Intelligent prediction and process compensation integrated system and method for size deviation of aluminum alloy castings

Abstract

The application discloses an aluminum alloy casting size deviation intelligent prediction and process compensation integrated system and method, comprising a data acquisition module, a simulation analysis module, a deviation prediction module, a process compensation module for deviation grading, and compensation size verification according to a formula, iterative adjustment until the standard is met, and a UG integrated module; the data acquisition module supports specific format model input, collects relevant parameters and cleanses data, the simulation analysis module constructs a high-precision simulation model based on ProCast and outputs relevant data, the deviation prediction module constructs a multi-model integrated architecture, extracts features and trains a model to ensure prediction accuracy, the process compensation module grades deviations according to the CT7 standard, calculates compensation sizes and iteratively verifies, and the UG integrated module packages the system into a plug-in to construct a visual interface to realize integrated operation without switching software.

Description

Technical Field

[0001] This invention relates to the field of metal casting technology, and in particular to an integrated system and method for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings. Background Technology

[0002] The aerospace field has stringent requirements for the dimensional accuracy of aluminum alloy metal castings. In the traditional manufacturing process, due to the problems of thermal shrinkage and stress deformation during the solidification process of the castings, it is usually necessary to adjust the mold size through multiple mold repairs to meet the accuracy requirements. This model has three major problems: (1) High cost: The cost of manpower and material resources for a single mold repair exceeds RMB 100,000, and multiple mold repairs lead to a significant increase in the total cost; (2) Long cycle: The cycle from a single mold repair to the trial mold is about 7-15 days, and the total cycle for 5 mold repairs is 45-60 days, which cannot meet the needs of rapid delivery; (3) Reliance on experience: The mold repair parameters are mostly judged manually, lacking quantitative basis, and are prone to the situation of "the more you repair, the worse it gets".

[0003] In existing technologies, ProCast-based casting simulation methods can simulate the deformation trend of castings, but it is difficult to directly quantify and predict the specific deviation values ​​of key dimensions in castings. Machine learning methods mostly use a single model, resulting in limited prediction accuracy. Furthermore, the simulation analysis, deviation prediction, and mold compensation processes are disconnected, requiring manual switching between multiple software programs, which is inefficient. Therefore, there is an urgent need for an integrated system that combines "casting simulation - multi-model machine learning - 3D visualization - UG integration". Summary of the Invention

[0004] The purpose of this invention is to provide an integrated system and method for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings, so as to solve the problems of insufficient accuracy, disconnected processes, and low efficiency in the prior art.

[0005] The technical solution of the present invention is an integrated system for intelligent prediction and process compensation of dimensional deviations of aluminum alloy castings, including a data acquisition module. The data acquisition module is used to acquire the three-dimensional model, structural parameters, process parameters, material data and historical test data of the aluminum alloy castings, and to preprocess and handle outliers of the acquired data. The preprocessing includes three-dimensional model format recognition, unit unification, geometric integrity check, missing value filling, unit conversion and numerical range verification.

[0006] The simulation analysis module is used to establish a casting simulation model, perform mesh modeling, boundary condition setting, and simulation calculations of filling, solidification, and thermal stress, and output temperature field, solidification time, stress field, nodal displacement, and model data after deformation.

[0007] The deviation prediction module includes a feature engineering unit and a multi-model integrated prediction unit. The feature engineering unit is used to extract deviation-related features from the 3D model, structural parameters, process parameters, material data and simulation results and perform standardization processing. The multi-model integrated prediction unit is used to predict the standardized deviation-related features and output the dimensional deviation values ​​of key dimensions of the aluminum alloy casting.

[0008] The process compensation module is used to classify the deviation according to the dimensional deviation value, calculate the mold compensation dimension, and perform simulation verification and iterative adjustment on the compensated mold model until the compensated dimensional deviation meets the preset tolerance standard.

