Laser cutting speed control method based on multi-source analysis

Abstract

The present invention discloses a laser cutting speed control method based on multi-source analysis, which relates to the technical field related to laser cutting. The method includes: searching in the control database of the laser cutting speed control platform; extracting cutting edge images to obtain multiple historical cutting edge image sets; taking the same cutting position as the clustering target to perform cross-record extraction; identifying the mapping relationship between K cutting positions and target speed defect edge values; limiting the cutting speed tolerance bandwidth; generating a target cutting speed tolerance bandwidth sequence, and authenticating any two adjacent target bandwidths for overlap; monitoring and identifying the cutting speed change state, and performing laser cutting speed control. The method solves the technical problems that the existing laser cutting control relies on a single data source, ignores the influence of burrs and ablation in actual processing, and leads to poor cutting speed control accuracy and stability, and achieves the technical effect of improving the cutting speed control accuracy and cutting efficiency.

Description

Technical Field

[0001] The present application relates to the field related to laser cutting technology, and specifically to a laser cutting speed control method based on multi-source analysis. Background Art

[0002] As a high-precision and high-efficiency processing technology, laser cutting is widely used in many fields such as automobiles, aviation, electronics, and medical devices. With the rapid development of the industrial field and the diversification of market demand, higher requirements are placed on the control of laser cutting speed, especially in the control of cutting speed. Cutting speed is not only directly related to production efficiency, but also affects cutting quality and material utilization. However, traditional laser cutting speed control methods mostly rely on a single sensor or parameter, which is often difficult to cope with complex processing environments and changeable material properties. It is impossible to accurately control the laser cutting speed, resulting in burrs on the workpiece.

[0003] Therefore, at the current stage, there is a technical problem in the laser cutting speed control related technologies that rely on a single data source and ignore the impact of burrs and ablation in actual processing, resulting in poor cutting speed control accuracy and stability. Summary of the invention

[0004] The present application provides a laser cutting speed control method based on multi-source analysis, and adopts technical means such as edge image extraction and mapping relationship recognition to solve the technical problems existing in the existing laser cutting control, such as reliance on a single data source, ignoring the influence of burrs and ablation in actual processing, resulting in poor cutting speed control accuracy and stability, thereby achieving the technical effect of improving the cutting speed control accuracy and cutting efficiency.

[0005] The present application provides a laser cutting speed control method based on multi-source analysis, the method comprising: searching in a control database of a laser cutting speed control platform based on a target cutting object to obtain a plurality of historical cutting speed control records and a cutting speed tolerance bandwidth; traversing the plurality of historical cutting speed control records to extract cutting edge images to obtain a plurality of historical cutting edge image sets; taking the same cutting position as a clustering target, performing cross-record extraction on the plurality of historical cutting edge image sets to generate K historical cutting edge image sets, wherein the K historical cutting edge image sets correspond one-to-one to the K cutting positions of the target cutting object, and each historical cutting edge image has a cutting speed identifier; based on the K historical cutting edge image sets, the K The mapping relationship between the cutting position and the target speed defect edge value is identified to generate a cutting position-target speed defect edge value endpoint mapping relationship; the cutting speed tolerance bandwidth is limited according to the cutting position-target speed defect edge value endpoint mapping relationship to generate K target cutting speed tolerance bandwidths; based on the cutting order of the K cutting positions, a target cutting speed tolerance bandwidth sequence is generated, and any two adjacent target cutting speed tolerance bandwidths in the target cutting speed tolerance bandwidth sequence are overlapped and authenticated. If the authentication fails, a speed change calibration point is set at the corresponding cutting position; a speed monitoring sensor is set at the speed change calibration point to monitor and identify the cutting speed change state, and the laser cutting speed is controlled according to the monitoring and identification results.

[0006] In a possible implementation, the mapping relationship between the K cutting positions and the target speed defect edge values ​​is identified based on the K historical cutting edge image sets to generate a cutting position-target speed defect edge value mapping relationship, and the following processing is performed: edge defect identification is performed on the K historical cutting edge image sets to obtain K edge defect feature sets; speed defect edge identification is performed on the K edge defect feature sets in combination with the cutting speed identifier to generate K speed defect edge value sets; cross-defect fusion is performed on the K speed defect edge value sets respectively to generate K target speed defect edge value endpoints; based on the mapping of the K target speed defect edge value endpoints with the corresponding edge positions, the cutting position-target speed defect edge value endpoint mapping relationship is generated.

