Damage time prediction system of laser irradiation metal material

Abstract

The invention relates to the technical field of engineering optics, and discloses a damage time prediction system for a laser irradiation metal material. The system comprises a dynamic temperature distribution acquisition module, a feature vector extraction module, a parameter association network construction module, a damage time prediction module and a learning optimization prediction model module. Compared with static heat conduction modeling and single parameter analysis in the prior art, the technical problem that high-precision damage time prediction cannot be achieved under the complex conditions that dynamic laser input parameters fluctuate greatly and metal materials are microcosmically non-uniform is solved. According to the method, high-precision prediction of the dynamic damage time of the metal material is realized through dynamic nonlinear heat conduction modeling, multi-modal feature fusion, knowledge graph correlation analysis and reinforcement learning optimization, so that the problems that the prediction error is large and parameters are not adaptive to dynamic changes are avoided, and the accuracy and applicability of laser irradiation damage time prediction are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of engineering optics, and in particular relates to a damage time prediction system for laser irradiated metal materials. Background Art

[0002] At present, there are many deficiencies in the damage time prediction technology of laser irradiated metal materials. For example, the existing technology mainly relies on static heat conduction models and single laser parameter inputs, and fails to fully consider complex factors such as the dynamic fluctuations of laser input energy, the heterogeneity of the microstructure of metal materials, and the phase change heat dissipation during the melting process, resulting in low damage time prediction accuracy. In addition, when faced with external disturbances such as dynamic laser power density and spot shape changes, the existing methods lack adaptive capabilities and are difficult to accurately reflect the dynamic evolution of the temperature field. For example, under high-power density laser irradiation conditions, the temperature gradient on the surface of metal materials changes dramatically, and the temperature distribution has highly nonlinear characteristics. Traditional methods often ignore key parameters such as temperature-dependent thermal conductivity and reflectivity changes, and cannot achieve high-precision modeling. At the same time, existing prediction methods are usually based on fixed parameters and static formulas, and fail to combine real-time feedback mechanisms for dynamic correction, resulting in large errors in practical applications, making it difficult to meet the needs of high-precision damage time prediction. Therefore, there is an urgent need for a method that can achieve high-precision prediction of damage time under dynamic laser irradiation conditions, combining the nonlinear thermal conduction characteristics of metal materials and the dynamic temperature field distribution, so as to improve the adaptability and prediction accuracy of the laser irradiation system under complex working conditions. Summary of the invention

[0003] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to propose a damage time prediction system for laser irradiated metal materials, aiming to solve the technical problem that static heat conduction modeling and single parameter analysis in the prior art cannot achieve high-precision damage time prediction, especially under complex conditions such as large fluctuations in dynamic laser input parameters and microscopic non-uniformity of metal materials.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solution: The present invention provides a damage time prediction system for laser irradiated metal materials, comprising:

[0005] Dynamic temperature distribution acquisition module, used to collect laser power density P(t), spot radius r(t), metal material density ρ m Specific heat capacity C p (T m ), thermal conductivity k m (T m );

[0006] Construct a dynamic nonlinear heat conduction equation and solve it to obtain the dynamic temperature distribution T m(x, y, t), where the formula for the dynamic nonlinear heat conduction equation is:

[0007]

[0008] in, is the gradient of the temperature field, indicating the direction and rate of temperature change with spatial position; Q laser (x,y,t) is the laser input heat energy, in W / m 2 ;Q ref (x,y,t) is the laser reflection energy loss, unit is W / m 2 ;Q phase (x, y, t) is the latent heat loss of phase change, in W / m 2 , x and y are the two-dimensional spatial coordinates on the surface of the metal material, t is the time variable, and m represents the material mass point;

[0009] Feature vector extraction module is used to extract the dynamic temperature distribution T m (x, y, t) and laser power density P(t) and spot radius r(t) fuse multimodal features and generate a dynamic fusion feature vector F fusiom , the fusion formula is:

[0010] F fusion =α P ·P(t)+α r ·r(t)+α T ·T amb (t)+α m ∫ A T m (x,y,t)dxdy

[0011] Among them, α P ,α r ,α T ,α m is the feature weight coefficient; ∫ A T m (x,y,t)dxdy is the integral of the temperature field in the irradiated area A, in K·m 2 ; T amb )t_ is the ambient temperature, which is collected in real time by the ambient temperature sensor, and the unit is K;

