cutting force data and signal characteristic data; The tool anomaly detection module is used to build a tool anomaly detection model, input the direct detection data set and the indirect detection data set of the tool into the tool anomaly detection model, and obtain the tool anomaly detection result.

[0022] By adopting the above-mentioned technical scheme, the present application uses the tool anomaly detection system to perform data collection, data analysis, direct detection and indirect detection fusion on the CNC machine tool tools to finally obtain the anomaly detection results, making the detection scope wider and more comprehensive. Even if an unexpected situation causes a certain data collection failure, it will not affect the final detection result, thereby improving the accuracy and reliability of tool anomaly detection.

[0023] In a third aspect, the present application provides a computer-readable storage medium that adopts the following technical solution: A computer-readable storage medium stores a computer program that can be loaded by a processor and executed by a method for detecting tool abnormality of a numerically controlled machine tool as described in the first aspect.

[0024] By adopting the above technical solution, the present application processes and analyzes the collected data through a computer program, thereby improving the accuracy and intelligence of tool anomaly detection.

[0025] In summary, the present application includes at least one of the following beneficial technical effects: 1. This application directly detects the tool through image data and indirectly detects the tool through signal data. The direct detection data set and the indirect detection data set are combined to predict the abnormal detection results, making the detection range more extensive and comprehensive. Even if an unexpected situation causes a certain data collection failure, it will not affect the final detection result, thereby improving the accuracy and reliability of tool abnormality detection; 2. This application obtains various data of the tool in the processing control stage by sampling the tool signal. Extracting the signal characteristics during the processing process can make a more accurate analysis of the tool state and processing conditions, so as to improve the accuracy of tool abnormality detection; 3. This application constructs an initial model for tool anomaly detection, and uses tool anomaly detection result samples and corresponding direct detection data set samples and indirect detection data set samples to train the initial model. The deep learning method is used to enable the model to learn the feature relationship between input samples and output samples, so that the tool anomaly detection results can be obtained based on the input direct detection data set and indirect detection data set, thereby improving the accuracy and intelligence of tool anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a method flow chart of an embodiment of the present application; Figure 2 is a system structure diagram of an embodiment of the present application; In the figure, 101 is an image data acquisition module; 102 is a direct detection data set acquisition module; 103 is a tool signal data acquisition module; 104 is an indirect detection data set acquisition module; 105 is a tool anomaly detection module. DETAILED DESCRIPTION

[0027] The following is combined with Figure 1 -Attached Figure 2 , further details of this application are given.

[0028] Since the accessories of pneumatic tools and electric tools are small in size and have a relatively precise structure, they require a high degree of precision in CNC machine tool processing. When the CNC machine tool's tool breaks, bends, or moves, it will have a serious impact on the processing accuracy. In the prior art, a broken tool detection device is usually used to sample and detect the tool's control signal. However, the tool abnormality information obtained in this way has defects such as a single type and limited information volume, resulting in poor accuracy and reliability of tool abnormality detection. Therefore, the present application provides a CNC machine tool tool abnormality detection method, referring to Figure 1 ,include: S1. Acquire tool image data acquired by an image acquisition device.

[0029] Specifically, the image acquisition device may be a camera or a laser imaging device, etc., which acquires video or laser imaging during the tool processing process to obtain tool image data.

[0030] S2. Obtain a direct detection data set of the tool according to the tool image data.

[0031] Specifically, image analysis of tool image data can obtain parameters in the tool processing process as a direct detection data set, which includes offset data, breakage data, and runout data.

[0032] S3. Obtain tool signal data collected by the signal acquisition device.

[0033] Specifically, the signal acquisition device may use a sensor to sample the power signal, load signal, and cutting force signal of the tool according to a preset sampling frequency and number of sampling points.

[0034] S4. Obtain an indirect detection data set of the tool according to the tool signal data.

[0035] Specifically, the indirect detection data set can be obtained by signal processing and analysis of the sampling results, which includes power data, load data, cutting force data and signal feature data. S5. Construct a tool anomaly detection model, input the direct detection data set and the indirect detection data set of the tool into the tool anomaly detection model, and obtain the tool anomaly detection result.

[0036] Specifically, constructing a tool anomaly detection model can obtain anomaly detection results through the detected tool direct detection data set and indirect detection data set, combining direct detection and indirect detection.

