Intelligent optimization method and system for aluminum alloy die casting forming process

Abstract

The invention discloses an intelligent optimization method and system for an aluminum alloy die casting forming process, and relates to the technical field of intelligent optimizing.The intelligent optimization method comprises the steps that the multi-element content in an aluminum alloy melt is collected in real time, and the multi-element content is combined into a three-dimensional material gene vector; inputting the three-dimensional material gene vector into a pre-trained graph neural network, and calculating to obtain an injection speed compensation coefficient and a boost pressure compensation coefficient; performing space-time alignment on the geometric center coordinate of the defect and die temperature field and pressure time sequence data recorded in the die casting process, inputting a reinforcement learning agent model constructed through a depth Q network algorithm, and reconstructing an evolution path of an internal defect area; and according to the defect formation time point and position point identified in the evolution path, reversely correcting the weight parameter of the graph neural network, and updating. According to the method, the content of multiple elements is fused into the three-dimensional material gene vector, and millisecond-level cooperative compensation of the injection speed and the boost pressure is achieved through the dynamic topology modeling capacity of the pre-training graph neural network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent optimization, in particular to an intelligent optimization method and system for an aluminum alloy die casting forming process. BACKGROUND

[0002] The intelligent control of aluminum alloy die casting process has made significant progress in recent years. Spectral analysis has achieved millisecond-level online detection of melt multi-element content; graph neural networks have gradually matured in the application of industrial process modeling, and can handle high-dimensional process parameter mapping problems; the combination of industrial CT and deep learning can achieve sub-millimeter-level identification of internal defects in die castings. The optimization capability of reinforcement learning algorithm in dynamic system control has also been verified through multiple industrial cases, providing a technical foundation for process closed-loop control.

[0003] The existing technical system has deficiencies, and the dynamic coupling effect between melt composition fluctuations and die casting parameters lacks real-time modeling capability. Traditional solutions rely on static process windows or offline calibration models, which cannot capture the nonlinear effects of multi-element interactions in aluminum melt on flow / solidification behavior. When raw material batches fluctuate, fixed parameter strategies lead to a sharp increase in gas entrapment and shrinkage cavity defect rates, becoming a bottleneck for high-precision die casting quality control. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an intelligent optimization method for an aluminum alloy die casting forming process to solve the problem of process parameter mismatch caused by real-time fluctuations in the multi-element content of the aluminum melt during the die casting process.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides an intelligent optimization method for an aluminum alloy die casting forming process, which includes: real-time acquisition of multi-element content in an aluminum melt, combination of the multi-element content into a three-dimensional material gene vector; inputting the three-dimensional material gene vector into a pre-trained graph neural network to calculate a injection speed compensation coefficient and a boost pressure compensation coefficient; adjusting the real-time injection speed and boost pressure set values of the die casting machine based on the injection speed compensation coefficient and the boost pressure compensation coefficient, and performing die casting operation to generate a die cast forming casting; scanning the die cast forming casting to obtain a three-dimensional tomographic image, identifying an internal defect area in the three-dimensional tomographic image through a deep learning segmentation algorithm, and generating a four-dimensional defect center coordinate; spatiotemporal alignment of the defect geometric center coordinate with the mold temperature field and pressure time series data recorded during the die casting process, inputting into a deep Q network algorithm to construct a reinforcement learning agent model, and reconstructing the evolution path of the internal defect area; According to the defect formation time and location points identified in the evolution path, the weight parameters of the graph neural network are reversely corrected and updated.

[0007] As a preferred solution of the intelligent optimization method of the aluminum alloy die casting molding process of the present invention, wherein: the multi-element content is combined into a three-dimensional material gene vector, the specific steps are as follows: The axial magnetic field drives the melt to form a steady-state melt flow, breaks the surface oxide film, and discretizes the steady-state melt flow into a uniform melt droplet array under inert gas shearing; Apply vortex phase-modulated pulsed laser to excite a toroidal plasma on a uniform melt droplet array, and collect the spectral signals generated by the plasma to calculate the multi-element content; Input the multi-element content into the aluminum-based alloy phase diagram database, call the phase diagram parameters, build the element interaction matrix according to the atomic properties and extract the element interaction main eigenvalues; The multi-element contents, phase diagram parameters and element interaction principal eigenvalues ​​are combined into a three-dimensional material gene vector.

[0008] As a preferred solution of the intelligent optimization method of the aluminum alloy die casting molding process of the present invention, wherein: the injection speed compensation coefficient and the boost pressure compensation coefficient are obtained by calculation, the specific steps are as follows: The three-dimensional material gene vector is input into a pre-trained graph neural network to generate a dynamic physical constraint graph structure, map the vector dimensions into graph nodes, and construct edge weights based on multi-scale physical coupling relationships; Applying quantum projection-driven graph convolution operations to dynamic physical constraint graph structures, global features are generated through aggregation of complex feature space transformations; The global features are decoded by two branches of thermodynamics and fluid mechanics to output the injection velocity compensation coefficient and the boost pressure compensation coefficient respectively.

[0009] As a preferred solution of the intelligent optimization method of the aluminum alloy die casting molding process of the present invention, wherein: the die casting operation is performed to generate the die casting molded casting, and the specific steps are as follows: Calculating the injection speed dynamic gain coefficient and the boost pressure dynamic gain coefficient based on the injection speed compensation coefficient and the boost pressure compensation coefficient; Combine the dynamic gain coefficient of injection speed and the dynamic gain coefficient of boost pressure with the basic process parameters to solve the linkage constraints and generate the adjusted setting values ​​of injection speed and boost pressure; The adjusted injection speed and boost pressure setting values ​​are synchronously sent to the servo valve group through the real-time bus to execute the die-casting operation in a coordinated manner to generate die-cast castings.

