Intelligent slicing and scanning path planning method and system for selective laser melting

Abstract

This invention belongs to the field of laser selective melting, and particularly relates to an intelligent slicing and scanning path planning method and system for laser selective melting. The method first collects the attribute parameters and process requirements of the metal component, constructs a geometric model, and achieves adaptive layering through deep learning combined with a genetic algorithm. Second, it automatically calibrates the scanning priority based on a local priority strategy. Third, it introduces a helical-cycloidal hybrid path algorithm to construct a path self-switching model, and optimizes and trains it in a simulation environment. Fourth, it uses a reinforcement learning algorithm to construct a processing control model, enabling precise scanning of the metal component by the processing head. During processing, sensors monitor and provide feedback on the temperature field, stress distribution, and surface condition in real time, adjusting the processing path and control parameters accordingly. Finally, an interactive interface is provided to display, set, and monitor and alarm for abnormal processes in real time, improving the processing efficiency and accuracy of laser selective melting.

Description

Technical Field

[0001] The invention belongs to the field of laser selective melting, and in particular relates to an intelligent slicing and scanning path planning method and system for laser selective melting. Background Art

[0002] Selective laser melting (SLM) technology plays a key role in the field of advanced manufacturing, especially in the production of complex curved metal parts. However, for complex geometric shapes, traditional slicing and path planning technologies face many challenges, such as difficulty in accurately controlling deformation caused by thermal effects, increased surface roughness, and low processing efficiency.

[0003] For example, the patent with publication number CN115837470A discloses a laser selective melting path planning method, device and equipment, including: obtaining three-dimensional laser selective melting data; determining the airflow direction and airflow angle of the wind field; based on the airflow direction and airflow angle, sorting the slice data and partition data corresponding to the three-dimensional laser selective melting data against the wind field to obtain a first scanning path; obtaining the partition angle, and based on the airflow direction, airflow angle and partition angle, sorting the fill line data in the partition data against the wind field to obtain a second scanning path; determining the laser selective melting path based on the first scanning path and the second scanning path.

[0004] For example, the patent with authorization announcement number CN114713844B discloses a metal selective laser melting forming method and system. The metal selective laser melting forming system includes: a regional spot optical path system, a Gaussian spot optical path system and a terminal controller; the metal selective laser melting forming method includes: the terminal controller obtains a three-dimensional digital model of the part to be formed; the terminal controller performs layered slicing on the three-dimensional digital model of the part to obtain layered slicing data; obtains the printing mode of the part, and performs path planning to obtain laser scanning path planning data; based on the laser scanning path planning data, controls the regional spot optical path system or controls the regional spot optical path system and the Gaussian spot optical path system to print the part.

[0005] The above existing technologies have the following problems: The limitations of the current technology lie in the fixed-mode slicing strategy and simple path planning, the lack of adaptability to complex structures and dynamic management of thermal effects, resulting in unstable quality of finished products and failure to meet the needs of precision manufacturing. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention proposes an intelligent slicing and scanning path planning method and system for laser selective melting. The method first collects the property parameters and process requirements of metal components, constructs a geometric model, and realizes adaptive stratification through deep learning combined with genetic algorithm. Secondly, the scanning priority is automatically calibrated based on the local priority strategy. Thirdly, a spiral-cycloid hybrid path algorithm is introduced to construct a path self-switching model, and the training is optimized in a simulation environment. Fourthly, a reinforcement learning algorithm is used to construct a processing control model to realize the precise processing scanning of the metal component by the processing head. During the processing, the temperature field, stress distribution and processing surface state are monitored and fed back by sensors in real time, and the processing path and control parameters are adjusted in real time. Finally, an interactive interface is set to realize the real-time display, setting and abnormal monitoring and alarm of processing process parameters, thereby improving the processing efficiency and accuracy of laser selective melting.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The intelligent slicing and scanning path planning method for selective laser melting includes the following specific steps:

[0009] S1. Collect the property parameters and process requirement parameters of the metal components to be processed, use the property parameters of the metal components to build a metal geometric model, and use the intelligent stratification strategy constructed by deep learning combined with genetic algorithm to adaptively stratify the metal geometric model according to the constructed geometric model, and use the process requirement parameters to adaptively adjust the stratification process;

[0010] S2. Based on the layered slicing results in S1, the attribute parameter characteristics of each layer of metal component slices after stratification are obtained, and the configured local priority strategy is used to automatically calibrate the scanning priority value for each scanning area in each layer of slices according to the attribute parameters of each layer of metal component slices;

[0011] S3. Introduce the spiral-cycloid hybrid path algorithm, construct a path self-switching model, and use the acquired regional scanning priority and metal component attribute parameters and process requirement parameters to optimize the constructed path self-switching model in a simulation environment to obtain a trained path self-switching model;

[0012] S4. constructing and training a processing control model based on a reinforcement learning algorithm and processing control parameters, and using the acquired path self-switching model and processing control model to control the processing scanning of the processing head on the metal component to be processed;

[0013] S5. Use the sensor integrated on the processing head to monitor the temperature field, stress distribution and processing surface status in real time during the processing, and use the monitoring algorithm configured in the sensor to analyze the monitoring parameters. Feedback the analysis results to the path self-switching model and the processing control model to self-correct the processing path and adjust the processing control parameters in real time.

[0014] Specifically, the specific process of S1 includes:

[0015] S101, collecting attribute parameters, process requirement parameters and historical stratification strategies of metal components, and preprocessing the collected parameters to keep them in a unified input format;

[0016] S102, constructing o-xyz three-dimensional coordinates with the center of gravity of the metal component as the origin, inputting the pre-processed property parameters of the metal component into the three-dimensional modeling software to construct a three-dimensional virtual model of the metal component in the three-dimensional coordinates, and marking the property parameters on the constructed three-dimensional virtual model;

[0017] S103, constructing a stratification strategy model using a BP neural network combined with a genetic algorithm, constructing a model input according to the acquired attribute parameters, process requirement parameters and historical stratification strategies, and constructing a model output as a stratification strategy, wherein the stratification strategy includes the number of layers L, the layer thickness H, the stratification slicing angle A and the real-time speed v during the movement of the laser processing head;

[0018] S104, inputting the constructed input and output into the hierarchical strategy model for training, and obtaining a trained hierarchical strategy model;

[0019] S105. Integrate the trained layering strategy model into the three-dimensional modeling software, perform virtual layered slicing on the three-dimensional virtual model of the metal component, and use the PID algorithm to generate a corresponding control signal for the virtual slicing process. Use the sensor to transmit the control signal to the corresponding processing head to control the processing head to perform layered slicing on the metal component.