[0007] In a possible implementation, edge defect recognition is performed on the K historical cutting edge image sets to obtain K edge defect feature sets, and the following processing is performed: multiple sample historical cutting edge image sets and multiple sample edge defect feature sets are obtained as sample data; the sample data is divided into a training sample set and a test sample set according to a preset division ratio, the training sample set is used to train a framework built based on a convolutional neural network, the mapping relationship between edge cutting images and edge defect features is learned, and the learning results are tested using the test sample set, and the network parameters are updated according to the test accuracy until the output test accuracy meets the preset test accuracy requirement, thereby generating a trained defect recognition network layer; the defect recognition network layer is used to perform edge defect recognition on the K historical cutting edge image sets to obtain the K edge defect feature sets.

[0008] In a possible implementation, the K edge defect feature sets are combined with the cutting speed identifier to perform speed defect edge identification to generate K speed defect edge value sets, and the following processing is also performed: edge defect feature clustering analysis is performed on the K edge defect feature sets respectively to generate K edge defect feature clusters; the mode of multiple cutting speeds of multiple cutting defect feature sets corresponding to each edge defect feature cluster in the K edge defect feature clusters is extracted as the speed defect edge value to generate K speed defect edge value sets.

[0009] In a possible implementation, the K speed defect edge value sets are respectively subjected to cross-defect fusion to generate K target speed defect edge value endpoints, and the following processing is also performed: the K speed defect edge value sets are traversed to perform maximum value extraction to generate K first target speed defect edge value endpoints; the K speed defect edge value sets are traversed to perform minimum value extraction to generate K second target speed defect edge value endpoints; based on the K first target speed defect edge value endpoints and the K second target speed defect edge value endpoints, the K target speed defect edge value endpoints are generated.

[0010] In a possible implementation, a speed monitoring sensor is set at the speed change calibration point to monitor and identify the changing state of the cutting speed, and the laser cutting speed is controlled according to the monitoring and identification results. The following processing is also performed: a steady-state monitoring window is set, and the monitoring data of the speed monitoring sensor at the speed change calibration point within the steady-state monitoring window is extracted to generate a monitoring data set; an outlier analysis is performed on the monitoring data set to generate a first outlier degree; it is determined whether the first outlier degree is less than or equal to a preset outlier degree threshold, and if not, an auxiliary control instruction is generated to call the laser cutting speed control platform to perform auxiliary control on the laser cutting speed.

[0011] In a possible implementation, the monitoring data set is subjected to an outlier analysis to generate a first outlier degree, and the following processing is also performed: the monitoring data set is traversed to perform mean calculation to obtain a first monitoring mean; taking the first monitoring mean as a starting point, iterating with a preset concentration step to obtain first monitoring data, and performing initial concentration identification on the first monitoring mean to obtain an initial concentration; determining whether the initial concentration of the first monitoring mean is greater than or equal to the first concentration of the first monitoring data, and if not, taking the first monitoring data as stage monitoring data; after multiple iterations, taking the stage monitoring data corresponding to the maximum concentration value as target monitoring data; counting the amount of outlier data whose distance to the target monitoring data is greater than the preset concentration step, and comparing the amount of outlier data with the total amount of data in the monitoring data set to generate a first outlier degree.

[0012] In a possible implementation, a target cutting speed tolerance bandwidth sequence is generated based on the cutting order of the K cutting positions, and any two adjacent target cutting speed tolerance bandwidths in the target cutting speed tolerance bandwidth sequence are overlapped and authenticated, and the following processing is performed: if the authentication is passed, the median of the overlapping target cutting speed tolerance bandwidth is used as the cutting speed of the two adjacent cutting positions, and the cutting speed is sent to the laser cutting speed control platform for speed control.

