Method for predicting damage threshold of laser-induced quartz material and related device

Abstract

The invention discloses an electronic device parasitic parameter network analysis method based on point cloud deep learning and a related device, and the method comprises the steps: obtaining physical characteristic data in a laser-induced quartz material process, and the physical characteristic data comprises laser wavelength, pulse width, photon energy and material characteristics; and inputting the physical characteristic data into the trained neural network model, and predicting the damage threshold of the laser-induced quartz material. According to the machine learning method fusing physical information, unification of data efficiency and physical consistency is achieved, prediction precision and model expandability are remarkably improved, and the method is used for accurately predicting the damage threshold value of the quartz material under the specific laser condition, so that the laser processing technology is optimized, and the processing precision and efficiency of the quartz material are improved.

Description

Technical Field

[0001] The present invention relates to the fields of materials science, laser processing and machine learning, and in particular to a method and related device for predicting the damage threshold of laser-induced quartz materials based on physical information and machine learning algorithms. Background Art

[0002] Laser technology, due to its advantages such as high energy density, precise controllability, and non-contact processing, has been widely used in materials processing, optical device manufacturing, and micro-nano manufacturing. During laser processing, the material's damage threshold is a critical parameter that influences both processing results and material properties. It directly determines the minimum energy density at which ablation or damage occurs under specific laser conditions. Accurately measuring and predicting the damage threshold has important scientific and engineering implications for optimizing laser processing techniques, ensuring process safety, and improving material utilization.

[0003] Currently, the main methods for measuring the laser-induced ablation threshold of materials include plasma radiation and scatterometry. The plasma radiation method detects the plasma radiation characteristics generated by a material under laser irradiation to determine the material's ablation onset and offers high sensitivity. The scatterometry method, on the other hand, analyzes changes in the material's surface structure under laser irradiation and uses the light scattering characteristics to determine the ablation threshold. The ablation area extrapolation method measures the ablation threshold by fitting the relationship between the area and power of ablated materials at different femtosecond laser powers. Although these methods can measure the ablation threshold under specific laser conditions, such as wavelength and pulse width, and reveal the influence of laser parameters, the laser ablation threshold has complex relationships with wavelength, pulse width, and material band gap. Establishing a model correlating the laser-induced ablation threshold with laser parameters and material properties through experimental measurement requires extensive equipment and measurement work. Furthermore, revealing the complex relationships between the laser ablation threshold, laser parameters, and material properties is extremely difficult. Therefore, establishing a broadly applicable model of the laser-induced ablation threshold of materials is difficult.

[0004] Physical models based on the microscopic mechanisms of laser-induced ablation provide important theoretical support for threshold prediction. Such models typically involve multiple complex processes in the interaction between laser and material, such as multiphoton absorption, avalanche ionization, and the transfer of plasma energy to the lattice. However, due to the coupling of multiple physical processes, the complexity of nonlinear effects, and the nonideal behavior of materials, purely physical models face significant challenges in estimating the laser-induced damage threshold. In addition, the complexity of laser parameters and their high dependence on input parameters (such as thermophysical properties and optical absorption coefficient) further increase the accuracy and computational complexity of the models. Any deviation in the input parameters will significantly affect the prediction results.