[0037] The implementation principle of the embodiment of the present application is: directly detect the tool through image data, and indirectly detect the tool through signal data at the same time, combine the direct detection data set and the indirect detection data set to predict the abnormal detection results, so that the detection range is wider and more comprehensive. Even if an unexpected situation causes a certain data collection failure, it will not affect the final detection result, thereby improving the accuracy and reliability of tool abnormality detection.

[0038] Since the directly detected data set includes bias data, damage data and jitter data, step S2 of the embodiment of the present application includes: S11, obtaining the actual position of the tool according to the tool image data.

[0039] Specifically, the actual position of the tool is accurately located through image recognition algorithms, including the actual length and actual radius.

[0040] S12, obtaining standard tool data, and obtaining a standard tool position according to the standard tool data.

[0041] Specifically, the tool standard data is acquired according to the preset processing standard, and the tool standard data includes the tool standard position, that is, the theoretical standard length and standard radius at the moment of image data acquisition during the processing.

[0042] S13. Obtain a tool length difference and a tool radius difference according to the tool standard position and the tool actual position, and construct the length difference and the radius difference as offset data.

[0043] Specifically, the standard position and the actual position are compared and analyzed, the difference between the actual length and the standard length is taken as the length difference, the difference between the actual radius and the standard radius is taken as the radius difference, the length difference and the radius difference are the length offset and the radius offset, and the length offset and the radius offset are constructed as offset data, such as whether there is a length offset, the specific value of the length offset, etc.

[0044] S221, identifying the actual appearance of the tool according to the tool image data.

[0045] Specifically, the actual appearance of the tool can be obtained by visually inspecting the tool image data.

[0046] S222, acquiring standard tool data, and obtaining a standard appearance of the tool according to the standard tool data.

[0047] Specifically, the tool standard data is obtained according to the preset processing standard. The tool standard data includes the tool standard appearance. Since the tools used in different processing procedures are different, the tool standard appearance also changes accordingly.

[0048] S223. Compare the actual appearance of the tool with the standard appearance of the tool to obtain damage data.

[0049] Specifically, tool damage may include cracks, notches or blunting, which will directly affect the processing quality and tool life. Therefore, when comparing the actual appearance of the tool with the standard appearance of the tool, it is necessary to detect the cracks, notches and blunting of the tool, and use the detection results as damage data, such as whether there is a crack, the location and size of the crack, etc.

[0050] S331, obtaining tool image data collected at different times.

[0051] Specifically, during the machining process, tool image data is continuously collected according to a preset cycle to obtain tool image data collected at different times.

[0052] S332, obtaining jitter data according to tool image data collected at different times.

[0053] Specifically, tool runout is the radial runout during tool rotation. Excessive runout may affect machining accuracy or even cause tool damage. Therefore, runout data can be identified based on image data at different times, such as whether tool runout exists and the runout amplitude.

[0054] S4. Add the bias data, damage data and runout data to the direct detection data set.

[0055] Specifically, the direct detection data set is capable of directly identifying the actual condition of the tool during the machining process through the tool image.

[0056] Since the indirect detection data set of the embodiment of the present application includes power data, load data, cutting force data and signal characteristic data, step S4 includes: S41, sampling the power signal, load signal, and cutting force signal of the tool according to a preset sampling frequency and number of sampling points.

[0057] Specifically, the sampling frequency refers to the number of points per second at which the analog signal is sampled by A / D. The number of sampling points refers to the number of data points collected in one sampling process. The relevant control signals of the tool are sampled by presetting the sampling frequency and the number of sampling points.

[0058] S42. Obtain power data according to the sampling result of the power signal.

[0059] Specifically, the power signal is sampled using a power sensor, which is used to monitor the power changes of CNC machine tools. Abnormal power changes may indicate that the tool is broken or worn, so the power data is obtained based on the power signal sampling result to detect tool abnormalities.

[0060] S43. Obtain load data according to the sampling result of the load signal.

[0061] Specifically, the load signal is sampled using a load sensor, which is used to monitor the load condition of the CNC machine tool spindle during the machining process. Excessive load may mean that there is an abnormality in the tool or improper machining conditions. Therefore, the load data is obtained based on the sampling result of the load signal to detect tool abnormalities.

[0062] S44. Obtain cutting force data according to the sampling result of the cutting force signal.