[0010] As a preferred scheme of the intelligent optimization method of the aluminum alloy die casting forming process, the four-dimensional defect center coordinates are generated, and the specific steps are as follows, The die casting forming castings are subjected to dual-energy spectrum tomography to obtain three-dimensional tomographic images, and the three-dimensional tomographic images are subjected to tomographic reconstruction processing to generate three-dimensional volume data; The three-dimensional volume data are processed by a deep learning segmentation algorithm to identify the internal defect area; The geometric center coordinates of the internal defect area are extracted and fused with the die casting filling time to generate four-dimensional defect center coordinates.

[0011] As a preferred scheme of the intelligent optimization method of the aluminum alloy die casting forming process, the four-dimensional defect center coordinates are generated, and the specific steps are as follows, The four-dimensional defect center coordinates are fused with the mold temperature field and pressure time sequence data recorded during the die casting process in the mold coordinate system, and are subjected to space-time alignment to construct five-dimensional space-time field data; The five-dimensional space-time field data are input into a deep Q network algorithm to construct a reinforcement learning agent model, and evolution action sequences are generated by exploring physical constraint strategies; The evolution action sequences are regularized and inverted by using multi-field coupled partial differential equations to reconstruct the evolution path of the internal defect area.

[0012] As a preferred scheme of the intelligent optimization method of the aluminum alloy die casting forming process, the four-dimensional defect center coordinates are generated, and the specific steps are as follows, Based on the reconstructed evolution path of the internal defect area, a holographic defect energy field is constructed in a five-dimensional anti-De Sitter space-time, and the defect formation time point is mapped to the time coordinate and the position point is mapped to the boundary space coordinate; The holographic defect energy field is input into a non-commutative differential operator, and the space-time invariants are extracted as quantum geometric features by a quantum covariant gradient; A pre-trained string theory matrix model is used to perform dimension renormalization processing on the quantum geometric features to generate a correction tensor of the graph neural network weight; The correction tensor is injected into the backpropagation flow of the graph neural network to perform updating.

[0013] In a second aspect, the present application provides an intelligent optimization system for an aluminum alloy die casting forming process, comprising a collection element module, an intelligent compensation module, a linkage execution module, a defect analysis module, a time tracing inversion module and a reverse updating module; The collection element module is used to collect the multi-element content in the aluminum alloy melt in real time, and combine the multi-element content into a three-dimensional material gene vector; The intelligent compensation module is configured to input a three-dimensional material gene vector into a pre-trained graph neural network to calculate a ram speed compensation coefficient and a boost pressure compensation coefficient. The linkage execution module is configured to adjust real-time ram speed and boost pressure set values of the die casting machine based on the ram speed compensation coefficient and the boost pressure compensation coefficient, and perform a die casting operation to generate a die casting formed casting. The defect analysis module is configured to scan a three-dimensional tomographic image of the die casting formed casting, identify an internal defect area in the three-dimensional tomographic image through a deep learning segmentation algorithm, and generate four-dimensional defect center coordinates. The time tracing inversion module is configured to perform space-time alignment of the defect geometric center coordinates and mold temperature field and pressure time sequence data recorded in the die casting process, input a deep Q network algorithm to construct a reinforcement learning agent model, and reconstruct an evolution path of the internal defect area. The reverse updating module is configured to reverse correct weight parameters of the graph neural network according to a defect formation time point and a position point identified in the evolution path, and update the weight parameters.

[0014] In a third aspect, the present application provides a computer device comprising a memory and a processor, and the memory stores a computer program, wherein the computer program is executed by the processor to implement any step of the intelligent optimization method of the aluminum alloy die casting forming process according to the first aspect of the present application.

[0015] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement any step of the intelligent optimization method of the aluminum alloy die casting forming process according to the first aspect of the present application.

[0016] The present application has the following beneficial effects: the multi-element content is fused into a three-dimensional material gene vector, the dynamic topology modeling capability of the pre-trained graph neural network is used to realize millisecond-level collaborative compensation of the ram speed and the boost pressure, the multi-scale physical coupling mechanism of the graph convolution is used to accurately compensate the speed compensation coefficient and the pressure compensation amount, and the real-time control of the die casting machine is directly driven. The physical interpretable mapping relationship between the material gene and the process parameter is simultaneously constructed to provide real-time optical diagnosis level decision support for intelligent die casting. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0018] Figure 1A flow chart of an intelligent optimization method for an aluminum alloy die casting forming process.

[0019] Figure 2 A module diagram of an intelligent optimization system for an aluminum alloy die casting forming process.

[0020] Figure 3 A flow chart of a graph neural network compensation calculation.

[0021] Figure 4 A flow chart of an evolutionary path reconstruction. DETAILED DESCRIPTION

[0022] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited by the specific embodiments disclosed below.

[0024] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent or alternative to other embodiments.

[0025] REFERENCE Figures 1-4 For one embodiment of the present application, the embodiment provides an intelligent optimization method for an aluminum alloy die casting forming process, comprising the following steps: S1, real-time acquisition of multi-element content in aluminum alloy melt, combination of multi-element content into three-dimensional material gene vector.