[0005] Machine learning algorithms are powerful and effective methods for building models that correlate input parameters and output quantities using known data. They are particularly well-suited for building models that correlate complex variables and nonlinear pipelines. However, using purely data-driven machine learning algorithms to build threshold models for laser-induced material ablation not only requires massive amounts of experimental data but also often requires expensive or even scarce experimental equipment to cover a wide range of parameters such as wavelength and pulse width. The generalization performance of the resulting models also faces challenges. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and related device for predicting the damage threshold of laser-induced quartz materials. In view of the problems in the existing technology that laser-induced damage threshold modeling relies on massive experiments, some parameters are difficult to obtain, and the modeling complexity caused by the coupling of multiple physical processes, the present invention proposes a machine learning method that integrates physical information, achieves the unity of data efficiency and physical consistency, significantly improves the prediction accuracy and model scalability, and is used to accurately predict the damage threshold of quartz materials under specific laser conditions, thereby optimizing the laser processing technology and improving the processing accuracy and efficiency of quartz materials.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions: A method for predicting the damage threshold of laser-induced quartz material comprises the following steps: Acquiring physical property data during the laser-induced quartz material process, wherein the physical property data includes laser wavelength, pulse width, photon energy, and material properties; The physical property data is fed into the trained neural network model to predict the damage threshold of the laser-induced quartz material. The training process of the neural network model is specifically as follows: Constructing a data set and dividing the data set into a training set, a test set, and a validation set; the data set includes laser wavelength, pulse width, photon energy, material properties, and corresponding damage threshold data; Use the SciANN library to generate a neural network model; Establish physical constraint conditions, embed the physical constraint conditions into the loss function of the neural network model, train the neural network model based on the data set, and obtain a trained neural network model; the loss function includes a data loss part and a physical constraint loss part, wherein the data loss part is used to minimize the error between the predicted value of the neural network model and the experimental data, and the physical constraint loss part is used to ensure that the output result of the neural network model meets the physical constraint conditions.

[0008] Furthermore, the constructing of the data set is specifically as follows: Under preset experimental conditions, the damage threshold data of laser-induced quartz materials at different laser wavelengths and pulse widths were obtained; Analyzing the damage threshold data, determining a dependency relationship between the damage threshold data and the laser wavelength and pulse width, and constructing a corresponding data set based on the dependency relationship; The preset experimental conditions are specifically as follows: the laser power is set in the range of 5 mW to 20 mW, the stable output is monitored by a high-precision power meter, the experimental environment temperature is controlled at 25±1°C, the relative humidity is controlled at 40%±5%, the laser focusing conditions are kept consistent between different experimental groups, the light spot has a Gaussian distribution, the focus position matches the sample surface, the laser is incident on the surface of the laser-induced quartz material sample in a vertical or near-vertical manner to ensure stable transmission and uniform irradiation of the laser energy, and after irradiation is completed, ultrasonic cleaning is performed using anhydrous ethanol for no less than 15 minutes, and the sample is naturally dried after cleaning to ensure that the surface of the laser-induced quartz material sample is clean and in a consistent state.

[0009] Furthermore, the physical constraint condition includes a first physical constraint condition and a second physical constraint condition; The first physical constraint condition is used to constrain the relationship between the laser wavelength and the pulse width, and the first physical constraint condition includes limiting the sum of the second-order derivatives of the laser wavelength and the pulse width to zero or the sum of higher orders to zero; The second physical constraint condition adopts a regional classification labeling method based on physical constraints, which is used to segment the damage threshold according to photon energy and multi-photon absorption order, and divide the data set into different regional labels to enhance the prediction ability of the neural network model for different regions.

[0010] Furthermore, the neural network model adopts a fully connected multi-layer perceptron structure.

[0011] Furthermore, the physical constraint loss part in the loss function uses mean square error or root mean square error to calculate the derivatives of the laser wavelength and pulse width and their deviations from zero to ensure effective embedding of physical constraint conditions.

[0012] Furthermore, when training the neural network model, a gradient descent optimization algorithm, an Adam optimizer, an RMSprop optimizer, or an Adagrad optimizer is used.

[0013] Furthermore, when training the neural network model, the learning rate and training rounds are automatically selected through the hyperparameter tuning method.

[0014] A laser-induced quartz material damage threshold prediction system, comprising: Data acquisition module: used to obtain physical property data of laser-induced quartz material during the process, the physical property data including laser wavelength, pulse width, photon energy and material properties; Prediction module: used to input physical property data into the trained neural network model to predict the damage threshold of laser-induced quartz materials; The training process of the neural network model is specifically as follows: Constructing a data set and dividing the data set into a training set, a test set, and a validation set; the data set includes laser wavelength, pulse width, photon energy, material properties, and corresponding damage threshold data; Use the SciANN library to generate a neural network model; Establish physical constraint conditions, embed the physical constraint conditions into the loss function of the neural network model, train the neural network model based on the data set, and obtain a trained neural network model; the loss function includes a data loss part and a physical constraint loss part, wherein the data loss part is used to minimize the error between the predicted value of the neural network model and the experimental data, and the physical constraint loss part is used to ensure that the output result of the neural network model meets the physical constraint conditions.

