Laser drilling process parameter optimization method based on artificial intelligence

Abstract

The present invention discloses a method for optimizing laser drilling process parameters based on artificial intelligence, which relates to the field of intelligent control technology. The method comprises: interacting with target scenes, extracting substrate features and equipment features, and outputting scene task features. Based on the principle of big data, the scene task features are used as index constraints to screen a dual-target sample process parameter set. According to the dual-target sample process parameter set, optimization initialization is performed, near-source sample space and far-source sample space are configured, and heterogeneous parallel optimization is performed to obtain the optimal control parameter set. Based on the simulation method, a two-dimensional numerical model of the target scene is constructed, and an isolated temperature field analysis is performed on the optimal control parameter set to obtain the isolated temperature field. Multiple isolated temperature fields are combined for time series joint analysis, path optimization is performed, the optimal path is output, and laser drilling control is performed with the optimal control parameter set and the optimal path as process parameters. Thereby achieving the technical effect of reducing optimization costs and improving optimization efficiency.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a method for optimizing laser drilling process parameters based on artificial intelligence. Background Art

[0002] Laser drilling technology is a high-precision and high-efficiency processing method. The selection of laser drilling process parameters has a crucial impact on the processing effect. Common process parameters include laser power, pulse frequency, scanning speed, focus position, etc. There is a complex nonlinear relationship between these parameters. The processing requirements of different materials and workpieces are different. Traditional empirical methods and trial and error methods are difficult to quickly and accurately find the optimal parameter combination. There are technical problems such as high trial and error costs and low optimization efficiency. Summary of the invention

[0003] The present invention provides a laser drilling process parameter optimization method based on artificial intelligence to solve the technical problems of high trial and error cost and low optimization efficiency in the prior art, and achieves the technical effect of reducing optimization cost and improving optimization efficiency.

[0004] The present invention provides a method for optimizing laser drilling process parameters based on artificial intelligence, wherein the method comprises:

[0005] Interactive target scene, extracting substrate features based on drilling task information, extracting equipment features based on scene laser equipment information, and outputting the substrate features and the equipment features as scene task features; obtaining historical drilling records based on the big data principle, and taking the scene task features as index constraints, traversing and screening to obtain a dual-target sample process parameter set, wherein the dual-target sample process parameter set includes a near-source sample process parameter set and a far-source sample process parameter set; performing optimization initialization according to the dual-target sample process parameter set, configuring the near-source sample space and the far-source sample space, and performing heterogeneous parallel optimization to obtain an optimal control parameter set, wherein a sample shuttle path is set up between the near-source sample space and the far-source sample space; constructing a two-dimensional numerical model of the target scene based on a simulation method and combining the scene task features, and performing an isolated temperature field analysis on the optimal control parameter set to obtain an isolated temperature field; performing a time-series-based joint analysis on multiple isolated temperature fields according to the drilling task information, performing path optimization, outputting an optimal path, and performing laser drilling control with the optimal control parameter set and the optimal path as process parameters.

[0006] In a feasible implementation, the interactive target scene, the substrate features are extracted based on the punching task information, and the device features are extracted based on the scene laser device information, including:

[0007] Interact with the target scene, parse the punching task document, and collect the punching task information; based on the punching task information, obtain the intrinsic characteristics of the substrate and the finished product characteristics of the substrate, and store them as the substrate characteristics; based on the scene laser equipment information, obtain the laser parameters, and store them as the equipment characteristics, wherein the equipment characteristics at least include equipment wavelength characteristics and equipment power characteristics.

[0008] In a feasible implementation, historical drilling records are obtained based on the principle of big data, and the scene task features are used as index constraints to traverse and screen to obtain a dual-target sample process parameter set, including:

[0009] Based on the record source mark, the historical punching records are divided into a local record set and a big data record set; taking the scene task feature as the index constraint, the local record set and the big data record set are traversed respectively to obtain the near-source sample process parameter set and the far-source sample process parameter set; the near-source sample process parameter set and the far-source sample process parameter set are output as the dual-target sample process parameter set.