[0026] Further, the surface oxide film is broken by driving the melt to form a steady-state melt flow by an axial magnetic field, and the steady-state melt flow is discretized into a uniform melt droplet array under inert gas shearing; Specifically, a vertical Lorentz force is applied to the conductive melt by an axial magnetic field, inducing the formation of a self-organizing rotating flow field inside the melt, which is stretched along the axial direction under the action of the Coriolis effect and develops into a steady-state vortex structure. The radial velocity gradient generated by the steady-state vortex structure exceeds the threshold of the oxide film binding energy, realizing the rupture and peeling of the oxide layer. The broken melt is sheared by inert gas jetted by a coaxial annular nozzle, and the inert gas shearing flow tears the continuous melt into a uniform melt droplet array.

[0027] It should be noted that the oxide film binding energy threshold is set according to the aluminum alloy surface aluminum oxide crystal cleavage surface binding strength determination experiment, and the example value is 2.4 joules per square meter.

[0028] The vortex phase modulated pulsed laser is focused on the rotation phase zero point of the uniform melt droplet array, the pulsed laser spot diameter matches the spacing of the melt droplet array, the energy density of the pulsed laser is improved, the melt droplet array absorbs the pulsed laser energy to cause multi-photon ionization, the multi-photon ionization promotes the outer edge electron avalanche breakdown of the melt droplet array, generates a ring-shaped plasma, the ring-shaped plasma expands and cools, the bound state electrons are excited to emit atomic characteristic spectrum lines, and a high-resolution fiber spectrometer synchronously collects the emission spectrum data of the ring-shaped plasma in the cooling stage. The emission spectrum data is compared with the aluminum alloy material standard spectrum database to calculate the silicon iron copper manganese zinc multi-element percentage content. Specifically, the vortex phase modulated pulsed laser is focused on the rotation phase zero point of the uniform melt droplet array, the pulsed laser spot diameter matches the spacing of the melt droplet array, the energy density of the pulsed laser is improved, the melt droplet array absorbs the pulsed laser energy to cause multi-photon ionization, the multi-photon ionization promotes the outer edge electron avalanche breakdown of the melt droplet array, generates a ring-shaped plasma, the ring-shaped plasma expands and cools, the bound state electrons are excited to emit atomic characteristic spectrum lines, and a high-resolution fiber spectrometer synchronously collects the emission spectrum data of the ring-shaped plasma in the cooling stage. The emission spectrum data is compared with the aluminum alloy material standard spectrum database to calculate the silicon iron copper manganese zinc multi-element percentage content.

[0029] The multi-element content is input into the aluminum-based alloy phase diagram database, the phase diagram parameters are called, the element interaction matrix is constructed according to the atomic properties, and the element interaction main eigenvalue is extracted; Specifically, the aluminum-based alloy phase diagram database receives the silicon iron copper manganese zinc multi-element percentage content, the aluminum-based alloy phase diagram database indexes the binary eutectic point solidification shrinkage rate, the interdendritic spacing and the solubility parameter matched with the multi-element percentage content, the atomic property database calls the electronegativity atomic radius valence electron number data of silicon iron copper manganese zinc, the electronegativity atomic radius valence electron number data is calculated according to the Pauling rule, the orbital hybridization energy and the lattice distortion energy of the center of the two elements are calculated, the element interaction matrix is formed, and the maximum modulus eigenvalue of the element interaction matrix is extracted as the element interaction main eigenvalue.

[0030] It should be noted that the aluminum-based alloy phase diagram database refers to a metallurgical database that stores binary phase diagrams, multi-element phase diagrams, and solidification process phase equilibrium points, solidification shrinkage rates, interdendritic spacings, and solubility parameters of alloy systems composed of aluminum and silicon, iron, copper, manganese, and zinc. Figure Three The atomic property database refers to a quantum chemical database that stores atomic numbers, electron configurations, orbital hybridization, electronegativity, atomic radius, valence electron number, and first ionization energy of silicon, iron, copper, manganese, and aluminum.

[0031] The multi-element content, phase diagram parameters, and element interaction main eigenvalues are combined into a three-dimensional material gene vector.

[0032] Specifically, the multi-element content is transmitted to the first dimension position of the three-dimensional material gene vector, the phase diagram parameter called by the aluminum-based alloy phase diagram database is embedded into the second dimension position of the three-dimensional material gene vector, and the element interaction main eigenvalue is positioned to the last dimension position of the three-dimensional material gene vector. The format of the three-dimensional material gene vector conforms to the fixed-length floating-point number array specification, and the generation of the three-dimensional material gene vector completes the spatial combination mapping of the multi-element content, the phase diagram parameter and the element interaction main eigenvalue.

[0033] S2, input the three-dimensional material gene vector into the pre-trained graph neural network to calculate the injection speed compensation coefficient and the supercharging pressure compensation coefficient.

[0034] Further, the three-dimensional material gene vector is input into the pre-trained graph neural network to generate a dynamic physical constraint graph structure, map the vector dimension to a graph node, and construct an edge weight according to a multi-scale physical coupling relationship; It should be noted that the three-dimensional material gene vector is used as a node feature to initialize the graph neural network, the edge weight of the graph neural network is initialized according to the melt flow physical topology connection rule, the graph convolution operation aggregates the thermal conductivity coefficient and the rheological stress state of adjacent nodes layer by layer, the output full connection layer maps the refining features to a double-target space composed of the injection speed compensation coefficient and the supercharging pressure compensation coefficient, the training process minimizes the loss function composed of the speed compensation coefficient residual, the pressure compensation coefficient residual and the defect inversion consistency constraint, the back propagation executes the random gradient descent optimization with Nesterov momentum, and finally the graph neural network with high-precision injection speed compensation coefficient and supercharging pressure compensation coefficient prediction capability is obtained.