[0015] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method for predicting the laser-induced quartz material damage threshold are implemented.

[0016] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for predicting a laser-induced quartz material damage threshold.

[0017] Compared with the prior art, the present invention has the following beneficial technical effects: By combining experimental data with physical prior knowledge and utilizing the nonlinear fitting capabilities of machine learning, the present invention provides a method for predicting the laser-induced destruction threshold of quartz materials. This method not only improves the accuracy of the prediction, but also provides a scientific basis and technical support for optimizing laser processing technology and improving material properties.

[0018] Specifically, by integrating physical information with machine learning algorithms and leveraging the nonlinear fitting capabilities of neural networks, the accuracy of damage threshold predictions has been significantly improved. Through effective data modeling and the embedding of physical constraints, the present invention reduces reliance on large amounts of experimental data, thereby reducing the cost of experimental measurements. The introduction of physical constraints ensures that the predictions of the neural network model on unknown data conform to physical laws, improving the generalization and reliability of the neural network model. Accurate damage threshold prediction provides a scientific basis for optimizing laser processing techniques, helping to improve processing efficiency and material properties. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings in the specification are used to provide further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0020] Figure 1 This is the microstructure of the laser-induced damage threshold hole in an embodiment of the present invention.

[0021] Figure 2 lnPavg and D 2 The straight line fitting result is obtained, and the damage threshold is calculated based on this relationship.

[0022] Figure 3 It is the damage threshold result obtained through multiple measurements under specific experimental conditions.

[0023] Figure 4 Figure 3 shows the damage threshold measurement results and their linear fitting curves corresponding to different pulse widths at a wavelength of 800 nm.

[0024] Figure 5 These are the experimental results of the damage threshold measured under different wavelength conditions.

[0025] Figure 6 A dataset constructed based on experimental results and literature data for training and testing of machine learning.

[0026] Figure 7 This is the performance percentage error result on the test set after the neural network model training is completed.

[0027] Figure 8 This is a flow chart of the method for predicting the laser-induced quartz material damage threshold of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Example 1 See also Figure 8 A method for predicting the damage threshold of laser-induced quartz material comprises the following steps: Acquiring physical property data during the laser-induced quartz material process, wherein the physical property data includes laser wavelength, pulse width, photon energy, and material properties; The physical property data is fed into the trained neural network model to predict the damage threshold of the laser-induced quartz material. The training process of the neural network model is specifically as follows: Under specific experimental conditions, obtaining damage threshold data of laser-induced quartz materials at different laser wavelengths and pulse widths; analyzing the damage threshold data to determine the dependency between the damage threshold data and the laser wavelength and pulse width; and constructing a corresponding data set based on the dependency, dividing the data set into a training set, a test set, and a validation set; the data set includes laser wavelength, pulse width, photon energy, and material properties, as well as the corresponding damage threshold data; The SciANN library is used to generate a neural network model, wherein the neural network model adopts a fully connected multilayer perceptron structure; Establishing physical constraints, the physical constraints including a first physical constraint and a second physical constraint; the first physical constraint constraining the relationship between laser wavelength and pulse width, including limiting the sum of the second-order derivatives of the laser wavelength and pulse width to zero or the sum of higher orders to zero; the second physical constraint employing a physical constraint-based regional classification labeling method to segment the damage threshold according to photon energy and multiphoton absorption order, thereby dividing the data set into different regional labels to enhance the prediction capability of the neural network model for different regions; The physical constraint condition is embedded in the loss function of the neural network model, and the neural network model is trained based on the data set to obtain a trained neural network model; the loss function includes a data loss part and a physical constraint loss part, wherein the data loss part is used to minimize the error between the predicted value of the neural network model and the experimental data, and the physical constraint loss part is used to ensure that the output result of the neural network model meets the physical constraint condition; the physical constraint loss part in the loss function uses the mean square error or the root mean square error to calculate the derivative and the deviation from zero of the laser wavelength and pulse width to ensure the effective embedding of the physical constraint condition, and when training the neural network model, a gradient descent optimization algorithm, an Adam optimizer, an RMSprop optimizer or an Adagrad optimizer is used, and the learning rate and training rounds are automatically selected by a hyperparameter tuning method.