[0010] In a feasible implementation, according to the dual-objective sample process parameter set, optimization initialization is performed, near-source sample space and far-source sample space are configured, and heterogeneous parallel optimization is performed to obtain the optimal control parameter set, including:

[0011] The near-source sample space and the far-source sample space are initialized with the near-source sample process parameter set and the far-source sample process parameter set respectively, wherein the near-source sample space is smaller than the far-source sample space; the sample shuttle path is set up between the near-source sample space and the far-source sample space; sample shuttle constraints are configured, and a heterogeneous optimization algorithm is selected to perform parallel optimization on the near-source sample space and the far-source sample space in combination with a parameter evaluation function; samples are shared through the sample shuttle path based on the sample shuttle constraints, and an optimal control parameter set is output.

[0012] In a feasible implementation, based on a simulation method and in combination with the scene task characteristics, a two-dimensional numerical model of the target scene is constructed, and the two-dimensional numerical model includes a heat conduction mathematical model and a fluid flow mathematical model.

[0013] In a feasible implementation, according to the punching task information, a plurality of the isolated temperature fields are combined to perform a joint analysis based on time series, perform path optimization, and output an optimal path, including:

[0014] A global evaluation function is constructed, wherein the global evaluation function includes a path trajectory evaluation factor, a temperature field distribution evaluation factor, and a substrate energy evaluation factor; based on a preset path planning strategy, a multimode path set is obtained; based on the intrinsic characteristics of the substrate, a temperature attenuation model is obtained, and a time-series-based joint analysis is performed on the multimode path set in combination with the temperature attenuation model and the isolated temperature field; an evaluation is performed according to the global evaluation function, and the optimal path is screened and obtained.

[0015] In a feasible implementation, a sample shuttle constraint is configured, a parameter evaluation function is combined, a heterogeneous optimization algorithm is selected to optimize the near-source sample space and the far-source sample space in parallel, and samples are shared through the sample shuttle path based on the sample shuttle constraint, and an optimal control parameter set is output, which also includes:

[0016] Based on the punching task information, the parameter evaluation function is constructed; with the parameter evaluation function as the cost function, the near-source sample space and the far-source sample space are iteratively optimized; when there are samples that satisfy the sample shuttling constraint, the sample shuttling path is activated to share the optimal solution sample; the iterative optimization process is repeated until the global optimal sample satisfies the punching task information requirements, and the output is the optimal control parameter set.

[0017] In a feasible implementation, the method further includes:

[0018] The historical punching records are interacted, and a poor solution domain is determined through association analysis; a penalty space and a penalty function are configured based on the poor solution domain; and heterogeneous parallel optimization is performed by combining the penalty space and the penalty function.

[0019] The invention discloses a laser drilling process parameter optimization method based on artificial intelligence, comprising: an interactive target scene, extracting substrate features based on drilling task information, extracting device features based on scene laser device information, and outputting substrate features and device features as scene task features; acquiring historical drilling records based on the big data principle, and taking scene task features as index constraints, traversing and screening to acquire a dual-target sample process parameter set, wherein the dual-target sample process parameter set includes a near-source sample process parameter set and a far-source sample process parameter set; performing optimization initialization according to the dual-target sample process parameter set, configuring a near-source sample space and a far-source sample space, and performing heterogeneous parallel optimization to acquire an optimal control parameter set, wherein a sample shuttle path is set up between the near-source sample space and the far-source sample space; constructing a two-dimensional numerical model of the target scene based on a simulation method in combination with scene task features, and performing isolated temperature field analysis on the optimal control parameter set to acquire an isolated temperature field; performing a time-series-based joint analysis in conjunction with a plurality of isolated temperature fields according to the drilling task information, performing path optimization, outputting an optimal path, and performing laser drilling control with the optimal control parameter set and the optimal path as process parameters. The artificial intelligence-based laser drilling process parameter optimization method disclosed in the present invention solves the technical problems of high trial and error cost and low optimization efficiency, and achieves the technical effect of reducing optimization cost and improving optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic flow chart of the laser drilling process parameter optimization method based on artificial intelligence of the present invention.

[0021] Figure 2 It is a schematic diagram of the process of obtaining the optimal control parameter set in the laser drilling process parameter optimization method based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0022] The technical solution provided in the embodiments of the present invention is to solve the technical problems of high trial and error cost and low optimization efficiency in the prior art. The overall idea adopted is as follows:

[0023] First, in an interactive target scene, substrate features are extracted based on the punching task information, and device features are extracted based on the scene laser device information, and the substrate features and the device features are output as scene task features.

