Three-dimensional five-axis laser cutting intelligent equipment state monitoring method, medium and equipment

Abstract

The invention provides a three-dimensional five-axis laser cutting intelligent equipment state monitoring method, a medium and equipment, and relates to the technical field of machine tool state monitoring, the method comprises the following steps: training an abnormal point identification model based on offline physical examination current data to obtain an abnormal point identification model of a scene; collecting current data and motion signal data of the corresponding shaft; inputting the segmented current data into an abnormal point identification model to capture abnormal points, and calculating an abnormal score based on model evaluation; visualizing the segmented and sliced current data to obtain a current picture, and calculating the similarity between the current picture and the target data picture; and carrying out weighted calculation on the abnormal scores and the similarity, calculating a health state score of each segment, and determining abnormal segments according to the health state scores so as to locate a fault position. According to the method, the conditions of the ABCXYZ axes of the equipment can be accurately monitored under the condition that a sensor and annotation data do not need to be additionally added.

Description

Technical Field

[0001] The present invention relates to the technical field of machine tool status monitoring, and in particular to a three-dimensional five-axis laser cutting intelligent equipment status monitoring method, medium and equipment. Background Art

[0002] The three-dimensional five-axis laser cutting intelligent equipment is the epitome of the company's core technologies. It can achieve high-precision processing of complex surfaces, polyhedrons and special-shaped structures through the coordinated control of each axis. It is a key equipment for solving the complex surface processing needs that cannot be completed by traditional three-axis processing.

[0003] However, in actual use, 3D and 5-axis products often require continuous operation for extended periods of time due to production requirements. Therefore, how to monitor equipment status, provide early warning of potential axis failures, and avoid economic losses caused by downtime has become a key focus in O&M scenarios.

[0004] However, the existing patent literature does not yet involve equipment status detection methods for three-dimensional five-axis laser cutting intelligent equipment. If the scope is expanded to all industrial scenarios, the existing technical paths can be roughly divided into two categories: existing predictive maintenance mainly includes supervised learning based on a large amount of labeled data of "part status variables-remaining service life" and unsupervised algorithms based on status detection of part-related variables.

[0005] Existing predictive maintenance algorithms primarily focus on predicting the service life of parts. This requires labeling the status and remaining service life of the relevant parts at different points in time, which is both labor-intensive and costly. Furthermore, given the numerous parts and brands involved in existing scenarios, and the potential for subsequent part brand replacement, following this technical approach to experiment and label the service life of parts in existing scenarios would be costly. Furthermore, if parts are subsequently replaced, the existing labeling data and models would become meaningless, making the overall method less transferable.

[0006] Existing unsupervised algorithms all rely on methods like Fourier transforms to extract time-frequency features, followed by neural network methods. However, due to the specificity of the current scenario, algorithms based on time-frequency feature extraction have limited robustness and are unable to effectively monitor device status. Summary of the Invention

[0007] The purpose of this invention is to provide a method, medium, and device for monitoring the status of a 3D 5-axis laser cutting intelligent device, aiming to accurately locate fault locations in 3D 5-axis laser cutting intelligent devices in scenarios with limited labeled data. The specific technical solution is as follows:

[0008] A method for monitoring the status of a three-dimensional five-axis laser cutting intelligent device, characterized in that the method comprises the following steps:

[0009] S100: Before the formal physical examination, generate offline physical examination current data based on the baking machine G code, and train an abnormal point recognition model based on the offline physical examination current data to obtain an abnormal point recognition model for the scenario;

[0010] S200, generating physical examination data based on the baking machine G code, collecting current data and motion signal data of the corresponding axis, and segmenting the current data according to the motion signal data;

[0011] S300: Input the segmented current data into an outlier recognition model to capture outliers, output the index of the outlier, and calculate the proportion of points predicted by the outlier recognition model in each segment to obtain an anomaly score based on model evaluation;

[0012] S400, visualizing the segmented and sliced current data to obtain a current image, and calculating the similarity between the current image and the target data image;

[0013] S500 , performing weighted calculation on the anomaly score and the similarity to calculate the health status score of each segment, and determining the segment with an abnormality based on the health status score to locate the fault position.