[0012] The dynamic fusion feature vector F fusiom As input for subsequent knowledge graph construction;

[0013] The parameter association network construction module is used to build a dynamic association network based on the knowledge graph. The input data of the dynamic association network includes the dynamic fusion feature vector F fusiomThe associated node data of the dynamic associated network include laser power density P(t), spot radius r(t), specific heat capacity C p (T m ), thermal conductivity k m (T m ); The dynamic association network outputs the optimized dynamic association weight feature vector Q;

[0014] The damage time prediction module is used to construct a damage time prediction model to obtain the predicted damage time T d ;

[0015] The input data of the damage time prediction model includes the dynamic association weight feature vector Q and the dynamic fusion feature vector F after knowledge graph optimization. fusiom ;

[0016] The formula of the damage time prediction model is:

[0017] T d =W1·F fusion +W2·Q+b

[0018] Among them, W1 is the dynamic fusion feature weight coefficient, W2 is the association weight feature weight coefficient, and b is the bias term of the damage time prediction model;

[0019] Learning optimization prediction model module is used to predict the damage time T d and the damage time T actually measured by the sensor real Set the reward function R for reinforcement learning:

[0020] R=-|T real -T d |

[0021] The weight coefficients W1 and W2 in the damage time prediction model are optimized according to the reward function R combined with the PPO algorithm.

[0022] Preferably, the dynamic temperature distribution acquisition module includes: laser power density P(t), which is collected in real time by a laser power sensor in units of W / cm 2 , used to describe the power of the laser per unit area; the spot radius r(t), measured by the built-in optical monitoring sensor of the laser, in cm; the density of the metal material ρ m , obtained through laboratory measurement or reference material technical parameters, unit is kg / m 3 Specific heat capacity C p (T m ), obtained through the material thermal property testing device, the unit is J / (kg·K); thermal conductivity k m (T m), measured and fitted by experiment, unit is W / (m·K); temperature distribution T m (x, y, t), collected in real time using an infrared thermal imager, in K, representing the dynamic temperature distribution on the material surface.

[0023] Preferably, the dynamic temperature distribution acquisition module includes:

[0024] Q laser (x,y,t) is the laser input heat energy, in W / m 2 , the calculation formula is:

[0025]

[0026] Among them, η abs is the laser energy absorption rate, σ is the spot diffusion parameter, and exp is the exponential function.

[0027] Preferably, in the parameter association network construction module, the dynamic association relationship between the association nodes of the dynamic association network is described by the following formula:

[0028] G=(N,E),E ij =f(correlation strength(N i ,N j ))

[0029] Where N is the set of associated nodes in the dynamic associated network; E ij is the dynamic association weight between the associated nodes, optimized by the graph neural network, G is the dynamic association network graph, N i is the ith associated node of the dynamic associated network, N j is the j-th associated node of the dynamic associated network, and f is the correlation strength function, which is used to quantify the association strength between the ith associated node and the j-th associated node.

[0030] Preferably, in the damage time prediction module, the training input data used in the training process of the damage time prediction model is the standard historical dynamic association weight feature vector Q and the dynamic fusion feature vector F fusiom ,The training output data is the actual damage time measured by historical experiments.

[0031] Preferably, in the dynamic temperature distribution acquisition module, Q ref (x,y,t(is the laser reflection energy loss, unit is W / m 2 , the calculation formula is:

[0032] Q ref (x,y,t(=R m (T m )·exp(-γ r ·T m)

[0033] Among them, R m (T m (is the reflectivity, γ r is the coefficient of reflectivity changing with temperature;

[0034] Q phase (x,y,t(is the latent heat loss of phase change, unit is W / m 2 , the calculation formula is:

[0035]

[0036] Among them, L m is the latent heat of fusion, which is measured by a preset scanning calorimeter; T solid , T liquid are the solid phase and liquid phase temperatures, respectively, which are directly obtained from the preset material technical standard table, f m To indicate the melting ratio of metal materials.

[0037] Preferably, in the learning optimization prediction model module, the weight coefficients W1 and W2 in the damage time prediction model are optimized, and the formula used is:

[0038]

[0039] Among them, W new ,W αld are the weight coefficients after optimization and before optimization respectively; α is the learning rate, which is used to control the step size of weight adjustment; is the gradient of the reward function R.

