A method for shaping a precision open aluminum material

Abstract

The present application provides a kind of precision opening aluminum material shaping method, belong to aluminum material processing technical field, the present application uses response surface parameter optimizer to adjust sand blasting parameter, material is placed in gradual transition fillet shaping die, start ultrasonic vibration auxiliary system to execute multi-step progressive shaping, in the shaping process, stress wave signal is collected by acoustic emission signal crack monitor and feature parameter is extracted, input crack recognition classifier, when determining that there is crack initiation risk, call rebound compensation prediction model based on artificial intelligence to calculate elastic rebound compensation value according to yield strength value, elastic modulus value, thickness value and shaping force initial value of current batch material and correct shaping force, after shaping, low temperature annealing treatment is carried out, the technical problem that the size precision fluctuation caused by the rebound amount difficult to accurately predict due to batch difference of material during shaping of precision opening aluminum material is solved.

Description

Technical Field

[0001] This invention belongs to the field of aluminum processing technology, and more specifically, relates to a method for shaping precision-opened aluminum materials. Background Technology

[0002] Precision-formed aluminum profiles are widely used in aerospace, electronics and communications, and precision instruments. Traditional forming processes typically employ fixed-parameter mold pressing combined with surface pretreatment and subsequent heat treatment to achieve dimensional accuracy. However, in existing technologies, due to fluctuations in the yield strength, elastic modulus, and other mechanical properties of different batches of aluminum, using uniform forming force parameters leads to significant differences in elastic springback, making it difficult to guarantee dimensional accuracy after forming. Current methods mainly rely on empirical formulas to estimate springback or repeated mold trials to adjust parameters. However, empirical formulas cannot capture the nonlinear effects of material properties, and mold trials are inefficient and still suffer from batch-to-batch consistency issues. In other words, existing technologies present a technical problem: the springback amount is difficult to predict accurately during the forming process of precision-formed aluminum profiles due to batch-to-batch material variations, resulting in dimensional accuracy fluctuations. Summary of the Invention

[0003] In view of this, the present invention provides a method for shaping precision-opening aluminum materials, which can solve the technical problem in the prior art where the springback amount of precision-opening aluminum materials is difficult to predict accurately due to batch differences in materials during the shaping process, resulting in fluctuations in dimensional accuracy.

[0004] This invention is implemented as follows: A method for shaping precision-opening aluminum profiles includes the following steps: tearing the precision-opening aluminum profile and then grinding off the burrs to remove any residue from the surface; sandblasting the precision-opening aluminum profile with diamond under air pressure, adjusting the sandblasting angle and moving speed using a response surface parameter optimizer; placing the sandblasted precision-opening aluminum profile in a gradually transitioning rounded corner shaping mold, activating an ultrasonic vibration-assisted system, and performing multi-step progressive shaping; during the multi-step progressive shaping process, collecting stress wave signals from the precision-opening aluminum profile using an acoustic emission signal crack monitor, extracting stress wave signal characteristic parameters, and inputting them into a crack identification classifier; when the crack identification classifier determines that there is a risk of crack initiation, calling a springback compensation prediction model to calculate the elastic springback compensation value of the current batch of material and correcting the shaping force; after the multi-step progressive shaping is completed, performing low-temperature annealing on the precision-opening aluminum profile to eliminate residual tensile stress, and then sending it to an oxidation process to complete surface treatment.

[0005] The response surface parameter optimizer is used to establish a mathematical mapping relationship between multiple sandblasting parameters and surface roughness and removal rate. It fits the interaction of parameters through a second-order response surface model and uses a genetic algorithm to globally optimize and obtain the optimal parameter combination of sandblasting angle and moving speed.

[0006] The inputs of the response surface parameter optimizer include sandblasting air pressure, sand particle size, initial spray angle, and initial moving speed. The outputs are the optimized sandblasting angle and optimized moving speed, which are used to adjust the sandblasting equipment.

[0007] The gradient transition rounded corner shaping mold refers to a mold with a rounded corner structure that gradually transitions in curvature radius at the corner of the opening. The mold structure is obtained through a topology optimization-driven variable density design method.

[0008] The gradual transition rounded corner shaping mold establishes a multi-objective optimization problem of mold structure stiffness and mass based on the SIMP material interpolation model. The moving asymptote method is used to iteratively update the element density distribution, minimize the mold mass and optimize the stiffness distribution of the contact area under the condition of satisfying the strength constraint.

[0009] The ultrasonic vibration-assisted system refers to installing an ultrasonic vibration generator on the gradually transitioning rounded corner forming mold, which reduces the friction coefficient between the precision opening aluminum material and the gradually transitioning rounded corner forming mold through high-frequency micro-amplitude vibration and promotes dislocation movement.

[0010] The multi-step progressive shaping refers to breaking down a single shaping process into progressive steps, with the shaping amount decreasing in each step, thereby avoiding sudden stress changes by gradually accumulating deformation.

[0011] The acoustic emission signal crack monitor is used to collect stress wave signals generated in real time during the multi-step progressive shaping process of precision-opened aluminum material. The mechanical waves are converted into electrical signals and amplified by a piezoelectric ceramic sensor.

[0012] The stress wave signal characteristic parameters are obtained by decomposing the stress wave signal into multiple frequency bands through wavelet packet transform, extracting the energy ratio, peak signal amplitude and duration of each frequency band as stress wave signal characteristic parameters, and combining Kalman filtering to eliminate environmental noise interference.

[0013] The crack identification classifier uses a support vector machine algorithm to construct a binary classification model. The stress wave signal feature parameters are input into the trained support vector machine classifier, and the result of the judgment of normal deformation state or crack initiation state is output.

[0014] The springback compensation prediction model has a feedforward architecture consisting of an input layer, three hidden layers, and an output layer. The input layer receives the material yield strength value, elastic modulus value, aluminum thickness value before shaping, and initial shaping force value. The output layer outputs the elastic springback compensation value.

[0015] In the rebound compensation prediction model, the first hidden layer uses a modified linear unit activation function, the second hidden layer embeds an adaptive gating unit, and the third hidden layer is directly connected to the input layer through residual connections to preserve the original feature information.

[0016] The gating parameters in the adaptive gating unit are dynamically adjusted through a gating weight function. The gating weight function calculates a balance value based on the yield strength value, elastic modulus value, and initial forming force value of the current batch of materials. The gating opening degree is set according to different weight adjustment functions based on the balance value.

[0017] The rebound compensation prediction model uses mean squared error as the loss function, an adaptive moment estimation optimization algorithm to update network weight parameters, a cosine annealing strategy to dynamically adjust the learning rate, and an early stopping mechanism to terminate training when the validation set loss does not decrease continuously.

[0018] In the rebound compensation prediction model, the neuron activation threshold of the second hidden layer is dynamically set through a threshold adjustment function. The threshold adjustment function calculates the adjustment coefficient based on the material yield strength, the aluminum thickness before shaping, and the initial value of the shaping force, and sets different neuron activation thresholds according to the adjustment coefficient.