[0009] The UG integration module encapsulates data acquisition, simulation calling, deviation prediction, process compensation, and result display functions into a UG / NX plugin based on the NX Open interface, and enables model viewing, deviation result display, compensation result viewing, and data export through a visual interface.

[0010] Furthermore, the outlier removal uses the interquartile range method, expressed as:

[0011]

[0012]

[0013] In the formula, Interquartile range (IQR) is used to measure the dispersion of data. Lower quartiles; Upper quartile; Interval Beyond that is the range for outlier removal.

[0014] Furthermore, the mesh modeling unit adopts a multi-threaded parallel processing method, and performs mesh generation on the casting, gate, riser and H13 steel mold in the MeshCAST module of ProCast to form a discrete mesh model for casting filling, solidification and thermal stress simulation; wherein, the multi-threaded parallel processing method is used to schedule the mesh generation tasks of the casting body, gate, riser and mold in parallel to improve the mesh generation efficiency.

[0015] Furthermore, the simulation calculation includes a number of calculation steps and a stopping condition. The number of calculation steps is 8000, and the stopping condition is that the temperature in each region of the casting does not exceed 555℃. After the simulation is completed, the maximum stress, average solidification time, and nodal displacement data are output, and the nodal displacements are mapped to the original mesh model to generate a deformed STL model. The STL model is then imported into the UG integrated plugin for deviation feature extraction and compensation analysis.

[0016] Furthermore, the feature standardization adopts the StandardScaler method, expressed as:

[0017]

[0018] In the formula, These are the original eigenvalues. The characteristic mean, The characteristic standard deviation, These are the standardized feature values.

[0019] Furthermore, the features in the deviation prediction module are evaluated using the information gain method to select features that have a significant impact on dimensional deviations, such as the wall thickness variation coefficient, pouring temperature, mold preheating temperature, maximum stress, and solidification time. A bar chart of feature contribution is displayed on the plug-in interface. The wall thickness variation coefficient is obtained by statistically analyzing local wall thickness sampling values ​​of the casting. Its feature importance is evaluated using the information gain method, and the information gain (IGA) is calculated according to the following formula:

[0020]

[0021] In the formula, For the information gain of feature A, The entropy of dataset D, Let A be the conditional entropy of dataset D given feature A.

[0022] Furthermore, the process compensation module also includes mold compensation, and the mold compensation dimension calculation formula is expressed as:

[0023]

[0024] In the formula, For mold size compensation, Design dimensions for castings, For alloy shrinkage rate, The value is the prediction deviation correction value. When the deviation is negative, the absolute value is taken, and when the deviation is positive, the negative value is taken.

[0025] Furthermore, the deviation is graded into 7 levels, and the colors corresponding to the 7 levels of deviation are as follows: ≤0.1mm (dark blue), 0.1-0.2mm (cyan), 0.2-0.3mm (green), 0.3-0.5mm (yellow-green), 0.5-0.6mm (yellow), 0.6-0.8mm (orange), and >0.8mm (red).

[0026] A method integrating intelligent prediction and process compensation for dimensional deviations in aluminum alloy castings is disclosed. This method sequentially activates a data acquisition module, a simulation analysis module, a deviation prediction module, a process compensation module, and a UG integration module to complete the fully automated process from model input to process compensation. The method includes the following steps:

[0027] Step 1: Import the 3D model of the aluminum alloy casting and collect the casting material parameters, structural parameters, casting process parameters and historical test data;

[0028] Step 2: Based on the data obtained in Step 1, establish a casting simulation model, mesh the casting, gate, riser and mold, and set boundary conditions and material parameters;

[0029] Step 3: Based on the simulation model established in Step 2, perform filling, solidification and thermal stress simulation calculations to obtain temperature field, solidification time, stress field, nodal displacement and model data after deformation.

[0030] Step four: Extract deviation prediction features based on the simulation results obtained in step three, and standardize the deviation prediction features.

[0031] Step 5: Input the features processed in Step 4 into the deviation prediction model to obtain the deviation prediction values ​​of the key dimensions of the aluminum alloy casting.