[0040] The beneficial effect of the present invention is that compared with the static heat conduction modeling and single parameter analysis in the prior art, especially under the complex conditions of large fluctuations in dynamic laser input parameters and microscopic non-uniformity of metal materials, the technical problem of being unable to achieve high-precision damage time prediction, this application achieves high-precision prediction of dynamic damage time of metal materials through dynamic nonlinear heat conduction modeling, multimodal feature fusion, knowledge graph association analysis and reinforcement learning optimization, thereby avoiding the problems of large prediction errors and parameters not adapting to dynamic changes, and improving the accuracy and applicability of laser irradiation damage time prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 A schematic flow chart of a first embodiment of a damage time prediction system for laser irradiated metal materials provided by the present invention.

[0043] Figure 2 A schematic diagram of a device for predicting damage time of laser-irradiated metal materials provided by the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments 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 creative work are within the scope of protection of the present invention.

[0045] Embodiment 1: Figure 1 , which is a flow chart of the first embodiment of the damage time prediction system for laser irradiated metal materials of the present invention, and provides the first embodiment of the damage time prediction system for laser irradiated metal materials of the present invention.

[0046] In a first embodiment, the damage time prediction system for laser irradiated metal materials includes:

[0047] Dynamic temperature distribution acquisition module, used to collect laser power density P(t), spot radius r(t(, metal material density ρ m Specific heat capacity C p (T m (、Thermal conductivity k m (T m (;

[0048] Construct a dynamic nonlinear heat conduction equation and solve it to obtain the dynamic temperature distribution T m (x, y, t), where the formula for the dynamic nonlinear heat conduction equation is:

[0049]

[0050] in, is the gradient of the temperature field, indicating the direction and rate of temperature change with spatial position; Q laser (x,y,t) is the laser input heat energy, in W / m 2 ;Q ref (x,y,t) is the laser reflection energy loss, unit is W / m 2 ;Q phase (x, y, t) is the latent heat loss of phase change, in W / m 2, x and y are the two-dimensional spatial coordinates on the surface of the metal material, t is the time variable, and m represents the material mass point;

[0051] It should be noted that the dynamic nonlinear heat conduction equation in the present invention is established based on the real-time correlation between the temperature of the metal material and the physical parameters, including the laser input energy Q laser , reflected heat dissipation Q ref and latent heat of phase change Q phase , which not only takes into account the change of temperature with time The temperature field gradient is also integrated The driving force for the heat conduction process, for example, when the laser power density P(t) increases, the temperature gradient in the central area increases rapidly, resulting in an accelerated rate of thermal energy diffusion to the surrounding area.

[0052] It should be noted that Q phase (x, y, t) is the latent heat loss of phase change, in W / m 2 , the calculation formula is:

[0053]

[0054] Among them, L m is the latent heat of fusion, which is measured by a preset scanning calorimeter; T solid , T liquid are the solid phase and liquid phase temperatures, respectively, which are directly obtained from the preset material technical standard table, f m To indicate the melting ratio of metal materials.

[0055] It should be understood that under high-power laser irradiation conditions, the melting process of metal materials is nonlinear, and the melting ratio f m The latent heat of phase change Q varies with the dynamic change of temperature. phase This is manifested as significant energy dissipation in the temperature distribution, especially at the solid phase temperature T solid and liquidus temperature T liquid The energy requirements for metal melting must be accurately modeled. For example, for 7075 aluminum alloy, when the laser irradiation causes the material temperature to reach the range of 750K to 890K, the latent heat of phase transformation L m About 3.97×10 5 J / kg, at this time, part of the input laser energy will be used to overcome the energy required for material melting, thus affecting the overall temperature distribution.