[0035] Specifically, the dynamic physical constraint graph structure is generated, the vector dimension is mapped to a graph node, and the edge weight is constructed according to the multi-scale physical coupling relationship, and the expression is: ; In the formula, denotes the edge weight of the source node and the target node , denotes the source node index, denotes the target node index, denotes the Softplus activation function, denotes the spatial scale hierarchical index, denotes the sum of all spatial scale hierarchies, denotes the scale weight coefficient of the spatial scale hierarchy , denotes the spatial correlation function of the spatial scale hierarchy , denotes the Euclidean distance between nodes, denotes the integral item gain coefficient, denotes integration over scale factor interval, denotes scale factor maximum, denotes scale factor minimum, denotes Gaussian kernel function, denotes spatial scale hierarchy, decay function, denotes scale factor, denotes differential variable of scale factor.

[0036] It should be noted that the scale weight coefficients of the spatial scale hierarchy are obtained by backpropagation training of the graph neural network, and example values are , , ; the integral term gain coefficient is obtained by Bayesian optimization from multi-physical field coupling matching experimental data, and example value range is 0.25-0.35.

[0037] A quantum projection driven graph convolution operation is applied to the dynamic physical constraint graph structure, and global features are generated by complex feature space transformation aggregation; Specifically, the quantum projection operator acts on the node embedding complex vector of the dynamic physical constraint graph structure, and the node complex vector is projected to the entangled Hilbert space through unitary matrix transformation, the linear combination features of the entangled Hilbert space are phase modulated by Pauli gate rotation, the entangled state features after phase modulation perform point multiplication operation of the complex domain graph convolution kernel, the complex domain graph convolution kernel matches the adjacency tensor of the dynamic physical constraint graph structure to complete local feature aggregation, and the long-range correlation component is extracted by Fourier frequency domain filtering, and the long-range correlation component is restored to the coordinate space by inverse quantum Fourier transform, and the restored features are projected to the real number subspace to reduce dimension to generate global features.

[0038] Specifically, the global features are decoded in two branches of thermodynamics and fluid mechanics, and the pressure injection velocity compensation coefficient and the supercharging pressure compensation coefficient are output respectively, and the expressions are: ; ; In the formula, denotes the pressure injection velocity compensation coefficient, denotes the reference value of the pressure injection velocity compensation coefficient, denotes the scaling coefficient of the pressure injection velocity compensation coefficient, denotes the sigmoid function, denotes the full connection neural network for velocity decision, denotes the global feature reorganization function, denotes the global feature vector, ​denotes a pressurization pressure compensation coefficient, denotes a reference value of the pressurization pressure compensation coefficient, denotes a scaling coefficient of the pressurization pressure compensation coefficient, denotes a hyperbolic tangent function, denotes a fully connected neural network for pressure decision, denotes a normalization function.

[0039] It should be noted that the injection speed compensation coefficient is determined by industrial grade die casting test calibration and feedback optimization, and the example value range is 0.82-1.18; the reference value of the injection speed compensation coefficient is based on the fluid mechanics calculation of the critical filling speed of the material, and the example fixed reference value is always 0.8; the scaling coefficient of the injection speed compensation coefficient is obtained by fitting the wall thickness-flow nonlinear equation, and the example fixed value is 0.4; the pressurization pressure compensation coefficient is optimized according to the mold locking force-internal pressure coupling model, and the example value range is 0.55-1.42; the reference value of the pressurization pressure compensation coefficient corresponds to the material solidification shrinkage pressure value, and the example fixed reference value is 1.0; the scaling coefficient of the pressurization pressure compensation coefficient is set by statistical regression analysis of shrinkage hole defects, and the example fixed value is 0.5.

[0040] S3, based on the injection speed compensation coefficient and the pressurization pressure compensation coefficient, adjusting the real-time injection speed and pressurization pressure set value of the die casting machine, and performing die casting operation to generate die castings.

[0041] Specifically, based on the injection speed compensation coefficient and the pressurization pressure compensation coefficient, the injection speed dynamic gain coefficient and the pressurization pressure dynamic gain coefficient are calculated, and the expression is: ; ; In the formula, denotes the injection speed dynamic gain coefficient, denotes a heat-sensitive coefficient, denotes a hydraulic oil temperature difference, denotes a nonlinear effect of the injection speed compensation coefficient adjustment, denotes the pressurization pressure dynamic gain coefficient, denotes a pressure deviation absolute value, denotes a deviation, denotes a pressure, denotes a natural logarithm function, denotes a nonlinear effect of the pressurization pressure compensation coefficient adjustment.

[0042] It should be noted that the speed dynamic gain coefficient is based on the hydraulic oil temperature-viscosity characteristic experiment calibration, and is optimized to 1.6-2.3 in the oil temperature range of 50-100°C; the heat-sensitive coefficient is obtained according to the Arrhenius equation fitting of the viscosity-temperature curve, and is fixed at 0.1; and the boost pressure dynamic gain coefficient is optimized through step response test of the pressure servo valve, and is taken as 0.42.

[0043] Specifically, the injection speed dynamic gain coefficient and the boost pressure dynamic gain coefficient are combined with the basic process parameters to solve the linkage constraint, to generate the adjusted injection speed and boost pressure set values, and the expression is: ; In the formula, indicates the minimum value, indicates the adjusted injection speed set value, indicates the basic injection speed, indicates the L2 norm operator, indicates the gain-adjusted speed reference, indicates the square deviation of the adjusted injection speed set value and the gain-adjusted speed reference, indicates the pressure term optimization weight coefficient, indicates the adjusted boost pressure set value, indicates the basic boost pressure, indicates the gain-adjusted pressure reference, indicates the square deviation of the adjusted boost pressure set value and the gain-adjusted pressure reference, indicates the constraint condition identifier, indicates the servo valve wear coefficient, indicates the speed change amount, indicates the time interval, indicates the absolute value constraint of the speed change rate, indicates the pressure gradient, indicates the gradient operator, indicates the material flow coefficient, indicates the 0.7 power of the adjusted injection speed set value.

