Aluminum alloy metal mold low-pressure casting defect detection method

Abstract

The invention discloses an aluminum alloy metal mold low-pressure casting defect detection method, and relates to the technical field of alloy casting, and the method comprises the following steps: extracting a feature vector of a pre-processed prior defect picture and a feature vector of a pre-processed casting image, respectively calculating the similarity between the current feature vector and different reference feature vectors, and calculating the similarity between the current feature vector and the different reference feature vectors; and counting the occurrence frequency of the current defect, generating a casting decision, training a machine learning model for feeding back the current defect, and testing the machine learning model. The product quality and the production efficiency are improved through similarity calculation of the feature vectors and training of the machine learning model, manual intervention and errors are reduced by using the machine learning model and automatic decision generation, the defect detection and defect processing speed is increased, and the casting decision through statistics and weight calculation is more accurate. The importance and frequency of different defect types can be reflected more accurately, and a more effective quality management strategy can be formulated.

Description

Technical Field

[0001] The invention relates to the technical field of alloy casting, and in particular to a method for detecting defects in low-pressure casting of an aluminum alloy metal mold. Background Art

[0002] In recent years, alloy casting technology has continued to develop. Based on traditional non-destructive testing technology, alloy casting technology has developed in the direction of intelligence, efficiency and precision. With the application and integration of various advanced technologies, the level of casting quality management will continue to improve, providing reliable casting quality assurance for various industrial applications.

[0003] At present, a Chinese invention patent with publication number CN 117233179A discloses a defect detection method and system for aluminum alloy casting products. The method collects X-Ray detection images of the aluminum alloy casting products to be inspected by X-ray detection equipment, analyzes the X-Ray detection images, and obtains defect detection results. However, the related technology does not intelligently identify the defect type, lacks accuracy, does not generate decisions based on the results of defect detection, and lacks scientificity. Summary of the invention

[0004] The technical problem solved by the present invention is that the related art does not intelligently identify defect types, lacks accuracy, does not generate decisions based on defect detection results, and lacks scientificity.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for detecting defects in low-pressure casting of aluminum alloy metal molds, comprising the following steps:

[0006] Step S1, obtaining a priori defect picture, preprocessing the priori defect picture, extracting a feature vector of the preprocessed priori defect picture, and recording the feature vector of the preprocessed priori defect picture as a reference feature vector;

[0007] Step S2, acquiring a casting image, preprocessing the casting image, extracting a feature vector of the preprocessed casting image, and recording the feature vector of the preprocessed casting image as a current feature vector;

[0008] Step S3, respectively calculating the similarity between the current feature vector and different reference feature vectors, retaining the similarity with the largest value, and obtaining the defect corresponding to the current feature vector according to the similarity with the largest value, recording it as the current defect, counting the number of occurrences of the current defect, calculating the weight of the number of occurrences of different defect types in the total number of occurrences of the prior defect types, and generating a casting decision based on the weight of the number of occurrences of different defect types in the total number of occurrences of the prior defect types;

[0009] Step S4: training a machine learning model for feeding back the current defect, and testing the machine learning model until the accuracy reaches a first threshold.

[0010] As a preferred embodiment of the method for detecting defects in low-pressure casting of aluminum alloy metal molds according to the present invention, step S1 includes the following sub-steps:

[0011] Step S11, obtaining a priori defect images, wherein the a priori defect images include a priori pore image, a priori shrinkage image, a priori crack image, a priori sand hole image, a priori segregation image, a priori deformation image, and a priori mold sticking image;

[0012] Step S12, filtering and graying the priori defect image respectively;

[0013] Step S13, extracting the feature vector of the preprocessed prior defect image, and recording the feature vector of the preprocessed prior defect image as a reference feature vector, wherein the reference feature vector includes a reference pore feature vector, a reference crack feature vector, a reference sand hole feature vector, a reference segregation feature vector, a reference deformation feature vector and a reference mold sticking feature vector.

[0014] As a preferred embodiment of the method for detecting defects in low-pressure casting of aluminum alloy metal molds according to the present invention, step S2 includes the following sub-steps:

[0015] Step S21, acquiring a casting image, filtering, enhancing and graying the casting image, and performing rotation correction based on a priori defect images;

[0016] Step S21, extracting a feature vector of the preprocessed casting image, and recording the feature vector of the preprocessed casting image as a current feature vector.