[0032] Step 6: Calculate the process compensation amount based on the deviation prediction value obtained in Step 5, and generate the compensated process parameters or the compensated casting model.

[0033] Step 7: Display and export the deviation prediction results, compensation results, and related model files through the UG integration module;

[0034] In this process, after the user selects the model to be processed and inputs the relevant parameters through the UG integrated plugin, the computer program automatically calls each module to complete the above steps in sequence.

[0035] Compared with existing technologies, this invention has the following advantages: It supports input of 3D model formats such as STEP, IGES, STL, Parasolid, and prt, and can collect and clean casting structural parameters, material parameters, process parameters, and historical inspection data; it constructs filling, solidification, and thermal stress simulation models based on ProCast, outputting temperature field, solidification time, stress field, nodal displacement, and deformed model data; it adopts a multi-model integrated architecture for feature extraction, model training, and deviation prediction, improving prediction accuracy; it performs deviation grading according to CT7 tolerance requirements and preset deviation ranges, and displays different deviation levels using color mapping; it performs compensation dimension calculation and iterative verification, and integrates data acquisition, simulation analysis, deviation prediction, process compensation, and result display through a UG plugin. Verification has shown that the RMSE predicted by this invention is less than 0.1mm, which is about 30% more accurate than the single LightGBM model and meets the CT7 tolerance requirements. The number of mold repairs has been reduced from 3 to 5 times to 1 to 2 times, the mold repair cost has been reduced by more than 60%, and the production cycle has been shortened from about 45 days to about 20 days, significantly improving processing efficiency. At the same time, the high deviation area is displayed through CT7 color mapping and 3D preview, which makes it easier for engineers to quickly formulate process compensation solutions. Attached Figure Description

[0036] Figure 1 This is an overall architecture diagram of the integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings provided in this embodiment of the invention;

[0037] Figure 2 This is a 3D visualization interface diagram of the prediction system in an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram of the CT7 tolerance standard color mapping in an embodiment of the present invention;

[0039] Figure 4 A comparison chart of the number of mold repairs required for the invention of an integrated system for intelligent prediction and process compensation of dimensional deviations and traditional methods under different task numbers;

[0040] Figure 5 This is a comparison chart of the production cycle of the integrated system for intelligent prediction and process compensation of dimensional deviation of the present invention and traditional methods. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0042] The following is combined Figures 1 to 5 The present invention will be described below. This embodiment is only for explaining the present invention and does not limit the scope of protection of the present invention.

[0043] This invention provides an integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings, including a data acquisition module, a simulation analysis module, a deviation prediction module, a process compensation module, and a UG integration module.

[0044] The data acquisition module is used to collect 3D models, structural parameters, process parameters, material data, and historical inspection data of aluminum alloy castings. It performs preprocessing and outlier removal on the data. Structural parameters refer to the structural components of the casting, such as the size of the columns and chamfers. The preprocessing includes 3D model format recognition, unit standardization, geometric integrity checking, missing value imputation, unit conversion, and numerical range verification. Outlier removal uses the interquartile range method. The range of outliers is Data outside this range will be removed.

[0045] The simulation analysis module is used to establish a casting simulation model, perform mesh modeling, boundary condition setting, and simulation calculations for filling, solidification, and thermal stress. It outputs temperature field, solidification time, stress field, nodal displacement, and deformed model data. The simulation analysis module uses ProCast with multi-threaded parallel processing for mesh modeling, boundary condition setting, and simulation calculations. Mesh modeling is completed in the MeshCAST module, meshing the casting, gating, riser, and H13 steel mold. The multi-threaded parallel processing method is used to schedule the meshing tasks of the casting body, gating, riser, and mold in parallel to improve meshing efficiency. The simulation calculation includes filling, solidification, and thermal stress calculations, with 8000 steps. The stopping condition is that the temperature in each region of the casting is ≤555℃. After the simulation is completed, stress, solidification time, and nodal displacement data are output, and the nodal displacements are mapped to the original mesh model to generate a deformed STL model. This STL model is imported into the UG integrated plugin for deviation feature extraction and compensation analysis.