[0014] Furthermore, the outlier recognition model includes the rocket classifier model and the autoencoder model. The rocket classifier model training process is as follows:

[0015] Generate random convolution kernels by randomizing parameters: randomly select the convolution kernel length from a preset list to randomize the size, randomly sample the random weight parameters from the normal distribution, and randomly generate the bias parameters in the [-1,1] uniform distribution. Randomly generate the expansion coefficient according to the length of the input time series to expand the receptive field of the convolution kernel by expansion to cover the entire sequence;

[0016] Convolution calculation is performed on time series: first, zero padding is performed before and after the input sequence to fill the data, and 0 is inserted between the convolution kernel weights according to the expansion coefficient to expand the coverage of the kernel for expansion processing. Finally, the sequence is traversed, the weighted term of the convolution kernel and the input sequence is calculated, and the bias term is superimposed to implement the convolution calculation;

[0017] Extract features from the convolution result: Perform feature extraction through maximum pooling operation and positive example ratio, and combine the maximum value and positive example ratio of each convolution kernel into a feature vector;

[0018] Linear classifier training: A simple and efficient ridge regression model is used to train the feature vector, and finally a rocket classifier is obtained;

[0019] The AutoEncoder model training process is as follows: normalize and slice the data, train the model based on the fully connected layer combined with the MSE function, predict the model results based on the test data after model training, and perform distribution analysis and visualization of the predicted MSE loss. By examining the loss distribution, the threshold is cut off, and finally the AutoEcoder model and the corresponding threshold value are output.

[0020] Furthermore, in step S300, the anomaly score calculation formula based on model evaluation is as follows:

[0021] score=w1* score_r+w2*score_a

[0022] score_r=n_r / N

[0023] score_a=n_a / N

[0024] Where score is the anomaly score based on model evaluation, w1 and w2 are calculation weights, score_r is the anomaly score evaluated by the rocketClassifier model, score_a is the anomaly score evaluated by the AutoEncoder model, n_r is the number of anomalies captured by the rocketClassifier model, n_a is the number of anomalies captured by the AutoEncoder model, and N is the total number of currents in this round of physical examinations.

[0025] Furthermore, in step S400 , the step of visualizing the segmented and sliced current data is as follows: based on the matplotlib module, the current data is plotted and displayed on a canvas of specified coordinates and size.

[0026] Furthermore, in step S400, calculating the similarity between the current image and the target data image includes the following steps:

[0027] S410, determine whether it is the first round of factory physical examination, if yes, proceed to step S420, if not, proceed to step S430;

[0028] S420: Use the second segment of the first round of factory physical examination current data as the standard segment for all 2i segments, and the first segment as the standard segment for all 2i+1 segments, to obtain a target data image and output and store it, where i is an integer; perform vector representation on the current image based on pixel features, and calculate the similarity between the visualized vector representations of all segments and the visualized vector representations of the corresponding standard segments;

[0029] S430 , performing vector representation on the slice data of the current round of physical examination according to the number of segments split, and calculating similarity with the image vector representation of the corresponding number of segments of the target data image.

[0030] Furthermore, the similarity calculation includes the following steps:

[0031] S440, determining whether the current data is A-axis data or C-axis data, if not, proceeding to step S450, if yes, proceeding to step S460;

[0032] S450, directly performing cosine similarity calculation on the standard data of the vector standard and the vector results of the visualization image of the current round of physical examination data, and outputting the cosine similarity;

[0033] S460, converting the vectorized standard data and the current visualization data of the current examination into image data;

[0034] S470: Block-process the standard data image and the fragment image data of the current round of physical examination, divide the two images to be processed into non-overlapping local blocks, and process them step by step through a sliding window; at the same time, set the SSIM window size and weight parameters;

[0035] S480, calculate local SSIM values: first calculate the local mean, local variance, and local covariance of each block, and then perform brightness similarity, contrast similarity, and structural similarity based on the above results and the SSIM calculation formula to finally obtain the local SSIM value.