[0013] It is proposed that through the laser cutting speed control method based on multi-source analysis proposed in this application, based on the target cutting object, a search is performed in the control database of the laser cutting speed control platform to obtain multiple historical cutting speed control records and cutting speed tolerance bandwidth; multiple historical cutting speed control records are traversed to extract cutting edge images to obtain multiple historical cutting edge image sets; multiple historical cutting edge image sets are cross-record extracted with the same cutting position as the clustering target to generate K historical cutting edge image sets, and each historical cutting edge image has a cutting speed identifier; based on the K historical cutting edge image sets, the mapping between K cutting positions and target speed defect edge values ​​is The mapping relationship between the cutting position and the target speed defect edge value endpoint is identified to generate a mapping relationship between the cutting position and the target speed defect edge value endpoint; the cutting speed tolerance bandwidth is limited according to the mapping relationship between the cutting position and the target speed defect edge value endpoint to generate K target cutting speed tolerance bandwidths; based on the cutting order of the K cutting positions, a target cutting speed tolerance bandwidth sequence is generated, and any two adjacent target cutting speed tolerance bandwidths in the target cutting speed tolerance bandwidth sequence are overlapped for authentication. If the authentication fails, a speed change calibration point is set at the corresponding cutting position; a speed monitoring sensor is set at the speed change calibration point to monitor and identify the cutting speed change state, and the laser cutting speed is controlled according to the monitoring and identification results. The technical problem that the existing laser cutting control relies on a single data source, ignores the influence of burrs and ablation in actual processing, and leads to poor cutting speed control accuracy and stability is solved, and the technical effect of improving the cutting speed control accuracy and cutting efficiency is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A schematic flow chart of a laser cutting speed control method based on multi-source analysis provided in an embodiment of the present application;

[0016] Figure 2 A schematic diagram of a flow chart for generating a cutting position-target speed defect edge value mapping relationship in a laser cutting speed control method based on multi-source analysis provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0020] The present application embodiment provides a laser cutting speed control method based on multi-source analysis, such as Figure 1 As shown, the method includes:

[0021] Step S100, searching the control database of the laser cutting speed control platform based on the target cutting object to obtain a plurality of historical cutting speed control records and cutting speed tolerance bandwidth. A search is performed in the control database of the laser cutting speed control platform to obtain multiple historical cutting speed control records and cutting speed tolerance bandwidths of the target cutting object. Specifically, the target cutting object refers to the specific material or workpiece that needs to be laser cut at present. The laser cutting speed control platform is a comprehensive platform that integrates the laser cutting speed control algorithm and the database management system, which is responsible for receiving cutting tasks, analyzing the characteristics of the cutting object, selecting the appropriate cutting speed, and monitoring the cutting process. The control database is the core component of the laser cutting speed control platform, which is used to store and manage historical data, algorithm parameters, cutting conditions and other information related to cutting speed control. The historical cutting speed control records are the speed control parameters and results used in the past when laser cutting similar target cutting objects. Each record contains information such as the characteristics of the cutting object, the cutting speed used, and the cutting quality evaluation. The cutting speed tolerance bandwidth refers to the interval in which the cutting speed is allowed to fluctuate within a certain range for a specific cutting object while ensuring the cutting quality. The determination of the tolerance bandwidth is usually based on the analysis of the historical cutting speed control records and the in-depth understanding of the characteristics of the cutting object, which helps to deal with various uncertain factors that may occur in the actual cutting process, such as material non-uniformity and fluctuations in equipment performance.

[0022] Step S200, traverse the multiple historical cutting speed control records to extract cutting edge images, and obtain multiple historical cutting edge image sets. Access multiple historical cutting speed control records stored in the control database one by one, extract the corresponding cutting edge image for each historical cutting speed control record, and use the extracted multiple cutting edge images to form multiple historical cutting edge image sets, wherein the cutting edge image is important visual information reflecting the cutting effect, which shows the shape, flatness, roughness and other characteristics of the edge of the workpiece after cutting.

[0023] Step S300, performing cross-record extraction on the multiple historical cutting edge image sets with the same cutting position as the clustering target to generate K historical cutting edge image sets, wherein the K historical cutting edge image sets correspond one-to-one to the K cutting positions of the target cutting object, and each historical cutting edge image has a cutting speed identifier. Taking the same cutting position as the clustering target, multiple historical cutting edge image sets are cross-recorded extracted to generate K historical cutting edge image sets. Specifically, in the laser cutting process, the cutting position refers to the specific area or point on the workpiece that is irradiated and cut by the laser beam. Taking the same cutting position as the clustering target, the images corresponding to the same cutting position of the target cutting object are screened out from multiple historical cutting edge image sets, including image feature extraction and matching, to ensure that the image at the same position of the target cutting object is accurately extracted, and the historical cutting edge images are divided into K sets according to the image features, each set corresponding to a cutting position of the target cutting object, and finally K historical cutting edge image sets are obtained, each set contains all the historical cutting edge images related to the cutting position, and each historical cutting edge image is added with a cutting speed mark. When we analyze a historical cutting edge image set, we can know the cutting speed corresponding to all the images in the set, thereby establishing the association between the cutting position, cutting speed and cutting edge image.