[0017] As a preferred solution of the aluminum alloy metal mold low pressure casting defect detection method described in the present invention, the correction logic of the rotation correction is:

[0018] Acquire a casting image and a prior defect image, construct an inertia ellipse of the casting image and an inertia ellipse of the prior defect image respectively, grayscale the inertia ellipse of the casting image and extract the moment of inertia of the inertia ellipse of the grayscaled casting image, grayscale the inertia ellipse of the prior defect image and extract the moment of inertia of the inertia ellipse of the grayscaled prior defect image, respectively set the major axis directions of the inertia ellipse of the casting image and the inertia ellipse of the prior defect image, calculate the rotation angles of the major axis directions of the inertia ellipse of the casting image and the inertia ellipse of the prior defect image, and calibrate the rotation angles.

[0019] As a preferred solution of the aluminum alloy metal mold low-pressure casting defect detection method described in the present invention, the filtering processing obtains a blurred image according to a Gaussian filtering algorithm, sets different parameters respectively, restores the edge of the blurred image, selects the optimal parameters, and traverses the blurred image, and the parameters include a spatial scale operator, an edge strength operator and the number of iterations.

[0020] As a preferred solution of the aluminum alloy metal mold low-pressure casting defect detection method described in the present invention, the casting image includes a surface image and an internal image, the surface image is obtained by an industrial camera, and the internal image is obtained by X-ray irradiation.

[0021] As a preferred embodiment of the method for detecting defects in low-pressure casting of aluminum alloy metal molds according to the present invention, step S3 includes the following sub-steps:

[0022] Step S31, obtaining a current feature vector, a reference pore feature vector, a reference crack feature vector, a reference sand hole feature vector, a reference segregation feature vector, a reference deformation feature vector and a reference mold sticking feature vector;

[0023] Step S32, respectively calculating the similarity between the current feature vector and the reference pore feature vector, the reference crack feature vector, the reference sand hole feature vector, the reference segregation feature vector, the reference deformation feature vector and the reference mold sticking feature vector;

[0024] Step S33, sorting the similarities in descending order, retaining the similarity with the largest value, and obtaining the defect corresponding to the current feature vector according to the similarity with the largest value, and recording it as the current defect;

[0025] Step S34, counting the number of times the current defect occurs, calculating the weights of different defect types in the overall defect types, and generating a casting decision based on the weights of different defect types in the overall defect types.

[0026] As a preferred solution of the aluminum alloy metal mold low pressure casting defect detection method described in the present invention, the calculation expression of the similarity is:

[0027]

[0028] The calculation expression of the weight is:

[0029]

[0030] Among them, XS is the similarity, x i is the i-th current feature vector of the casting image, y i is the i-th reference feature vector of the prior defect image, n is the number of casting images, Q tis the weight of the t-th prior defect type, r t is the number of occurrences of the t-th prior defect type, is the total number of occurrences of the prior defect type.

[0031] As a preferred embodiment of the defect detection method for low-pressure casting of aluminum alloy metal molds described in the present invention, the generation logic of the casting decision is as follows: obtaining the weights of different defect types in the overall defect type, comparing the weights with the first weight threshold and the second weight threshold, when the weight is less than the first weight threshold, the casting decision is to ignore the defect type, when the weight is greater than or equal to the first weight threshold and less than the second weight defect type, the casting decision is to use the first value as a gradient to adjust the production steps corresponding to the defect until the weight of the defect is less than the first weight threshold, when the weight is greater than or equal to the second weight threshold, the casting decision is to use the second value as a gradient to adjust the production steps corresponding to the defect until the defect is less than the first weight threshold;

[0032] When the current defect is a pore, the pouring speed is reduced with the first value as a gradient or the second value as a gradient. When the current defect is a crack, the core temperature is increased with the first value as a gradient or the second value as a gradient. When the current defect is a sand hole, the sand blowing speed is increased with the first value as a gradient or the second value as a gradient. When the current defect is segregation, the stirring speed is increased with the first value as a gradient or the second value as a gradient. When the current defect is deformation, the draft angle is increased with the first value as a gradient or the second value as a gradient. When the current defect is mold sticking, the alloy liquid temperature is reduced with the first value as a gradient or the second value as a gradient.