[0044] It should be noted that the basic process parameters refer to five core control variables of the preset reference injection speed value, the reference boost pressure value, the maximum clamp force threshold, the fast injection stroke length and the alloy pouring temperature; the maximum clamp force threshold is set according to the safety margin of the product area and the internal pressure peak value, and is taken as 23000 kilonewtons.

[0045] It should be noted that the pressure term optimization weight coefficient is determined by double objective Pareto front analysis, and the example fixed value is 0.7; the servo valve wear coefficient is calibrated according to the servo valve current-displacement curve slope attenuation rate, and the example value is 0.35; the material flow coefficient is fitted based on the high shear rate test of melt rheometer, and the example value is 0.15±0.02.

[0046] The adjusted injection speed and boost pressure set value are synchronously issued to the servo valve group linkage to perform die casting operation through the real-time bus, and the die casting molded castings are generated.

[0047] Specifically, the adjusted injection speed set value and boost pressure set value are packaged as EtherCAT real-time bus communication frames; the EtherCAT real-time bus communication frames are synchronously transmitted to the injection speed control servo valve controller and the boost pressure control servo valve controller at a period of less than 500 microseconds, the injection speed control servo valve controller analyzes the set value to drive the injection speed control servo valve core displacement, and the boost pressure control servo valve controller adjusts the boost pressure control servo valve opening degree in response to the set value; the injection speed control servo valve displacement instruction links the die casting machine injection cylinder plunger advancing speed, and the boost pressure control servo valve opening degree instruction links the boost accumulator oil pressure output; the injection cylinder plunger advancing speed and the boost accumulator oil pressure are accurately matched at the set value in the high injection stroke stage, the molten aluminum alloy completes mold cavity filling under the linkage control of the injection pressure and the filling speed, the liquid metal realizes sequential solidification in the final pressure stage maintained by the boost pressure control servo valve, and the ejection mechanism separates the mold to generate the die casting molded castings after the metal solidification.

[0048] S4, scanning the die casting molded castings to obtain three-dimensional tomographic images, identifying the internal defect area in the three-dimensional tomographic images through a deep learning segmentation algorithm, and generating four-dimensional defect center coordinates.

[0049] Further, the die casting molded castings are subjected to dual-energy spectral tomography to obtain three-dimensional tomographic images, and the three-dimensional tomographic images are subjected to tomographic reconstruction processing to generate three-dimensional volume data. Specifically, the die-cast casting is fixed to the rotary table of the dual-energy tomography device, the rotary table drives the die-cast casting to rotate at a uniform speed for 360 degrees, the high-energy X-ray source and the low-energy X-ray source of the dual-energy tomography device are alternately pulsed to emit, the high-energy X-ray source pulse penetrates the die-cast casting and is received by the high-energy linear array detector, the low-energy X-ray source pulse penetrates the die-cast casting and is received by the low-energy linear array detector, the dual-energy tomography device records the projection data of the high-energy linear array detector and the low-energy linear array detector, the projection data is input into the FDK back-projection reconstruction algorithm to perform attenuation coefficient matrix calculation, the attenuation coefficient matrix is corrected for scattering artifacts by Hilbert space filtering, the corrected attenuation coefficient matrix is mapped to three-dimensional space gray matter voxel, and the three-dimensional space gray matter voxel is arranged according to the Cartesian coordinate system to generate three-dimensional body data.

[0050] It should be noted that the attenuation coefficient matrix refers to a two-dimensional numerical array obtained by dual-energy tomography projection data reconstruction, which represents the X-ray attenuation ability of the material at each spatial position inside the casting.

[0051] The three-dimensional body data is processed by a deep learning segmentation algorithm to identify the internal defect area. Specifically, the three-dimensional body data is input into the deep learning segmentation algorithm, the deep learning segmentation algorithm includes a three-dimensional encoder-decoder structure, the three-dimensional encoder stage adopts a three-dimensional convolution kernel to extract defect features in layers, and the spatial resolution is reconstructed by three-dimensional deconvolution operation, the output layer of the deep learning segmentation algorithm applies a Sigmoid activation function to generate a defect probability map, a binary defect mask is formed after segmentation, a three-dimensional connected domain labeling algorithm is executed to identify the internal defect area, the geometric center coordinates of the internal defect area are extracted, and four-dimensional defect center coordinates are generated by fusing the die casting filling time.

[0052] S5, the defect geometric center coordinates are spatiotemporally aligned with the mold temperature field and pressure time series data recorded during the die casting process, input through a deep Q network algorithm to construct a reinforcement learning agent model, and the evolution path of the internal defect area is reconstructed.

[0053] Further, the four-dimensional defect center coordinates are fused with the mold temperature field and pressure time series data recorded during the die casting process in the mold coordinate system, spatiotemporally aligned, and a five-dimensional spatiotemporal field data is constructed. Specifically, a space reference datum is set at the origin of the mold coordinate system, the spatial components of the four-dimensional defect center coordinates are converted to Cartesian coordinates of the mold coordinate system, the mold temperature field time series data recorded during the die casting process are rigidly registered to the mold coordinate system, the pressure time series data are filled in the time axis by linear interpolation to form a time-continuous pressure field, the pressure field and the mold temperature field are spatiotemporally synchronized by interpolation at the grid points of the mold coordinate system, the timestamps of the four-dimensional defect center coordinates are associated with the synchronization time nodes of the pressure field, the data points of the synchronization time nodes of the pressure field are matched with the temperature field at the grid points of the mold coordinate system in terms of spatial position, the matched defect center coordinates, data points of the mold temperature field and data points of the pressure field are combined into five-dimensional data points, and the five-dimensional data points are arranged in lexicographic order to generate five-dimensional spatiotemporal field data.