[0024] Step S400 , identifying the mapping relationship between the K cutting positions and the target speed defect edge values ​​based on the K historical cutting edge image sets, and generating a cutting position-target speed defect edge value endpoint mapping relationship. Specifically, in the laser cutting process, the choice of cutting speed directly affects the quality of the cutting edge. If the cutting speed is too fast or too slow, defects such as burrs, slag, and ablation may occur on the cutting edge. The target speed defect edge value refers to the quantitative index of the cutting edge defect corresponding to each cutting position at different cutting speeds, which is obtained by feature extraction and quantitative evaluation of the cutting edge image. Based on the K historical cutting edge image sets that have been obtained, feature extraction and defect analysis are performed on the historical cutting edge images in each set. For example, features related to cutting edge defects are extracted through edge detection, texture analysis, etc. These feature data are used to identify the mapping relationship between each cutting position and the target speed defect edge value. For example, regression analysis is used to fit the relationship between the cutting position, cutting speed, and cutting edge defect value, and finally a cutting position-target speed defect edge value endpoint mapping relationship is generated, which describes the cutting speed range that should be selected to obtain the best cutting edge quality at different cutting positions. Each cutting position corresponds to one or more endpoints of the target speed defect edge value, and these endpoints constitute the acceptable and unacceptable cutting speed boundaries at different cutting positions.

[0025] In a possible implementation, step S400 further includes step S410, performing edge defect recognition on the K historical cutting edge image sets to obtain K edge defect feature sets. For each historical cutting edge image set, edge defect recognition is performed, for example, edge detection algorithms (such as Canny edge detection, Sobel edge detection, etc.) are used to identify the edge portion in the image, and then detailed edge features are further extracted through morphological analysis, texture analysis and other techniques, including edge roughness, unevenness, color change, etc. For each cutting position of the target cutting object, a corresponding edge defect feature set is obtained, that is, K edge defect feature sets are obtained, each set contains edge defect feature information of all historical cutting edge images at the cutting position. It also includes step S420, performing speed defect edge recognition on the K edge defect feature sets in combination with the cutting speed identifier to generate K speed defect edge value sets. On the basis of the edge defect feature set that has been obtained, the cutting speed identifier is further used to identify the edge defect conditions at different speeds, and a defect edge value set related to the speed is generated. Specifically, in combination with the cutting speed identifier in each edge defect feature set, the feature data is associated with the corresponding cutting speed, the influence of the cutting speed on the edge defect is analyzed, and the speed defect edge is identified, and the significant change points or trends of the edge defects at different cutting speeds are identified. Finally, K speed defect edge value sets are generated, each set corresponds to a cutting position, and contains the defect edge values ​​corresponding to different cutting speeds at the position. It also includes step S430, respectively performing cross-defect fusion on the K speed defect edge value sets to generate K target speed defect edge value endpoints. Among them, cross-defect fusion refers to integrating and integrating the defect edge values ​​at different speeds, obtaining comprehensive information about the speed defect edge value at each cutting position, so as to generate K target speed defect edge value endpoints, which reflect the distribution of the speed defect edge values ​​at different cutting positions. It also includes step S440, based on mapping the K target speed defect edge value endpoints with the corresponding edge positions, generating the cutting position-target speed defect edge value endpoint mapping relationship. The obtained target speed defect edge value endpoints are associated with specific cutting positions to form a complete mapping relationship. Specifically, the target speed defect edge value endpoints are mapped with their corresponding cutting positions, which contains information about each cutting position and its corresponding target speed defect edge value endpoint (including upper and lower limits). This mapping relationship is used to guide the selection and control of the cutting speed. When faced with a new cutting task, this mapping relationship can be consulted according to the characteristics of the target cutting object and the required cutting quality, so as to quickly determine the cutting speed range that should be selected at different cutting positions.

[0026] In a possible implementation, step S410 further includes step S411, obtaining multiple sample historical cutting edge image sets and multiple sample edge defect feature sets as sample data. The sample historical cutting edge image set refers to the cutting edge images captured in different cutting tasks in the past, which record the detailed information such as the shape and texture of the edge during the cutting process; the sample edge defect feature set refers to the feature data related to edge defects extracted in the process of analyzing the historical cutting edge images, which is used to describe the quality status of the cutting edge. It also includes step S412, dividing the sample data into a training sample set and a test sample set according to a preset division ratio, using the training sample set to train the framework built based on the convolutional neural network, learning the mapping relationship between the edge cutting image and the edge defect feature, and using the test sample set to test the learning results, and updating the network parameters according to the test accuracy until the output test accuracy meets the preset test accuracy requirements, and generating a trained defect recognition network layer. Among them, the defect recognition network layer is a network model built based on a convolutional neural network (CNN), which can automatically learn and extract features related to edge defects in the image, such as edge roughness, unevenness, fracture, etc. It also includes step S413, using the defect recognition network layer to perform edge defect recognition on the K historical cutting edge image sets to obtain the K edge defect feature sets. The K historical cutting edge image sets are sent as input data to the defect recognition network layer. The defect recognition network layer will perform edge defect recognition on each image in each image set one by one, identify whether there are edge defects in the image, and extract feature information related to these defects. These feature information constitute K edge defect feature sets, which are used to describe the characteristics and degree of edge defects in the image.