[0056] Feature vector extraction module is used to extract the dynamic temperature distribution T m (x, y, t) and laser power density P(t) and spot radius r(t) fuse multimodal features and generate a dynamic fusion feature vector F fusiom , the fusion formula is:

[0057] F fusion =α P ·P(t)+α r ·r(t)+α T ·T amb (t)+α m ∫ A T m (x,y,t)dxdy

[0058] Among them, α P ,α r ,α T ,α m is the feature weight coefficient; ∫ A T m (x,y,t)dxdy is the integral of the temperature field in the irradiated area A, in K·m 2 ; T amb (t) is the ambient temperature, which is collected in real time by the ambient temperature sensor, in K;

[0059] The dynamic fusion feature vector F fusiom As input for subsequent knowledge graph construction;

[0060] It should be noted that the dynamic fusion feature vector F fusion By weighted summing of multi-modal features, the relationship between laser input energy and material thermal conduction is fully reflected; the temperature field integral term ∫ A T m (x,y,t)dxdy describes the overall level of temperature distribution in the laser irradiation area, and can quantify the thermal effect of the laser input energy diffusing on the material surface. Unlike a single temperature peak, this item can more accurately describe the uniformity and spatial differences of temperature distribution through regional integration.

[0061] It should be understood that the ambient temperature T amb (t) has an impact on the temperature distribution. In high or low temperature environments, the thermal conductivity of the material will change, thus affecting the prediction accuracy of the damage time.

[0062] For example, at a room temperature of 300K, the laser power P(t) = 100 W / cm 2 When the spot radius r(t) = 1 cm, the temperature field integral ∫ A T m (x,y,t)dxdy is 5.2×10 3 K·m 2 , at this time, the dynamic fusion feature vector F fusion The calculation result is:

[0063] Ffusion =α P 100+α r 1+α T 300+α m 5200

[0064] By adjusting the weight coefficient α P ,α r ,α T ,α m , which can make F fusion Optimally reflects the influence of each feature on the damage time.

[0065] The parameter association network construction module is used to build a dynamic association network based on the knowledge graph. The input data of the dynamic association network includes the dynamic fusion feature vector F fusiom The associated node data of the dynamic associated network include laser power density P(t), spot radius r(t), specific heat capacity C p (T m ), thermal conductivity k m (T m ); The dynamic association network outputs the optimized dynamic association weight feature vector Q;

[0066] It should be noted that the parameter association network constructs a dynamic association relationship between nodes through the knowledge graph, so that the influence relationship between multiple parameters can be intuitively represented, and the weights between different nodes will be dynamically adjusted according to their contribution to the damage time prediction.

[0067] It can be understood that the weight feature vector of the dynamic association network is obtained through the data relationship and weight optimization between nodes, which can reflect the influence of each feature on the prediction result. Compared with the traditional static weight allocation method, the dynamic association network can adaptively adjust the weight distribution when the input data changes.

[0068] It should be understood that the node data do not exist independently, but are interrelated systems that jointly affect the temperature distribution and damage behavior of metal materials. For example, an increase in laser power density P(t) will cause the temperature distribution to rise rapidly, thereby affecting the thermal conductivity k. m (T m ) and thermal diffusion processes.

[0069] For example, when the laser power density P(t) = 150 W / cm 2 , spot radius r(t) = 0.5 cm, specific heat capacity of the material C p (T m )=950J / (kg·K), thermal conductivity k m (T m)=140W / (m·K), the dynamic association network inputs these node data and outputs the dynamic association weight eigenvector Q=[0.35,0.25,0.20,0.20], among which the laser power density P(t) has the largest weight, indicating that it has the most significant impact on the temperature field distribution and the final damage time.

[0070] Through the above steps, the dynamic association network not only optimizes the association weights between nodes, but also provides more accurate input data for the subsequent damage time prediction model, thereby improving the reliability and accuracy of the prediction results.

[0071] The damage time prediction module is used to construct a damage time prediction model to obtain the predicted damage time T d ;

[0072] The input data of the damage time prediction model includes the dynamic association weight feature vector Q and the dynamic fusion feature vector F after knowledge graph optimization. fusiom ;

[0073] The formula of the damage time prediction model is:

[0074] T d =W1·F fusion +W2·Q+b

[0075] Among them, W1 is the dynamic fusion feature weight coefficient, W2 is the association weight feature weight coefficient, and b is the bias term of the damage time prediction model;

[0076] It should be noted that the model can comprehensively consider the laser input energy, ambient temperature, temperature distribution and the dynamic correlation between the features by weighted summing the multimodal fusion features and the associated features optimized by the knowledge graph, ensuring more accurate prediction results; the dynamic fusion feature vector mainly reflects the overall influence of the laser input parameters and the temperature field distribution, while the dynamic association weight vector further optimizes the association weights between the features, ensuring that the influence of each feature on the prediction result is dynamically adaptive.