[0046] The deviation prediction module includes a feature engineering unit and a multi-model integrated prediction unit. The feature engineering unit extracts deviation-related features from the 3D model, structural parameters, process parameters, material data, and simulation results, and performs standardization processing. The multi-model integrated prediction unit predicts the standardized deviation-related features and outputs the dimensional deviation values ​​of key dimensions in the aluminum alloy casting, as well as features such as the wall thickness variation coefficient, pouring temperature, mold preheating temperature, maximum stress, and solidification time. Standardization is performed. The multi-model ensemble prediction unit fuses the outputs of multiple prediction models to obtain the size bias value. Feature importance is evaluated using the information gain method. The feature contribution is displayed in the plugin interface. In the formula, These are the original eigenvalues. The characteristic mean, The characteristic standard deviation, These are the standardized feature values.

[0047] The process compensation module is used to classify deviations based on dimensional deviation values, calculate mold compensation dimensions, and perform compensation verification and iterative optimization adjustments on the compensated mold model until the compensated dimensional deviation meets the preset tolerance standard, with the deviation being less than or equal to 0.1mm. The mold compensation dimension formula is as follows: ,in For mold size compensation, Design dimensions for castings, For alloy shrinkage rate, To predict the deviation correction value, the absolute value is taken when the deviation is negative, and the negative value is taken when the deviation is positive. The deviation is graded into 7 levels, with colors in the following order: dark blue, cyan, green, yellowish-green, yellow, orange, and red. The colors corresponding to the 7 deviation levels are: ≤0.1mm (dark blue), 0.1-0.2mm (cyan), 0.2-0.3mm (green), 0.3-0.5mm (yellowish-green), 0.5-0.6mm (yellow), 0.6-0.8mm (orange), and >0.8mm (red).

[0048] The UG integration module encapsulates data acquisition, simulation call, deviation prediction, process compensation, and result display functions into a UG / NX plugin based on the NX Open interface. It enables model viewing, deviation result display, compensation CSV result viewing, and data export through a visual interface, and displays the running progress through ProgressBar for compensation verification.

[0049] This invention also provides an integrated method for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings, comprising the following steps:

[0050] Step 1: Import the 3D model of the aluminum alloy casting and collect material parameters, structural parameters, process parameters and historical test data;

[0051] Step two: Preprocess the collected data and remove outliers;

[0052] Step 3: Build a casting simulation model in ProCast and complete mesh generation, boundary condition setting, and simulation calculation;

[0053] Step 4: Extract and standardize the deviation prediction features based on the simulation results;

[0054] Step 5: Input the standardized features into the multi-model integration prediction unit to obtain the predicted size deviation value;

[0055] Step 6: Calculate the mold compensation dimensions based on the predicted dimensional deviation values ​​and perform simulation verification;

[0056] Step 7: Display and export the deviation prediction results, compensation results, and related files through the UG integration module;

[0057] Through the above embodiments, the present invention can realize the integrated processing of aluminum alloy casting dimensional deviation prediction, mold compensation and UG visualization operation.

[0058] Example 1: Deviation Prediction and Compensation for ZL105 Castings

[0059] This embodiment uses a ZL105 aluminum alloy casting for aerospace applications as an example for explanation. The casting's external dimensions are... ,Include , and Three typical wall thickness regions.