[0036] Furthermore, in step S500, the step of determining the abnormal segment based on the health status score to locate the fault position is as follows: determining the abnormal interval based on the first three segments with the lowest health status score, locating the stroke of the G code, and then calculating the index starting point with the largest proportion based on the sliding window of a set number of points to determine the fault position of the corresponding axis.

[0037] Furthermore, step S500 also includes outputting an abnormal prompt: for the part based on the image similarity score value lower than the set value, it is marked as an abnormal fragment, and the fragment information is output; for the fragment whose abnormal proportion predicted by the abnormal point recognition model exceeds the set proportion, it is judged as abnormal, and the top three fragments with the highest abnormal proportion are output.

[0038] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the three-dimensional five-axis laser cutting intelligent equipment status monitoring method as described above.

[0039] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the three-dimensional five-axis laser cutting intelligent device status monitoring method as described above are implemented.

[0040] The present invention provides a method, medium, and device for monitoring the status of a three-dimensional five-axis laser cutting intelligent device, which has the following beneficial effects:

[0041] The present invention generates offline physical examination current data based on the baking machine G code before the formal physical examination, and trains the abnormal point recognition model based on the offline physical examination current data to obtain the abnormal point recognition model of the scenario; generates physical examination data based on the baking machine G code, collects current data and motion signal data of the corresponding axis, and segments the current data according to the motion signal data; inputs the segmented current data into the abnormal point recognition model to capture the abnormal points, outputs the index of the abnormal points, calculates the proportion of the points predicted by the abnormal point recognition model in each segment to obtain the abnormal score based on model evaluation; visualizes the current data after segmentation to obtain a current image, and calculates the similarity between the current image and the target data image; performs weighted calculation on the abnormal score and similarity, calculates the health status score of each segment, determines the abnormal segment according to the health status score to locate the fault position; can realize accurate monitoring of the status of the ABCXYZ axis of the equipment without the need for additional sensors and labeled data. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of a method for monitoring the status of a three-dimensional five-axis laser cutting intelligent device provided by the present invention;

[0043] Figure 2 is an overall flow chart of an embodiment of the present invention;

[0044] Figure 3 1 is a flow chart of outlier recognition model training according to an embodiment of the present invention;

[0045] Figure 4 1 is a flow chart of similarity calculation according to an embodiment of the present invention;

[0046] Figure 5 It is a structural block diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the accompanying drawings provided by the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the drawings are all in a very simplified form and are not in exact proportions. They are only used to facilitate and clearly illustrate the purpose of the embodiments of the present invention.

[0048] Example 1

[0049] This embodiment provides a three-dimensional five-axis laser cutting intelligent equipment status monitoring method, see Figure 1 、 2 As shown, the method includes the following steps:

[0050] S100. Before the formal physical examination, generate offline physical examination current data based on the baking machine G code, specifically including offline physical examination current data of the A, B, C, Y, Z axes and the X gantry axis, and train an abnormal point recognition model based on the offline physical examination current data to obtain an abnormal point recognition model for the scene.

[0051] In one embodiment, the outlier identification model includes a rocket classifier model and an autoencoder model.