[0054] The five-dimensional spatiotemporal field data are input into a deep Q network algorithm to construct a reinforcement learning agent model, and an evolutionary action sequence is generated by exploration with a physical constraint strategy. It should be noted that the five-dimensional spatiotemporal field data are input into the input layer of the deep Q network algorithm, the deep Q network algorithm uses a three-dimensional convolution kernel to process the spatial dimension and a long short-term memory network to process the time dimension; the action space is defined as discrete adjustment values of the injection velocity compensation coefficient and the boost pressure compensation coefficient, the physical constraint strategy is used to constrain the action selection by a boundary limiting function, the target network adopts a soft update mechanism, and the training process is optimized by exploration with a greedy strategy to generate a reinforcement learning agent model supporting the physical constraint strategy.

[0055] Specifically, the five-dimensional spatiotemporal field data are input into the reinforcement learning agent model constructed by the deep Q network algorithm, the deep Q network algorithm of the reinforcement learning agent model extracts spatiotemporal features in combination with a convolution long short-term memory network, the physical constraint strategy of the deep Q network algorithm is embedded in an action boundary limiting function such as a truncation function, the action boundary limiting function of the physical constraint strategy forces the action selection to meet the solidification shrinkage rate and pressure gradient constraints of the die casting process, and the exploration strategy of the reinforcement learning agent model iteratively selects actions by a greedy method to generate an evolutionary action sequence.

[0056] The evolutionary action sequence is regularized and inverted by using a multi-field coupled partial differential equation to reconstruct the evolution path of the internal defect region.

[0057] Specifically, the evolution action sequence input multi-field coupled partial differential equation inversion framework, the multi-field coupled partial differential equation includes Navier-Stokes equation describing melt flow, Fourier heat conduction equation describing temperature propagation and solidification kinetics equation describing phase change process, the injection speed compensation coefficient of the evolution action sequence drives the boundary condition update of the Navier-Stokes equation, and the supercharging pressure compensation coefficient synchronously updates the solidification pressure term of the solidification kinetics equation, the time discretization step is aligned with the time resolution of the five-dimensional space-time field data to form an inversion time grid, the regularization constraint adopts Tikhonov norm to punish the non-physical fluctuations of the evolution path, and the Newton-Raphson iteration method is used to solve the implicit format of the multi-field coupled partial differential equation. According to the temperature gradient calculated by the heat conduction equation, the solidification front position is updated, the spatial coincidence degree of the solidification front position and the internal defect area is taken as the inversion convergence criterion, and the converged solidification front displacement path is output as the evolution path of the reconstructed internal defect area.

[0058] S6, according to the defect formation time point and position point identified in the evolution path, the weight parameters of the graph neural network are reversely corrected and updated.

[0059] Further, based on the reconstructed evolution path of the internal defect area, a holographic defect energy field is constructed in a five-dimensional anti-de Sitter space-time, the defect formation time point is mapped to a time coordinate, and the position point is mapped to a boundary space coordinate; Specifically, the time point data of the reconstructed evolution path of the internal defect area is mapped to the time coordinate dimension of the five-dimensional anti-de Sitter space-time, the position point data of the reconstructed evolution path of the internal defect area is projected to the boundary space coordinate dimension of the five-dimensional anti-de Sitter space-time, the time coordinate dimension and the boundary space coordinate dimension form a discrete space-time point set in the five-dimensional anti-de Sitter space-time, each discrete space-time point set is associated with a defect energy density value, the negative curvature manifold relationship between the boundary space coordinate points is calculated through the hyperbolic distance function, the negative curvature manifold relationship is converted into a quantum entanglement state of the five-dimensional anti-de Sitter space-time body interval, the quantum entanglement state evolves in accordance with the Einstein field equation constraint, and the evolution process generates a scalar field distribution of the five-dimensional anti-de Sitter space-time whole domain. The scalar field distribution is expressed as a holographic defect energy field.

[0060] Input the holographic defect energy field into a non-commutative differential operator, and extract space-time invariants as quantum geometric features through quantum covariant gradients; Specifically, the holographic defect energy field is input into the domain of a noncommutative differential operator, the noncommutative differential operator performs a discrete form of a Weyl algebra operation to generate a noncommutative differential form, a quantum covariant derivative is established with a spin connection structure of a Riemannian manifold, the quantum covariant derivative acts on a scalar value of the holographic defect energy field to generate a quantum covariant gradient field, the quantum covariant gradient field is integrated along an Abelian gauge group orbit to calculate a curvature two-form, the curvature two-form is mapped to a de Rham cohomology group through a Chen-Weil mapping, an above-closed chain of the de Rham cohomology group extracts an invariant closed form satisfying a Cartan-Killing equation, an intrinsic symmetry of the invariant closed form generates a spacetime invariant cluster, and the spacetime invariant cluster is output as a quantum geometric feature.