[0060] First, the STEP format 3D model of the casting was imported using the UG / NX plugin, and the geometric feature information of the model was extracted using Python OCC, extracting 286 surface features, including the area, centroid coordinates, and surface type of each surface. Among them, the centroid coordinates of a certain plane in the 5mm wall thickness area are (113.7, 82.5, 60.0), and the surface type is planar. The collected material parameters include: liquidus temperature of ZL105 alloy 615℃, solidus temperature 555℃, and thermal conductivity 120W / (m·K); the collected process parameters include: pouring temperature 730℃, mold preheating temperature 300℃, and pouring speed 0.8m / s. The data was cleaned to remove missing values, and two stress anomalies (the maximum stress in the original data was 55MPa, exceeding the reasonable range of [15,40]) were removed using the IQR method.

[0061] The collected data underwent preprocessing and outlier removal. Preprocessing included model format identification, unit standardization, geometric integrity checks, missing value imputation, and numerical range verification. In this embodiment, the data had no missing values, and the stress data was analyzed using the interquartile range method for outlier detection. The formula is as follows:

[0062]

[0063] The normal range is:

[0064] Calculations show that the normal range for stress data is: The original data contained two stress anomalies, with the maximum stress being... Those that exceed this range are therefore removed or corrected.

[0065] Subsequently, a casting simulation model was built in ProCast's MeshCAST module, and volumetric meshing was performed on the casting, gating, riser, and H13 steel mold. The gating diameter was 20mm, and the riser size was... The mesh size was set to 1.5mm for areas with a 5mm wall thickness and 3mm for other areas, with a total of approximately 800,000 elements. Mesh quality checks showed a twist rate of 0.72, a Jacobian determinant of 0.35, and no negative volume elements. Twist rate evaluates the degree to which the element shape deviates from the ideal shape, while the Jacobian determinant evaluates the degree of deformation and effectiveness of element mapping. The absence of negative volume elements indicates a valid mesh topology, which can be used for subsequent filling, solidification, and thermal stress calculations.

[0066] The boundary conditions were set to gravity casting, with the gravity direction along the -Z axis; the initial casting temperature was 730℃, and the mold preheating temperature was 300℃; the air convection heat transfer coefficient was 25W / (m²·K), and the heat transfer coefficient at the casting-mold interface was experimentally calibrated to 800W / (m²·K). The simulation calculation steps were set to 8000 steps, and the stopping condition was that the temperature in any region of the casting did not exceed 555℃. After the simulation was completed, the maximum stress of 35MPa, the average solidification time of 280s, the nodal displacement data, and the deformed STL model were output.

[0067] In the deviation prediction stage, 28 deviation-related features were extracted from the 3D model, process parameters, material data, and simulation results, including mean wall thickness, coefficient of variation of wall thickness, surface area, cooling factor, pouring temperature, mold preheating temperature, material shrinkage rate, maximum stress, and solidification time. For example, the mean wall thickness of the 5mm wall thickness region is 5mm, the area is 120mm², and the cooling factor is 120 / (5²) = 4.8.

[0068] The extracted features are standardized using the StandardScaler method, with the following formula:

[0069]

[0070] Meanwhile, the information gain method is used to evaluate feature importance, with the following formula:

[0071]

[0072] In this embodiment, a pre-trained multi-model ensemble prediction model is loaded, with LightGBM accounting for 50% of the weight, RandomForest for 25%, and GradientBoosting for 25%. After inputting the standardized features into the model, the predicted values ​​of key region size deviations are output: the average deviation for a 5mm wall thickness region is -0.23mm, the deviation for an R5 rounded corner region is -0.15mm, and the deviation for a φ10mm through-hole region is -0.18mm.

[0073] Five-fold cross-validation was performed on the prediction model, and the results were: Root Mean Square Error (RMSE) = 0.08 mm, Mean Absolute Error (MAE) = 0.06 mm, and Coefficient of Determination (R²) = 0.96. Here, RMSE reflects the overall magnitude of the prediction error; MAE reflects the average deviation between the predicted and measured values; and the R² reflects the model's ability to explain changes in actual deviations. These results indicate that the prediction model in this embodiment has high prediction accuracy.