[0020] Specifically, the specific steps of training the hierarchical strategy model in S104 include:

[0021] S1041. Use computer software to extract component attribute parameter vectors marked on the three-dimensional virtual model At the same time, the process requirement parameters are used as the target control variables in the training process and numerically preprocessed, and the process requirement attribute parameter vector is constructed using the preprocessed process requirement parameters , the component attribute parameter vector and the process requirement attribute parameter vector are cascaded to form a joint input vector ;in represents the metal component attribute parameter of the ith type, and n represents the number of categories of the metal component attribute parameter; represents the process requirement attribute parameter of the jth category, and m represents the number of categories of process requirement attribute parameters;

[0022] S1042, constructing a BP neural network using three convolutional layers and two fully connected layers connected in series, inputting the constructed joint input vector into the BP neural network, and using the collected process requirement attribute parameters and the set target process requirement attribute parameters to obtain a mean square error as a training error, training the constructed BP neural network, and outputting an initial stratification strategy;

[0023] S1043, constructing an initial population of the genetic algorithm using the initial stratification strategy and the historical stratification strategy; constructing the qth chromosome input by the genetic algorithm using the obtained initial population , and set the following constraint set S for each variable in the chromosome:

[0024] ,

[0025] in, represents the number of slice layers in the qth chromosome, represents the slice thickness parameter in the qth chromosome, represents the layered slicing angle parameter in the qth chromosome, represents the speed parameter of the processing head moving process in the qth chromosome, In turn, they represent the minimum number of slice layers and the maximum number of slice layers. represents the minimum slice thickness and the maximum slice thickness respectively. represents the minimum stratification slicing angle and the maximum stratification slicing angle respectively, In turn, they represent the minimum processing head moving process speed and the maximum processing head moving process speed;

[0026] S1044. Use the initial population to construct the fitness function corresponding to the genetic algorithm , the specific formula is:

[0027] ,

[0028] in, represents the processing time corresponding to the qth chromosome, is an indicator function, indicating that when the qth chromosome corresponds to the processing time Exceeding the preset maximum time The value is 1 when , otherwise it is 0. Indicates that when the qth chromosome corresponds to the processing time The penalty coefficient corresponding to exceeding the preset maximum time, represents the processing cost corresponding to the qth chromosome, Indicates the quality score corresponding to the qth chromosome, It represents the penalty term for violating the sth constraint in the constraint set S corresponding to the qth chromosome. If the sth constraint in the constraint set S is not satisfied, the term is 1, otherwise it is 0. They represent the weighted coefficients of the fitness function in turn;

[0029] S1045. Set the iteration cycle and stop condition, input the constructed qth chromosome, constraint set S, and fitness function into the genetic algorithm, iterate the above S1041-S1044 process, and when the iteration cycle or stop condition is met, stop the iteration, output the obtained optimal stratification strategy, and feed back the obtained optimal stratification strategy to the constructed BP neural network for fine-tuning.

[0030] Specifically, the specific steps of S2 include:

[0031] S201, in the 3D modeling software, gridding each layer of slices after layered slicing using a fixed-size square grid, numerically labeling the divided grids, and obtaining attribute parameter features in each labeled grid;

[0032] S202, inputting the acquired attribute parameter features and corresponding labels in each grid into a clustering algorithm, performing a clustering operation, and merging grid labels belonging to the same category after clustering to obtain each scanning area of ​​the corresponding layered slice;

[0033] S203, construct a local priority model, and input the acquired attribute parameter features of each scanning area corresponding to the layered slices into the local priority model for training, and integrate the trained local priority model into the three-dimensional virtual model of the metal component in S105 to automatically mark the corresponding scanning area priorities of the three-dimensional virtual model slices after layered slicing.

[0034] Specifically, the specific steps of local priority model training include:

[0035] S2031. Collect historical attribute parameters and process requirement parameters of the slicing areas of the same type of metal components, and use expert experience to mark the processing priority of each scanning area according to the collected parameters to obtain priority label data; the priorities are divided into 1-9, where 1-3 are low priorities, 4-6 are medium priorities, and 7-9 are high priorities;

[0036] S2032. Use support vector regression SVR to build a local priority model, and input the obtained priority label data set into the local priority model for training to obtain a trained local priority model.

[0037] Specifically, the path self-switching model includes a spiral path planning sub-model, a cycloid path sub-model, a spiral-cycloid path hybrid planning sub-model and a path planning priority working switching condition, and the path planning priority working switching condition specifically includes:

[0038] S3031, when the priority of the scanning area corresponding to the layered slice is low priority, the spiral path planning submodel in the path self-switching model is used to perform scanning path planning;

[0039] S3032, when the priority of the scanning area corresponding to the layered slice is medium priority, the spiral-cycloid path hybrid planning submodel in the path self-switching model is used to perform scanning path planning;

[0040] S3033. When the priority of the scanning area corresponding to the layered slice is a high priority, the scanning path is planned using the cycloid path sub-model in the path self-switching model.

[0041] Specifically, the specific steps of constructing the S4 processing control model include:

[0042] S401. Construct processing control constraints based on metal component property parameters, process requirement parameters and processing control parameters. The specific conditions are:

[0043] ,

[0044] in, represents the laser output power of the laser processing head at time t, represents the optimal power threshold corresponding to the laser output power of the laser processing head at time t, represents the thickness of the metal component after layered processing at time t, It represents the target thickness of the metal component layered processing at time t, represents the path planning sub-model of the laser processing head's travel path at time t, The path planning sub-model represents the path target execution of the laser processing head at time t, represents the strength of the processed metal component at time t, Indicates the target strength of processed metal components, It represents the speed of the laser processing head at time t, Indicates the target speed of the metal laser processing head during processing;

[0045] S402: According to the constraint conditions in S401, construct the processing control trigger conditions, which specifically include:

[0046] when and When the processing control constraint condition is not met, the first processing control condition is triggered;

[0047] when and When the processing control constraint condition is not met, the second processing control condition is triggered;

[0048] when and When the processing control constraint condition is not met, the third processing control condition is triggered;

[0049] when and When the processing control constraint condition is not met, the fourth processing control condition is triggered;

[0050] when and When the processing control constraint condition is not met, the fifth processing control condition is triggered;

[0051] When two or more processing control constraint conditions in S402 are not satisfied, the sixth processing control condition is triggered.