[0019] The low-temperature annealing process refers to placing the precision-opened aluminum material, which has undergone multi-step progressive shaping, in a temperature environment for heat preservation, so that the residual tensile stress inside the precision-opened aluminum material can be released through atomic diffusion and dislocation rearrangement.

[0020] This invention constructs a springback compensation prediction model that takes the material's yield strength, elastic modulus, aluminum thickness before forming, and initial forming force as inputs. It utilizes the multi-layer nonlinear mapping capability of a feedforward neural network to learn the complex relationship between material performance parameters and springback. An adaptive gating unit is embedded in the second hidden layer to dynamically adjust the information transmission weights based on the characteristics of the current batch of materials, enabling the model to adaptively output precise elastic springback compensation values ​​for materials with different mechanical properties. Combined with a crack monitor that collects stress wave signals in real time and uses a crack identification classifier to determine the risk of crack initiation, the springback compensation prediction model is immediately invoked when a risk is detected to calculate the compensation amount for the current batch and correct the forming force, achieving dynamic parameter adjustment based on the actual material properties. In summary, this invention solves the technical problem mentioned in the background art where the difficulty in accurately predicting the springback during the forming process of precision-opened aluminum materials due to batch-to-batch material differences causes dimensional accuracy fluctuations. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 Convergence curve of the genetic algorithm for response surface parameter optimization.

[0023] Figure 3 This is a diagram showing the energy distribution of stress wave signals across different frequency bands.

[0024] Figure 4 The training loss curve for the rebound compensation prediction model.

[0025] Figure 5 The diagram shows the iterative solution process of the two-layer game optimization model.

[0026] Figure 6 This is a diagram showing the distribution of the opening width deviation after shaping. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0028] like Figure 1 The diagram shown is a flowchart of a precision-shaped aluminum profile with openings provided by this invention. The method includes the following steps:

[0029] S01. After tearing the precision-opened aluminum material, grind off the burrs and remove all residue from the surface of the precision-opened aluminum material.

[0030] S02. Precision-opening aluminum materials are sandblasted using 120-mesh diamond abrasive at 0.3MPa air pressure, and the sandblasting angle and moving speed are adjusted by a response surface parameter optimizer.

[0031] S03. Place the sandblasted precision-opened aluminum material into the gradient transition rounded corner shaping mold, start the ultrasonic vibration auxiliary system and perform multi-step progressive shaping;

[0032] S04. During the multi-step progressive shaping process, the stress wave signal of the precision-opened aluminum material is collected by the acoustic emission signal crack monitor, the characteristic parameters of the stress wave signal are extracted and input into the crack identification classifier.

[0033] S05. When the crack identification classifier determines that there is a risk of crack initiation, the springback compensation prediction model is called to calculate the elastic springback compensation value of the current batch of materials and correct the forming force.

[0034] S06. After the multi-step progressive shaping is completed, the precision-opened aluminum material is subjected to low-temperature annealing to eliminate residual tensile stress, and then sent to the oxidation process to complete the surface treatment.

[0035] The response surface parameter optimizer is used to establish a mathematical mapping relationship between multiple sandblasting parameters and surface roughness and removal rate. Inputs include sandblasting air pressure, sand particle size, initial spray angle, and initial moving speed. The optimizer fits the parameter interactions through a second-order response surface model and uses a genetic algorithm for global optimization to obtain the optimal parameter combination. The outputs are the optimized sandblasting angle and optimized moving speed, thus improving surface roughness control accuracy and ensuring removal efficiency. The sandblasting air pressure value is 0.3. The sand particle size is 120 mesh, the initial spray angle is 45 to 60 degrees, and the initial moving speed is 50. Up to 100 The optimized sandblasting angle is used to adjust the nozzle angle of the sandblasting equipment in step S02, and the optimized moving speed is used to adjust the moving mechanism speed of the sandblasting equipment in step S02.

[0036] The aforementioned gradient transition rounded corner shaping mold refers to a mold designed with a curvature radius starting from 0.5 at the corner of the opening. Gradually transition to 2 The rounded corner mold structure is obtained through a topology optimization-driven variable density design method. Based on the SIMP material interpolation model, a multi-objective optimization problem of mold structure stiffness and mass is established. The gradient of design variables is calculated through sensitivity analysis, and the unit density distribution is iteratively updated using the moving asymptote method. Under the condition of satisfying strength constraints, the mold mass is minimized and the stiffness distribution of the contact area is optimized, which improves the pressure uniformity of the multi-step progressive forming process and reduces the mold inertia. Finally, a non-homogeneous structure mold design scheme with both lightweight and high performance is obtained, which reduces the stress concentration factor from 3 to 5 times to less than 1.5 times.

[0037] The ultrasonic vibration-assisted system refers to the system installed on the gradually transitioning rounded corner shaping mold with a frequency of 20. Up to 40 An ultrasonic vibration generator with amplitude controlled at 5 Up to 15 Within the range, high-frequency micro-amplitude vibration reduces the friction coefficient between the precision-opened aluminum material and the gradually transitioning rounded corner forming mold and promotes dislocation movement, thereby dispersing local stress peaks and reducing resistance to plastic deformation.

[0038] The multi-step progressive shaping refers to breaking down a single shaping process into 3 to 5 progressive steps, with the shaping amount decreasing in each step. The shaping amount in the first step accounts for 40% to 50% of the total deformation amount, and the shaping amount in each subsequent step decreases in turn. By gradually accumulating deformation, stress abrupt changes are avoided, making the stress distribution at the opening edge more uniform.

[0039] The acoustic emission signal crack monitor is used to collect stress wave signals generated in real time during the multi-step progressive shaping process of precision-opened aluminum profiles, with a sampling frequency set to 500 kHz. Up to 1 The sensor is positioned near the edge of the opening. up to 8 The position is determined by converting mechanical waves into electrical signals using a piezoelectric ceramic sensor, which then amplifies the signals.

[0040] The stress wave signal feature parameters are decomposed into 8 frequency bands through wavelet packet transform. The energy ratio, peak value and duration of each frequency band are extracted as stress wave signal feature parameters. Combined with Kalman filtering to eliminate environmental noise interference, the stress wave signal feature parameters are used as input to the crack identification classifier to determine crack initiation.

[0041] The crack identification classifier uses a support vector machine algorithm to construct a binary classification model. The stress wave signal feature parameters are input into the trained support vector machine classifier, and the output is the judgment result of normal deformation state or crack initiation state. This enables millisecond-level online detection and localization of cracks. When a crack feature signal is detected, the forming force is immediately adjusted to prevent expansion, reducing the crack defect rate to below 0.5% and eliminating the risk of subsequent batch scrapping.