[0024] Next, based on the principle of big data, historical punching records are obtained, and the scene task characteristics are used as index constraints to traverse and filter to obtain a dual-target sample process parameter set, wherein the dual-target sample process parameter set includes a near-source sample process parameter set and a far-source sample process parameter set.

[0025] Then, according to the dual-target sample process parameter set, optimization initialization is performed, the near-source sample space and the far-source sample space are configured, and heterogeneous parallel optimization is performed to obtain the optimal control parameter set, wherein a sample shuttle path is set up between the near-source sample space and the far-source sample space.

[0026] Next, based on the simulation method, a two-dimensional numerical model of the target scene is constructed in combination with the scene task characteristics, and an isolated temperature field analysis is performed on the optimal control parameter set to obtain an isolated temperature field.

[0027] Subsequently, according to the punching task information, a plurality of the isolated temperature fields are combined to perform a time-series-based joint analysis, perform path optimization, and output an optimal path.

[0028] Finally, laser drilling control is performed using the optimal control parameter set and the optimal path as process parameters.

[0029] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0030] Example

[0031] Figure 1 The figure is a flow chart of the laser drilling process parameter optimization method based on artificial intelligence of the present invention, wherein the method comprises:

[0032] Interactive target scene, extracting substrate features based on punching task information, extracting device features based on scene laser device information, and outputting the substrate features and the device features as scene task features.

[0033] Optionally, collect drilling task information to obtain specific features related to the substrate, where substrate features include substrate type, thickness, hardness, material, surface treatment, etc. Collect information about laser equipment in the scene and extract relevant equipment features. Equipment features include power, wavelength, focal length, spot diameter, cooling method, etc. of the laser equipment. Combine substrate features and equipment features to form a complete scene task feature to guide the execution of the drilling task.

[0034] By extracting and integrating substrate features and equipment features, the execution environment of the punching task can be more accurately understood and controlled, thereby improving the accuracy and efficiency of the punching task.

[0035] In some embodiments, the interactive target scene, extracting substrate features based on punching task information, and extracting device features based on scene laser device information include:

[0036] Interact with the target scene, parse the punching task document, and collect the punching task information; based on the punching task information, obtain the intrinsic characteristics of the substrate and the finished product characteristics of the substrate, and store them as the substrate characteristics; based on the scene laser equipment information, obtain the laser parameters, and store them as the equipment characteristics, wherein the equipment characteristics at least include equipment wavelength characteristics and equipment power characteristics.

[0037] Optionally, in an interactive target scenario, documents related to the drilling task are read and parsed to extract task requirements and parameters, including, for example, what material to drill holes on, hole diameter, hole depth, number of holes, hole distribution and other specific parameters.

[0038] Among them, the intrinsic characteristics of the substrate refer to the physical and chemical properties of the substrate, including the type of substrate (such as steel plate, aluminum plate, plastic, etc.), material composition (specific material composition of the substrate, such as carbon steel, alloy, etc.), thickness, hardness, etc. The finished material characteristics refer to the characteristics that the substrate should have after processing, such as the accuracy requirements that the substrate needs to achieve after processing (such as hole accuracy, hole edge integrity, hole inner wall smoothness, etc.).

[0039] Optionally, based on the scene laser device information, laser parameters are obtained, and the laser parameters include device wavelength characteristics and device power characteristics. Among them, the laser wavelength of the laser drilling device determines its penetration depth and energy transfer efficiency in the material, and the laser power of the laser drilling device determines its drilling speed and depth.

[0040] Optionally, the laser parameters also include pulse frequency, focal length, etc., where pulse frequency refers to the number of pulses emitted by the laser drilling device per second, usually in Hertz (Hz). Focal length refers to the distance between the focus of the laser beam and the surface of the workpiece after focusing, which affects the accuracy and effect of laser drilling.

[0041] Optionally, the device characteristics characterize the performance and parameter adjustable range of the scene laser device in the target scene. By obtaining the device characteristics, it can be ensured that the subsequently acquired process parameters are suitable for the current target scene.

[0042] Based on the principle of big data, historical punching records are obtained, and the scene task characteristics are used as index constraints to traverse and filter to obtain a dual-target sample process parameter set, wherein the dual-target sample process parameter set includes a near-source sample process parameter set and a far-source sample process parameter set.