[0052] See Figure 3 As shown, the rocket Classifer model training process is as follows:

[0053] Generate random convolution kernels by randomizing parameters: randomly select the convolution kernel length from a preset list to randomize the size, randomly sample the random weight parameters from the normal distribution, and randomly generate the bias parameters in the [-1,1] uniform distribution. Randomly generate the expansion coefficient according to the length of the input time series to expand the receptive field of the convolution kernel by expansion to cover the entire sequence;

[0054] Convolution calculation is performed on time series: first, zero padding is performed before and after the input sequence to fill the data, and 0 is inserted between the convolution kernel weights according to the expansion coefficient to expand the coverage of the kernel for expansion processing. Finally, the sequence is traversed, the weighted term of the convolution kernel and the input sequence is calculated, and the bias term is superimposed to implement the convolution calculation;

[0055] Extract features from the convolution result: extract features through max-pooling operation and PPV (Positive Example Proportion), and combine the maximum value and PPV of each convolution kernel into a feature vector;

[0056] Linear classifier training: A simple and efficient ridge regression model (RidgeClassifier) is used to train the feature vector, and finally a rocket classifier is obtained.

[0057] The AutoEncoder model training process is as follows: the data is normalized and sliced, and the model is trained based on a fully connected layer (N=5) combined with the MSE function. After the model is trained, the model results are predicted based on the test data, and the distribution of the predicted MSE loss is analyzed and visualized. By examining the distribution of the loss, the threshold is cut off, and finally the AutoEcoder model and the corresponding threshold value are output.

[0058] S200 , generating physical examination data based on the baking machine G code, collecting current data and motion signal data of the corresponding axis, and segmenting the current data according to the motion signal data.

[0059] Specifically, the current data is segmented based on the 0-1 motion signal, and only the current data in the motion state is collected. To avoid the impact of data collection, the current of the last segment collected in this round will be filtered and not saved or processed in all subsequent operations.

[0060] S300: Input the segmented current data into an outlier recognition model to capture outliers, output the index of the outlier, and calculate the proportion of points predicted by the outlier recognition model in each segment to obtain an anomaly score based on model evaluation.

[0061] In one embodiment, the anomaly score calculation formula based on model evaluation is as follows:

[0062] score=w1* score_r+w2*score_a

[0063] score_r=n_r / N

[0064] score_a=n_a / N

[0065] Where score is the anomaly score based on model evaluation, w1 and w2 are calculation weights, score_r is the anomaly score evaluated by the rocketClassifier model, score_a is the anomaly score evaluated by the AutoEncoder model, n_r is the number of anomalies captured by the rocketClassifier model, n_a is the number of anomalies captured by the AutoEncoder model, and N is the total number of currents in this round of physical examinations.

[0066] S400 , visualizing the segmented and sliced current data to obtain a current image, and calculating the similarity between the current image and the target data image.

[0067] Specifically, the resulting visualization image is used for subsequent image similarity calculations. The steps are: Using the Matplotlib module, the current data is plotted on a canvas of specified coordinates and size. However, due to performance concerns, the visualization results are not stored directly in .png format. Instead, they are stored as array data in memory using the PIL module.

[0068] In one embodiment, calculating the similarity between the current image and the target data image includes the following steps:

[0069] S410, determine whether it is the first round of factory physical examination, if yes, proceed to step S420, if not, proceed to step S430;

[0070] To determine whether the first round of factory inspection has been completed, the register on the lower computer PLC side is read. When the device leaves the factory, the register value of a specific point will be 0. The value can be used to determine whether the inspection has been completed. When the customer performs the inspection and calls this service, the register value on the PLC side will be reset to 1.

[0071] The purpose of the first round of factory inspection is to use the values of the first round of inspection as standard data. The actual current of each machine may vary slightly. Using the data of the first inspection as the standard data will be more accurate and eliminate the impact of individual differences in machines.

[0072] S420: Use the second segment of the first round of factory physical examination current data as the standard segment for all 2i segments, and the first segment as the standard segment for all 2i+1 segments, to obtain a target data image and output it for storage, where i is an integer; perform vector representation on the current image based on pixel features to avoid memory usage and accelerate the algorithm of the overall process, and calculate the similarity between the visualized vector representations of all segments and the visualized vector representations of the corresponding standard segments;

[0073] S430 , performing vector representation on the slice data of the current round of physical examination according to the number of segments split, and calculating similarity with the image vector representation of the corresponding number of segments of the target data image.