[0063] Specifically, the cutting force signal is sampled using a force sensor, which is used to monitor changes in the cutting force of the tool during the machining process. Abnormal changes in the cutting force may indicate abnormal tool control, so the cutting force data is obtained based on the sampling results of the cutting force signal.

[0064] S45. Obtain signal characteristic data according to the sampling results of the power signal, the load signal and the cutting force signal.

[0065] Specifically, signal feature extraction can mine valuable information in the signal and distinguish it from noise and interference to improve the accuracy and efficiency of subsequent processing or analysis. Therefore, the embodiment of the present application performs feature extraction on the power signal, load signal and cutting force signal respectively, and uses the feature extraction results as signal feature data.

[0066] In the specific implementation process, signal feature extraction can be carried out in a variety of ways. According to the characteristics of the signal and application requirements, the following signal feature extraction methods can be used: Time domain feature extraction, feature extraction based on the change of the signal in time, extracting information such as mean, variance, peak, peak-to-valley difference. Frequency domain feature extraction, converting the signal to the frequency domain through Fourier transform or wavelet transform, extracting information such as frequency components and spectrum morphology. Statistical feature extraction, statistical analysis of the signal, extracting statistics such as mean, standard deviation, skewness, and kurtosis. Spectral feature extraction, extracting the power spectrum density or autocorrelation function of the signal, etc., to obtain the spectral characteristics of the signal. Time-frequency domain feature extraction, using short-time Fourier transform, wavelet packet analysis and other methods, jointly analyzes the signal in the time domain and frequency domain. Empirical mode decomposition, used to analyze complex, nonlinear and non-stationary signals, the eigenmode function obtained by EMD can represent the characteristics of the signal and retain local information. Wavelet transform, suitable for the analysis of non-stationary signals, can simultaneously analyze the time domain and frequency domain information of the signal. Deep learning methods extract features by building deep convolutional neural networks or recurrent neural networks. Adaptive signal decomposition, such as empirical mode decomposition, can decompose signals into some natural mode functions, which is suitable for the analysis of nonlinear and non-stationary signals.

[0067] In another embodiment, in addition to power signals, load signals and cutting force signals, current signals, voltage signals, acoustic emission signals, acceleration signals and other signals can also be sampled, and current data, voltage data, acoustic emission data, acceleration data, etc. can be obtained based on the sampling results, and the above data can be added to the indirect detection data set to make the data in the indirect detection data set more comprehensive.

[0068] The embodiment of the present application constructs an initial model for tool anomaly detection, and uses a deep learning method to enable the model to learn the characteristic relationship between input samples and output samples, so that the tool anomaly detection result can be obtained according to the input direct detection data set and indirect detection data set, thereby improving the accuracy and intelligence of tool anomaly detection. Therefore, step S5 of the embodiment of the present application specifically includes: S51. Construct an initial model for tool anomaly detection.

[0069] Specifically, an initial model for tool anomaly detection is constructed based on LSTM, and model parameters of the initial model for tool anomaly detection are configured. The model parameters include but are not limited to the number of LSTM units, activation function, and input sequence length.

[0070] S52, obtaining a tool anomaly detection result sample in a tool anomaly detection database and a corresponding direct detection data set sample and an indirect detection data set sample.

[0071] Specifically, the tool abnormality detection database includes direct detection data set samples and indirect detection data set samples that are statistically recorded during previous abnormality detection of tools, as well as tool abnormality detection result samples obtained by comprehensive analysis of the direct detection data set samples and the indirect detection data set samples. The tool abnormality detection result samples include several abnormal samples and several normal samples.

[0072] S53, taking the filling tool abnormality detection result samples as output features, and taking the corresponding direct detection data set samples and indirect detection data set samples as input features, to construct a sample set.

[0073] Specifically, direct detection dataset samples, indirect detection dataset samples and corresponding tool anomaly detection result samples are selected to construct a sample set to reflect the characteristic relationship between the direct detection dataset samples, indirect detection dataset samples and tool anomaly detection result samples.

[0074] S54. Train the initial model for tool anomaly detection according to the sample set to obtain a trained tool anomaly detection model.