[0043] In some embodiments, historical drilling records are obtained based on the big data principle, and the scene task features are used as index constraints to traverse and screen to obtain a dual-target sample process parameter set, including:

[0044] Based on the record source mark, the historical punching records are divided into a local record set and a big data record set; taking the scene task feature as the index constraint, the local record set and the big data record set are traversed respectively to obtain the near-source sample process parameter set and the far-source sample process parameter set; the near-source sample process parameter set and the far-source sample process parameter set are output as the dual-target sample process parameter set.

[0045] Optionally, historical punching records are divided based on the record source tag, where the historical punching records include all previously completed punching task records, including process parameters and punching results. In addition, each record will have a tag indicating the source of the record. The local record set represents historical punching records from the local system. The big data record set represents historical punching records from a wider source of the big data platform.

[0046] Optionally, the local record set is traversed, and records with high similarity to the current task are screened out according to the scene task characteristics, and stored as the near-source sample process parameter set. In other words, the near-source sample process parameter set is a set of local records that meet the index constraints. Further, also according to the scene task characteristics, the big data record set is traversed to screen out records with high similarity, and the far-source sample process parameter set is obtained, that is, the set of big data records that meet the index constraints.

[0047] Optionally, the near-source sample process parameter set and the far-source sample process parameter set are integrated to form a dual-target sample process parameter set. The dual-target sample process parameter set includes process parameter sets from both the local and big data platforms that meet the current task characteristics, and has relatively rich and diverse process parameter data.

[0048] According to the dual-target sample process parameter set, optimization initialization is performed, a near-source sample space and a far-source sample space are configured, and heterogeneous parallel optimization is performed to obtain an optimal control parameter set, wherein a sample shuttle path is set up between the near-source sample space and the far-source sample space.

[0049] Optionally, first, two sample spaces are initialized based on the near-source sample process parameter set and the far-source sample process parameter set in the dual-target sample process parameter set. Each sample point in the sample space represents a set of process parameter samples. Then, a data transmission path is established between the near-source sample space and the far-source sample space, allowing the sample data between the two spaces to interact and migrate under certain conditions, thereby facilitating the search for the global optimum in the subsequent optimization process.

[0050] Optionally, heterogeneous parallel optimization is performed based on the near-source sample space and the far-source sample space, which means that differentiated optimization algorithms are assigned to the above-mentioned near-source sample space and the far-source sample space, and iterative optimization is performed separately, and the optimization results of the near-source sample space and the far-source sample space are integrated, and the optimal control parameter set is selected, and the output is used for the execution of the actual punching task.

[0051] In some embodiments, Figure 2 As shown, according to the dual-objective sample process parameter set, optimization initialization is performed, near-source sample space and far-source sample space are configured, and heterogeneous parallel optimization is performed to obtain the optimal control parameter set, including:

[0052] The near-source sample space and the far-source sample space are initialized with the near-source sample process parameter set and the far-source sample process parameter set respectively, wherein the near-source sample space is smaller than the far-source sample space; the sample shuttle path is set up between the near-source sample space and the far-source sample space; sample shuttle constraints are configured, and a heterogeneous optimization algorithm is selected to perform parallel optimization on the near-source sample space and the far-source sample space in combination with a parameter evaluation function; samples are shared through the sample shuttle path based on the sample shuttle constraints, and an optimal control parameter set is output.

[0053] Optionally, first, two multidimensional parameter spaces are constructed, in which multiple dimensions of the parameter space correspond to multiple process parameters, and the spatial point position in the multidimensional parameter space represents a process parameter combination. Then, the near-source sample space is initialized based on the near-source sample process parameter set, and the multiple groups of process parameters in the near-source sample process parameter set are mapped to the near-source sample space; the far-source sample space is initialized based on the far-source sample process parameter set, and the multiple groups of process parameters in the far-source sample process parameter set are mapped to the far-source sample space.

[0054] Optionally, a sample shuttle path is set up between the near-source sample space and the far-source sample space, and the sample shuttle path is used to transmit data in the two sample spaces to ensure that the samples between the two spaces can flow and share with each other. Exemplarily, the path can be implemented by a software module, and data is exchanged by means of shared memory, data bus or communication protocol. The establishment of the sample shuttle path enables the near-source sample space and the far-source sample space to share the intermediate results of the optimization process in real time, thereby improving the efficiency and accuracy of the overall optimization.