[0074] In one embodiment, see Figure 4 As shown, the calculation of similarity includes the following steps:

[0075] S440, determining whether the current data is A-axis data or C-axis data, if not, proceeding to step S450, if yes, proceeding to step S460;

[0076] S450, directly performing cosine similarity calculation on the standard data of the vector standard and the vector results of the visualization image of the current round of physical examination data, and outputting the cosine similarity;

[0077] cosine_similarity = (A·B) / (||A|| * ||B||)

[0078] Where cosine_similarity represents cosine similarity, A and B represent two vectors, A·B represents the inner product of vector A and vector B, and ||A|| and ||B|| represent the modulus of vector A and vector B.

[0079] S460, converting the vectorized standard data and the current visualization data of the current examination into image data;

[0080] S470: Block the standard data image and the fragment image data of the current round of physical examination, divide the two images to be processed into non-overlapping local blocks, and process them step by step through the sliding window; at the same time, set the SSIM window size and weight parameters, and the default brightness, contrast, and structure weights are all 1;

[0081] S480, Calculate Local SSIM Value: First, calculate the local mean, local variance, and local covariance of each block. Based on these results and the SSIM calculation formula, perform brightness similarity, contrast similarity, and structural similarity to obtain the local SSIM value. The SSIM calculation formula is prior art and will not be further described here.

[0082] S500 , performing weighted calculation on the anomaly score and the similarity to calculate the health status score of each segment, and determining the segment with an abnormality based on the health status score to locate the fault position.

[0083] In one embodiment, the steps for determining the abnormal segments based on the health status score to locate the fault position are as follows: determining the abnormal interval based on the first three segments with the lowest health status scores, locating the G-code stroke, and then calculating the index starting point with the largest proportion based on a sliding window of a set number of points (for example, 500 points) to determine the fault position of the corresponding axis.

[0084] In one embodiment, step S500 also includes outputting an abnormality prompt: for parts based on image similarity scores lower than a set value (for example, lower than 0.7), they are marked as abnormal fragments, and the fragment information is output; for fragments whose abnormality ratio predicted by the abnormal point recognition model exceeds a set ratio (for example, 20%), they are judged as abnormal, and the top three fragments with the highest abnormality ratio are output.

[0085] The present invention provides a three-dimensional five-axis laser cutting intelligent equipment status monitoring method. Before a formal physical examination, offline physical examination current data is generated based on the baking machine G code, and an abnormal point recognition model is trained based on the offline physical examination current data to obtain an abnormal point recognition model for the scene; physical examination data is generated based on the baking machine G code, current data and motion signal data of the corresponding axis are collected, and the current data is segmented according to the motion signal data; the segmented current data is input into the abnormal point recognition model to capture the abnormal points, the index of the abnormal points is output, and the proportion of the points predicted by the abnormal point recognition model in each segment is calculated to obtain an abnormal score based on model evaluation; the current data after segmentation is visualized to obtain a current image, and the similarity between the current image and the target data image is calculated; the abnormal score and the similarity are weighted to calculate the health status score of each segment, and the abnormal segment is determined according to the health status score to locate the fault position; it can realize accurate monitoring of the status of the ABCXYZ axis of the equipment without the need for additional sensors and labeled data.

[0086] Example 2

[0087] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for monitoring the status of a three-dimensional five-axis laser cutting intelligent device described above are implemented.

[0088] The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium may also include a combination of the above types of memory.

[0089] Example 3

[0090] This embodiment provides a computer device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the three-dimensional five-axis laser cutting intelligent device status monitoring method described above are implemented.