[0075] Specifically, the sample set is used to train the initial model of tool anomaly detection, wherein the direct detection dataset samples and the indirect detection dataset samples are used as input data, and the tool anomaly detection result samples are used as output data, so that the model can learn the feature relationship between the direct detection dataset samples and the indirect detection dataset samples and the tool anomaly detection result samples.

[0076] S55, inputting the direct detection data set and the indirect detection data set of the tool into the trained tool anomaly detection model to obtain the tool anomaly detection result.

[0077] Specifically, the tool anomaly detection model predicts the corresponding tool anomaly detection result according to the learned feature relationship and the input direct detection data set and indirect detection data set, and outputs the tool anomaly detection result.

[0078] The embodiment of the present application also tests the trained model through a test set, and adjusts the model parameters according to the mean square error of tool anomaly detection, so as to obtain a trained tool anomaly detection model, improve the accuracy of the tool anomaly detection model, and thus improve the accuracy of tool anomaly detection. Therefore, step S54 specifically includes: S541, dividing the sample set into a training set and a test set, and presetting the number of training rounds.

[0079] Specifically, 70% of the sample set is usually used as a training set and 30% as a test set.

[0080] S542, inputting the tool anomaly detection result samples in the training set and the corresponding direct detection data set samples and indirect detection data set samples into the tool anomaly detection initial model, and performing training according to a preset number of training rounds.

[0081] Specifically, during training, the direct detection data set samples and the indirect detection data set samples of the training set are used as input data, and the tool anomaly detection samples are used as output data to train the tool anomaly detection initial model.

[0082] S543: When the initial model for tool anomaly detection reaches a preset number of training rounds, a trained tool anomaly detection model is obtained.

[0083] S544, inputting the direct detection data set samples and the indirect detection data set samples in the test set into the trained tool anomaly detection model to obtain tool anomaly detection result prediction samples.

[0084] Specifically, after the model is trained, the output results of the model are tested using input data in a sample set of known output data.

[0085] S545. Adjust the model parameters of the tool anomaly detection model according to the tool anomaly detection result samples in the test set and the corresponding tool anomaly detection result prediction samples, and obtain a trained tool anomaly detection model according to the adjusted model parameters.

[0086] Specifically, the tool anomaly detection mean square error is obtained based on the tool anomaly detection samples in the test set and the corresponding tool anomaly detection prediction samples, the model parameters of the tool anomaly detection model are adjusted based on the tool anomaly detection mean square error, and the trained tool anomaly detection model is obtained based on the adjusted model parameters.

[0087] In the specific implementation process, mean-square error (MSE) is a measure of the difference between the estimated quantity and the estimated quantity, that is, the difference between the tool anomaly detection sample and the corresponding tool anomaly detection prediction sample. The difference between the tool anomaly detection sample and the corresponding tool anomaly detection prediction sample is judged based on the mean-square error, so as to judge whether the trained tool anomaly detection meets the requirements. When the mean-square error of the tool anomaly detection is greater than the preset error value, the model parameters of the tool anomaly detection model are adjusted cyclically until the mean-square error of the tool anomaly detection is less than the preset error value. When the mean-square error of the tool anomaly detection is less than the preset error value, the trained tool anomaly detection model will be obtained according to the current model parameters.

[0088] The implementation principle of the embodiment of the present application is: by constructing an initial model for tool anomaly detection, and using tool anomaly detection result samples and corresponding direct detection data set samples and indirect detection data set samples to train the initial model, and using deep learning methods to enable the model to learn the feature relationship between input samples and output samples, so that the tool anomaly detection results can be obtained based on the input direct detection data set and indirect detection data set, thereby improving the accuracy and intelligence of tool anomaly detection.

[0089] like Figure 2 As shown, the embodiment of the present application also provides a CNC machine tool tool abnormality detection system, comprising: The image data acquisition module 101 is used to acquire tool image data acquired by an image acquisition device. Specifically, the image acquisition device includes a camera or a laser imaging device, etc. After the image acquisition device acquires the tool image data, it uploads it to the image data acquisition module 101 .