[0052] Intelligent slicing and scanning path planning system for laser selective melting, including: data processing module, intelligent stratification module, priority calibration module;

[0053] A data processing module is used to collect and process the property parameters and process requirement parameters of the metal component to be processed, so that the collected data is kept in a unified format;

[0054] Intelligent stratification module, used to perform adaptive stratification of metal components by using deep learning combined with genetic algorithm; the intelligent stratification module includes a geometric modeling unit and a strategy stratification unit;

[0055] The geometric modeling unit is used to construct a metal geometric model using computer software according to the property parameters of the pre-processed metal component to be processed; the strategy stratification unit is used to adaptively stratify the metal geometric model according to the constructed geometric model using the configured intelligent stratification strategy, and to optimize the stratification process in real time using the process requirement parameters;

[0056] A priority calibration module, used for automatically calibrating a scanning priority value for each scanning area in each layer of slices based on the stratification result of the strategy stratification unit; the priority calibration module includes an area division unit and a priority calibration unit;

[0057] The area division unit is used to use sensors to extract the attribute parameter characteristics of each layer of slices, and use the configured clustering algorithm to divide the area on each layer of slices; the priority calibration unit is used to calibrate the scanning priority value of the divided area according to the divided area and the attribute parameter characteristics and process requirement parameters within the area, using the configured local priority strategy.

[0058] Specifically, the intelligent slicing and scanning path planning system for laser selective melting also includes: a path planning module, a processing control module and a monitoring feedback module;

[0059] The path planning module is used to construct a path self-switching model based on the spiral-cycloid hybrid path algorithm and perform optimization simulation training on the constructed model in a simulation environment;

[0060] A processing control module is used to build a processing control model based on a reinforcement learning algorithm to control the processing scanning of the metal component by the processing head; the processing control module includes a control model building unit and a processing scanning control unit;

[0061] A control model building unit is used to build a processing control model based on a reinforcement learning algorithm and processing control parameters; a processing scanning control unit is used to control a processing head to perform processing scanning on a metal component to be processed by using a path self-switching model and a processing control model;

[0062] The monitoring feedback module is used to monitor the processing process in real time and adjust the processing path and processing control parameters in real time according to the monitoring results; the monitoring feedback module includes a monitoring unit and an analysis feedback unit;

[0063] The monitoring unit is used to use the sensors on the processing head to monitor the temperature field, stress distribution and processing surface status data in real time; the analysis feedback unit is used to analyze the monitoring parameters using the monitoring algorithm, and feed back the analysis results to the path self-switching model and the processing control model to perform path self-correction and real-time adjustment of the processing control parameters.

[0064] A computer-readable storage medium stores computer instructions, which, when executed, execute an intelligent slicing and scanning path planning method for laser selective melting.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. Aiming at the deficiencies of the prior art, the present invention firstly constructs an intelligent stratification strategy to automatically generate the optimal stratification slicing strategy, thereby ensuring the accuracy of stratification slicing. Secondly, the local priority strategy constructed can automatically assign a corresponding priority to each scanning area according to the slice attribute parameters, and according to the assigned priority, the configured path self-switching model is used to perform different path strategy processing on different priority processing areas. This process ensures the targeted processing of the corresponding priority scanning area, improves the stability of the overall structure and the printing efficiency, and optimizes the thermal effect management of complex metal component processing.

[0067] 2. In view of the deficiencies in the prior art, the present invention combines the path self-switching model with the processing control model to enable precise adjustment of the scanning action of the processing head and optimize the processing process. At the same time, the sensor monitors the temperature field, stress distribution and surface state in real time, combines the instant analysis of the monitoring algorithm, and feeds back to the model, thereby realizing real-time dynamic correction of the processing path and control parameters, and improving the stability and quality control capability of the finished product. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a flow chart of the intelligent slicing and scanning path planning method for selective laser melting in Example 1 of the present invention;

[0069] Figure 2 This is a logical structure diagram of the hierarchical strategy model training in Example 1 of the present invention;

[0070] Figure 3 This is a module architecture diagram of the intelligent slicing and scanning path planning system for selective laser melting according to Example 2 of the present invention. DETAILED DESCRIPTION

[0071] Example 1

[0072] See also Figure 1 , an embodiment of the present invention provides: an intelligent slicing and scanning path planning method for laser selective melting, comprising the following steps:

[0073] S1. Collect the property parameters and process requirement parameters of the metal components to be processed, use the property parameters of the metal components to build a metal geometric model, and use the intelligent stratification strategy constructed by deep learning combined with genetic algorithm to adaptively stratify the metal geometric model according to the constructed geometric model, and use the process requirement parameters to adaptively adjust the stratification process;

[0074] S2. Based on the layered slicing results in S1, the attribute parameter characteristics of each layer of metal component slices after stratification are obtained, and the configured local priority strategy is used to automatically calibrate the scanning priority value for each scanning area in each layer of slices according to the attribute parameters of each layer of metal component slices;

[0075] S3. Introduce the spiral-cycloid hybrid path algorithm, construct a path self-switching model, and use the acquired regional scanning priority and metal component attribute parameters and process requirement parameters to optimize the constructed path self-switching model in a simulation environment to obtain a trained path self-switching model;

[0076] S4. constructing and training a processing control model based on a reinforcement learning algorithm and processing control parameters, and using the acquired path self-switching model and processing control model to control the processing scanning of the processing head on the metal component to be processed;

[0077] S5. Using the sensor integrated on the processing head, the temperature field, stress distribution and processing surface state during the processing are monitored in real time, and the monitoring algorithm configured in the sensor is used to analyze the monitoring parameters, and the analysis results are fed back to the path self-switching model and the processing control model to self-correct the processing path and adjust the processing control parameters in real time;

[0078] In this embodiment, a variety of monitoring algorithms are configured in the integrated sensor in S5. Specifically, the infrared temperature monitoring algorithm configured in the temperature sensor is used to monitor the temperature field of the processing area in real time, including the surface and internal temperature of the material, and the output power of the laser processing head is adjusted through the monitored temperature feedback. The support vector machine (SVM), random forest or neural network monitoring algorithm configured in the strain sensor is used to monitor the stress distribution of the processed metal component in real time, and the processing strength of the metal component is adjusted through feedback based on the monitored stress distribution. The roughness, flatness and defects of the surface state of the processed metal component are monitored in real time through the visual sensor, and the surface state of the processed metal component is evaluated using the configured VIT-based defect detection scoring algorithm, and an evaluation score is output, which corresponds to the quality score in S1044. The processing thickness of the processed metal component is detected by an acoustic wave detection sensor configured with a non-destructive detection algorithm, and the detection result is fed back to the processing control model to adjust the processing process.