[0077] Learning optimization prediction model module is used to predict the damage time T d and the damage time T actually measured by the sensor real Set the reward function R for reinforcement learning:

[0078] R=-|T real -T d |

[0079] The weight coefficients W1 and W2 in the damage time prediction model are optimized according to the reward function R combined with the PPO algorithm.

[0080] It should be noted that the reward function enables the model to gradually learn the optimal mapping relationship between input features and output results by minimizing the prediction error, thereby improving the accuracy of the prediction.

[0081] It can be understood that when the predicted damage time is closer to the actual damage time, the reward value R is larger, and the weight update direction of the model is more conducive to reducing the error. This mechanism ensures that the model is continuously optimized towards the optimal solution during the training process.

[0082] It should be understood that the PPO algorithm makes the model training process more stable by limiting the amplitude of each weight update to avoid overfitting or weight update oscillation, while improving the accuracy of damage time prediction. For example, in a training, the actual damage time T measured by the sensor is real = 3050s, initial output of the prediction model Reward function R = -|3050-3100| = -50; combined with the PPO algorithm, the model calculates the reward gradient and adjusts the weight parameters W1 and W2, ultimately making the next round of prediction results closer to the actual value. For example, Optimized to 3052s. Through the above optimization steps, this module effectively reduces the damage time prediction error and improves the adaptability and prediction accuracy of the prediction model under complex laser irradiation conditions.

[0083] In addition, the present invention also provides a laser irradiated metal material damage time prediction device, please refer to Figure 2, a damage time prediction device for laser irradiated metal materials includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the damage time prediction system for laser irradiated metal materials in the above-mentioned embodiment 1. A damage time prediction device for laser irradiated metal materials in the embodiment of the present invention may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. A damage time prediction device for laser irradiated metal materials is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention. A damage time prediction device for laser irradiation of metal materials may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM) 1004. Various programs and data required for the operation of a damage time prediction device for laser irradiation of metal materials are also stored in RAM 1004. The processing device 1001, ROM 1002, and RAM 1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow a damage time prediction device for laser irradiated metal materials to communicate wirelessly or wired with other devices to exchange data. Although a damage time prediction device for laser irradiated metal materials having various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0084] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements a damage time prediction system for laser irradiated metal materials as described above. The computer program product provided by the present invention can solve the technical problem of damage time prediction for laser irradiated metal materials. Compared with the prior art, the beneficial effects of the computer program product provided by the present invention are the same as the beneficial effects of the damage time prediction system for laser irradiated metal materials provided by the above embodiment, which will not be described in detail here.

[0085] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the system of the embodiments disclosed in the present invention are executed.

[0086] It should be understood that the various parts disclosed in the present invention can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0087] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A damage time prediction system for laser irradiated metal materials, characterized in that the system include: Dynamic temperature distribution acquisition module, used to collect laser power density P(t), spot radius r(t), metal material density ρ m Specific heat capacity C p (T m ), thermal conductivity k m (T m ); Construct a dynamic nonlinear heat conduction equation and solve it to obtain the dynamic temperature distribution T m (x, y, t), where the formula for the dynamic nonlinear heat conduction equation is: in, is the gradient of the temperature field, indicating the direction and rate of temperature change with spatial position; Q laser (x,y,t) is the laser input heat energy, in W / m 2 ;Q ref (x,y,t) is the laser reflection energy loss, unit is W / m 2 ;Q phase (x, y, t) is the latent heat loss of phase change, in W / m 2 , x and y are the two-dimensional spatial coordinates on the surface of the metal material, t is the time variable, and m represents the material mass point; Feature vector extraction module is used to extract the dynamic temperature distribution T m (x, y, t) and laser power density P(t) and spot radius r(t) fuse multimodal features and generate a dynamic fusion feature vector F fusiom , the fusion formula is: F fusion =a P ·P)t_+α r ·r(t)+α T ·T amb 9t)+a m ·∫ A T m )x,y,t_dxdy Among them, α P ,α r ,α T ,α m is the feature weight coefficient; ∫ A T m (x,y,t)dxdy is the integral of the temperature field in the irradiated area A, in K·m 2 ; T amb )t_ is the ambient temperature, which is collected in real time by the ambient temperature sensor, and the unit is K; The dynamic fusion feature vector F fusiom As input for subsequent knowledge graph construction; The parameter association network construction module is used to build a dynamic association network based on the knowledge graph. The input data of the dynamic association network includes the dynamic fusion feature vector F fusiom The associated node data of the dynamic associated network include laser power density P)t_, spot radius r)t_, specific heat capacity C p )T m ), thermal conductivity k m (T m ); The dynamic association network outputs the optimized dynamic association weight feature vector Q; The damage time prediction module is used to construct a damage time prediction model to obtain the predicted damage time T d ; The input data of the damage time prediction model includes the dynamic association weight feature vector Q and the dynamic fusion feature vector F after knowledge graph optimization. fusiom ; The formula of the damage time prediction model is: T d =W1·F fusion +W2·Q+b Among them, W1 is the dynamic fusion feature weight coefficient, W2 is the association weight feature weight coefficient, and b is the bias term of the damage time prediction model; Learning optimization prediction model module is used to predict the damage time T d and the damage time T actually measured by the sensor real Set the reward function R for reinforcement learning: R=-|T real -T d | The weight coefficients W1 and W2 in the damage time prediction model are optimized according to the reward function R combined with the PPO algorithm.