[0074] During the process compensation stage, deviations are graded based on predicted deviation values. The predicted deviation for the 5mm wall thickness area is -0.23mm, with an absolute value of 0.23mm, falling within the 0.2-0.3mm range, corresponding to a green rating. The mold compensation dimensions are calculated using the following formula:

[0075]

[0076] Therefore, the compensation dimensions for the 5mm wall thickness region are:

[0077]

[0078] The compensation dimension for the R5 rounded corner area is:

[0079]

[0080] The compensation dimensions for the φ10mm through-hole area are:

[0081]

[0082] The compensated mold model was re-imported into ProCast for simulation verification. The verification results showed that the dimensional deviation of the critical area was reduced to between -0.05mm and 0.03mm, which met the control target of ≤0.1mm for critical dimensional deviation set in this embodiment and complied with the tolerance requirements corresponding to CT7 grade.

[0083] During the UG / NX integration application phase, users open the system plugin in NX11, import the casting model via "Select STP," call the feature engineering units via "Extract Features," and call the pre-trained model via "Predict Deviation," displaying the deviation values ​​for each surface. For example, a 5mm wall thickness surface displays a predicted deviation of -0.23mm in the results table, with a green background. After switching to the "3D Visualization" tab, the model is displayed using color mapping according to the CT7 deviation level. Users can view the deviation distribution in different areas through rotation, scaling, and panning. After clicking "Export Results," the system generates a CSV file with fields including surface ID, surface type, design dimensions, predicted deviation, compensation dimensions, and CT7 level, for subsequent process archiving.

[0084] To verify the effectiveness of the system of the present invention, a comparative test was conducted using traditional mold repair methods, single simulation methods, and the system of the present invention. The test subjects were 10 sets of complex ZL105 aluminum alloy castings, and the results are shown in Table 1:

[0085] Table 1

[0086]

[0087] The results above show that the system of the present invention can realize the integrated processing of aluminum alloy casting dimensional deviation prediction, mold compensation and UG / NX visualization operation. It is superior to traditional methods and single simulation methods in terms of mold repair times, production cycle, mold repair cost and pass rate, thus verifying the effectiveness and practicality of the technical solution of the present invention.

[0088] In addition to the embodiments described above, this invention can also be used for predicting dimensional deviations and compensating for process defects in other aluminum alloy castings. All other embodiments obtained based on the technical solution of this invention without inventive effort are within the scope of protection of this invention.

Claims

1. An integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings, characterized in that, include: The data acquisition module is used to acquire the three-dimensional model, structural parameters, process parameters, material data and historical test data of aluminum alloy castings, and to preprocess and handle outliers of the acquired data. The preprocessing includes three-dimensional model format recognition, unit unification, geometric integrity check, missing value filling, unit conversion and numerical range verification. The simulation analysis module is used to establish a casting simulation model, perform mesh modeling, boundary condition setting, and simulation calculations of filling, solidification, and thermal stress, and output temperature field, solidification time, stress field, nodal displacement, and model data after deformation. The deviation prediction module includes a feature engineering unit and a multi-model integrated prediction unit. The feature engineering unit is used to extract deviation-related features from the 3D model, structural parameters, process parameters, material data and simulation results and perform standardization processing. The multi-model integrated prediction unit is used to predict the standardized deviation-related features and output the dimensional deviation values ​​of key dimensions of the aluminum alloy casting. The process compensation module is used to classify the deviation according to the dimensional deviation value, calculate the mold compensation size, and perform simulation verification and iterative adjustment on the compensated mold model until the compensated dimensional deviation meets the preset tolerance standard. The UG integration module encapsulates data acquisition, simulation calling, deviation prediction, process compensation, and result display functions into a UG / NX plugin based on the NX Open interface, and enables model viewing, deviation result display, compensation result viewing, and data export through a visual interface.

2. The integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings according to claim 1, characterized in that, The outlier removal method uses the interquartile range (ICM) method, which is expressed as follows: ； ； In the formula, Interquartile range (IQR) is used to measure the dispersion of data. Lower quartiles; Upper quartile; interval Beyond that is the range for outlier removal.