[0027] In a possible implementation, step S420 further includes step S421, performing edge defect feature clustering analysis on the K edge defect feature sets respectively to generate K edge defect feature clusters. Data processing is performed on each edge defect feature set using a clustering analysis algorithm to merge data points with similar defect features into the same cluster, and each cluster represents one or a class of specific edge defect features. Specifically, when clustering analysis is performed on the K edge defect feature sets, the feature data in each set will be regarded as input data, and the similarity between them will be calculated based on the feature values ​​of these data (such as the size, shape, position, etc. of the defect), and similar features will be merged into the same cluster. After generating K edge defect feature clusters, the feature data in each cluster will have similar defect characteristics. It also includes step S422, extracting the mode of multiple cutting speeds of multiple cutting defect feature sets corresponding to each edge defect feature cluster in the K edge defect feature clusters as speed defect edge values, and generating K speed defect edge value sets. Specifically, the mode of the cutting speeds corresponding to each feature cluster (i.e., the speed value with the highest frequency of occurrence) is extracted as the speed defect edge value. The mode can represent the most typical cutting speed under the feature cluster, that is, at this cutting speed, the edge defect corresponding to the feature cluster is most likely to appear. K speed defect edge values ​​are obtained, which reveal the cutting speed that is most likely to cause defects under different types of edge defects. Finally, the K speed defect edge values ​​are combined into K speed defect edge value sets.

[0028] In a possible implementation, step S430 further includes step S431, traversing the K speed defect edge value sets to extract the maximum value and generate K first target speed defect edge value endpoints. Traversing the K speed defect edge value sets, finding the maximum speed defect edge value therefrom, and using these maximum values ​​as the first target speed defect edge value endpoints to generate K first target speed defect edge value endpoints. Specifically, the first target speed defect edge value endpoint represents the cutting speed that is most likely to cause serious defects under the specific edge defect feature cluster. It also includes step S432, traversing the K speed defect edge value sets to extract the minimum value and generate K second target speed defect edge value endpoints. Traversing each speed defect edge value set, and finding the minimum speed defect edge value from each set, these minimum values ​​are used as the second target speed defect edge value endpoints, and generating K second target speed defect edge value endpoints. Specifically, the second target speed defect edge value endpoint represents the relatively safe lower limit of the cutting speed under the specific edge defect feature cluster, that is, cutting above this speed can avoid certain types of edge defects. The method further includes step S433, generating the K target speed defect edge value endpoints based on the K first target speed defect edge value endpoints and the K second target speed defect edge value endpoints. Each pair of corresponding first target speed defect edge value endpoints and second target speed defect edge value endpoints are combined to form K target speed defect edge value endpoints, each endpoint pair representing a speed control range under a specific edge defect feature cluster.

[0029] Step S500, the cutting speed tolerance bandwidth is limited according to the cutting position-target speed defect edge value endpoint mapping relationship to generate K target cutting speed tolerance bandwidths. The original cutting speed tolerance bandwidth is finely limited and adjusted by using the established cutting position-target speed defect edge value endpoint mapping relationship to obtain more accurate K target cutting speed tolerance bandwidths that conform to actual processing. Specifically, the target speed defect edge value endpoints corresponding to each cutting position are found, and the original cutting speed tolerance bandwidth is adjusted according to these endpoints to make the cutting speed closer to those speed values ​​that can avoid or reduce edge defects, thereby improving product quality. After limiting, we obtain K new and more accurate target cutting speed tolerance bandwidths. The K target cutting speed tolerance bandwidths are optimized for each cutting position to ensure that the generation of edge defects is minimized while maintaining production efficiency.