[0042] The springback compensation prediction model has a feedforward architecture comprising an input layer, three hidden layers, and an output layer. The input layer receives four parameters: the material yield strength, the elastic modulus, the aluminum thickness before shaping, and the initial shaping force. The first hidden layer has 128 neurons and uses a modified linear unit activation function. The second hidden layer has 64 neurons and embeds an adaptive gating unit. The third hidden layer has 32 neurons. The output layer outputs the elastic springback compensation value. The input layer and the third hidden layer are directly connected through residual connections to retain the original feature information. The material yield strength and the elastic modulus are measured by tensile testing. The aluminum thickness before shaping is obtained by micrometer measurement. The initial shaping force is preset according to the specifications of the precision-opened aluminum material. The elastic springback compensation value is used to correct the shaping force in step S05.

[0043] The steps for establishing the training dataset for the springback compensation prediction model include: collecting material mechanical property test data of different batches of precision-opened aluminum materials, recording the yield strength, elastic modulus, and Poisson's ratio of each batch; conducting shaping tests on each batch of materials, recording the aluminum material thickness before shaping, the initial shaping force, and the actual dimensions after shaping; measuring the key dimensions of the precision-opened aluminum material after shaping using a coordinate measuring machine, and calculating the actual springback value as label data; using the material yield strength, elastic modulus, aluminum material thickness before shaping, and initial shaping force as input features, and the corresponding actual springback value as output label, constructing a training dataset containing 500 to 800 samples, and dividing it into a training set and a validation set in a 7:3 ratio.

[0044] The steps for training the rebound compensation prediction model include: using mean squared error as the loss function to measure the deviation between the predicted rebound amount and the actual rebound amount; using an adaptive moment estimation optimization algorithm to update the network weight parameters; setting the initial learning rate to 0.001 and dynamically adjusting it using a cosine annealing strategy; setting the batch size to 32; and setting the training epochs to 200 epochs; evaluating the model performance using a validation set after each training epoch; triggering an early stopping mechanism to terminate training when the validation set loss does not decrease for 10 consecutive epochs; and saving the model parameters corresponding to the minimum loss value as the final rebound compensation prediction model after training.

[0045] The gating parameters in the adaptive gating unit are dynamically adjusted through a gating weight function. This function calculates an equilibrium value based on the yield strength, elastic modulus, and initial forming force of the current batch of material. The equilibrium value is expressed as the ratio of the yield strength to the standard yield strength multiplied by the elastic modulus normalization weight factor, plus the ratio of the initial forming force to the rated forming force multiplied by the forming force normalization weight factor. The standard yield strength is 200. The rated shaping force is 5000. The normalization weighting factor for the elastic modulus is 0.6, and the normalization weighting factor for the shaping force is 0.4. When the equilibrium value is in the range [0.8, 1.0), a linear weighting adjustment function is used, and the gate opening is set to 0.7 to 0.9. When the equilibrium value is in the range [1.0, 1.2], a constant weighting adjustment function is used, and the gate opening is set to 0.5. When the equilibrium value is greater than 1.2, an exponentially decaying weighting adjustment function is used, and the gate opening decreases to below 0.3 as the equilibrium value increases. The yield strength value of the current batch of material is obtained from step S05, the elastic modulus value of the current batch is obtained from step S05, and the initial value of the shaping force of the current batch is obtained from step S05.

[0046] Furthermore, it may also include: introducing a two-layer game optimization model in step S05 to coordinate the shaping force adjustment strategy. The upper-layer model aims to minimize the dimensional deviation after shaping. The objective function inputs include the shaping force adjustment amount, the elastic rebound compensation value, and the mold contact stiffness value. The output is the predicted dimensional deviation value. The constraints are that the shaping force adjustment amount does not exceed 20% of the initial shaping force value and the elastic rebound compensation value is within 0.02. Up to 0.15 Within the specified range; the lower-level model aims to minimize the crack initiation risk index. The objective function inputs include the stress concentration factor, the shaping force adjustment, and the ultrasonic vibration amplitude, and the output is the crack risk assessment value. The constraints are that the stress concentration factor is less than 2.0 and the ultrasonic vibration amplitude is within 5. Up to 15 Within the scope; the two objective functions achieve game equilibrium through the shaping force adjustment amount as a coupling term. When the upper-level model increases the shaping force adjustment amount to reduce the predicted value of the size deviation, it will cause the crack risk assessment value of the lower-level model to increase. The optimal shaping force adjustment strategy is obtained by solving through Nash equilibrium. The mold contact stiffness value is obtained by finite element simulation calculation. The stress concentration coefficient is determined by the design parameters of the gradually transition rounded corner shaping mold. The ultrasonic vibration amplitude is output by the ultrasonic vibration auxiliary system. The optimal shaping force adjustment strategy is used to correct the shaping force in step S05.

[0047] In the described two-layer game optimization model, there is a correlation mechanism between the forming force adjustment coupling term and the elastic rebound compensation value output by the springback compensation prediction model. The correlation principle is that the elastic rebound compensation value is one of the input parameters of the objective function of the upper-layer model. When the elastic rebound compensation value is large, the upper-layer model tends to increase the forming force adjustment to overcome the springback effect. However, the increase in the forming force adjustment will be transmitted to the lower-layer model through the coupling term, leading to increased stress concentration. The lower-layer model feeds back the signal of rising crack risk assessment value, which constrains the forming force adjustment range of the upper-layer model. The two-layer model finds a balance solution that satisfies both dimensional accuracy requirements and controls crack risk through iterative game. The correlation mechanism deeply integrates the intelligent prediction capability of the springback compensation prediction model with the multi-objective coordination capability of the two-layer game optimization model, avoiding the problem of simply pursuing dimensional accuracy leading to an increase in crack defect rate or excessively conservatively controlling crack risk at the expense of dimensional accuracy. Through bidirectional feedback of the model output parameters, the forming quality and structural integrity are synergistically improved, enabling the multi-step progressive forming process to adapt to the differences in material characteristics of different batches and maintain stable comprehensive performance output during dynamic adjustment.

[0048] The low-temperature annealing process refers to placing the precision-opened aluminum material, after its multi-step shaping process, in an environment with a temperature of 150°C to 200°C for 30 minutes. Up to 60 This process allows the residual tensile stress inside the precision-opened aluminum material to be released through atomic diffusion and dislocation rearrangement. After annealing, the residual stress value on the surface of the precision-opened aluminum material is reduced by more than 60%, effectively suppressing the tendency of crack propagation during subsequent oxidation.