[0033] As a preferred embodiment of the method for detecting defects in low-pressure casting of aluminum alloy metal molds according to the present invention, step S4 includes the following sub-steps:

[0034] Step S41, obtaining a casting image and a current defect corresponding to the casting image, taking the casting image and the current defect corresponding to the casting image as a sample set, and dividing the sample set into a test set and a training set according to a certain ratio;

[0035] Step S42, with the predicted current defect as the output, the actual current defect as the prediction target, and minimizing the sum of the prediction accuracies of all predicted current defects as the training target, the machine learning model is trained until the prediction accuracy reaches a first threshold.

[0036] The beneficial effects of the present invention are as follows: the image is converted into a feature vector through preprocessing and feature extraction, and then the defect type is identified through similarity calculation. Through the similarity calculation of the feature vector and the training of the machine learning model, the defects in the casting can be more accurately identified and classified, thereby improving product quality and production efficiency. The machine learning model and automated decision generation can reduce human intervention and errors, speed up defect detection and defect processing, and increase the accuracy of defect detection through continuous training and testing of the machine learning model. The casting decision based on statistics and weight calculation can more accurately reflect the importance and frequency of different defect types, which is conducive to the formulation of more effective quality management strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A schematic diagram of the basic flow of a method for detecting defects in low-pressure casting of an aluminum alloy metal mold provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0039] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a method for detecting defects in low-pressure casting of an aluminum alloy metal mold, comprising the following steps: A method for detecting defects in low-pressure casting of an aluminum alloy metal mold, comprising the following steps:

[0040] Step S1, obtaining a priori defect picture, preprocessing the priori defect picture, extracting a feature vector of the preprocessed priori defect picture, and recording the feature vector of the preprocessed priori defect picture as a reference feature vector;

[0041] Step S2, acquiring a casting image, preprocessing the casting image, extracting a feature vector of the preprocessed casting image, and recording the feature vector of the preprocessed casting image as a current feature vector;

[0042] Step S3, respectively calculating the similarity between the current feature vector and different reference feature vectors, retaining the similarity with the largest value, and obtaining the defect corresponding to the current feature vector according to the similarity with the largest value, recording it as the current defect, counting the number of occurrences of the current defect, calculating the weight of the number of occurrences of different defect types in the total number of occurrences of the prior defect types, and generating a casting decision based on the weight of the number of occurrences of different defect types in the total number of occurrences of the prior defect types;

[0043] Step S4: training a machine learning model for feeding back the current defect, and testing the machine learning model until the accuracy reaches a first threshold.

[0044] The present invention converts images into feature vectors through preprocessing and feature extraction, and then identifies defect types through similarity calculation. Through similarity calculation of feature vectors and training of machine learning models, defects in castings can be more accurately identified and classified, thereby improving product quality and production efficiency. Machine learning models and automated decision generation are used to reduce human intervention and errors, speed up defect detection and defect processing, and increase the accuracy of defect detection through continuous training and testing of machine learning models. Casting decisions calculated through statistics and weights can more accurately reflect the importance and frequency of different defect types, which is conducive to formulating more effective quality management strategies.

[0045] The step S1 includes the following sub-steps:

[0046] Step S11, obtaining a priori defect images, wherein the a priori defect images include a priori pore image, a priori shrinkage image, a priori crack image, a priori sand hole image, a priori segregation image, a priori deformation image, and a priori mold sticking image;

[0047] Step S12, filtering and graying the priori defect image respectively;

[0048] Step S13, extracting the feature vector of the preprocessed prior defect image, and recording the feature vector of the preprocessed prior defect image as a reference feature vector, wherein the reference feature vector includes a reference pore feature vector, a reference crack feature vector, a reference sand hole feature vector, a reference segregation feature vector, a reference deformation feature vector and a reference mold sticking feature vector.