[0061] The pre-trained string theory matrix model is used for dimension renormalization processing of the quantum geometric feature, to generate a correction tensor of the graph neural network weight; It should be noted that a Gaussian unit ensemble random matrix of the D0-membrane matrix model is initialized, a super-symmetry action functional of the string theory matrix model is introduced, a matrix configuration is generated by a hybrid Monte Carlo sampling algorithm, the super-symmetry is verified by solving an eigen-spectrum through a Dirac equation, a conjugate gradient method is used to optimize a gradient of the action functional, a convergence criterion is based on a super-symmetry conservation charge deviation degree, a converged matrix configuration is stored to complete the pre-training, and the string theory matrix model is obtained.

[0062] Specifically, a gauge field operator of a quantum geometric feature is input into a pre-trained string theory matrix model, the string theory matrix model performs an operator product expansion through a non-commutative algebraic structure of an IIB type superstring theory, the operator product expansion generates a regularization flow invariant to a gauge, a dimension regularization parameter adjusts a degree of freedom contraction of the quantum geometric feature, a feature subspace after the degree of freedom contraction is mapped to a conformal field boundary through an AdS / CFT duality, a super-symmetry Yang-Mills field of the conformal field boundary generates a renormalization weight, the renormalization weight is decomposed into a three-order tensor form according to a feature mode, and the three-order tensor form is represented as a correction tensor.

[0063] The correction tensor is injected into a back propagation flow of a graph neural network to perform updating.

[0064] Specifically, a weight gradient tensor of a current training step is calculated in a back propagation flow of a graph neural network, a graph neural network weight correction tensor is matched with a dimension of the weight gradient tensor through a dimension alignment operation, the matched graph neural network weight correction tensor is completely assigned and covers an original weight gradient tensor, and the covered weight gradient tensor is submitted to a stochastic gradient descent optimizer with Nesterov momentum to perform parameter updating according to a learning rate scaling coefficient and a momentum accumulation amount, and a parameter updating result is synchronized to the graph neural network to complete updating.

[0065] The embodiment also provides an intelligent optimization system of an aluminum alloy die casting forming process, comprising: an acquisition element module, an intelligent compensation module, a linkage execution module, a defect analysis module, a time tracing inversion module and a reverse updating module; the acquisition element module is configured to acquire the content of multiple elements in the aluminum alloy melt in real time, and combine the content of multiple elements into a three-dimensional material gene vector; the intelligent compensation module is configured to input the three-dimensional material gene vector into a pre-trained graph neural network, and calculate a ram speed compensation coefficient and a boost pressure compensation coefficient; the linkage execution module is configured to adjust the real-time ram speed and boost pressure set value of the die casting machine based on the ram speed compensation coefficient and the boost pressure compensation coefficient, and perform a die casting operation to generate a die casting formed casting; the defect analysis module is configured to scan the die casting formed casting to obtain a three-dimensional tomographic image, identify an internal defect area in the three-dimensional tomographic image through a deep learning segmentation algorithm, and generate a four-dimensional defect center coordinate; the time tracing inversion module is configured to perform space-time alignment on the defect geometric center coordinate and mold temperature field and pressure time sequence data recorded in the die casting process, input a deep Q network algorithm to construct a reinforcement learning agent model, and reconstruct an evolution path of the internal defect area; and the reverse updating module is configured to reverse correct the weight parameters of the graph neural network according to the defect formation time point and position point identified in the evolution path, and update the weight parameters.

[0066] The embodiment also provides a computer device suitable for the intelligent optimization method of the aluminum alloy die casting forming process, comprising: a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the intelligent optimization method of the aluminum alloy die casting forming process proposed in the above embodiment.

[0067] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is configured to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0068] The embodiment also provides a storage medium on which a computer program is stored, the computer program being executed by a processor to implement the intelligent optimization method for realizing the aluminum alloy die casting forming process proposed in the above embodiment; the storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0069] To sum up, the present application realizes millisecond-level collaborative compensation of the injection speed and the boost pressure by fusing the multi-element content into a three-dimensional material gene vector and using the dynamic topology modeling capability of a pre-trained graph neural network. The multi-scale physical coupling mechanism of graph convolution is used to accurately compensate the speed compensation coefficient and the pressure compensation amount, thereby directly driving the real-time control of the die casting machine. A physically interpretable mapping relationship between the material gene and the process parameters is simultaneously constructed to provide real-time optical diagnosis level decision support for intelligent die casting.

[0070] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. An intelligent optimization method for aluminum alloy die casting forming process, characterized by: include, Real-time collection of multi-element content in aluminum alloy melt, combining the multi-element content into a three-dimensional material gene vector; The 3D material gene vector is input into the pre-trained graph neural network to calculate the injection velocity compensation coefficient and the boost pressure compensation coefficient; Based on the injection speed compensation coefficient and the boost pressure compensation coefficient, adjusting the real-time injection speed and boost pressure setting values ​​of the die casting machine, and performing the die casting operation to generate the die-casting casting; Scan the die-cast casting to obtain a 3D tomographic image. Use a deep learning segmentation algorithm to identify internal defect areas in the 3D tomographic image and generate the 4D defect center coordinates. The four-dimensional defect center coordinates are spatiotemporally aligned with the mold temperature field and pressure time series data recorded during the die-casting process. The input is used to construct a reinforcement learning agent model through a deep Q-network algorithm to reconstruct the evolution path of the internal defect area. According to the defect formation time and location points identified in the evolution path, the weight parameters of the graph neural network are reversely corrected and updated.