[0049] In the rebound compensation prediction model, the neuron activation threshold of the second hidden layer is dynamically set through a threshold adjustment function. This function calculates an adjustment coefficient based on the material yield strength, the aluminum thickness before shaping, and the initial shaping force of the current input sample. The adjustment coefficient is expressed as the square root of the ratio of the material yield strength to the reference yield strength multiplied by a thickness normalization coefficient, plus the logarithm of the ratio of the initial shaping force to the baseline shaping force multiplied by a force normalization coefficient. The reference yield strength is 180... The reference shaping force value is 4500. The thickness normalization coefficient is set to 0.5, and the force normalization coefficient is set to 0.3. When the adjustment coefficient is in the range of [0.5, 0.8), the neuron activation threshold is set to 0.3 to 0.5. When the adjustment coefficient is in the range of [0.8, 1.2], the neuron activation threshold is set to 0.2. When the adjustment coefficient is greater than 1.2, the neuron activation threshold is set to 0.1 to 0.15, so that the rebound compensation prediction model has higher sensitivity to high-strength materials or large deformation conditions.

[0050] The specific implementation methods of the above steps are described in detail below.

[0051] The specific implementation of step S01 is to first perform a tearing process on the precision-cut aluminum material to form an opening structure. The tearing process is carried out by mechanical shearing to create a cut at a predetermined position on the aluminum material. The cut depth is determined according to the thickness of the aluminum material, and the thickness is less than 3 mm. The incision depth should be 80% to 90% of the thickness, and the thickness should be greater than 3. The cutting depth should be 70% to 85% of the thickness. After tearing, use an angle grinder or sandpaper to polish the burrs on the cut edges. During polishing, maintain an angle of 30 to 45 degrees between the polishing tool and the aluminum surface, and control the polishing speed at 800°. Up to 1200 After polishing, use compressed air to blow away any residue from the surface, setting the air pressure to 0.2. Up to 0.3 The blowing time lasts for 10 to 15 seconds to ensure surface cleanliness. The purpose of tearing and deburring is to provide a uniform base surface for subsequent sandblasting and to avoid burrs affecting the sandblasting effect.

[0052] The specific implementation of step S02 involves setting the input parameter of the response surface parameter optimizer to a sandblasting air pressure value of 0.3. The sand particle size is 120 mesh, the initial spray angle is 45 to 60 degrees, and the initial moving speed is 50. Up to 100 The response surface parameter optimizer constructs a three-factor, three-level response surface model based on the Box-Behnken experimental design method. Experimental data is obtained by testing surface roughness and removal rate under different parameter combinations. A second-order polynomial response surface function is fitted using the least squares method to establish the mathematical relationship between parameters and response values. This second-order polynomial response surface function includes linear, interaction, and quadratic terms to capture the nonlinear interactions between parameters. After the response surface model is established, a genetic algorithm is called for global optimization. The genetic algorithm population size is set to 100 individuals, the number of iterations is set to 50 generations, the crossover probability is set to 0.8, and the mutation probability is set to 0.05. The fitness function comprehensively considers the two objectives of minimizing surface roughness and maximizing removal rate. A weighted summation method is used to transform the multi-objective problem into a single-objective optimization problem. The weight coefficients are set according to actual process requirements: surface roughness weight 0.6 and removal rate weight 0.4. After the genetic algorithm iteration is completed, the optimized blasting angle and optimized moving speed are output. The optimized blasting angle is typically between 52 and 58 degrees, and the optimized moving speed is typically between 65 degrees. Up to 85 In this context, the role of the response surface parameter optimizer is to avoid the problem that single-factor adjustments cannot achieve the optimal balance between surface quality and processing efficiency through multi-parameter coupling optimization.

[0053] The specific implementation of step S03 involves placing the sandblasted precision-opening aluminum material onto the lower die surface of the gradient transition rounded corner forming mold according to the positioning reference. The position of the precision-opening aluminum material is fixed by positioning pins and pressure plates to prevent displacement during the forming process. The ultrasonic vibration auxiliary system is then activated to cause the gradient transition rounded corner forming mold to generate 20... Up to 40 The high-frequency vibration is controlled at an amplitude of 5 by adjusting the power of the ultrasonic generator. Up to 15 Within this range, the introduction of ultrasonic vibration is based on the principle of acousto-plastic effect to reduce the yield strength and flow stress of precision-opening aluminum materials. The acousto-plastic effect promotes dislocation disengagement and dislocation slip through ultrasonic energy, reducing the resistance to plastic deformation. The multi-step progressive shaping process is divided into 3 to 5 progressive steps executed sequentially. In the first step of shaping, the upper die pressing amount is set to 40% to 50% of the total deformation amount, and the holding time is set to 3 to 5 seconds before returning. In the second step of shaping, the upper die pressing amount is set to 50% to 60% of the remaining deformation amount, and the holding time is set to 2 to 4 seconds before returning. The shaping amount in each subsequent step is reduced sequentially until the design size requirements are met. The interval time between each shaping step is set to 5 to 10 seconds to allow partial release of internal stress in the material. The purpose of the multi-step progressive shaping is to avoid excessive stress concentration at the opening edge due to a single large deformation, which could lead to the initiation of microcracks, by gradually accumulating deformation.

[0054] The specific implementation of step S04 is that during the multi-step progressive shaping process, the acoustic emission signal crack monitor uses a 500... Up to 1 The stress wave signal is continuously acquired at a sampling frequency. The piezoelectric ceramic sensor converts the mechanical wave into a voltage signal, which is then amplified by a preamplifier by 40°. Up to 60 The amplified stress wave signal is converted into a digital signal by an analog-to-digital converter and stored in a buffer. The stress wave signal in the buffer is decomposed by wavelet packet transform, and the Daubechies wavelet is selected as the mother wavelet function. The decomposition layer is set to 3 layers to obtain sub-signals of 8 frequency bands. The energy proportion of each frequency band sub-signal is calculated as the first type of feature parameter. The peak amplitude of the stress wave signal is extracted as the second type of feature parameter. The duration of the stress wave signal exceeding the threshold voltage is extracted as the third type of feature parameter. The threshold voltage is set to 3 to 5 times the root mean square value of the background noise voltage. The three types of feature parameters are combined to form a stress wave signal feature parameter vector. The stress wave signal feature parameter vector is processed by Kalman filtering before being input into the crack identification classifier to eliminate environmental noise and electromagnetic interference. The system state equation of the Kalman filter describes the time evolution of the feature parameters, and the observation equation describes the relationship between the sensor measurement value and the true feature parameters. The optimal value of the feature parameters is estimated by recursion to reduce measurement uncertainty. The combined use of wavelet packet transform and Kalman filtering can effectively extract the feature signal of crack initiation and suppress the influence of interference signals.