[0079] S6. Set up an interactive interface to display and set the processing parameters of S1 to S6 in real time, and use integrated sensors to monitor and alarm abnormalities in the processing in real time.

[0080] Furthermore, the specific steps of S1 in this embodiment include:

[0081] S101, collect the attribute parameters, process requirement parameters and historical stratification strategy of the metal component, and pre-process the collected parameters to keep them in a unified input format; in this embodiment, the attribute parameters of the metal component include: metal component geometric parameters, material type, density, thermal conductivity, melting point and yield strength; the process requirement parameters include the surface state of the metal component, mechanical property requirements and dimensional tolerance; the mechanical property requirements include the tensile strength (the maximum stress of the material under tensile load until it breaks), compressive strength (the ability to withstand compressive load), bending strength (strength under bending load) and fatigue strength (the maximum stress level of the material under repeated loads without breaking); the geometric parameters of the metal component include geometric dimensions, shape and topological features; the geometric dimensions include length, width, height, diameter, wall thickness, characteristic dimensions, and the topological features include holes, protrusions, depressions, edge types, surface curvature and stratification slicing angle; the process requirement parameters include temperature field, stress distribution and processing surface state data; the historical stratification strategy of this embodiment is the number of layers L, layer thickness T, stratification slicing angle A and speed v data of the historical stratification of the same type of metal components during the movement of the processing head;

[0082] S102, constructing o-xyz three-dimensional coordinates with the center of gravity of the metal component as the origin, inputting the pre-processed property parameters of the metal component into the three-dimensional modeling software to construct a three-dimensional virtual model of the metal component in the three-dimensional coordinates, and marking the property parameters on the constructed three-dimensional virtual model;

[0083] S103, constructing a stratification strategy model using a BP neural network combined with a genetic algorithm, constructing a model input according to the acquired attribute parameters, process requirement parameters and historical stratification strategies, and constructing a model output as a stratification strategy, wherein the stratification strategy includes the number of layers L, the layer thickness H, the stratification slicing angle A and the real-time speed v during the movement of the laser processing head;

[0084] S104, inputting the constructed input and output into the hierarchical strategy model for training, and obtaining a trained hierarchical strategy model;

[0085] For further information, see Figure 2 In this embodiment, S104, the specific steps of training the hierarchical strategy model include:

[0086] S1041. Use computer software to extract component attribute parameter vectors marked on the three-dimensional virtual model At the same time, the process requirement parameters are used as the target control variables in the training process and numerically preprocessed, and the process requirement attribute parameter vector is constructed using the preprocessed process requirement parameters , the component attribute parameter vector and the process requirement attribute parameter vector are cascaded to form a joint input vector ;in represents the metal component attribute parameter of the ith type, and n represents the number of categories of the metal component attribute parameter; represents the process requirement attribute parameter of the jth category, and m represents the number of categories of process requirement attribute parameters;

[0087] S1042, constructing a BP neural network using three convolutional layers and two fully connected layers connected in series, inputting the constructed joint input vector into the BP neural network, and using the collected process requirement attribute parameters and the set target process requirement attribute parameters to obtain a mean square error as a training error, training the constructed BP neural network, and outputting an initial stratification strategy; in this embodiment, the target process requirement attribute parameters are set by the production personnel according to specific production requirements;

[0088] S1043, constructing an initial population of the genetic algorithm using the initial stratification strategy and the historical stratification strategy; constructing the qth chromosome input by the genetic algorithm using the obtained initial population , and set the following constraint set S for each variable in the chromosome:

[0089] ,

[0090] in, represents the number of slice layers in the qth chromosome, represents the slice thickness parameter in the qth chromosome, represents the layered slicing angle parameter in the qth chromosome, represents the speed parameter of the processing head moving process in the qth chromosome, In turn, they represent the minimum number of slice layers and the maximum number of slice layers. represents the minimum slice thickness and the maximum slice thickness respectively. represents the minimum stratification slicing angle and the maximum stratification slicing angle respectively, In turn, they represent the minimum processing head moving process speed and the maximum processing head moving process speed;

[0091] In addition to selecting the above-mentioned process requirement parameters as the initial stratification strategy input, the processing movement process speed in the processing control parameters is also selected in this embodiment because this parameter is the variable that most directly affects the heat affected zone, part quality and path. By limiting this parameter, the temperature distribution, stress distribution and processing quality of the processing area corresponding to each layered slice can be well controlled; in this embodiment , , and Obtained and determined by expert experience method;

[0092] S1044. Use the initial population to construct the fitness function corresponding to the genetic algorithm , the specific formula is:

[0093] ,

[0094] in, represents the processing time corresponding to the qth chromosome, is an indicator function, indicating that when the qth chromosome corresponds to the processing time Exceeding the preset maximum time The value is 1 when , otherwise it is 0. Indicates that when the qth chromosome corresponds to the processing time The penalty coefficient corresponding to exceeding the preset maximum time, represents the processing cost corresponding to the qth chromosome, Indicates the quality score corresponding to the qth chromosome, It represents the penalty term for violating the sth constraint in the constraint set S corresponding to the qth chromosome. If the sth constraint in the constraint set S is not satisfied, the term is 1, otherwise it is 0. In turn, they represent the corresponding weighting coefficients; in this embodiment, the corresponding quality score is obtained by using the quality detection algorithm integrated in the integrated sensor arranged in S5;

[0095] S1045. Set the iteration cycle and stop condition, input the constructed qth chromosome, constraint set S, and fitness function into the genetic algorithm, iterate the above S1041-S1044 process, and when the iteration cycle or stop condition is met, stop the iteration, output the obtained optimal stratification strategy, and feed back the obtained optimal stratification strategy to the BP neural network for fine-tuning. Through the online learning mechanism, let the stratification strategy model gradually adapt to the corresponding stratification strategy to improve the accuracy of metal construction stratification. The iteration cycle and stop condition of this embodiment are set through a specific experimental process. For example, the iteration cycle is 100, and the stop condition is that when 90% of the constructed chromosomes violate the constraint set S, it is considered that the algorithm has converged and the iteration is stopped.