[0049] The step S2 includes the following sub-steps:

[0050] Step S21, acquiring a casting image, filtering, enhancing and graying the casting image, and performing rotation correction based on a priori defect images;

[0051] Step S21, extracting a feature vector of the preprocessed casting image, and recording the feature vector of the preprocessed casting image as a current feature vector.

[0052] The correction logic of the rotation correction is:

[0053] Acquire a casting image and a prior defect image, construct an inertia ellipse of the casting image and an inertia ellipse of the prior defect image respectively, grayscale the inertia ellipse of the casting image and extract the moment of inertia of the inertia ellipse of the grayscaled casting image, grayscale the inertia ellipse of the prior defect image and extract the moment of inertia of the inertia ellipse of the grayscaled prior defect image, respectively set the major axis directions of the inertia ellipse of the casting image and the inertia ellipse of the prior defect image, calculate the rotation angles of the major axis directions of the inertia ellipse of the casting image and the inertia ellipse of the prior defect image, and calibrate the rotation angles.

[0054] In the specific implementation, the image is converted into a feature vector through preprocessing and feature extraction, and then the defect type is identified through similarity calculation, which is conducive to accurately identifying and classifying defects in castings.

[0055] The filtering process obtains a blurred image according to a Gaussian filtering algorithm, sets different parameters respectively, restores the edge of the blurred image, selects the optimal parameters, and traverses the blurred image. The parameters include a spatial scale operator, an edge strength operator, and the number of iterations.

[0056] The casting image includes a surface image and an internal image. The surface image is obtained by an industrial camera, and the internal image is obtained by X-ray irradiation.

[0057] The step S3 includes the following sub-steps:

[0058] Step S31, obtaining a current feature vector, a reference pore feature vector, a reference crack feature vector, a reference sand hole feature vector, a reference segregation feature vector, a reference deformation feature vector and a reference mold sticking feature vector;

[0059] Step S32, respectively calculating the similarity between the current feature vector and the reference pore feature vector, the reference crack feature vector, the reference sand hole feature vector, the reference segregation feature vector, the reference deformation feature vector and the reference mold sticking feature vector;

[0060] Step S33, sorting the similarities in descending order, retaining the similarity with the largest value, and obtaining the defect corresponding to the current feature vector according to the similarity with the largest value, and recording it as the current defect;

[0061] Step S34, counting the number of times the current defect occurs, calculating the weights of different defect types in the overall defect types, and generating a casting decision based on the weights of different defect types in the overall defect types.

[0062] The calculation expression of the similarity is:

[0063]

[0064] The calculation expression of the weight is:

[0065]

[0066] Among them, XS is the similarity, x i is the i-th current feature vector of the casting image, y i is the i-th reference feature vector of the prior defect image, n is the number of casting images, Q t is the weight of the t-th prior defect type, r t is the number of occurrences of the t-th prior defect type, is the total number of occurrences of the prior defect type.

[0067] In specific implementation, through the similarity calculation of feature vectors and the training of machine learning models, defects in castings can be more accurately identified and classified, effectively improving product quality and production efficiency.

[0068] The generation logic of the casting decision is: obtain the weights of different defect types in the overall defect types, compare the weights with the first weight threshold and the second weight threshold, when the weight is less than the first weight threshold, the casting decision is to ignore the defect type, when the weight is greater than or equal to the first weight threshold and less than the second weight defect type, the casting decision is to adjust the production steps corresponding to the defect with the first value as a gradient until the weight of the defect is less than the first weight threshold, when the weight is greater than or equal to the second weight threshold, the casting decision is to adjust the production steps corresponding to the defect with the second value as a gradient until the defect is less than the first weight threshold;

[0069] When the current defect is a pore, the pouring speed is reduced with the first value as a gradient or the second value as a gradient. When the current defect is a crack, the core temperature is increased with the first value as a gradient or the second value as a gradient. When the current defect is a sand hole, the sand blowing speed is increased with the first value as a gradient or the second value as a gradient. When the current defect is segregation, the stirring speed is increased with the first value as a gradient or the second value as a gradient. When the current defect is deformation, the draft angle is increased with the first value as a gradient or the second value as a gradient. When the current defect is mold sticking, the alloy liquid temperature is reduced with the first value as a gradient or the second value as a gradient.