[0055] The specific implementation of step S05 is as follows: After the crack identification classifier receives the feature parameter vector of the stress wave signal, it performs binary classification using the support vector machine algorithm. The support vector machine algorithm selects the radial basis function as the kernel function to map the feature parameter vector to a high-dimensional space. The kernel function parameters are optimized and determined through a grid search method, and the penalty coefficient is set between 10 and 100. The support vector machine classifier outputs the judgment result of normal deformation state or crack initiation state. When the judgment result is crack initiation state, the springback compensation prediction model calling process is triggered. The input layer of the springback compensation prediction model receives the material yield strength value, elastic modulus value, and other parameters of the current batch of materials. The initial value of the aluminum material thickness and the initial value of the forming force before shaping are determined by the following methods: the yield strength and elastic modulus of the material are obtained through tensile testing according to national standard testing methods; the initial value of the aluminum material thickness before shaping is obtained by averaging measurements taken at three different locations on the aluminum material using a micrometer; and the initial value of the forming force is determined by referring to a table based on the specifications and material grade of the precision-opening aluminum material. The input parameters are processed by 128 neurons in the first hidden layer and then output through a modified linear unit activation function. This modified linear unit activation function sets negative outputs to zero and retains positive outputs to introduce nonlinear characteristics. The adaptive gating units embedded in the 64 neurons of the second hidden layer are calculated based on the gating weight function. The calculated balance value dynamically adjusts the gating opening degree, which controls the information flow to achieve adaptive response to different material properties. The 32 neurons in the third hidden layer receive the original feature information from the input layer through residual connections to avoid the gradient vanishing problem in deep networks. After the output layer outputs the elastic rebound compensation value, it calls a two-layer game optimization model to calculate the optimal shaping force adjustment strategy. The upper model of the two-layer game optimization model uses the elastic rebound compensation value, mold contact stiffness value, and shaping force adjustment value as inputs to calculate the predicted value of dimensional deviation. The lower model uses the stress concentration factor, shaping force adjustment value, and ultrasonic vibration amplitude as inputs to calculate the crack risk assessment value. The two-layer model is solved iteratively through Nash equilibrium. During the iteration process, the upper-layer model adjusts the forming force adjustment amount based on the crack risk assessment value fed back by the lower-layer model, and the lower-layer model updates the stress distribution state based on the forming force adjustment amount transmitted by the upper-layer model. The iteration terminates when the change in the forming force adjustment amount is less than 0.5% in two consecutive iterations or the number of iterations reaches 20. After the iteration is completed, the optimal forming force adjustment strategy output is used to correct the forming force in the current step. The corrected forming force is applied to the hydraulic system of the gradually transitioning rounded corner forming mold to achieve precise force control. The introduction of the two-layer game optimization model realizes multi-objective collaborative optimization of dimensional accuracy and crack control.

[0056] The specific implementation of step S06 involves removing the precision-opening aluminum material from the gradient transition rounded corner forming mold after the multi-step progressive shaping is completed, and placing it in a resistance heating furnace or box annealing furnace for low-temperature annealing. The annealing temperature is set to 150℃ to 200℃, the temperature control accuracy is ±5℃, and the holding time is set to 30 minutes. Up to 60 The heat preservation time is determined based on the wall thickness of the precision-opening aluminum material; for a wall thickness less than 2 mm... Keep warm for 30 minutes. Up to 40 Wall thickness greater than 2 The insulation time is set at 45 minutes. Up to 60 During the low-temperature annealing process, the residual tensile stress inside the precision-opening aluminum material is released through thermally activated atomic diffusion and dislocation climb. The low-temperature annealing temperature is lower than the recrystallization temperature of the aluminum material to avoid grain growth that could lead to a decrease in mechanical properties. After annealing, the material is cooled to room temperature by furnace cooling at a rate controlled between 20°C and 40°C / h to avoid rapid cooling that could generate new thermal stress. The residual stress value on the surface of the annealed precision-opening aluminum material was measured by X-ray diffraction, verifying a reduction of over 60%. The precision-opening aluminum material after low-temperature annealing is then sent to an anodizing bath for anodic oxidation. The anodic solution used is a sulfuric acid solution with a concentration of 180%. Up to 220 The oxidation temperature was controlled between 18℃ and 22℃, and the current density was set to 1.2. Up to 1.8 The oxidation time is determined according to the required oxide film thickness. The purpose of the low-temperature annealing treatment is to eliminate residual tensile stress and avoid stress corrosion that accelerates crack propagation during subsequent oxidation.

[0057] It should be noted that the key technical ideas of this invention include the following aspects. The first key technical idea is to use a response surface methodology (RSM) combined with a genetic algorithm to perform coupled optimization of multiple parameters in the sandblasting process. Compared with traditional single-factor experimental optimization methods, the RSM can establish a complete mathematical model between multiple parameters and surface quality indicators and capture parameter interactions. The global optimization capability of the genetic algorithm avoids getting trapped in local optima, enabling the sandblasting process to achieve the best balance between surface roughness control and removal efficiency, providing a uniform surface state foundation for subsequent shaping processes. The second key technical idea is to design a gradually transitioning rounded corner shaping mold and implement multi-step progressive shaping in conjunction with an ultrasonic vibration-assisted system. Compared with the traditional single-step shaping method, the gradually transitioning rounded corner structure disperses stress concentration areas through continuous changes in the radius of curvature. The multi-step progressive shaping strategy decomposes large deformations into several small deformations that are gradually accumulated to avoid stress abrupt changes. Ultrasonic vibration reduces material flow stress through the acoustic-plastic effect. The synergistic effect of these three factors significantly reduces the stress concentration coefficient at the opening edge and effectively inhibits the initiation of microcracks. The third key technical approach is to construct a joint decision-making mechanism between a springback compensation prediction model and a two-layer game optimization model. Compared with traditional empirical compensation methods, the springback compensation prediction model automatically extracts the complex nonlinear relationship between material properties and springback amount through a deep learning architecture. The two-layer game optimization model decouples the dimensional accuracy target and the crack control target into upper and lower layer models and solves the optimal compromise solution through Nash equilibrium. The two models achieve bidirectional information transmission through the coupling term of the forming force adjustment, so that the forming process can simultaneously meet the dimensional requirements and structural integrity requirements during dynamic adjustment. The synergistic effect of the above three key technical approaches is reflected in the optimization of the entire process chain. The optimization of sandblasting parameters ensures the consistency of surface quality and provides a stable foundation for shaping. The gradient rounded corner mold, combined with multi-step shaping and ultrasonic vibration, reduces stress concentration and inhibits crack initiation from the source. The joint decision-making of the intelligent prediction model and the game optimization model achieves a multi-objective balance between elastic rebound compensation and crack risk control. The three form a complete technical system from surface pretreatment to deformation control and intelligent decision-making. Compared with the independent optimization of each process in the existing technology, the collaborative optimization strategy of this invention achieves an overall improvement in shaping quality through parameter transfer and feedback adjustment between processes.