[0090] The direct detection data set acquisition module 102 is used to obtain a direct detection data set of the tool according to the tool image data, and the direct detection data set includes offset data, breakage data and runout data. Specifically, it includes: obtaining the actual position of the tool according to the tool image data; obtaining the standard data of the tool, and obtaining the standard position of the tool according to the standard data of the tool; obtaining the length difference and radius difference of the tool according to the standard position of the tool and the actual position of the tool, and constructing the length difference and radius difference as offset data; adding the offset data to the direct detection data set. Identifying the actual appearance of the tool according to the tool image data; obtaining the standard data of the tool, and obtaining the standard appearance of the tool according to the standard data of the tool; obtaining the breakage data by comparing the actual appearance of the tool with the standard appearance of the tool; and adding the breakage data to the direct detection data set. Obtaining the tool image data collected at different times; obtaining the runout data according to the tool image data collected at different times; and adding the runout data to the direct detection data set.

[0091] The tool signal data acquisition module 103 is used to acquire tool signal data acquired by a signal acquisition device. Specifically, the signal acquisition device includes various sensors, and the sensors acquire tool signal data and upload the acquired tool signal data to the tool signal data acquisition module 103 .

[0092] The indirect detection data set acquisition module 104 is used to obtain the indirect detection data set of the tool according to the tool signal data, and the indirect detection data set includes power data, load data, cutting force data and signal characteristic data. Specifically, it includes: sampling the power signal, load signal and cutting force signal of the tool according to the preset sampling frequency and sampling point number; obtaining power data according to the sampling result of the power signal; obtaining load data according to the sampling result of the load signal; obtaining cutting force data according to the sampling result of the cutting force signal; obtaining signal characteristic data according to the sampling results of the power signal, load signal and cutting force signal.

[0093] The tool anomaly detection module 105 is used to construct a tool anomaly detection model, input the direct detection data set and the indirect detection data set of the tool into the tool anomaly detection model, and obtain the tool anomaly detection result. Specifically, it includes: constructing an initial model for tool anomaly detection; obtaining the tool anomaly detection result samples and the corresponding direct detection data set samples and indirect detection data set samples in the tool anomaly detection database, wherein the tool anomaly detection result samples include a number of abnormal samples and a number of normal samples; using the tool anomaly detection result samples as output features, and using the corresponding direct detection data set samples and indirect detection data set samples as input features to construct a sample set; training the initial model for tool anomaly detection according to the sample set to obtain a trained tool anomaly detection model; inputting the direct detection data set and the indirect detection data set of the tool into the trained tool anomaly detection model to obtain the tool anomaly detection result.

[0094] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: S1, obtaining tool image data acquired by an image acquisition device; S2. Obtaining a direct detection data set of the tool according to the tool image data, wherein the direct detection data set includes offset data, breakage data and runout data; S3, acquiring tool signal data collected by a signal acquisition device; S4, obtaining an indirect detection data set of the tool according to the tool signal data, wherein the indirect detection data set includes power data, load data, cutting force data and signal characteristic data; S5. Construct a tool anomaly detection model, input the direct detection data set and the indirect detection data set of the tool into the tool anomaly detection model, and obtain the tool anomaly detection result.

[0095] In addition, the processor in the computer device executes all the steps of the above-mentioned CNC machine tool tool abnormality detection method when executing the computer program.

[0096] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM).

[0097] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0098] The embodiments of this specific implementation are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. The same components are represented by the same figure marks. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for detecting abnormality of a CNC machine tool tool, characterized in that: include: Acquire tool image data acquired by an image acquisition device; Obtaining a direct detection data set of the tool according to the tool image data, wherein the direct detection data set includes offset data, breakage data and runout data; Acquire tool signal data collected by a signal acquisition device; Obtaining an indirect detection data set of the tool according to the tool signal data, wherein the indirect detection data set includes power data, load data, cutting force data and signal characteristic data; A tool anomaly detection model is constructed, and the direct detection data set and indirect detection data set of the tool are input into the tool anomaly detection model to obtain the tool anomaly detection result.

2. A method for detecting abnormality of a CNC machine tool according to claim 2, characterized in that: The direct detection data set of the tool is obtained according to the tool image data, and the direct detection data set includes offset data, breakage data and runout data, including: Obtain the actual position of the tool according to the tool image data; Obtaining tool standard data, and obtaining the tool standard position according to the tool standard data; The tool length difference and radius difference are obtained according to the standard position of the tool and the actual position of the tool, and the length difference and radius difference are constructed as offset data; Add bias data to the direct detection dataset.