3. The integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings according to claim 1, characterized in that, The mesh modeling unit adopts a multi-threaded parallel processing method. In the MeshCAST module of ProCast, the casting, gate, riser and H13 steel mold are meshed to form a discrete mesh model for casting filling, solidification and thermal stress simulation. The multi-threaded parallel processing method is used to schedule the meshing tasks of the casting body, gate, riser and mold in parallel to improve the meshing efficiency.

4. The integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings according to claim 1 or 3, characterized in that, The simulation calculation includes a number of calculation steps and a stopping condition. The number of calculation steps is 8000, and the stopping condition is that the temperature in each region of the casting does not exceed 555℃. After the simulation is completed, the maximum stress, average solidification time, and nodal displacement data are output, and the nodal displacements are mapped to the original mesh model to generate a deformed STL model. The STL model is then imported into the UG integrated plugin for deviation feature extraction and compensation analysis.

5. The integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings according to claim 1, characterized in that, The feature standardization adopts the StandardScaler method, which is expressed as follows: ； In the formula, These are the original eigenvalues. The characteristic mean, The characteristic standard deviation, These are the standardized feature values.

6. The integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings according to claim 4, characterized in that, The features in the deviation prediction module are evaluated using the information gain method, and features that have a significant impact on dimensional deviations, such as the wall thickness variation coefficient, pouring temperature, mold preheating temperature, maximum stress, and solidification time, are selected and displayed as a bar chart of feature contribution on the plug-in interface. The wall thickness variation coefficient is obtained by statistically analyzing the local wall thickness sampling values ​​of the casting. Its feature importance is evaluated using the information gain method, and the information gain (IGA) is calculated according to the following formula: ； In the formula, For the information gain of feature A, The entropy of dataset D, Let A be the conditional entropy of dataset D given feature A.

7. The integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings according to claim 5, characterized in that, The process compensation module also includes mold compensation, and the formula for calculating the mold compensation dimension is as follows: ； In the formula, For mold size compensation, Design dimensions for castings, For alloy shrinkage rate, The value is the prediction deviation correction value. When the deviation is negative, the absolute value is taken, and when the deviation is positive, the negative value is taken.

8. The integrated system for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings according to claim 6, characterized in that, The deviation is classified into 7 levels, and the corresponding colors for the 7 levels of deviation are: ≤0.1mm (dark blue), 0.1-0.2mm (cyan), 0.2-0.3mm (green), 0.3-0.5mm (yellow-green), 0.5-0.6mm (yellow), 0.6-0.8mm (orange), and >0.8mm (red).

9. A method for intelligent prediction and process compensation of dimensional deviations in aluminum alloy castings, characterized in that, The method is executed by a computer program and includes the following steps in sequence: Step 1: Import the 3D model of the aluminum alloy casting and collect the casting material parameters, structural parameters, casting process parameters and historical test data; Step 2: Based on the data obtained in Step 1, establish a casting simulation model, mesh the casting, gate, riser and mold, and set boundary conditions and material parameters; Step 3: Based on the simulation model established in Step 2, perform filling, solidification and thermal stress simulation calculations to obtain temperature field, solidification time, stress field, nodal displacement and model data after deformation. Step four: Extract deviation prediction features based on the simulation results obtained in step three, and standardize the deviation prediction features. Step 5: Input the features processed in Step 4 into the deviation prediction model to obtain the deviation prediction values ​​of the key dimensions of the aluminum alloy casting. Step 6: Calculate the process compensation amount based on the deviation prediction value obtained in Step 5, and generate the compensated process parameters or the compensated casting model. Step 7: Display and export the deviation prediction results, compensation results, and related model files through the UG integration module; In this process, after the user selects the model to be processed and inputs the relevant parameters through the UG integrated plugin, the computer program automatically calls each module to complete the above steps in sequence.

Images