2. The intelligent optimization method for aluminum alloy die casting molding process according to claim 1, characterized in that: The specific steps of combining the multi-element contents into a three-dimensional material gene vector are as follows: The axial magnetic field drives the melt to form a steady-state melt flow, breaks the surface oxide film, and discretizes the steady-state melt flow into a uniform melt droplet array under inert gas shear. Apply vortex phase-modulated pulsed laser to excite a toroidal plasma on a uniform melt droplet array, and collect the spectral signals generated by the plasma to calculate the multi-element content; Input the multi-element content into the aluminum-based alloy phase diagram database, call the phase diagram parameters, build the element interaction matrix according to the atomic properties and extract the element interaction main eigenvalues; The multi-element contents, phase diagram parameters and element interaction principal eigenvalues ​​are combined into a three-dimensional material gene vector.

3. The intelligent optimization method for aluminum alloy die casting molding process according to claim 2, characterized in that: The injection speed compensation coefficient and the boost pressure compensation coefficient are obtained by calculation. The specific steps are as follows: The three-dimensional material gene vector is input into a pre-trained graph neural network to generate a dynamic physical constraint graph structure, map the vector dimensions into graph nodes, and construct edge weights based on multi-scale physical coupling relationships; Applying quantum projection-driven graph convolution operations to dynamic physical constraint graph structures, global features are generated through aggregation of complex feature space transformations; The global features are decoded by two branches of thermodynamics and fluid mechanics to output the injection velocity compensation coefficient and the boost pressure compensation coefficient respectively.

4. The intelligent optimization method for aluminum alloy die casting molding process according to claim 3, characterized in that: The die-casting operation is performed to generate a die-cast casting, and the specific steps are as follows: Calculating the injection speed dynamic gain coefficient and the boost pressure dynamic gain coefficient based on the injection speed compensation coefficient and the boost pressure compensation coefficient; Combine the dynamic gain coefficient of injection speed and the dynamic gain coefficient of boost pressure with the basic process parameters to solve the linkage constraints and generate the adjusted injection speed and boost pressure setting values; The adjusted injection speed and boost pressure setting values ​​are synchronously sent to the servo valve group through the real-time bus to execute the die-casting operation in a coordinated manner to generate die-cast castings.

5. The intelligent optimization method for aluminum alloy die casting molding process according to claim 4, characterized in that: The specific steps of generating the four-dimensional defect center coordinates are as follows: Perform dual-energy spectrum tomography on the die-cast casting to obtain a three-dimensional tomographic image, and perform tomographic reconstruction on the three-dimensional tomographic image to generate three-dimensional volume data; Process 3D volume data using deep learning segmentation algorithms to identify internal defect areas; The geometric center coordinates of the internal defect area are extracted and integrated with the die casting filling time to generate the four-dimensional defect center coordinates.

6. The intelligent optimization method for aluminum alloy die casting molding process according to claim 5, characterized in that: The specific steps of reconstructing the evolution path of the internal defect area are as follows: In the mold coordinate system, the four-dimensional defect center coordinates are integrated with the mold temperature field and pressure time series data recorded during the die-casting process to perform spatiotemporal alignment and construct five-dimensional spatiotemporal field data. The five-dimensional space-time field data is input into the deep Q-network algorithm to build a reinforcement learning agent model, and the evolutionary action sequence is generated through physical constraint strategy exploration; The evolutionary action sequence is regularized and inverted using multi-field coupled partial differential equations to reconstruct the evolution path of the internal defect area.

7. The intelligent optimization method for aluminum alloy die casting molding process according to claim 6, characterized in that: The weight parameters of the reverse correction graph neural network are updated. The specific steps are as follows: Based on the reconstructed evolution path of the internal defect area, a holographic defect energy field is constructed in the five-dimensional anti-de Sitter space-time. The defect formation time point is mapped as the time coordinate, and the position point is mapped as the boundary space coordinate. The holographic defect energy field is input into a non-commutative differential operator, and the space-time invariants are extracted as quantum geometric features through quantum covariant gradients. A pre-trained string theory matrix model is used to renormalize the quantum geometric features and generate a modified tensor of the graph neural network weights. Inject the correction tensor into the backpropagation flow of the graph neural network to perform the update.

8. An intelligent optimization system for an aluminum alloy die-casting forming process, based on the intelligent optimization method for an aluminum alloy die-casting forming process according to any one of claims 1 to 7, characterized in that: Including acquisition element module, intelligent compensation module, linkage execution module, defect analysis module, time inversion module and reverse update module; The element collection module is used to collect the multi-element content in the aluminum alloy melt in real time and combine the multi-element content into a three-dimensional material gene vector; The intelligent compensation module is used to input the three-dimensional material gene vector into the pre-trained graph neural network to calculate the injection speed compensation coefficient and the boost pressure compensation coefficient; The linkage execution module is used to adjust the real-time injection speed and boost pressure setting values ​​of the die-casting machine based on the injection speed compensation coefficient and the boost pressure compensation coefficient, and execute the die-casting operation to generate the die-cast casting; The defect analysis module is used to scan the die-cast casting to obtain a three-dimensional tomographic image, identify the internal defect area in the three-dimensional tomographic image through a deep learning segmentation algorithm, and generate the four-dimensional defect center coordinates; The time-tracing inversion module is used to align the coordinates of the defect geometric center with the mold temperature field and pressure time series data recorded during the die-casting process, and input the reinforcement learning agent model constructed through the deep Q-network algorithm to reconstruct the evolution path of the internal defect area; The reverse update module is used to reversely correct and update the weight parameters of the graph neural network according to the defect formation time point and location point identified in the evolution path.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the intelligent optimization method for the aluminum alloy die casting forming process described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent optimization method for the aluminum alloy die casting forming process according to any one of claims 1 to 7 are implemented.

Images