[0096] S105, integrating the trained layering strategy model into the three-dimensional modeling software, performing virtual layered slicing on the three-dimensional virtual model of the metal component, and generating corresponding control signals for the virtual slicing process using a PID algorithm, and transmitting the control signals to the corresponding processing head using a sensor to control the processing head to perform layered slicing on the metal component;

[0097] Furthermore, the specific steps of S2 in this embodiment include:

[0098] S201, in the 3D modeling software, gridding each layer of slices after layered slicing using a fixed-size square grid, numerically labeling the divided grids, and obtaining attribute parameter features in each labeled grid; in this embodiment, the size of the grid needs to be pre-set according to the complexity and accuracy requirements of the parts to ensure that the detailed features can be captured without causing excessive calculation burden;

[0099] S202, input the acquired attribute parameter features and corresponding labels in each grid into the clustering algorithm, perform clustering operation, and merge the grid labels belonging to the same category after clustering to obtain each scanning area of ​​the corresponding layered slice; the corresponding clustering algorithms in this embodiment include K-means, DBSCAN, fuzzy clustering and hierarchical clustering;

[0100] S203, construct a local priority model, and input the acquired attribute parameter features of each scanning area corresponding to the layered slices into the local priority model for training, and integrate the trained local priority model into the three-dimensional virtual model of the metal component in S105 to automatically mark the corresponding scanning area priorities of the three-dimensional virtual model slices after layered slicing.

[0101] Furthermore, the specific steps of local priority model training in this embodiment include:

[0102] S2031. Collect historical attribute parameters and process requirement parameters of the slicing areas of the same type of metal components, and use expert experience to mark the processing priority of each area according to the collected parameters to obtain priority label data; the priorities are divided into 1-9, where 1-3 are low priorities, 4-6 are medium priorities, and 7-9 are high priorities;

[0103] S2032. Use support vector regression SVR to build a local priority model, and input the obtained priority label data set into the local priority model for training to obtain a trained local priority model.

[0104] Furthermore, the specific steps of constructing and training the path self-switching model in S3 of this embodiment specifically include:

[0105] S301, collecting property parameters, process requirement parameters and historical processing path parameters of the same type of metal components, and preprocessing the obtained parameters in a unified format;

[0106] S302, using the spiral path algorithm and the cycloid path algorithm to construct a spiral path planning sub-model, a cycloid path sub-model, and a spiral-cycloid path hybrid planning sub-model; using the acquired historical metal component attribute parameters, process requirement parameters, and historical processing path parameters, the three constructed sub-models are trained in parallel in an integrated manner to obtain the trained spiral path planning sub-model, the cycloid path sub-model, and the spiral-cycloid path hybrid planning sub-model;

[0107] S303, according to the priority of each layered slice corresponding to the scanning area obtained by the local priority model, setting the priority working switching conditions for the path planning processes of the spiral path planning sub-model, the cycloid path sub-model and the spiral-cycloid path mixed planning sub-model;

[0108] S304: Integrate the acquired spiral path planning sub-model, cycloid path sub-model, spiral-cycloid path hybrid planning sub-model and the set priority work switching conditions to acquire a path self-switching model.

[0109] Furthermore, in this embodiment, the path planning priority working switching conditions of the spiral path planning sub-model, the cycloid path sub-model and the spiral-cycloid path mixed planning sub-model specifically include:

[0110] S3031. When the priority of the scanning area corresponding to the layered slice is low priority, that is, when the priority is 1-3, the spiral path planning submodel in the path self-switching model is used to plan the scanning path; the low priority area in this embodiment is the area that does not require high thermal stress and structural integrity but pursues efficiency. These areas are usually relatively flat, uniform in thickness, simple in geometry, and low in stress concentration, such as large-area bases and internal filling areas; the spiral path planning submodel can maximize the printing speed and material deposition efficiency and reduce empty strokes; the spiral path planning submodel in this embodiment is constructed by the existing spiral path algorithm;

[0111] S3032. When the priority of the scanning area corresponding to the layered slice is medium priority, that is, when the priority is 4-6, the spiral-cycloid path hybrid planning submodel in the path self-switching model is used to plan the scanning path; in this embodiment, the medium priority area includes some areas that require efficiency but also need to control thermal stress, such as structures with certain complexity but not the most critical support or load-bearing. The spiral-cycloid path hybrid planning submodel can use a spiral path to quickly fill a large area locally, and switch to a cycloid path near key details or stress concentration points to balance efficiency and quality; in this embodiment, the spiral-cycloid path hybrid planning submodel is constructed in an integrated manner based on the existing spiral path algorithm and the cycloid path algorithm;

[0112] S3033, when the priority of the scanning area corresponding to the layered slice is a high priority, that is, when the priority is 7-9, the scanning path is planned by using the cycloid path submodel in the path self-switching model; the cycloid path submodel of this embodiment is constructed by the existing cycloid path algorithm;

[0113] In this embodiment, high-priority areas, i.e., areas that are sensitive to thermal stress management, have complex structures, or are easily deformed, include thin-walled structures, sharp angles, complex geometric features, and stress concentration areas. The cycloid path sub-model can control the uniform distribution of thermal effects, reduce thermal stress accumulation, and improve structural integrity.

[0114] Furthermore, the specific steps of constructing the processing control model in S4 of this embodiment include:

[0115] S401. Construct processing control constraints based on metal component property parameters, process requirement parameters and processing control parameters. The specific conditions are:

[0116] ,

[0117] in, represents the laser output power of the laser processing head at time t, represents the optimal power threshold corresponding to the laser output power of the laser processing head at time t, represents the thickness of the metal component after layered processing at time t, It represents the target thickness of the metal component layered processing at time t, represents the path planning sub-model of the laser processing head's travel path at time t, The path planning sub-model represents the path target execution of the laser processing head at time t, represents the strength of the processed metal component at time t, Indicates the target strength of processed metal components, It represents the speed of the laser processing head at time t, Indicates the target speed of the metal laser processing head during processing;

[0118] The processing control parameters of this embodiment include the moving speed of the processing head, that is, the laser scanning speed. The processing control parameters also include the laser output power and the incident angle of the laser processing head to the scanning area.