[0055] Optionally, configure sample shuttle constraints, which define the conditions and rules for sample shuttle to determine when to trigger sample shuttle. Specifically, sample shuttle constraints include period constraints and mutation constraints. The period constraint controls the timing of sample exchange by setting the number of iterations, such as sharing samples every 10 iterations. Mutation constraints stipulate that sample sharing is performed when a parameter change reaches a certain threshold.

[0056] Optionally, different optimization algorithms or different optimization parameters are selected in the near-source sample space and the far-source sample space for parallel optimization, making full use of the advantages of heterogeneous optimization and improving the overall optimization effect. Exemplarily, the near-source sample space uses an algorithm suitable for local optimization, such as a genetic algorithm (GA), and has a smaller optimization step length, which is conducive to accurate search, and the far-source sample space uses an algorithm suitable for global search, such as a particle swarm optimization algorithm (PSO), and has a longer optimization step length, which is conducive to finding possible optimal solutions in the global space and avoiding falling into local optimality.

[0057] Optionally, when the optimization iteration process satisfies the preset iteration constraints, the optimization results of the near-source sample space and the far-source sample space are integrated, and the optimal process parameter combination is selected as the final output. By integrating the optimization results of the two sample spaces, the optimal control parameter set obtained has higher adaptability and accuracy, can effectively guide the actual process, and improve the overall production efficiency.

[0058] The above method steps adopt heterogeneous optimization algorithms for parallel optimization, and utilize the advantages of different algorithms to perform both local fine optimization and global search optimization, thereby improving the comprehensive effect of process parameter optimization.

[0059] In some implementations, configuring sample shuttle constraints, combining parameter evaluation functions, selecting heterogeneous optimization algorithms to optimize the near-source sample space and the far-source sample space in parallel, and sharing samples through the sample shuttle path based on the sample shuttle constraints, and outputting an optimal control parameter set, further includes:

[0060] Based on the punching task information, the parameter evaluation function is constructed; with the parameter evaluation function as the cost function, the near-source sample space and the far-source sample space are iteratively optimized; when there are samples that satisfy the sample shuttling constraint, the sample shuttling path is activated to share the optimal solution sample; the iterative optimization process is repeated until the global optimal sample satisfies the punching task information requirements, and the output is the optimal control parameter set.

[0061] Optionally, during the optimization process, a parameter evaluation function is used as a cost function to evaluate the optimization effect of each iteration. The parameter evaluation function is set according to specific process requirements and performance indicators, including multiple dimensions such as drilling accuracy, processing speed, material loss, and heat-affected zone.

[0062] Optionally, based on the punching quality evaluation method for the substrate in the punching task information, a parameter evaluation function is constructed, and the quality evaluation indicators with higher attention in the target scene have higher evaluation weights, thereby ensuring that the obtained parameter evaluation function can better reflect the quality control preferences and needs of the target scene.

[0063] Optionally, during the optimization process, the near-source sample space and the far-source sample space are iteratively optimized multiple times using the parameter evaluation function as the cost function. When there are samples that satisfy the sample shuttle constraint, the sample shuttle pathway is activated to share the optimal solution samples. Specifically, the optimal solution samples are exchanged, replaced, shared, mutated, etc. between the near-source sample space and the far-source sample space through the sample shuttle pathway.

[0064] Furthermore, the shared optimal solution samples are used to repeat the iterative optimization process to further optimize the near-source sample space and the far-source sample space until the global optimal sample meets the punching task information requirements. Based on the optimal sample, the actual process parameters are output to obtain the optimal control parameter set.

[0065] Based on the simulation method, a two-dimensional numerical model of the target scene is constructed in combination with the scene task characteristics, and an isolated temperature field analysis is performed on the optimal control parameter set to obtain an isolated temperature field.

[0066] Optionally, according to the specific requirements of the scene task, select a suitable numerical modeling method, such as finite element method (FEM), finite difference method (FDM), etc., to construct a two-dimensional numerical model of the target scene. The two-dimensional numerical model is constructed based on the basic principles of laser drilling. Specifically, laser drilling includes the joint action of multiple processes such as melting, evaporation, splashing, and flow, involving laser physics, nonlinear optics, heat transfer, thermodynamics, fluid mechanics, gas dynamics, etc. When establishing a two-dimensional numerical model of the target scene, it is necessary to simplify the physical phenomena in the actual drilling accordingly to facilitate subsequent simulation calculations.