[0030] Step S600, based on the cutting order of the K cutting positions, a target cutting speed tolerance bandwidth sequence is generated, and any two adjacent target cutting speed tolerance bandwidths in the target cutting speed tolerance bandwidth sequence are overlapped and authenticated. If the authentication fails, a speed change calibration point is set at the corresponding cutting position. Based on the cutting order of the K cutting positions, a target cutting speed tolerance bandwidth sequence is generated, and any two adjacent target cutting speed tolerance bandwidths in the target cutting speed tolerance bandwidth sequence are overlapped and authenticated. If the authentication fails, a speed change calibration point is set at the corresponding cutting position. Specifically, according to the obtained K target cutting speed tolerance bandwidths and the cutting order of the K cutting positions, these target cutting speed tolerance bandwidths are arranged in the corresponding cutting position order to form an ordered target cutting speed tolerance bandwidth sequence. Overlap authentication is performed on any two adjacent target cutting speed tolerance bandwidths in the target cutting speed tolerance bandwidth sequence to check whether there is an intersection of speed ranges between the two adjacent tolerance bandwidths, that is, whether there is a speed value that belongs to both of the two adjacent tolerance bandwidths. If the two adjacent tolerance bandwidths have no overlapping parts, that is, the authentication fails, then there may be problems in the speed transition between the two cutting positions, and a speed change calibration point needs to be set to ensure a smooth transition of the cutting speed. The speed change calibration point is used to indicate the position where special attention needs to be paid to adjusting the cutting speed during the cutting process, so as to ensure the smoothness and accuracy of the speed when switching between different tolerance bandwidths, and avoid the situation where the speed suddenly changes or exceeds the allowable range when switching the tolerance bandwidth, thereby improving the cutting quality and production efficiency.

[0031] Step S700, a speed monitoring sensor is set at the speed change calibration point to monitor and identify the cutting speed change state, and the laser cutting speed is controlled according to the monitoring and identification results. Specifically, a speed monitoring sensor is installed at the speed change calibration point to detect the change state of the cutting speed in real time to ensure that the speed data is accurately captured at the key position. The speed monitoring sensor feeds back the real-time monitored cutting speed data to the control system, and the control system processes and analyzes these data to identify the change state of the cutting speed, including determining whether the cutting speed is within the preset tolerance bandwidth and whether there are abnormal fluctuations or mutations. According to the monitoring and identification results, the control system adjusts the laser cutting speed in real time. If the cutting speed deviates from the target range or abnormal fluctuations occur, the control system will automatically adjust the laser power, cutting speed or other related parameters to return the cutting speed to the preset tolerance bandwidth.

[0032] In a possible implementation, step S700 further includes step S710, setting a steady-state monitoring window, extracting the monitoring data of the speed monitoring sensor at the speed change calibration point within the steady-state monitoring window, and generating a monitoring data set. The steady-state monitoring window is a time range used to capture data when the cutting speed near the speed change calibration point reaches a steady state. During the laser cutting process, when the cutting speed reaches a preset steady-state value and remains stable, the speed monitoring sensor will record relevant speed data within this window. From the large amount of data recorded by the speed monitoring sensor, the data corresponding to the speed change calibration point and within the steady-state monitoring window are screened out, reflecting the actual situation when the cutting speed reaches a steady state under a specific cutting position and condition. Finally, the extracted monitoring data are sorted to generate a monitoring data set, which includes all relevant speed data within the steady-state monitoring window, and can be used for subsequent data analysis, speed control optimization, and fault diagnosis. It also includes step S720, performing outlier analysis on the monitoring data set to generate a first outlier degree. Among them, outlier analysis is an important branch of data mining, which is used to find abnormal values ​​or outliers in the data set that are obviously deviated from other data points and do not meet the general data pattern or behavior. Specifically, the monitoring data set extracted from the steady-state monitoring window is processed to identify and quantify the outlier data points therein, and an indicator reflecting the degree of outliers in the data set, namely the first outlier degree, is generated accordingly. It also includes step S730, judging whether the first outlier degree is less than or equal to the preset outlier degree threshold, if not, generating an auxiliary control instruction, calling the laser cutting speed control platform to perform auxiliary control on the laser cutting speed. The preset outlier threshold is a value determined based on factors such as the process requirements of laser cutting, equipment performance, and historical data. It represents the maximum acceptable degree of outliers for cutting speed data and is used to determine whether the cutting process is stable. Specifically, if the first outlier is greater than the preset outlier threshold (i.e., the "no" case), it means that there are significant outliers in the cutting speed data, and the cutting process may be unstable or abnormal. The system needs to generate auxiliary control instructions and call the laser cutting speed control platform to perform auxiliary control on the laser cutting speed. For example, adjust the operating parameters of the laser cutting equipment to achieve auxiliary control of the cutting speed, thereby improving the stability of the cutting process, improving the cutting quality, and reducing potential problems caused by outliers.