[0091] like Figure 5As shown, the computer device may include: at least one processor 71, such as a CPU (Central Processing Unit), at least one communication interface 73, a memory 74, and at least one communication bus 72. The communication bus 72 is used to realize the connection and communication between these components. The communication interface 73 may include a display screen (Display) and a keyboard (Keyboard), and the optional communication interface 73 may also include a standard wired interface and a wireless interface. The memory 74 may be a high-speed RAM memory (Random Access Memory, volatile random access memory) or a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory 74 may optionally be at least one storage device located away from the aforementioned processor 71. The memory 74 stores application programs, and the processor 71 calls the program code stored in the memory 74 to execute any of the above method steps.

[0092] The communication bus 72 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The communication bus 72 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0093] Among them, the memory 74 may include a volatile memory (English: volatile memory), such as a random-access memory (English: random-access memory, abbreviated: RAM); the memory may also include a non-volatile memory (English: non-volatile memory), such as a flash memory (English: flash memory), a hard disk drive (English: hard disk drive, abbreviated: HDD) or a solid-state drive (English: solid-state drive, abbreviated: SSD); the memory 74 may also include a combination of the above types of memory.

[0094] The processor 71 may be a central processing unit (CPU), a network processor (NP), or a combination of a CPU and a NP.

[0095] The processor 71 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0096] Optionally, the memory 74 is further configured to store program instructions. The processor 71 can call the program instructions to implement the three-dimensional five-axis laser cutting intelligent device status monitoring method of the present invention.

[0097] Those skilled in the art should understand that the present invention can be implemented in many other specific forms without departing from the spirit and scope of the present invention. Based on the embodiments of the present invention, any changes and modifications made by ordinary technicians in the field of the present invention in accordance with the above disclosure are within the scope of protection of the claims.

Claims

1. A three-dimensional five-axis laser cutting intelligent equipment status monitoring method, characterized in that: The method comprises the following steps: S100: Before the formal physical examination, generate offline physical examination current data based on the baking machine G code, and train an abnormal point recognition model based on the offline physical examination current data to obtain an abnormal point recognition model for the scenario; S200, generating physical examination data based on the baking machine G code, collecting current data and motion signal data of the corresponding axis, and segmenting the current data according to the motion signal data; S300: Input the segmented current data into an outlier recognition model to capture outliers, output the index of the outlier, and calculate the proportion of points predicted by the outlier recognition model in each segment to obtain an anomaly score based on model evaluation; S400, visualizing the segmented and sliced current data to obtain a current image, and calculating the similarity between the current image and the target data image; S500 , performing weighted calculation on the anomaly score and the similarity to calculate the health status score of each segment, and determining the segment with an abnormality based on the health status score to locate the fault position.

2. The three-dimensional five-axis laser cutting intelligent equipment status monitoring method according to claim 1 is characterized in that: The outlier recognition model includes the Rocket Classifier model and the AutoEncoder model. The Rocket Classifier model training process is as follows: Generate random convolution kernels by randomizing parameters: randomly select the convolution kernel length from a preset list to randomize the size, randomly sample the random weight parameters from the normal distribution, and randomly generate the bias parameters in the [-1,1] uniform distribution. Randomly generate the expansion coefficient according to the length of the input time series to expand the receptive field of the convolution kernel by expansion to cover the entire sequence; Convolution calculation is performed on time series: first, zero padding is performed before and after the input sequence to fill the data, and 0 is inserted between the convolution kernel weights according to the expansion coefficient to expand the coverage of the kernel for expansion processing. Finally, the sequence is traversed, the weighted term of the convolution kernel and the input sequence is calculated, and the bias term is superimposed to implement the convolution calculation; Extract features from the convolution result: Perform feature extraction through maximum pooling operation and positive example ratio, and combine the maximum value and positive example ratio of each convolution kernel into a feature vector; Linear classifier training: A simple and efficient ridge regression model is used to train the feature vector, and finally a rocket classifier is obtained; The AutoEncoder model training process is as follows: normalize and slice the data, train the model based on the fully connected layer combined with the MSE function, predict the model results based on the test data after model training, and perform distribution analysis and visualization on the predicted MSE loss. By viewing the distribution of the loss, the threshold is truncated, and finally the AutoEcoder model and the corresponding threshold value of the model are output.