[0070] In specific implementation, the use of machine learning models and automated decision generation reduces human intervention and errors, and speeds up defect detection and defect handling.

[0071] The step S4 includes the following sub-steps:

[0072] Step S41, obtaining a casting image and a current defect corresponding to the casting image, taking the casting image and the current defect corresponding to the casting image as a sample set, and dividing the sample set into a test set and a training set according to a certain ratio;

[0073] Step S42, with the predicted current defect as the output, the actual current defect as the prediction target, and minimizing the sum of the prediction accuracies of all predicted current defects as the training target, the machine learning model is trained until the prediction accuracy reaches a first threshold.

[0074] In specific implementation, the accuracy of defect detection is increased by continuously training and testing machine learning models.

[0075] The present invention converts images into feature vectors through preprocessing and feature extraction, and then identifies defect types through similarity calculation. Through similarity calculation of feature vectors and training of machine learning models, defects in castings can be more accurately identified and classified, thereby improving product quality and production efficiency. Machine learning models and automated decision generation are used to reduce human intervention and errors, speed up defect detection and defect processing, and increase the accuracy of defect detection through continuous training and testing of machine learning models. Casting decisions calculated through statistics and weights can more accurately reflect the importance and frequency of different defect types, which is conducive to formulating more effective quality management strategies.

[0076] It should be understood by those skilled in the art that the embodiments of the present invention can be provided as methods, systems or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk. These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0077] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting defects in low-pressure casting of aluminum alloy metal molds, characterized in that: The following steps are involved: Step S1, obtaining a priori defect picture, preprocessing the priori defect picture, extracting a feature vector of the preprocessed priori defect picture, and recording the feature vector of the preprocessed priori defect picture as a reference feature vector; Step S2, acquiring a casting image, preprocessing the casting image, extracting a feature vector of the preprocessed casting image, and recording the feature vector of the preprocessed casting image as a current feature vector; Step S3, respectively calculating the similarity between the current feature vector and different reference feature vectors, retaining the similarity with the largest value, and obtaining the defect corresponding to the current feature vector according to the similarity with the largest value, recording it as the current defect, counting the number of occurrences of the current defect, calculating the weight of the number of occurrences of different defect types in the total number of occurrences of the prior defect types, and generating a casting decision based on the weight of the number of occurrences of different defect types in the total number of occurrences of the prior defect types; Step S4: training a machine learning model for feeding back the current defect, and testing the machine learning model until the accuracy reaches a first threshold.

2. A method for detecting defects in low-pressure casting of aluminum alloy metal molds according to claim 1, characterized in that: The step S1 includes the following sub-steps: Step S11, obtaining a priori defect images, wherein the a priori defect images include a priori pore image, a priori shrinkage image, a priori crack image, a priori sand hole image, a priori segregation image, a priori deformation image, and a priori mold sticking image; Step S12, filtering and graying the priori defect image respectively; Step S13, extracting the feature vector of the preprocessed prior defect image, and recording the feature vector of the preprocessed prior defect image as a reference feature vector, wherein the reference feature vector includes a reference pore feature vector, a reference crack feature vector, a reference sand hole feature vector, a reference segregation feature vector, a reference deformation feature vector and a reference mold sticking feature vector.

3. A method for detecting defects in low-pressure casting of aluminum alloy metal molds as claimed in claim 1, characterized in that: The step S2 includes the following sub-steps: Step S21, acquiring a casting image, filtering, enhancing and graying the casting image, and performing rotation correction based on a priori defect images; Step S21, extracting a feature vector of the preprocessed casting image, and recording the feature vector of the preprocessed casting image as a current feature vector.

4. A method for detecting defects in low-pressure casting of aluminum alloy metal molds as claimed in claim 3, characterized in that: The correction logic of the rotation correction is: Acquire a casting image and a prior defect image, construct an inertia ellipse of the casting image and an inertia ellipse of the prior defect image respectively, grayscale the inertia ellipse of the casting image and extract the moment of inertia of the inertia ellipse of the grayscaled casting image, grayscale the inertia ellipse of the prior defect image and extract the moment of inertia of the inertia ellipse of the grayscaled prior defect image, respectively set the major axis directions of the inertia ellipse of the casting image and the inertia ellipse of the prior defect image, calculate the rotation angles of the major axis directions of the inertia ellipse of the casting image and the inertia ellipse of the prior defect image, and calibrate the rotation angles.