3. A method for detecting abnormality of a CNC machine tool according to claim 3, characterized in that: The direct detection data set of the tool is obtained according to the tool image data, and the direct detection data set includes offset data, damage data and runout data, and also includes: Identify the actual appearance of the tool based on the tool image data; Acquire tool standard data, and obtain tool standard appearance according to the tool standard data; The damage data is obtained by comparing the actual appearance of the tool with the standard appearance of the tool; Add the damaged data to the direct detection dataset.

4. A method for detecting abnormality of a CNC machine tool according to claim 3, characterized in that: The direct detection data set of the tool is obtained according to the tool image data, and the direct detection data set includes offset data, damage data and runout data, and also includes: Obtain tool image data collected at different times; The runout data is obtained according to the tool image data collected at different times; Add the beating data to the direct detection dataset.

5. The method for detecting abnormality of a CNC machine tool tool according to claim 1, characterized in that: The indirect detection data set of the tool is obtained according to the tool signal data, and the indirect detection data set includes power data, load data, cutting force data and signal characteristic data, including: The power signal, load signal and cutting force signal of the tool are sampled according to the preset sampling frequency and number of sampling points; Obtaining power data according to a sampling result of the power signal; Obtaining load data according to a sampling result of the load signal; Obtaining cutting force data according to the sampling result of the cutting force signal; Signal characteristic data is obtained according to the sampling results of power signal, load signal and cutting force signal.

6. A method for detecting abnormality of a CNC machine tool according to claim 5, characterized in that: The signal characteristic data is obtained according to the power signal, the load signal and the cutting force signal, including: Feature extraction is performed on power signal, load signal and cutting force signal respectively; The feature extraction result is used as signal feature data.

7. A method for detecting abnormality of a CNC machine tool tool according to any one of claims 1 to 6, characterized in that: The tool anomaly detection model is constructed, and the direct detection data set and the indirect detection data set of the tool are input into the tool anomaly detection model to obtain the tool anomaly detection result, specifically including: Construct an initial model for tool anomaly detection; Acquire a tool abnormality detection result sample in a tool abnormality detection database and a corresponding direct detection data set sample and an indirect detection data set sample, wherein the tool abnormality detection result sample includes a plurality of abnormal samples and a plurality of normal samples; The sample of the abnormal detection result of the filling tool is used as the output feature, and the corresponding direct detection data set sample and indirect detection data set sample are used as the input feature to construct a sample set; The initial model for tool anomaly detection is trained according to the sample set to obtain a trained tool anomaly detection model; The direct detection data set and the indirect detection data set of the tool are input into the trained tool anomaly detection model to obtain the tool anomaly detection result.

8. A method for detecting abnormality of a CNC machine tool according to claim 7, characterized in that: The training of the initial model for tool anomaly detection according to the sample set to obtain a trained tool anomaly detection model specifically includes: Divide the sample set into a training set and a test set, and preset the number of training rounds; Input the tool anomaly detection result samples in the training set and the corresponding direct detection data set samples and indirect detection data set samples into the tool anomaly detection initial model, and train according to the preset number of training rounds; When the initial model for tool anomaly detection reaches a preset number of training rounds, a trained tool anomaly detection model is obtained; Input the direct detection data set samples and the indirect detection data set samples in the test set into the trained tool anomaly detection model to obtain the tool anomaly detection result prediction samples; The model parameters of the tool anomaly detection model are adjusted according to the tool anomaly detection result samples in the test set and the corresponding tool anomaly detection result prediction samples, and the trained tool anomaly detection model is obtained according to the adjusted model parameters.

9. A CNC machine tool tool abnormality detection system, characterized in that: include: An image data acquisition module, used to acquire tool image data acquired by an image acquisition device; A direct detection data set acquisition module, used to obtain a direct detection data set of the tool according to the tool image data, wherein the direct detection data set includes offset data, breakage data and runout data; A tool signal data acquisition module is used to acquire tool signal data collected by a signal acquisition device; An indirect detection data set acquisition module is used to obtain an indirect detection data set of a tool according to the tool signal data, wherein the indirect detection data set includes power data, load data, cutting force data and signal characteristic data; The tool anomaly detection module is used to build a tool anomaly detection model, input the direct detection data set and the indirect detection data set of the tool into the tool anomaly detection model, and obtain the tool anomaly detection result.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and executes a method for detecting tool abnormality of a numerical control machine tool as claimed in any one of claims 1 to 8.

Images