3. The three-dimensional five-axis laser cutting intelligent equipment status monitoring method according to claim 2 is characterized in that: In step S300, the anomaly score calculation formula based on model evaluation is as follows: score=w1* score_r+w2*score_a score_r=n_r / N score_a=n_a / N Where score is the anomaly score based on model evaluation, w1 and w2 are calculation weights, score_r is the anomaly score evaluated by the rocketClassifier model, score_a is the anomaly score evaluated by the AutoEncoder model, n_r is the number of anomalies captured by the rocketClassifier model, n_a is the number of anomalies captured by the AutoEncoder model, and N is the total number of currents in this round of physical examinations.

4. The three-dimensional five-axis laser cutting intelligent equipment status monitoring method according to claim 1 is characterized in that: In step S400 , the step of visualizing the segmented and sliced current data is as follows: based on the matplotlib module, the current data is plotted and displayed on a canvas of specified coordinates and size.

5. The three-dimensional five-axis laser cutting intelligent equipment status monitoring method according to claim 1 is characterized in that: In step S400, calculating the similarity between the current image and the target data image includes the following steps: S410, determine whether it is the first round of factory physical examination, if yes, proceed to step S420, if not, proceed to step S430; S420: Use the second segment of the first round of factory physical examination current data as the standard segment for all 2i segments, and the first segment as the standard segment for all 2i+1 segments, to obtain a target data image and output and store it, where i is an integer; perform vector representation on the current image based on pixel features, and calculate the similarity between the visualized vector representations of all segments and the visualized vector representations of the corresponding standard segments; S430 , performing vector representation on the slice data of the current round of physical examination according to the number of segments split, and calculating similarity with the image vector representation of the corresponding number of segments of the target data image.

6. The three-dimensional five-axis laser cutting intelligent equipment status monitoring method according to claim 5 is characterized in that: The calculation of similarity includes the following steps: S440, determining whether the current data is A-axis data or C-axis data, if not, proceeding to step S450, if yes, proceeding to step S460; S450, directly performing cosine similarity calculation on the standard data of the vector standard and the vector results of the visualization image of the current round of physical examination data, and outputting the cosine similarity; S460, converting the vectorized standard data and the current visualization data of the current examination into image data; S470: Block-process the standard data image and the fragment image data of the current round of physical examination, divide the two images to be processed into non-overlapping local blocks, and process them step by step through a sliding window; at the same time, set the SSIM window size and weight parameters; S480, calculate local SSIM values: first calculate the local mean, local variance, and local covariance of each block, and then perform brightness similarity, contrast similarity, and structural similarity based on the above results and the SSIM calculation formula to finally obtain the local SSIM value.

7. The three-dimensional five-axis laser cutting intelligent equipment status monitoring method according to claim 1 is characterized in that: In step S500, the steps of determining the abnormal segments based on the health status scores to locate the fault position are as follows: determining the abnormal interval based on the first three segments with the lowest health status scores, locating the G code stroke, and then calculating the index starting point with the largest proportion based on the sliding window of a set number of points to determine the fault position of the corresponding axis.

8. The three-dimensional five-axis laser cutting intelligent equipment status monitoring method according to claim 1 is characterized in that: Step S500 also includes outputting abnormal prompts: marking the part whose image similarity score is lower than the set value as an abnormal segment, and outputting the segment information; judging the segment whose abnormal proportion predicted by the abnormal point recognition model exceeds the set proportion as abnormal, and outputting the top three segments with the highest abnormal proportion.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the three-dimensional five-axis laser cutting intelligent equipment status monitoring method as described in any one of claims 1 to 8 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the three-dimensional five-axis laser cutting intelligent equipment status monitoring method according to any one of claims 1 to 8 are implemented.

Images