[0058] It should be noted that this invention also solves the following technical problem: stress concentration at the corner of the opening in precision-opened aluminum material during multi-step progressive forming process, leading to crack initiation. Traditional forming molds use a rounded corner structure with a fixed radius of curvature. During multi-step cumulative deformation, the stress concentration factor at the corner can reach 3 to 5 times, exceeding the tensile strength limit of the material, causing rapid crack propagation and product scrapping. This invention designs a gradually transitioning rounded corner forming mold with a radius of curvature gradually transitioning from 0.5mm to 2mm. It utilizes a topology optimization-driven variable density design method based on a SIMP material interpolation model to establish a multi-objective optimization problem of stiffness and mass. It uses a moving asymptote method to iteratively update the element density distribution to obtain a non-homogeneous structure mold, optimizing the stiffness distribution in the contact area and reducing the stress concentration factor to below 1.5 times. Combined with an ultrasonic vibration-assisted system, it generates 20... Up to 40 The low-frequency micro-amplitude vibration reduces the friction coefficient and promotes dislocation motion to disperse local stress peaks. At the same time, the stress wave signal is collected in real time by the acoustic emission signal crack monitor. The feature parameters are extracted by wavelet packet transformation and input into the support vector machine crack recognition classifier to achieve millisecond-level online detection. Once the risk of crack initiation is determined, the shaping force is adjusted immediately to avoid propagation, thereby effectively suppressing the generation of cracks at corner positions.

[0059] Specifically, the principle of this invention is as follows: The invention solves this technical problem by establishing an intelligent mapping mechanism between material mechanical property parameters and elastic rebound. Traditional empirical formulas only consider a single material parameter or assume linear relationships, failing to accurately describe the rebound behavior under the coupled effects of multiple parameters such as yield strength, elastic modulus, thickness, and forming force. This invention employs a three-layer hidden deep neural network architecture. Through multi-layer nonlinear transformations, it captures the interaction effects between parameters. The first hidden layer extracts basic features; the adaptive gating unit in the second hidden layer calculates the gating opening degree based on the material's yield strength and elastic modulus values ​​to achieve selective feature transfer; and the third hidden layer integrates deep features, with residual connections preserving the original input information to prevent gradient vanishing. During training, 500 to 800 sets of measured data are used to establish the input-output mapping relationship, enabling the model to learn the influence of batch-to-batch material differences on rebound. In application, the elastic rebound compensation value can be predicted based on the measured mechanical property parameters of the current batch, and the forming force can be corrected in real time, thereby offsetting the impact of material property fluctuations on dimensional accuracy and achieving stable forming quality between batches.

[0060] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0061] The specific implementation methods of steps S01, S03, and S06 are the same as those described above, and will not be repeated in detail here.

[0062] The specific implementation of step S02 is to use 120-mesh diamond abrasive at a depth of 0.3 mm. Precision-aperture aluminum profiles are sandblasted under air pressure, with the sandblasting angle and moving speed adjusted using a response surface parameter optimizer. The response surface parameter optimizer establishes a mathematical mapping relationship between multiple sandblasting parameters and surface roughness and removal rate. The formula for the second-order response surface model is as follows:

[0063] middle, Surface roughness, unit: ; The spray angle is expressed in degrees. For reference spray angle, the value is 60 degrees; Movement speed, unit: ; The reference movement speed is set to 100. ; The regression coefficients for the response surface model were obtained by fitting the experimental data using the least squares method, where... Units are All other coefficients are dimensionless. The formula for the removal rate response surface model is as follows:

[0064]

[0065] In the formula, Removal rate, in units of ; The regression coefficients for the removal rate response surface model were obtained by fitting the experimental data using the least squares method, where... Units are All other coefficients are dimensionless. The fitness function formula for global optimization in the genetic algorithm is expressed as follows:

[0066] ;

[0067] In the formula, The fitness function value; The target surface roughness is set to 3. ; The target removal rate is set to 50. ; These are weighting coefficients, with empirical values ​​of 0.6 and 0.4, respectively.

[0068] The specific implementation of step S04 is as follows: During the multi-step progressive shaping process, stress wave signals of the precision-opened aluminum material are collected using an acoustic emission signal crack monitor. Stress wave signal feature parameters are extracted and input into a crack identification classifier. The stress wave signal feature parameters are extracted using wavelet packet transform, and the formula for the energy proportion of each frequency band is expressed as follows:

[0069] ;

[0070] In the formula, For the first Energy percentage of each frequency band; For the first The first frequency band Wavelet packet coefficients; For the first The total number of wavelet packet coefficients in each frequency band; For the first The total number of wavelet packet coefficients in each frequency band; The value can range from 1 to 8; The frequency band index variable takes values ​​from 1 to 8. The crack identification classifier uses a support vector machine algorithm to construct a binary classification model, and the decision function is expressed as follows:

[0071]

[0072] In the formula, Output value for the classification decision function; For the first Lagrange multipliers of the support vectors; For the first The class label of each training sample is 1 for normal deformation state and negative 1 for crack initiation state; Here is the radial basis function kernel, and its expression is: ,in The kernel function bandwidth parameter is determined through cross-validation. For the first Support vectors; The feature vector of the sample to be classified; For bias terms; This represents the total number of support vectors. It is a symbolic function.

[0073] The specific implementation of step S05 is as follows: when the crack identification classifier determines that there is a risk of crack initiation, the springback compensation prediction model is invoked to calculate the elastic springback compensation value of the current batch of materials and correct the forming force. The formula for the output layer elastic springback compensation value of the springback compensation prediction model is expressed as follows:

[0074]

[0075] In the formula, This is the elastic rebound compensation value, in units of... ; For the output layer Each weight parameter, in units of ; For the third hidden layer The output value of each neuron is dimensionless; This is the output layer bias term, in units of ; The value ranges from 1 to 32. The formula for the equilibrium value of the gate weight function in the adaptive gate control unit is as follows:

[0076] ;

[0077] In the formula, This is the equilibrium value; This represents the yield strength value of the current batch of materials, in units of... ; This is the standard yield strength value, taken as 200. ; This is the normalization weighting factor for the elastic modulus, with a value of 0.6; This is the initial value of the shaping force for the current batch, in units of... ; The rated forming force is 5000. ; The normalization weighting factor for the shaping force is set to 0.4. The linear weighting adjustment function expression for the gate opening degree across different equilibrium value ranges is as follows: ,when Applicable when the value is between 0.8 and 1.0, among which Let gating degree be the gate opening degree; the constant weight adjustment function expression is: ,when Applicable to values ​​between 1.0 and 1.2; the expression for the exponentially decaying weight adjustment function is: ,when Applicable when the value is greater than 1.2.

[0078] The objective function of the upper-level model in the two-level game optimization model is expressed as follows: ;

[0079] In the formula, This represents the predicted value of the size deviation of the upper-level model; These are the predicted dimensions after shaping, in units of... ; The target dimension value, in units of ; This refers to the adjustment amount of the shaping force, in units of... ; This represents the mold contact stiffness value, in units of... The objective function of the lower-level model is expressed as follows:

[0080] ;

[0081] In the formula, This is the crack risk assessment value for the lower-level model; The stress concentration factor; The standard stress concentration factor is 2.0. This is the upper limit of the shaping force adjustment, which is 20% of the initial shaping force value, in units of... ; The amplitude of ultrasonic vibration is expressed in units of 1000 m / s. ; This is the upper limit of the ultrasonic vibration amplitude, with a value of 15. .