[0031] This invention combines experimental measurements with machine learning algorithms based on physical information. Machine learning algorithms possess powerful nonlinear modeling capabilities and can extract underlying patterns and regularities from extensive experimental data, revealing the complex relationships between damage thresholds and various factors (such as laser parameters and material properties). Especially when dealing with highly complex systems with multiple interactions, machine learning can effectively capture subtle variations that are difficult to describe with traditional physical models. Machine learning models trained solely on experimental data can face data scarcity and overfitting challenges, especially with small sample sizes. Therefore, incorporating physical information as constraints or prior knowledge within the model not only effectively reduces the difficulty of model training but also significantly improves prediction accuracy. This approach, by incorporating physical mechanisms (such as multiphoton absorption and plasma energy transfer) into the machine learning model, enables the model to more accurately reflect actual physical processes and provide results with greater physical explanatory power. The inclusion of physical information provides a rational framework for machine learning algorithms, helping to improve the model's reliability and generalization capabilities. This is particularly important in practical applications where uncertain experimental conditions or extreme cases are encountered, effectively avoiding errors caused by overreliance on experimental data. Establishing a laser-induced material damage threshold prediction model based on experimental measurements, integrating physical information and machine learning algorithms can not only significantly improve the prediction accuracy and generalization ability of the damage threshold, but also provide important theoretical support and technical guarantee for the optimization of laser processing technology and the improvement of material properties.

[0032] Example 2 A method for predicting the damage threshold of laser-induced quartz material comprises the following steps: 1. Data acquisition and dependency exploration: Obtain the damage threshold of the laser-induced quartz material to be predicted under different laser wavelengths and pulse widths.

[0033] Based on the laser wavelengths at a plurality of preset moments, a response speed curve of the laser-induced quartz material is obtained.

[0034] Analysis of the experimental data revealed a certain dependence between the destruction threshold and the laser wavelength and pulse width. Specifically, the destruction threshold showed an approximately linear growth trend with the increase of pulse width. At the same time, the relationship between laser wavelength and pulse width also gradually became clear, indicating that the two have a certain correlation in affecting the destruction threshold. In particular, the destruction threshold showed an approximately linear growth trend with the pulse width, and the relationship between laser wavelength and pulse width gradually became clear.

[0035] 2. Modeling the relationship between damage threshold, laser wavelength, and pulse width: Combined with the theoretical formulas in the relevant literature, the quantitative relationship between the destruction threshold and the laser wavelength and pulse width is derived. Regarding the theoretical formulas in the relevant literature, the explanation is as follows: the relationship between the destruction threshold and the laser wavelength and pulse width based on the present invention is derived from the physical modeling of the laser-induced ionization process, including the theoretical derivation of typical mechanisms such as avalanche ionization, multiphoton ionization and tunnel ionization. Under the condition dominated by multiphoton ionization, the destruction threshold is distributed in a step-by-step manner with the laser wavelength. The longer the wavelength, the more photons need to be absorbed, which leads to an increase in the destruction threshold; while in the same order interval, the destruction threshold changes slowly with the wavelength and is approximately linear. At the same time, according to the thermal diffusion control model, the destruction threshold of the material shows a logarithmic linear relationship with the laser pulse width. The longer the pulse width, the higher the destruction threshold. These relationships derived based on physical processes provide theoretical support for the physical constraint model constructed in the present invention, which is used to guide the design and training of the destruction threshold prediction model.