[0067] Optionally, an isolated temperature field analysis is performed on the two-dimensional numerical model of the target scene using the optimal control parameter set to obtain the distribution of the temperature field, wherein the isolated temperature field refers to the temperature distribution of the heated area of ​​the substrate near a single hole during the laser drilling process.

[0068] For example, first, input the optimal control parameter set into the two-dimensional numerical model and set the simulation conditions; then use numerical simulation software (such as COMSOL, ANSYS, etc.) to perform isolated temperature field analysis and calculate the temperature distribution of the substrate during laser drilling; finally, analyze the simulation results to obtain the isolated temperature field distribution during the drilling process. The isolated temperature field distribution includes the distribution of high temperature areas and low temperature areas, as well as the change of temperature gradient.

[0069] By combining simulation methods, building a two-dimensional numerical model of the target scene, and performing isolated temperature field analysis on the optimal control parameter set, the temperature distribution during laser drilling can be accurately evaluated, which helps to understand the heating conditions of the substrate under dense hole distribution, and then optimize the process parameters to reduce the deformation and creep of the substrate caused by excessive heat or uneven temperature distribution, which affects the quality of the finished product.

[0070] In some embodiments, based on a simulation method, a two-dimensional numerical model of the target scene is constructed in combination with the scene task characteristics, and the two-dimensional numerical model includes a heat conduction mathematical model and a fluid flow mathematical model.

[0071] Optionally, a two-dimensional numerical model of the target scene is constructed based on ideal assumptions, including assuming that the laser beam is an ideal fundamental mode Gaussian beam, assuming that there are no multiple reflections of the laser in the hole during the drilling process, assuming that the substrate vapor produced by evaporation is a transparent gas that does not absorb the laser energy, and assuming that the substrate is isotropic.

[0072] For example, a heat conduction mathematical model is established, including considering factors such as laser heating, substrate absorption, heat transfer, etc., and defining heat conduction equations and boundary conditions. A fluid flow mathematical model is established, including considering factors such as laser convection and substrate surface flow, and defining a flow model, a gas-liquid interface tracking model, and a force model.

[0073] According to the drilling task information, a plurality of the isolated temperature fields are combined to perform a timing-based joint analysis, perform path optimization, output an optimal path, and perform laser drilling control using the optimal control parameter set and the optimal path as process parameters.

[0074] In some embodiments, according to the punching task information, a plurality of the isolated temperature fields are combined to perform a joint analysis based on time series, perform path optimization, and output an optimal path, including:

[0075] A global evaluation function is constructed, wherein the global evaluation function includes a path trajectory evaluation factor, a temperature field distribution evaluation factor, and a substrate energy evaluation factor; based on a preset path planning strategy, a multimode path set is obtained; based on the intrinsic characteristics of the substrate, a temperature attenuation model is obtained, and a time-series-based joint analysis is performed on the multimode path set in combination with the temperature attenuation model and the isolated temperature field; an evaluation is performed according to the global evaluation function, and the optimal path is screened and obtained.

[0076] Optionally, a global evaluation function is constructed, in which the path trajectory evaluation factor considers factors such as the length, shape, and smoothness of the path to make the path more in line with the actual drilling requirements; the temperature field distribution evaluation factor is used to evaluate the uniformity and stability of the temperature field on the path to avoid excessive or uneven temperatures. The substrate energy evaluation factor is used to evaluate the degree of laser heating of the substrate to avoid excessive thermal damage. Excessive energy accumulation on the substrate may cause changes in the physical or chemical properties of the substrate.

[0077] Optionally, different path planning strategies are designed according to the punching task information and substrate characteristics, such as the shortest path, interval-constrained punching path (two adjacent points are at least a certain distance apart), etc. The multi-mode path set is obtained based on a variety of different path planning strategies.