2. A damage time prediction system for laser irradiated metal materials as claimed in claim 1, characterized in that: The dynamic temperature distribution acquisition module includes: laser power density P(t), which is collected in real time by the laser power sensor in units of W / cm 2 , used to describe the power of the laser per unit area; the spot radius r(t), measured by the built-in optical monitoring sensor of the laser, in cm; the density of the metal material ρ m , obtained through laboratory measurement or reference material technical parameters, unit is kg / m 3 Specific heat capacity C p (T m ), obtained through the material thermal property testing device, the unit is J / (kg·K); thermal conductivity k m (T m ), measured and fitted by experiment, unit is W / (m·K); temperature distribution T m (x, y, t), collected in real time using an infrared thermal imager, in K, representing the dynamic temperature distribution on the material surface.

3. A damage time prediction system for laser irradiated metal materials as claimed in claim 1, characterized in that: The dynamic temperature distribution acquisition module includes: Q laser (x,y,t) is the laser input heat energy, in W / m 2 , the calculation formula is: Among them, η abs is the laser energy absorption rate, σ is the spot diffusion parameter, and exp is the exponential function.

4. A damage time prediction system for laser irradiated metal materials as claimed in claim 1, characterized in that: In the parameter association network construction module, the dynamic association relationship between the association nodes of the dynamic association network is described by the following formula: G=(N,E),E ij =f(correlation strength(N i ,N j )) Where N is the set of associated nodes in the dynamic associated network; E ij is the dynamic association weight between the associated nodes, optimized by the graph neural network, G is the dynamic association network graph, N i is the ith associated node of the dynamic associated network, N j is the j-th associated node of the dynamic associated network, and f is the correlation strength function, which is used to quantify the association strength between the ith associated node and the j-th associated node.

5. The damage time prediction system for laser irradiated metal materials according to claim 1, characterized in that: In the damage time prediction module, the training input data used in the training process of the damage time prediction model are the standard historical dynamic association weight feature vector Q and the dynamic fusion feature vector F fusiom ,The training output data is the actual damage time measured by historical experiments.

6. The damage time prediction system for laser irradiated metal materials according to claim 1, characterized in that: In the dynamic temperature distribution acquisition module, Q ref (x,y,t) is the laser reflection energy loss, unit is W / m 2 , the calculation formula is: Q ref (x,y,t)=R m (T m )·exp(-γ r ·T m ) Among them, R m (T m ) is the reflectivity, γ r is the coefficient of reflectivity changing with temperature; Q phase (x, y, t) is the latent heat loss of phase change, in W / m 2 , the calculation formula is: Among them, L m is the latent heat of fusion, which is measured by a preset scanning calorimeter; T solid , T liquid are the solid phase and liquid phase temperatures, respectively, which are directly obtained from the preset material technical standard table, f m To indicate the melting ratio of metal materials.

7. The damage time prediction system for laser irradiated metal materials according to claim 1, characterized in that: In the learning optimization prediction model module, the weight coefficients W1 and W2 in the damage time prediction model are optimized using the formula: Among them, W new ,W αld are the weight coefficients after optimization and before optimization respectively; α is the learning rate, which is used to control the step size of weight adjustment; is the gradient of the reward function R.

Images