[0033] In a possible implementation, step S720 further includes step S721, traversing the monitoring data set to perform mean calculation to obtain a first monitoring mean. The first monitoring mean represents a central trend or average level of the monitoring data. It also includes step S722, starting from the first monitoring mean, iterating with a preset concentration step length, obtaining the first monitoring data, and identifying the initial concentration amount of the first monitoring mean to obtain the initial concentration amount. Among them, the preset concentration step length refers to the step length of a single move during iteration. The difference between the maximum and minimum values ​​of the monitoring data set can be calculated, and the preset ratio of the difference is used as the preset concentration step length. The initial concentration amount reflects the amount of data gathered around the first monitoring mean, and the amount of data in the area constructed with the first monitoring mean as the starting point and the preset concentration step length as the radius. It also includes step S723, judging whether the initial concentration amount of the first monitoring mean is greater than or equal to the first concentration amount of the first monitoring data. If not, the first monitoring data is used as the stage monitoring data. Among them, the first concentration is the concentration calculated directly for the first monitoring data itself, reflecting the central trend or main distribution point of this set of data. If the initial concentration of the first monitoring mean is less than the first concentration of the first monitoring data (i.e., the "No" situation), this means that the data distribution represented by the initial concentration is deviated from the center relative to the entire first monitoring data, or the data distribution may be more dispersed, and the system will use the first monitoring data as stage monitoring data. By marking the first monitoring data as stage monitoring data, the system can conduct a more in-depth study on it, such as identifying potential problems and analyzing the causes of abnormal values. It also includes step S724, after multiple iterations, the stage monitoring data corresponding to the maximum concentration is used as the target monitoring data. After multiple iterations, the system compares the concentration of all stage monitoring data, finds the value with the largest concentration, determines the stage monitoring data corresponding to the maximum concentration, and uses it as the target monitoring data, which can best reflect the overall data characteristics. It also includes step S725, counting the amount of outlier data whose distance to the target monitoring data is greater than the preset concentration step, and comparing the amount of outlier data with the total amount of data in the monitoring data set to generate a first outlier degree. Among them, outlier data refers to data points whose distance from the target monitoring data is greater than the preset step size in the set. Specifically, each data point in the monitoring data set is traversed, their distance from the target monitoring data is calculated, and the number of data points whose distance is greater than the preset step size in the set is counted, that is, the amount of outlier data. The first outlier degree is obtained by dividing the amount of outlier data by the total amount of data in the monitoring data set, which reflects the proportion of outlier data in the entire data set, that is, the prevalence of the outlier phenomenon.

[0034] In a possible implementation, step S700 further includes step S740, if the authentication is passed, the median of the overlapping target cutting speed tolerance bandwidth is used as the cutting speed of the two adjacent cutting positions, and the cutting speed is sent to the laser cutting speed control platform for speed control. If there is an overlapping part between two adjacent tolerance bandwidths, that is, the authentication is passed, it means that the cutting speed between the two cutting positions can be smoothly transitioned without additional speed adjustment, and the median of the overlapping target cutting speed tolerance bandwidth is used as the cutting speed of the two adjacent cutting positions and sent to the laser cutting speed control platform for speed control.