[0078] Optionally, based on the intrinsic characteristics of the substrate and the isolated temperature field data, a temperature decay model over time is established to analyze the impact of different paths on the temperature. The temperature decay model is a simplified model obtained by fitting based on the actual observation analysis results, such as an exponential decay model, a polynomial model or other functions suitable for describing temperature changes, which characterizes the mapping relationship between time and temperature.

[0079] Furthermore, the temperature attenuation model and the isolated temperature field are combined to perform a time-series joint analysis of the multi-mode path set to analyze the impact of different paths on temperature. Among them, for different paths, the distribution position and distribution time of the isolated temperature field are different. In other words, under different drilling paths, the hole position nodes may start heating and enter temperature attenuation (i.e., cooling process) at different time nodes, thus affecting the temperature field of the substrate as a whole.

[0080] Through joint analysis and path optimization based on timing, the optimal laser drilling path can be found according to the drilling task information and isolated temperature field data, thereby effectively controlling the temperature distribution during the drilling process and improving the drilling quality and efficiency.

[0081] Furthermore, the method further comprises:

[0082] The historical punching records are interacted, and a poor solution domain is determined through association analysis; a penalty space and a penalty function are configured based on the poor solution domain; and heterogeneous parallel optimization is performed by combining the penalty space and the penalty function.

[0083] Optionally, historical punching records are obtained interactively and analyzed to determine the inferior solution domain. The inferior solution domain refers to the parameter space area that causes performance degradation or does not meet the requirements under certain conditions. Then, according to the determined inferior solution domain, the penalty space and the corresponding penalty function are configured. Among them, the penalty space refers to the restriction or setting of the value range of certain parameters. The penalty function modifies the optimization objective function according to the degree to which the parameter value is within the inferior solution domain.

[0084] Furthermore, the parameter space is optimized and searched by combining the penalty space and the penalty function using the heterogeneous parallel optimization algorithm. This includes introducing the penalty space into the near-source sample space and the far-source sample space, and introducing the penalty function into the global evaluation function. Specifically, based on the penalty space, the near-source sample space and the far-source sample space are updated and initialized, that is, multiple samples in the penalty space are eliminated. The above steps help avoid falling into the inferior solution domain during the optimization process and ensure that the final control parameter set meets the requirements in all aspects.

[0085] In summary, the laser drilling process parameter optimization method based on artificial intelligence provided by the present invention has the following technical effects:

[0086] Through interactive target scenes, the substrate features are extracted based on the punching task information, and the equipment features are extracted based on the scene laser equipment information, and the substrate features and equipment features are output as scene task features; based on the principle of big data, historical punching records are obtained, and the scene task features are used as index constraints to traverse and screen to obtain the dual-target sample process parameter set, where the dual-target sample process parameter set includes the near-source sample process parameter set and the far-source sample process parameter set; according to the dual-target sample process parameter set, optimization initialization is performed, the near-source sample space and the far-source sample space are configured, and heterogeneous parallel optimization is performed to obtain the optimal control parameter set, where a sample shuttle path is set up between the near-source sample space and the far-source sample space; based on the simulation method, a two-dimensional numerical model of the target scene is constructed in combination with the scene task features, and an isolated temperature field analysis is performed on the optimal control parameter set to obtain the isolated temperature field; according to the punching task information, multiple isolated temperature fields are combined for time-series-based joint analysis, path optimization is performed, the optimal path is output, and laser punching control is performed with the optimal control parameter set and the optimal path as process parameters. Thus, the technical effect of reducing optimization cost and improving optimization efficiency is achieved.