[0036] Through mathematical modeling, it was found that this quantitative relationship can be accurately described by a linear or power function, and presents a stepped segmented distribution determined by the order of multiphoton absorption.

[0037] In particular, as the pulse width increases, the destruction threshold shows a clear correlation with the laser wavelength and pulse width, and under certain photon energy conditions, the change of the destruction threshold is approximately linear.

[0038] 3. Analysis of the impact of photon energy on the damage threshold: Further analysis of the effect of photon energy on the destruction threshold shows that as the photon energy decreases, the destruction threshold gradually increases; and within the same multi-photon absorption order range, as the photon energy increases, the destruction threshold changes little and changes approximately linearly.

[0039] A further improvement of the present invention is: The first physical constraint is the balance between laser wavelength and pulse width: Based on the theoretical equations of multiphoton ionization and avalanche ionization, which describe the ionization process in materials under laser irradiation, the quantitative relationship between the destruction threshold and laser parameters is revealed.

[0040] The logarithm of the material damage threshold and the logarithm of the pulse width exhibit an approximately linear or exponential relationship within a pulse width of 1 picosecond. This relationship provides a theoretical basis for constraining higher-order derivatives, and the linear relationship is verified by experimental data.

[0041] By constraining the high-order derivatives of laser wavelength and pulse width, we can ensure that the output of the neural network model is not only accurate in data fitting, but also maintains reasonable smoothness physically, avoiding physically unreasonable fluctuations, thereby improving the model's generalization ability and prediction accuracy.

[0042] A further improvement of the present invention is that the second physical constraint condition - the regional classification labeling method: In order to further improve the accuracy and physical interpretability of the prediction model, a regional classification labeling method based on physical constraints was introduced.

[0043] By using the destruction threshold data of femtosecond laser ablation of quartz at different wavelengths, an unsupervised learning clustering algorithm is applied to segment the destruction threshold data according to photon energy and multiphoton absorption order. The clustering algorithm is then used to divide the data set into different regional labels to enhance the prediction ability of the neural network model for different working areas.

[0044] This method uses threshold features to classify data, thereby integrating physical information. This physically constrained machine learning classification algorithm, through input parameters such as laser wavelength, damage threshold, and material band gap, can rationally partition and classify data under different physical conditions, providing physically meaningful labels for the neural network model, further improving the accuracy and interpretability of predictions.

[0045] A further improvement of the present invention lies in the generation of a dataset and the construction of a neural network model. Based on the aforementioned experimental data and established physical relationships, a dataset is constructed for training and validating the neural network model. The dataset includes input features such as laser wavelength, pulse width, photon energy, and material properties (such as refractive index, thermal conductivity, and band gap), along with corresponding damage threshold results.

[0046] The SciANN library is used to generate a neural network model with a fully connected multilayer perceptron (MLP) structure. The neural network model contains at least three hidden layers, each layer contains at least ten neurons, and the number of hidden layers and the number of neurons in each layer can be adaptively set according to specific needs.

[0047] A further improvement of the present invention is that the physical constraints are embedded in the loss function of the neural network model. During the construction of the neural network model, the physical constraints are embedded in the loss function of the neural network model as part of the loss function, which goes beyond the original loss design of minimizing data errors.

[0048] The loss function includes a data loss part, which is used to minimize the error between the predicted value of the neural network model and the experimental data, and a physical constraint loss part, which is used to ensure that the output result of the neural network model meets the physical constraint conditions.

[0049] The physical constraint loss part uses mean square error (MSE) or root mean square error (RMSE) to calculate the derivatives of laser wavelength and pulse width and their deviation from zero to ensure the effective embedding of physical constraints.

[0050] A further improvement of the present invention is that the generated data set is used to train the constructed neural network model, and a gradient descent method or other optimization algorithms are used to minimize the loss function.