5. A method for detecting defects in low-pressure casting of aluminum alloy metal molds as claimed in claim 3, characterized in that: The filtering process obtains a blurred image according to a Gaussian filtering algorithm, sets different parameters respectively, restores the edge of the blurred image, selects the optimal parameters, and traverses the blurred image. The parameters include a spatial scale operator, an edge strength operator, and the number of iterations.

6. A method for detecting defects in low-pressure casting of aluminum alloy metal molds as claimed in claim 4, characterized in that: The casting image includes a surface image and an internal image. The surface image is obtained by an industrial camera, and the internal image is obtained by X-ray irradiation.

7. The method for detecting defects in low-pressure casting of aluminum alloy metal molds according to claim 1, characterized in that: The step S3 includes the following sub-steps: Step S31, obtaining a current feature vector, a reference pore feature vector, a reference crack feature vector, a reference sand hole feature vector, a reference segregation feature vector, a reference deformation feature vector and a reference mold sticking feature vector; Step S32, respectively calculating the similarity between the current feature vector and the reference pore feature vector, the reference crack feature vector, the reference sand hole feature vector, the reference segregation feature vector, the reference deformation feature vector and the reference mold sticking feature vector; Step S33, sorting the similarities in descending order, retaining the similarity with the largest value, and obtaining the defect corresponding to the current feature vector according to the similarity with the largest value, and recording it as the current defect; Step S34, counting the number of times the current defect occurs, calculating the weights of different defect types in the overall defect types, and generating a casting decision based on the weights of different defect types in the overall defect types.

8. A method for detecting defects in low-pressure casting of aluminum alloy metal molds as claimed in claim 7, characterized in that: The calculation expression of the similarity is: The calculation expression of the weight is: Among them, XS is the similarity, x i is the i-th current feature vector of the casting image, y i is the i-th reference feature vector of the prior defect image, n is the number of casting images, Q t is the weight of the t-th prior defect type, r t is the number of occurrences of the t-th prior defect type, is the total number of occurrences of the prior defect type.

9. A method for detecting defects in low-pressure casting of aluminum alloy metal molds according to claim 7, characterized in that: The generation logic of the casting decision is: obtain the weights of different defect types in the overall defect types, compare the weights with the first weight threshold and the second weight threshold, when the weight is less than the first weight threshold, the casting decision is to ignore the defect type, when the weight is greater than or equal to the first weight threshold and less than the second weight defect type, the casting decision is to adjust the production steps corresponding to the defect with the first value as a gradient until the weight of the defect is less than the first weight threshold, when the weight is greater than or equal to the second weight threshold, the casting decision is to adjust the production steps corresponding to the defect with the second value as a gradient until the defect is less than the first weight threshold; When the current defect is a pore, the pouring speed is reduced with the first value as a gradient or the second value as a gradient. When the current defect is a crack, the core temperature is increased with the first value as a gradient or the second value as a gradient. When the current defect is a sand hole, the sand blowing speed is increased with the first value as a gradient or the second value as a gradient. When the current defect is segregation, the stirring speed is increased with the first value as a gradient or the second value as a gradient. When the current defect is deformation, the draft angle is increased with the first value as a gradient or the second value as a gradient. When the current defect is mold sticking, the alloy liquid temperature is reduced with the first value as a gradient or the second value as a gradient.

10. The method for detecting defects in low-pressure casting of aluminum alloy metal molds according to claim 1, characterized in that: The step S4 includes the following sub-steps: Step S41, obtaining a casting image and a current defect corresponding to the casting image, taking the casting image and the current defect corresponding to the casting image as a sample set, and dividing the sample set into a test set and a training set according to a certain ratio; Step S42, with the predicted current defect as the output, the actual current defect as the prediction target, and minimizing the sum of the prediction accuracies of all predicted current defects as the training target, the machine learning model is trained until the prediction accuracy reaches a first threshold.

Images