[0119] S402: According to the constraint conditions in S401, construct the processing control trigger conditions, which specifically include:

[0120] when and When the processing control constraint condition is not met, the first processing control condition is triggered;

[0121] when and When the processing control constraint condition is not met, the second processing control condition is triggered;

[0122] when and When the processing control constraint condition is not met, the third processing control condition is triggered;

[0123] when and When the processing control constraint condition is not met, the fourth processing control condition is triggered;

[0124] when and When the processing control constraint condition is not met, the fifth processing control condition is triggered;

[0125] When two or more processing control constraint conditions in S402 are not satisfied, the sixth processing control condition is triggered.

[0126] S403: construct the input state of the processing control model at the current time t based on the above metal component attribute parameters, process requirement parameters and processing control parameters , using the trigger condition in S402 to construct a joint execution action , construct a reward function based on the input state and the joint execution action ;in, It represents the incident angle from the laser processing head to the scanning area at the current time t, It represents the incident angle threshold from the laser processing head to the scanning area at the current time t; represents the laser power error gain coefficient, represents the metal component thickness error gain coefficient, represents the metal component strength error gain coefficient, Indicates the gain coefficient for executing the correct path planning priority work switching condition. That is, when the path planning priority work switching condition is correct, a positive gain reward of +1 is given, and when the path planning priority work switching condition is wrong, a negative gain reward of -1 is given. represents the incident angle error gain coefficient of the laser processing head;

[0127] Indicates the action executed according to the triggering first processing control condition, that is, when Lower than When the power is automatically increased through the processing control model, Satisfy process control constraints in real time;

[0128] Indicates the action executed according to the triggering of the second machining control condition, that is, automatically adjusted by the machining control model and , so that Satisfy process control constraints in real time;

[0129] Indicates the action executed according to the triggering of the third processing control condition, that is, when the path planning sub-model executed by the corresponding scanning area at the current moment does not match the path planning sub-model executed by the target of the scanning area, the path planning priority work switching condition is adjusted through the processing control model so that Satisfy the processing control constraints in real time. For example, the priority corresponding to the scanning area at the current moment is low priority, but the executed path planning sub-model is a path planning sub-model corresponding to medium or high priority;

[0130] Indicates the action executed according to the triggering of the fourth processing control condition, that is, automatically adjusted by the processing control model , and , so that Satisfy process control constraints in real time;

[0131] Indicates the action executed according to the triggering of the fifth machining control condition, that is, the automatic adjustment through the machining control model and , so that Satisfy process control constraints in real time;

[0132] Indicates the action executed according to the triggering of the fifth processing control condition. When the action is executed, the combination is executed , , , and , so that it satisfies the corresponding processing control constraints;

[0133] In this embodiment, the incident angle threshold from the laser processing head to the scanning area and the optimal power threshold corresponding to the laser output power of the laser processing head are both obtained by collecting historical property parameters of the same metal component, process requirement parameters and processing control parameters, and calculating them using a support vector machine regression model.

[0134] S404, constructing a processing control model using reinforcement learning, and inputting the constructed input state, joint execution action and reward function at the current time t into the processing control model for training to obtain a trained processing control model;

[0135] S405, using integrated sensors to collect processing control parameters and metal component property parameters in real time during the processing, and inputting the collected processing control parameters and metal component property parameters and the set process requirement parameters into the trained processing control model, and generating execution action adjustment amounts according to constraint conditions and trigger conditions;

[0136] S406, input the generated execution action adjustment amount into the PID algorithm to generate a corresponding control signal, and perform real-time adjustment and control on the processing path, processing speed and processing laser power of the processing head.

[0137] Example 2

[0138] See also Figure 3 , another embodiment provided by the present invention: an intelligent slicing and scanning path planning system for laser selective melting, comprising: a data processing module, an intelligent stratification module, and a priority calibration module;

[0139] A data processing module is used to collect and process the property parameters and process requirement parameters of the metal component to be processed, so that the collected data is kept in a unified format;

[0140] Intelligent stratification module, used to perform adaptive stratification of metal components by using deep learning combined with genetic algorithm; the intelligent stratification module includes a geometric modeling unit and a strategy stratification unit;

[0141] The geometric modeling unit is used to construct a metal geometric model using computer software according to the property parameters of the pre-processed metal component to be processed; the strategy stratification unit is used to adaptively stratify the metal geometric model according to the constructed geometric model using the configured intelligent stratification strategy, and to optimize the stratification process in real time using the process requirement parameters;

[0142] A priority calibration module, used for automatically calibrating a scanning priority value for each scanning area in each layer of slices based on the stratification result of the strategy stratification unit; the priority calibration module includes an area division unit and a priority calibration unit;

[0143] The area division unit is used to use sensors to extract the attribute parameter characteristics of each layer of slices, and use the configured clustering algorithm to divide the area on each layer of slices; the priority calibration unit is used to calibrate the scanning priority value of the divided area according to the divided area and the attribute parameter characteristics and process requirement parameters within the area, using the configured local priority strategy.

[0144] Furthermore, the intelligent slicing and scanning path planning system for laser selective melting also includes a path planning module, a processing control module, a monitoring feedback module and an interaction module;

[0145] The path planning module is used to construct a path self-switching model based on the spiral-cycloid hybrid path algorithm and perform optimization simulation training on the constructed model in a simulation environment;

[0146] A processing control module is used to build a processing control model based on a reinforcement learning algorithm to control the processing scanning of the metal component by the processing head; the processing control module includes a control model building unit and a processing scanning control unit;

[0147] A control model building unit is used to build a processing control model based on a reinforcement learning algorithm and processing control parameters; a processing scanning control unit is used to control a processing head to perform processing scanning on a metal component to be processed by using a path self-switching model and a processing control model;

[0148] The monitoring feedback module is used to monitor the processing process in real time and adjust the processing path and processing control parameters in real time according to the monitoring results; the monitoring feedback module includes a monitoring unit and an analysis feedback unit;

[0149] The monitoring unit is used to monitor the temperature field, stress distribution and processing surface status data in real time using the sensors on the processing head; the analysis feedback unit is used to analyze the monitoring parameters using the monitoring algorithm and feed back the analysis results to the path self-switching model and the processing control model to perform path self-correction and real-time adjustment of the processing control parameters;

[0150] An interactive module is used to provide an interface for users to interact with the system, perform parameter setting, display and abnormal alarm; the interactive module includes a parameter setting and display unit and an abnormal alarm unit;

[0151] Parameter setting and display unit, used for users to set and display metal component property parameters, process requirement parameters and real-time monitoring processing parameters; real-time monitoring processing parameters include temperature field, stress distribution and processing surface state data during metal component processing;

[0152] The abnormality alarm unit is used to monitor the abnormality of the processing process using the abnormality monitoring algorithm configured in the integrated sensor, and to issue a real-time alarm for the monitored abnormal results.