[0051] During the training process, the weight parameters of the neural network model are continuously updated iteratively, so that the neural network model gradually converges and can accurately predict the damage threshold.

[0052] After the training is completed, the final neural network model and its weights are saved and a model file is generated for subsequent application and deployment.

[0053] A further improvement of the present invention is to test the trained neural network model using a test set, and verify its generalization ability and physical consistency by evaluating the prediction accuracy of the neural network model on the test set.

[0054] The test results include the prediction error of the neural network model under different photon energies, laser wavelengths and pulse widths, ensuring that the neural network model not only fits the training data well, but also provides high-precision predictions on unseen data.

[0055] Example 3 Completed the destruction threshold measurement of different pulse widths under 800 nm and 1030 nm lasers, and completed the destruction threshold measurement at different laser wavelengths using OPA (optical parametric amplifier); Figure 1 This is a microstructure image observed under a microscope after the laser-induced quartz material damage threshold experiment. The image shows regularly arranged circular damage points, each of which corresponds to a laser irradiation position. The damage threshold is measured using an indirect measurement method - the damage diameter extrapolation method. By measuring the size of the damage point under different laser powers, lnPavg and D are fitted. 2 The relationship between the two results is 2ω0. 2About lnPavg and D 2 When D = 0, the average Pavg is obtained, and the damage threshold φth is calculated based on the average power. Figure 2 The relationship between the area of ​​the laser-induced damage spot and the logarithm of the average laser power is demonstrated. The experimental data shows a clear linear growth trend, consistent with theoretical expectations. The slope of the fitted curve can be used to estimate the size parameters of the laser spot, and the intersection of the curve with the horizontal axis can be further inferred to determine the laser damage threshold, providing basic physical parameters for subsequent modeling and analysis. Figure 3 The laser damage threshold results obtained from five repeated measurements under the same experimental conditions are shown. The measured values ​​range from 3.2 to 3.5 J / cm 2 The fluctuation range is small, which shows that the experiment has good repeatability and result stability. Figure 4 The effect of laser pulse width on the material ablation threshold is demonstrated. The results show that as the pulse width increases, the destruction threshold shows a clear upward trend, which is consistent with the characteristics of laser-material thermal interaction. Figure 5 The figure shows the changing trend of the material's damage threshold under the action of lasers of different wavelengths. As can be seen from the figure, the material's damage threshold increases with increasing wavelength, especially after 680nm, indicating that the material's response sensitivity to long-wavelength lasers decreases. Due to the greater energy penetration depth of long-wavelength lasers, the energy is more dispersed in the material, making it difficult to form a sufficiently high energy density on the surface, thereby increasing the damage threshold required for ablation.

[0056] In an embodiment of the present invention, the damage threshold measurement of different pulse widths under 800 nm and 1030 nm lasers is completed, specifically including: Two laser wavelengths, 800 nm and 1030 nm, were selected, and different pulse widths (exemplary range was 0.1 ps to 1 ps) were set for each.

[0057] An optical parametric amplifier (OPA) was used to adjust the laser output at different wavelengths.

[0058] The indirect measurement method, the damage diameter extrapolation method, is used to measure the diameter D of the damage point under different laser powers and record the corresponding average power P_avg.

[0059] Fit the measured data and plot lnP_avg vs. D 2 The relationship diagram, and the slope is 2ω0 obtained by linear regression 2 straight line.

[0060] When D=0, the calculated average P_avg is the damage threshold ϕ_th.

[0061] Subsequently, the collected dataset is organized into input-output pairs. The input data includes laser wavelength, pulse width, photon energy, and material properties, and the output data is the corresponding damage threshold. The dataset is divided into training, test, and validation sets. In this embodiment, the original dataset is divided into training, validation, and test sets at a ratio of 70%, 15%, and 15%, respectively, for model training, parameter tuning, and final performance evaluation. A neural network model with a fully connected multilayer perceptron (MLP) structure is constructed using the SciANN library. The number of hidden layers and neurons is adaptively set according to specific needs. In this embodiment, the fully connected multilayer perceptron (MLP) model constructed based on the SciANN library contains four hidden layers, with 128 neurons per layer, and a ReLU activation function. Figure 6 What is shown is a two-dimensional data set constructed based on experimental data and theoretical models, which is used for subsequent training of neural network models.