[0082] The formula for the neuron activation threshold adjustment coefficient in the second hidden layer of the rebound compensation prediction model is as follows:

[0083] ;

[0084] In the formula, This is the adjustment coefficient; This represents the yield strength of the material, in units of... ; The reference yield strength value is 180. ; This is the thickness normalization factor, with a value of 0.5; This is the initial value of the shaping force, in units of ; The baseline shaping force value is set at 4500. ; The normalization coefficient is 0.3. The neuron activation threshold takes values ​​within different regulation coefficient ranges, when... When the value is between 0.5 and 0.8, the neuron activation threshold is... The calculation formula is ;when When the value is between 0.8 and 1.2, the neuron activation threshold is... ;when When it is greater than 1.2, the neuron activation threshold ,in This represents the neuron activation threshold.

[0085] To better understand and implement this invention, the following is a specific application scenario of this invention, Example 2:

[0086] A technical team received a batch of precision-cut aluminum profiles for aircraft seats. This batch consisted of 300 pieces of 6061-T6 aluminum alloy, with an opening width of 12mm, a wall thickness of 2.5mm, and a length of 850mm. Traditional forming methods often resulted in problems such as cracking at the opening edges, excessive dimensional springback, and uneven surface quality when processing these types of aluminum profiles, leading to a pass rate of only about 78%. The technical team decided to use the precision-cut aluminum profile forming method of this invention to solve these problems.

[0087] First, the technical team conducted material mechanical property tests on 300 precision-cut aluminum profiles. Using a universal testing machine, the yield strength of this batch of materials was measured to be 215 MPa, the elastic modulus to be 69 GPa, and the Poisson's ratio to be 0.33. Subsequently, all aluminum profiles were torn according to step S01, with the tear depth controlled at 0.8 mm. The burrs on the tear edges were polished with 800-grit sandpaper, and after polishing, anhydrous ethanol was used to remove surface residues and the surface was dried.

[0088] In step S02, the technical team set the input parameters of the response surface parameter optimizer to: blasting air pressure 0.3 MPa, sand particle size 120 mesh, initial blasting angle range of 45 to 60 degrees, and initial moving speed range of 50 mm / s to 100 mm / s. By collecting 25 sets of experimental data and fitting a second-order response surface model, the regression coefficients of the surface roughness response surface model were obtained. =3.2μm, =-1.8、 =1.4、 =0.9、 =0.7、 =-0.5, the regression coefficient of the removal rate response surface model. =48mg / min =12、 =-8、 =-3、 =2、 =1.5. For example... Figure 2 As shown, the genetic algorithm converged after 150 iterations of global optimization, obtaining an optimized sandblasting angle of 53 degrees and an optimized moving speed of 72 mm / s. At this time, the surface roughness was controlled at 2.8 μm, and the removal rate reached 52 mg / min.

[0089] In step S03, the technical team placed the sandblasted aluminum material into a gradually transitioning rounded corner shaping mold. This mold, through topology optimization design, transitions the rounded corner radius from 0.5mm at the opening to 2mm on the inner side. The mold mass is reduced by 32% compared to traditional homogeneous molds. The contact area stiffness, calculated by finite element simulation, is 8500 N / mm, and the stress concentration factor is reduced from 4.2 to 1.4. An ultrasonic vibration-assisted system is activated, with the ultrasonic generator frequency set to 28kHz and the amplitude controlled at 10μm. The multi-step progressive shaping is performed in four steps. The first step accounts for 45% of the total deformation, while the subsequent three steps account for 30%, 15%, and 10% respectively, with an 8-second interval between each step to release instantaneous stress.

[0090] In step S04, the sampling frequency of the acoustic emission signal crack monitor is set to 800kHz, and the sensor is positioned 5mm outside the edge of the opening. During the first shaping process, the acquired stress wave signal is decomposed into 8 frequency bands by wavelet packet transform, as shown in Table 1. The energy proportion of each frequency band is extracted as a characteristic parameter.

[0091] Table 1 Energy percentage of each frequency band of stress wave signal

[0092]

[0093] like Figure 3 As shown, ten feature parameters, including the extracted energy percentage, peak signal amplitude of 134 dB, and duration of 0.032 s, are input into the crack identification classifier. The radial basis function bandwidth parameter of the support vector machine classifier is also included. With the value set to 0.8, the trained classifier contains 87 support vectors. During the second shaping step of the 118th aluminum piece, the classifier output value is -0.92, indicating that a crack has begun to appear.

[0094] In step S05, the technical team immediately invoked the springback compensation prediction model, inputting the current batch of material's yield strength (215 MPa), elastic modulus (69 GPa), aluminum thickness before shaping (2.5 mm), and initial shaping force (4800 N). The springback compensation prediction model was trained with 700 sets of samples, the training process as follows: Figure 4 As shown, the validation set loss reaches its minimum value of 0.0089 mm in the 182nd round, triggering the early stopping mechanism. The model outputs a springback compensation value of 0.087 mm. The current batch balance value is calculated based on the balance value in the adaptive gating unit. =1.15, which is within the range of the constant weight adjustment function, and the gating opening degree is set to 0.5.

[0095] Subsequently, a two-layer game optimization model is activated to coordinate the forming force adjustment strategy. The upper-layer model's objective function inputs include an initial value of 480N for the forming force adjustment, a springback compensation value of 0.087mm, and a mold contact stiffness value of 8500N / mm. The lower-layer model's objective function inputs include a stress concentration factor of 1.4, an initial value of 480N for the forming force adjustment, and an ultrasonic vibration amplitude of 10μm. Figure 5 As shown, after solving the Nash equilibrium, the upper and lower models reach a game equilibrium in the 15th iteration, and the optimal shaping force adjustment is 385N, with a corresponding shaping force correction value of 5185N. At this time, the predicted value of the size deviation of the upper model is reduced to 0.012, and the crack risk assessment value of the lower model is controlled at 0.68.

[0096] Throughout the multi-step progressive reshaping process, the activation threshold adjustment coefficient of neurons in the second hidden layer... The calculated value was 0.92, which is within the constant threshold range, and the neuron activation threshold was set to 0.2. After the shaping force was corrected, the next two shaping steps were completed, and the output value of the crack recognition classifier recovered to 0.76, with no further risk of crack initiation detected.

[0097] In step S06, the technical team placed all the aluminum materials in a temperature-controlled furnace for low-temperature annealing. The annealing temperature was set to 175℃ and the holding time was 45 minutes. After annealing, the residual stress value on the surface of the aluminum material decreased from 182MPa to 65MPa, a reduction of 64%. Subsequently, it was sent to the oxidation process to complete the surface treatment, and the oxide film thickness was controlled within the range of 8μm to 12μm.