[0035] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A laser cutting speed control method based on multi-source analysis, characterized in that: The method comprises: Based on the target cutting object, a control database of the laser cutting speed control platform is searched to obtain multiple historical cutting speed control records and cutting speed tolerance bandwidth; Traversing the plurality of historical cutting speed control records to extract cutting edge images, and obtaining a plurality of historical cutting edge image sets; Taking the same cutting position as a clustering target, cross-record extraction is performed on the multiple historical cutting edge image sets to generate K historical cutting edge image sets, wherein the K historical cutting edge image sets correspond one-to-one to the K cutting positions of the target cutting object, and each historical cutting edge image has a cutting speed identifier; Identify the mapping relationship between the K cutting positions and the target speed defect edge values ​​based on the K historical cutting edge image sets, and generate a cutting position-target speed defect edge value endpoint mapping relationship; The cutting speed tolerance bandwidth is limited according to the cutting position-target speed defect edge value endpoint mapping relationship to generate K target cutting speed tolerance bandwidths; Based on the cutting order of the K cutting positions, a target cutting speed tolerance bandwidth sequence is generated, and any two adjacent target cutting speed tolerance bandwidths in the target cutting speed tolerance bandwidth sequence are overlapped and authenticated. If the authentication fails, a speed change calibration point is set at the corresponding cutting position; A speed monitoring sensor is arranged at the speed change calibration point to monitor and identify the cutting speed change state, and the laser cutting speed is controlled according to the monitoring and identification result; If the authentication is passed, the median of the overlapping target cutting speed tolerance bandwidth is used as the cutting speed of two adjacent cutting positions, and the cutting speed is sent to the laser cutting speed control platform for speed control; The mapping relationship between the K cutting positions and the target speed defect edge values ​​is identified based on the K historical cutting edge image sets, and a cutting position-target speed defect edge value mapping relationship is generated, including: Performing edge defect recognition on the K historical cutting edge image sets to obtain K edge defect feature sets; Performing speed defect edge recognition on the K edge defect feature sets in combination with the cutting speed identifier to generate K speed defect edge value sets; Performing cross-defect fusion on the K speed defect edge value sets respectively to generate K target speed defect edge value endpoints; Based on mapping the K target speed defect edge value endpoints with the corresponding edge positions, generating the cutting position-target speed defect edge value endpoint mapping relationship; Wherein, edge defect recognition is performed on the K historical cutting edge image sets to obtain K edge defect feature sets, including: Acquire a plurality of sample historical cutting edge image sets and a plurality of sample edge defect feature sets as sample data; The sample data is divided into a training sample set and a test sample set according to a preset division ratio, the framework based on the convolutional neural network is trained by using the training sample set, the mapping relationship between the edge cutting image and the edge defect feature is learned, and the learning result is tested by using the test sample set, and the network parameters are updated according to the test accuracy until the output test accuracy meets the preset test accuracy requirement, thereby generating a defect recognition network layer that has been trained; Using the defect recognition network layer to perform edge defect recognition on the K historical cutting edge image sets to obtain the K edge defect feature sets; The K edge defect feature sets are combined with the cutting speed identifier to perform speed defect edge recognition to generate K speed defect edge value sets, including: Performing edge defect feature clustering analysis on the K edge defect feature sets respectively to generate K edge defect feature clusters; Extracting the modes of multiple cutting speeds of multiple cutting defect feature sets corresponding to each edge defect feature cluster in the K edge defect feature clusters as speed defect edge values, and generating K speed defect edge value sets; The K speed defect edge value sets are respectively subjected to cross-defect fusion to generate K target speed defect edge value endpoints, including: Traversing the K speed defect edge value sets to extract the maximum value, and generating K first target speed defect edge value endpoints; Traversing the K speed defect edge value sets to extract the minimum value, and generating K second target speed defect edge value endpoints; The K target speed defect edge value endpoints are generated based on the K first target speed defect edge value endpoints and the K second target speed defect edge value endpoints.

2. The laser cutting speed control method based on multi-source analysis according to claim 1, characterized in that: A speed monitoring sensor is arranged at the speed change calibration point to monitor and identify the cutting speed change state, and the laser cutting speed is controlled according to the monitoring and identification result. The method includes: Setting a steady-state monitoring window, extracting monitoring data of the speed monitoring sensor of the speed change calibration point within the steady-state monitoring window, and generating a monitoring data set; Performing outlier analysis on the monitoring data set to generate a first outlier degree; It is determined whether the first outlier degree is less than or equal to a preset outlier degree threshold value. If not, an auxiliary control instruction is generated to call the laser cutting speed control platform to perform auxiliary control on the laser cutting speed.

3. The laser cutting speed control method based on multi-source analysis according to claim 2, characterized in that: Performing outlier analysis on the monitoring data set to generate a first outlier degree, the method comprising: Traversing the monitoring data set to perform mean calculation to obtain a first monitoring mean; Taking the first monitoring mean as a starting point, iterating with a preset concentration step length to obtain first monitoring data, and identifying an initial concentration amount for the first monitoring mean to obtain an initial concentration amount; Determine whether the initial concentration of the first monitoring mean is greater than or equal to the first concentration of the first monitoring data; if not, use the first monitoring data as stage monitoring data; After multiple iterations, the stage monitoring data corresponding to the maximum concentration value is used as the target monitoring data; The amount of outlier data whose distance to the target monitoring data is greater than the preset centralized step size is counted, and the amount of outlier data is compared with the total amount of data in the monitoring data set to generate a first outlier degree.

Images