[0153] Example 3

[0154] A computer-readable storage medium stores computer instructions, which, when executed, execute an intelligent slicing and scanning path planning method for laser selective melting.

[0155] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention, and all of these are within the protection of the present invention.

Claims

1. An intelligent slicing and scanning path planning method for selective laser melting, characterized in that: include: S1. Collect and preprocess the property parameters and process requirement parameters of the metal components to be processed, use the preprocessed property parameters of the metal components to build a metal geometric model, and use the deep learning combined with the genetic algorithm to build an intelligent stratification strategy to adaptively stratify the metal geometric model according to the constructed geometric model, and use the preprocessed process requirement parameters to adaptively adjust the stratification process; S2. Based on the layered slicing results in S1, the attribute parameter characteristics of each layer of metal component slices after stratification are obtained, and the configured local priority strategy is used to automatically calibrate the scanning priority value for each scanning area in each layer of slices according to the attribute parameters of each layer of metal component slices; S3. Introduce the spiral-cycloid hybrid path algorithm, construct a path self-switching model, use the acquired regional scanning priority and metal component attribute parameters and process requirement parameters to optimize the constructed path self-switching model in a simulation environment, and obtain the trained path self-switching model; S4. Building and training a processing control model based on a reinforcement learning algorithm and processing control parameters, and using the acquired path self-switching model and processing control model to control the processing scanning of the processing head on the metal component to be processed; S5. Use the sensor integrated on the processing head to monitor the temperature field, stress distribution and processing surface status in real time during the processing, and use the monitoring algorithm configured in the sensor to analyze the monitoring parameters. Feedback the analysis results to the path self-switching model and the processing control model to self-correct the processing path and adjust the processing control parameters in real time.

2. The intelligent slicing and scanning path planning method for selective laser melting according to claim 1, characterized in that: The specific process of S1 includes: S101, collecting attribute parameters, process requirement parameters and historical stratification strategies of metal components, and preprocessing the collected parameters to keep them in a unified input format; S102, constructing o-xyz three-dimensional coordinates with the center of gravity of the metal component as the origin, inputting the pre-processed property parameters of the metal component into the three-dimensional modeling software to construct a three-dimensional virtual model of the metal component in the three-dimensional coordinates, and marking the property parameters on the constructed three-dimensional virtual model; S103, constructing a stratification strategy model using a BP neural network combined with a genetic algorithm, constructing a model input according to the acquired attribute parameters, process requirement parameters and historical stratification strategies, and constructing a model output as a stratification strategy, wherein the stratification strategy includes the number of layers L, the layer thickness H, the stratification slicing angle A and the real-time speed v during the movement of the laser processing head; S104, inputting the constructed input and output into the hierarchical strategy model for training, and obtaining a trained hierarchical strategy model; S105. Integrate the trained layering strategy model into the three-dimensional modeling software, perform virtual layered slicing on the three-dimensional virtual model of the metal component, and use the PID algorithm to generate a corresponding control signal for the virtual slicing process. Use the sensor to transmit the control signal to the corresponding processing head to control the processing head to perform layered slicing on the metal component.

3. The intelligent slicing and scanning path planning method for selective laser melting according to claim 2, characterized in that: The specific steps of training the hierarchical strategy model in S104 include: S1041. Use computer software to extract component attribute parameter vectors marked on the three-dimensional virtual model At the same time, the process requirement parameters are used as the target control variables in the training process and numerically preprocessed, and the process requirement attribute parameter vector is constructed using the preprocessed process requirement parameters , the component attribute parameter vector and the process requirement attribute parameter vector are cascaded to form a joint input vector ;in represents the metal component attribute parameter of the ith type, and n represents the number of categories of the metal component attribute parameter; represents the process requirement attribute parameter of the jth category, and m represents the number of categories of process requirement attribute parameters; S1042, constructing a BP neural network using three convolutional layers and two fully connected layers connected in series, inputting the constructed joint input vector into the BP neural network, and using the collected process requirement attribute parameters and the set target process requirement attribute parameters to obtain a mean square error as a training error, training the constructed BP neural network, and outputting an initial stratification strategy; S1043, constructing an initial population of the genetic algorithm using the initial stratification strategy and the historical stratification strategy; constructing the qth chromosome input by the genetic algorithm using the obtained initial population , and set the following constraint set S for each variable in the chromosome: ， in, represents the number of slice layers in the qth chromosome, represents the slice thickness parameter in the qth chromosome, represents the layered slicing angle parameter in the qth chromosome, represents the speed parameter of the processing head moving process in the qth chromosome, In turn, they represent the minimum number of slice layers and the maximum number of slice layers. represents the minimum slice thickness and the maximum slice thickness respectively. represents the minimum stratification slicing angle and the maximum stratification slicing angle respectively, In turn, they represent the minimum processing head moving process speed and the maximum processing head moving process speed; S1044. Use the initial population to construct the fitness function corresponding to the genetic algorithm , the specific formula is: ， in, represents the processing time corresponding to the qth chromosome, is an indicator function, indicating that when the qth chromosome corresponds to the processing time Exceeding the preset maximum time The value is 1 when , otherwise it is 0. Indicates that when the qth chromosome corresponds to the processing time The penalty coefficient corresponding to exceeding the preset maximum time, represents the processing cost corresponding to the qth chromosome, Indicates the quality score corresponding to the qth chromosome, It represents the penalty term for violating the sth constraint in the constraint set S corresponding to the qth chromosome. If the sth constraint in the constraint set S is not satisfied, the term is 1, otherwise it is 0. They represent the weighted coefficients of the fitness function in turn; S1045. Set the iteration cycle and stop condition, input the constructed qth chromosome, constraint set S, and fitness function into the genetic algorithm, iterate the above S1041-S1044 process, and when the iteration cycle or stop condition is met, stop the iteration and output the obtained optimal stratification strategy.