[0098] After the above process, the technical team conducted a full inspection of 300 precision-cut aluminum profiles, using a coordinate measuring machine to measure key dimensions, such as... Figure 6 As shown, the deviation of the opening width was controlled within ±0.025mm, and the deviation of the fillet radius was controlled within ±0.018mm, with an overall pass rate of 97.3%, which is 19.3 percentage points higher than that of the traditional method. The crack defect rate was reduced from 3.8% in the traditional method to 0.3%, and the average surface roughness was 2.6μm, with a significant improvement in surface quality uniformity.

[0099] The technological advancements of this invention compared to traditional shaping methods are mainly reflected in the following aspects. First, the response surface parameter optimizer establishes a mathematical mapping relationship between multiple sandblasting parameters and surface quality, achieving intelligent optimization of sandblasting process parameters. This avoids the surface quality instability problem caused by traditional empirical parameter tuning methods. Its principle lies in accurately describing parameter interactions using a second-order response surface model and globally searching for the optimal solution using a genetic algorithm. Second, the gradually transitioning rounded corner shaping mold uses a topology optimization design method to obtain a heterogeneous structure. By gradually dispersing stress concentration through the rounded corner curvature radius, it significantly reduces local stress peaks at the opening edge compared to traditional sharp-corner molds, fundamentally suppressing the driving force for crack initiation. Third, the high-frequency micro-amplitude vibration introduced by the ultrasonic vibration-assisted system alters the frictional characteristics between the aluminum material and the mold and promotes dislocation movement, reducing resistance to plastic deformation and making the stress distribution more uniform during the shaping process. This principle effectively solves the stress concentration problem caused by traditional static shaping. Furthermore, the acoustic emission signal crack monitor, combined with wavelet packet transform and support vector machine classifier, achieves millisecond-level online crack detection. Compared to traditional offline detection methods, it can immediately identify and trigger shaping force adjustment in the early stages of crack initiation, avoiding batch scrapping caused by crack propagation. The springback compensation prediction model learns the complex nonlinear relationship between material properties and springback amount through deep neural networks, and dynamically adjusts the model's sensitivity to different batches of materials using adaptive gating units, achieving accurate springback prediction and compensation. This solves the problem that traditional fixed compensation methods cannot adapt to batch differences in materials. The two-layer game optimization model treats dimensional accuracy and crack risk as mutually constraining dual objectives for collaborative optimization. By solving through Nash equilibrium, it avoids the unintended consequences of single-objective optimization, achieving an optimal balance between shaping quality and structural integrity. Finally, low-temperature annealing releases residual tensile stress through atomic diffusion and dislocation rearrangement mechanisms, effectively suppressing the tendency for crack propagation during oxidation, significantly improving product reliability compared to traditional non-annealing processes.

[0100] It should be noted that the variables involved in this invention are explained in detail in Table 2.

[0101] Table 2 Variable Explanation Table

[0102]

[0103] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for shaping precision-opened aluminum profiles, characterized in that, Includes the following steps: After tearing the precision-opening aluminum material, the burrs are polished off, and all residues on the surface of the precision-opening aluminum material are completely removed. Precision-cut aluminum profiles are sandblasted under air pressure using diamond abrasive. The sandblasting angle and moving speed are adjusted using a response surface parameter optimizer. The sandblasted aluminum profiles are then placed in a gradually transitioning rounded corner forming mold. An ultrasonic vibration-assisted system is activated to perform multi-step progressive forming. During the multi-step progressive forming process, an acoustic emission signal crack monitor collects stress wave signals from the precision-cut aluminum profiles. The characteristic parameters of the stress wave signals are extracted and input into a crack identification classifier. When the crack identification classifier determines that there is a risk of crack initiation, a springback compensation prediction model is called to calculate the elastic springback compensation value of the current batch of materials and correct the forming force. After the multi-step progressive shaping is completed, the precision-opening aluminum material is subjected to low-temperature annealing to eliminate residual tensile stress, and then sent to the oxidation process to complete the surface treatment.

2. The precision-opening aluminum profile shaping method according to claim 1, characterized in that, The response surface parameter optimizer is used to establish a mathematical mapping relationship between multiple sandblasting parameters and surface roughness and removal rate. It fits the parameter interaction through a second-order response surface model and uses a genetic algorithm to globally optimize and obtain the optimal parameter combination of sandblasting angle and moving speed.

3. The precision-opening aluminum profile shaping method according to claim 2, characterized in that, The inputs to the response surface parameter optimizer include sandblasting air pressure, sand particle size, initial spray angle, and initial moving speed. The outputs are the optimized sandblasting angle and optimized moving speed, which are used to adjust the sandblasting equipment.

4. The precision-opening aluminum profile shaping method according to claim 3, characterized in that, The aforementioned gradient transition rounded corner shaping mold refers to a mold with a rounded corner structure that gradually transitions in curvature radius at the opening corner position, and the mold structure is obtained through a topology optimization-driven variable density design method.

5. The precision-opening aluminum profile shaping method according to claim 4, characterized in that, The gradient transition rounded corner shaping mold establishes a multi-objective optimization problem of mold structure stiffness and mass based on the SIMP material interpolation model. The moving asymptote method is used to iteratively update the element density distribution, and the mold mass is minimized and the stiffness distribution of the contact area is optimized under the condition of satisfying the strength constraint.

6. The precision-opening aluminum profile shaping method according to claim 5, characterized in that, The ultrasonic vibration-assisted system refers to installing an ultrasonic vibration generator on a gradually transitioning rounded corner forming mold, which reduces the friction coefficient between the precision-opening aluminum material and the gradually transitioning rounded corner forming mold through high-frequency micro-amplitude vibration and promotes dislocation movement.

7. The precision-opening aluminum profile shaping method according to claim 6, characterized in that, The multi-step progressive shaping refers to breaking down a single shaping process into progressive steps, with the shaping amount decreasing in each step, thereby avoiding sudden stress changes by gradually accumulating deformation.

8. The precision-opening aluminum profile shaping method according to claim 7, characterized in that, The acoustic emission signal crack monitor is used to collect stress wave signals generated in real time during the multi-step progressive shaping process of precision-opened aluminum material. The mechanical wave is converted into an electrical signal and amplified by a piezoelectric ceramic sensor.

9. The precision-opening aluminum profile shaping method according to claim 8, characterized in that, The stress wave signal characteristic parameters are decomposed into multiple frequency bands through wavelet packet transform, and the energy ratio, peak value and duration of each frequency band are extracted as stress wave signal characteristic parameters. Kalman filtering is then used to eliminate environmental noise interference.

10. The precision-opening aluminum profile shaping method according to claim 9, characterized in that, The crack identification classifier uses a support vector machine algorithm to construct a binary classification model. The stress wave signal feature parameters are input into the trained support vector machine classifier, and the result of the judgment of normal deformation state or crack initiation state is output.

Images