[0062] During neural network model training, a gradient descent optimization algorithm is used, along with a set learning rate and number of training rounds. Hyperparameter tuning is used to optimize neural network model performance. The loss function includes both a data loss component and a physical constraint loss component. The physical constraint loss component uses the mean square error (MSE) or root mean square error (RMSE) to calculate the derivatives of the laser wavelength and pulse width, and their deviations from zero. After multiple training iterations, the neural network model gradually converges, ultimately achieving a trained neural network model.

[0063] The trained neural network model was validated using a test set to evaluate its prediction accuracy under different photon energies, laser wavelengths, and pulse widths. The results showed that the neural network model significantly reduced its prediction error on unknown data, and its output was consistent with physical laws, demonstrating the generalization ability and physical consistency of the neural network model. Figure 7 The prediction error distribution of the neural network model for the test set under different wavelength and pulse width combinations is shown. Overall, the error distribution is within ±10%, with most areas within ±5%, indicating that the constructed neural network model has good generalization ability and prediction accuracy.

[0064] Based on the trained neural network model, new laser parameters and material properties are input to accurately predict the damage threshold of quartz materials, providing a scientific basis and technical support for the optimization of laser processing technology.

[0065] By integrating physical information with machine learning algorithms, this paper constructs an efficient and accurate prediction model for the laser-induced damage threshold of quartz materials. This model not only improves prediction accuracy but also provides a scientific basis and technical support for optimizing laser processing techniques and improving material properties. It has broad application prospects and significant technical advantages.

[0066] Example 4 A laser-induced quartz material damage threshold prediction system, comprising: Data acquisition module: used to obtain physical property data of laser-induced quartz material during the process, the physical property data including laser wavelength, pulse width, photon energy and material properties; Prediction module: used to input physical property data into the trained neural network model to predict the damage threshold of laser-induced quartz materials; The training process of the neural network model is specifically as follows: Constructing a data set and dividing the data set into a training set, a test set, and a validation set; the data set includes laser wavelength, pulse width, photon energy, material properties, and corresponding damage threshold data; Use the SciANN library to generate a neural network model; Establish physical constraint conditions, embed the physical constraint conditions into the loss function of the neural network model, train the neural network model based on the data set, and obtain a trained neural network model; the loss function includes a data loss part and a physical constraint loss part, wherein the data loss part is used to minimize the error between the predicted value of the neural network model and the experimental data, and the physical constraint loss part is used to ensure that the output result of the neural network model meets the physical constraint conditions.

[0067] Example 5 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method for predicting the laser-induced quartz material damage threshold are implemented.

[0068] Example 6 This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the steps of the method for predicting the laser-induced quartz material damage threshold are implemented.

[0069] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0070] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0071] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0072] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit its scope of protection. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art may still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the scope of protection of the pending claims of the invention.

Claims

1. A method for predicting the damage threshold of laser-induced quartz material, characterized in that: The steps include: Acquiring physical property data during the laser-induced quartz material process, wherein the physical property data includes laser wavelength, pulse width, photon energy, and material properties; The physical property data is fed into the trained neural network model to predict the damage threshold of the laser-induced quartz material. The training process of the neural network model is specifically as follows: Constructing a data set and dividing the data set into a training set, a test set, and a validation set; the data set includes laser wavelength, pulse width, photon energy, material properties, and corresponding damage threshold data; Use the SciANN library to generate a neural network model; Establish physical constraint conditions, embed the physical constraint conditions into the loss function of the neural network model, train the neural network model based on the data set, and obtain a trained neural network model; the loss function includes a data loss part and a physical constraint loss part, wherein the data loss part is used to minimize the error between the predicted value of the neural network model and the experimental data, and the physical constraint loss part is used to ensure that the output result of the neural network model meets the physical constraint conditions.