4. The intelligent slicing and scanning path planning method for selective laser melting according to claim 3, characterized in that: The specific steps of S2 include: S201, in the 3D modeling software, gridding each layer of slices after layered slicing using a fixed-size square grid, numerically labeling the divided grids, and obtaining attribute parameter features in each labeled grid; S202, inputting the acquired attribute parameter features and corresponding labels in each grid into a clustering algorithm, performing a clustering operation, and merging grid labels belonging to the same category after clustering to obtain each scanning area of ​​the corresponding layered slice; S203, construct a local priority model, and input the acquired attribute parameter features of each scanning area corresponding to the layered slices into the local priority model for training, and integrate the trained local priority model into the three-dimensional virtual model of the metal component in S105 to automatically mark the corresponding scanning area priorities of the three-dimensional virtual model slices after layered slicing.

5. The intelligent slicing and scanning path planning method for selective laser melting according to claim 4, characterized in that: The specific steps of the local priority model training include: S2031. Collect historical attribute parameters and process requirement parameters of the slicing areas of the same type of metal components, and use expert experience to mark the processing priority of each scanning area according to the collected parameters to obtain priority label data; the priorities are divided into 1-9, where 1-3 are low priorities, 4-6 are medium priorities, and 7-9 are high priorities; S2032. Use support vector regression SVR to build a local priority model, and input the obtained priority label data set into the local priority model for training to obtain a trained local priority model.

6. The intelligent slicing and scanning path planning method for selective laser melting according to claim 5, characterized in that: The path self-switching model includes a spiral path planning sub-model, a cycloid path sub-model, a spiral-cycloid path hybrid planning sub-model and a path planning priority working switching condition, and the path planning priority working switching condition specifically includes: S3031, when the priority of the scanning area corresponding to the layered slice is low priority, the spiral path planning submodel in the path self-switching model is used to perform scanning path planning; S3032, when the priority of the scanning area corresponding to the layered slice is medium priority, the spiral-cycloid path hybrid planning submodel in the path self-switching model is used to perform scanning path planning; S3033. When the priority of the scanning area corresponding to the layered slice is a high priority, the scanning path is planned using the cycloid path sub-model in the path self-switching model.

7. The intelligent slicing and scanning path planning method for selective laser melting according to claim 6, characterized in that: The specific steps of constructing the processing control model of S4 include: S401. Construct processing control constraints based on metal component property parameters, process requirement parameters and processing control parameters. The specific conditions are: ， in, represents the laser output power of the laser processing head at time t, represents the optimal power threshold corresponding to the laser output power of the laser processing head at time t, represents the thickness of the metal component after layered processing at time t, It represents the target thickness of the metal component layered processing at time t, represents the path planning sub-model of the laser processing head's travel path at time t, The path planning sub-model represents the path target execution of the laser processing head at time t, represents the strength of the processed metal component at time t, Indicates the target strength of processed metal components, It represents the speed of the laser processing head at time t, Indicates the target speed of the metal laser processing head during processing; S402: According to the constraint conditions in S401, construct the processing control trigger conditions, which specifically include: when and When the processing control constraint condition is not met, the first processing control condition is triggered; when and When the processing control constraint condition is not met, the second processing control condition is triggered; when and When the processing control constraint condition is not met, the third processing control condition is triggered; when and When the processing control constraint condition is not met, the fourth processing control condition is triggered; when and When the processing control constraint condition is not met, the fifth processing control condition is triggered; When two or more processing control constraint conditions in S402 are not satisfied, the sixth processing control condition is triggered.

8. An intelligent slicing and scanning path planning system for selective laser melting, which is implemented based on the intelligent slicing and scanning path planning method for selective laser melting according to any one of claims 1 to 7, characterized in that: include: Data processing module, intelligent stratification module, priority calibration module; The data processing module is used to collect and process the property parameters and process requirement parameters of the metal component to be processed, so that the collected data is kept in a unified format; The intelligent stratification module is used to adaptively stratify metal components by using deep learning combined with genetic algorithms; The intelligent stratification module includes a geometric modeling unit and a strategy stratification unit; The geometric modeling unit is used to construct a metal geometric model using computer software according to the pre-processed property parameters of the metal component to be processed; the strategy stratification unit is used to adaptively stratify the metal geometric model according to the constructed geometric model using the configured intelligent stratification strategy, and to optimize the stratification process in real time using the process requirement parameters; The priority calibration module is used to automatically calibrate the scanning priority value for each scanning area in each layer of slices based on the stratification result of the strategy stratification unit; The priority calibration module includes a region division unit and a priority calibration unit; The area division unit is used to use sensors to extract the attribute parameter characteristics of each layer of slices, and use the configured clustering algorithm to divide the area on each layer of slices; the priority calibration unit is used to calibrate the scanning priority value of the divided area according to the divided area and the attribute parameter characteristics and process requirement parameters within the area, using the configured local priority strategy.

9. The intelligent slicing and scanning path planning system for selective laser melting according to claim 8, characterized in that: It also includes a path planning module, a processing control module and a monitoring feedback module; The path planning module is used to construct a path self-switching model according to the spiral-cycloid hybrid path algorithm, and perform optimization simulation training on the constructed model in a simulation environment; The processing control module is used to build a processing control model based on a reinforcement learning algorithm to control the processing scanning of the metal component by the processing head; the processing control module includes a control model building unit and a processing scanning control unit; The control model building unit is used to build a processing control model based on a reinforcement learning algorithm and processing control parameters; the processing scanning control unit is used to control the processing head to perform processing scanning on the metal component to be processed by using the path self-switching model and the processing control model; The monitoring feedback module is used to monitor the processing process in real time and adjust the processing path and processing control parameters in real time according to the monitoring results; the monitoring feedback module includes a monitoring unit and an analysis feedback unit; The monitoring unit is used to use the sensors on the processing head to monitor the temperature field, stress distribution and processing surface status data in real time; the analysis feedback unit is used to use the monitoring algorithm to analyze the monitoring parameters, and feed back the analysis results to the path self-switching model and the processing control model to perform path self-correction and real-time adjustment of the processing control parameters.

10. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the intelligent slicing and scanning path planning method for selective laser melting described in any one of claims 1 to 7 is executed.

Images