[0087] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. The laser drilling process parameter optimization method based on artificial intelligence is characterized by: The method comprises: Interactive target scene, extracting substrate features based on punching task information, extracting device features based on scene laser device information, and outputting the substrate features and the device features as scene task features; Based on the principle of big data, historical punching records are obtained, and the scene task characteristics are used as index constraints to traverse and filter to obtain a dual-target sample process parameter set, wherein the dual-target sample process parameter set includes a near-source sample process parameter set and a far-source sample process parameter set; According to the dual-target sample process parameter set, optimization initialization is performed, a near-source sample space and a far-source sample space are configured, and heterogeneous parallel optimization is performed to obtain an optimal control parameter set, wherein a sample shuttle path is set up between the near-source sample space and the far-source sample space; Based on the simulation method, a two-dimensional numerical model of the target scene is constructed in combination with the scene task characteristics, and an isolated temperature field analysis is performed on the optimal control parameter set to obtain an isolated temperature field; According to the drilling task information, a plurality of the isolated temperature fields are combined to perform a joint analysis based on timing, perform path optimization, output an optimal path, and perform laser drilling control using the optimal control parameter set and the optimal path as process parameters; Among them, the interactive target scene extracts substrate features based on the punching task information and extracts device features based on the scene laser device information, including: Interact with the target scene, parse the punching task document, and collect the punching task information; Based on the punching task information, the intrinsic characteristics of the substrate and the finished product characteristics of the substrate are obtained and stored as the substrate characteristics, wherein the intrinsic characteristics of the substrate refer to the physical and chemical properties of the substrate, including the type of substrate, material composition, thickness, and hardness; the finished product characteristics refer to the characteristics that the substrate should have after processing; Based on the scene laser device information, laser parameters are acquired and stored as the device characteristics, wherein the device characteristics at least include device wavelength characteristics and device power characteristics; Among them, based on the principle of big data, historical drilling records are obtained, and the scene task characteristics are used as index constraints to traverse and screen to obtain a dual-target sample process parameter set, including: Based on the record source mark, the historical punching records are divided into a local record set and a big data record set; Taking the scene task feature as an index constraint, traverse the local record set and the big data record set respectively to obtain the near-source sample process parameter set and the far-source sample process parameter set; Outputting the near-source sample process parameter set and the far-source sample process parameter set as the dual-target sample process parameter set; According to the dual-objective sample process parameter set, optimization initialization is performed, near-source sample space and far-source sample space are configured, and heterogeneous parallel optimization is performed to obtain the optimal control parameter set, including: Initializing the near-source sample space and the far-source sample space with the near-source sample process parameter set and the far-source sample process parameter set respectively, wherein the near-source sample space is smaller than the far-source sample space; Setting up the sample shuttle path between the near-source sample space and the far-source sample space; Configure sample shuttle constraints, combine parameter evaluation functions, select heterogeneous optimization algorithms to optimize the near-source sample space and the far-source sample space in parallel, and share samples through the sample shuttle path based on the sample shuttle constraints, and output an optimal control parameter set; Among them, heterogeneous parallel optimization is performed, including: Interact with the historical punching records and determine the inferior solution domain through correlation analysis; Based on the inferior solution domain, configuring a penalty space and a penalty function; The penalty space and the penalty function are combined to perform heterogeneous parallel optimization.

2. The method for optimizing laser drilling process parameters based on artificial intelligence according to claim 1, characterized in that: Based on the simulation method, a two-dimensional numerical model of the target scene is constructed in combination with the scene task characteristics. The two-dimensional numerical model includes a heat conduction mathematical model and a fluid flow mathematical model.

3. The method for optimizing laser drilling process parameters based on artificial intelligence according to claim 2, characterized in that: According to the drilling task information, a plurality of the isolated temperature fields are combined to perform a joint analysis based on time series, perform path optimization, and output an optimal path, including: Constructing a global evaluation function, wherein the global evaluation function includes a path trajectory evaluation factor, a temperature field distribution evaluation factor, and a substrate energy evaluation factor; Based on the preset path planning strategy, a multi-mode path set is obtained; Based on the intrinsic characteristics of the substrate, a temperature attenuation model is obtained, and the temperature attenuation model is combined with the isolated temperature field to perform a joint analysis on the multimode path set based on a time series; The optimal path is obtained by performing evaluation according to the global evaluation function.

4. The method for optimizing laser drilling process parameters based on artificial intelligence according to claim 1, characterized in that: The method further comprises: configuring sample shuttle constraints, combining parameter evaluation functions, selecting a heterogeneous optimization algorithm to optimize the near-source sample space and the far-source sample space in parallel, and sharing samples through the sample shuttle path based on the sample shuttle constraints, and outputting an optimal control parameter set. The method further comprises: Based on the punching task information, construct the parameter evaluation function; Taking the parameter evaluation function as a cost function, performing iterative optimization of the near-source sample space and the far-source sample space; When there is a sample that satisfies the sample shuttle constraint, the sample shuttle pathway is activated to share the optimal solution sample; The iterative optimization process is repeated until the global optimal sample meets the punching task information requirements, and the output is the optimal control parameter set.

Images