2. The method for predicting the damage threshold of laser-induced quartz material according to claim 1, characterized in that: The construction of the data set is specifically as follows: Under preset experimental conditions, the damage threshold data of laser-induced quartz materials at different laser wavelengths and pulse widths were obtained; Analyzing the damage threshold data, determining a dependency relationship between the damage threshold data and the laser wavelength and pulse width, and constructing a corresponding data set based on the dependency relationship; The preset experimental conditions are specifically as follows: the laser power is set in the range of 5 mW to 20 mW, the stable output is monitored by a high-precision power meter, the experimental environment temperature is controlled at 25±1°C, the relative humidity is controlled at 40%±5%, the laser focusing conditions are kept consistent between different experimental groups, the light spot has a Gaussian distribution, the focus position matches the sample surface, the laser is incident on the surface of the laser-induced quartz material sample in a vertical or near-vertical manner to ensure stable transmission and uniform irradiation of the laser energy, and after irradiation is completed, ultrasonic cleaning is performed using anhydrous ethanol for no less than 15 minutes, and the sample is naturally dried after cleaning to ensure that the surface of the laser-induced quartz material sample is clean and in a consistent state.

3. The method for predicting the damage threshold of laser-induced quartz material according to claim 1, characterized in that: The physical constraint condition includes a first physical constraint condition and a second physical constraint condition; The first physical constraint condition is used to constrain the relationship between the laser wavelength and the pulse width, and the first physical constraint condition includes limiting the sum of the second-order derivatives of the laser wavelength and the pulse width to zero or the sum of higher orders to zero; The second physical constraint condition adopts a regional classification labeling method based on physical constraints, which is used to segment the damage threshold according to photon energy and multi-photon absorption order, and divide the data set into different regional labels to enhance the prediction ability of the neural network model for different regions.

4. The method for predicting the damage threshold of laser-induced quartz material according to claim 1, characterized in that: The neural network model adopts a fully connected multi-layer perceptron structure.

5. The method for predicting the damage threshold of laser-induced quartz material according to claim 1, characterized in that: The physical constraint loss part in the loss function uses mean square error or root mean square error to calculate the derivatives of the laser wavelength and pulse width and their deviations from zero to ensure the effective embedding of physical constraints.

6. The method for predicting the damage threshold of laser-induced quartz material according to claim 1, characterized in that: When training the neural network model, use the gradient descent optimization algorithm, Adam optimizer, RMSprop optimizer, or Adagrad optimizer.

7. The method for predicting the laser-induced damage threshold of quartz material according to claim 6, characterized in that: When training a neural network model, the learning rate and number of training epochs are automatically selected through hyperparameter tuning methods.

8. A laser-induced quartz material damage threshold prediction system, characterized in that: include: Data acquisition module: used to obtain physical property data of laser-induced quartz material during the process, the physical property data including laser wavelength, pulse width, photon energy and material properties; Prediction module: used to input physical property data into the trained neural network model to predict the damage threshold of laser-induced quartz materials; The training process of the neural network model is specifically as follows: Constructing a data set and dividing the data set into a training set, a test set, and a validation set; the data set includes laser wavelength, pulse width, photon energy, material properties, and corresponding damage threshold data; Use the SciANN library to generate a neural network model; Establish physical constraint conditions, embed the physical constraint conditions into the loss function of the neural network model, train the neural network model based on the data set, and obtain a trained neural network model; the loss function includes a data loss part and a physical constraint loss part, wherein the data loss part is used to minimize the error between the predicted value of the neural network model and the experimental data, and the physical constraint loss part is used to ensure that the output result of the neural network model meets the physical constraint conditions.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method for predicting the laser-induced quartz material damage threshold as claimed in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting the laser-induced quartz material damage threshold as claimed in any one of claims 1 to 7 are implemented.

Images