Design method of pressure-regulating precision-casting pouring system based on machine learning

Abstract

The invention provides a machine learning-based pressure-regulating precision casting gating system design method, which comprises the following steps of: collecting pressure-regulating precision casting data to form a database; screening the parameters in the database to obtain a principal component influence factor; pouring process independent variables are input, and a pouring system model is output based on the principal component influence factors and the pressure maintaining condition in the solidification process; based on the gating system model, anti-gravity pressure regulating casting mold filling simulation is conducted through analogue simulation software, and the porosity is counted according to the simulation result; and comparing the gating system model with a gating system in a database, if the porosity is the minimum value, determining that the gating system model is a target gating system, and exiting the whole process, otherwise, re-optimizing the gating system model. For complex parts in different shapes, an optimal pouring system can be provided, the mold filling capacity in the pouring process is improved, and the forming quality and the casting yield of castings are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of high-temperature alloy anti-gravity pressure-regulated casting, and in particular to a method for designing a pressure-regulated precision casting pouring system based on machine learning. Background Art

[0002] Pressure-regulated precision casting technology is an anti-gravity casting process that uses pressure regulation to fill the mold with liquid metal and solidify it. It is mostly used for the preparation of thin-walled complex castings. Pressure-regulated casting technology is an advanced casting forming technology proposed on the basis of traditional anti-gravity casting technology. Its main features are to reduce the gas content of the molten metal by means of vacuum pretreatment, negative pressure filling, pressure-regulated solidification, positive pressure shrinkage compensation, etc., to achieve smooth and efficient filling, avoid the involvement of gas and inclusions, strengthen the solidification sequence of the casting, improve the shrinkage compensation effect, and thus significantly improve the strength and plasticity of the casting, providing space for improving material utilization efficiency and reducing component weight. A major advantage of pressure-regulated precision casting is that it can manufacture high-quality high-temperature alloy thin-walled castings, and the working surface of the casting can achieve the dimensional accuracy and surface roughness requirements similar to polished castings without mechanical processing or with less processing.

[0003] In the casting process, the design of the gating system is crucial to the quality of the casting. The gating system usually includes a pouring cup, a sprue, a runner, and an ingate to guide the molten metal. A suitable gating system will not only improve the filling capacity of the molten metal, but also provide a shrinkage compensation function when the molten metal solidifies; an inappropriate gating system design will lead to defects such as shrinkage and thermal cracking, and will also reduce the surface quality of the casting. Generally speaking, the design of the gating system is developed through a large number of trial-and-error experiments conducted by engineers based on decades of engineering experience. The cost of learning and application is high, which is not conducive to improving the performance of casting alloys either in the laboratory or in engineering practice.

[0004] In recent years, the rapid development of artificial intelligence (AI) has proposed new solutions for the optimization of gating system design. Combining computational simulation and artificial intelligence (especially machine learning) can better predict casting defects. Machine learning algorithms use data to train, verify, evaluate and optimize models to predict the research objectives of materials. Due to their unique flexibility, rapid response, adaptability, prediction accuracy and excellent generalization ability, they have achieved remarkable success in many high-temperature alloy studies. With the iterative development of algorithms, maximum likelihood has developed into two main categories: traditional machine learning models and deep learning models. Traditional machine learning models usually exhibit relatively simple architectures and limited data processing capabilities; deep learning models are developed on the basis of traditional machine learning models to solve more complex problems. They have complex architectures and powerful data processing capabilities, so they require greater computing resources, making gating system design verification complex and costly. Summary of the invention

[0005] In view of the defects in the prior art, the purpose of the present invention is to provide a pressure-regulated precision casting pouring system design method based on machine learning.

[0006] The present invention provides a method for designing a pressure-regulated precision casting pouring system based on machine learning, comprising:

[0007] Collect pressure-regulated precision casting data and form a database;

[0008] Screening the parameters in the database to obtain the main component influencing factors;

[0009] Input the pouring process independent variables, and output the pouring system model based on the main component influencing factors and the pressure holding condition of the solidification process;

[0010] Based on the pouring system model, simulation software is used to perform anti-gravity pressure-regulated casting filling simulation, and the porosity is calculated according to the simulation results;

[0011] The pouring system model is compared with the pouring system in the database. If the porosity is the minimum value, the pouring system model is the target pouring system and the whole process is exited. Otherwise, the pouring system model is optimized again.

[0012] Optionally, the pressure-regulated precision casting data is collected to form a database, wherein the independent variables in the quantity database include data from two processes, namely, filling and solidification, of the pressure-regulated precision casting.

[0013] Optionally, for the filling process, the independent variables include at least one of the pressure difference between the upper and lower cavities of the mold shell, the filling speed, the alloy liquid temperature, the mold shell temperature, the pouring system diameter, the pouring system length, and the heat exchange coefficient of the mold shell and the alloy liquid.

[0014] Optionally, for the solidification process, the independent variables are the holding pressure and the holding time.

[0015] Optionally, the parameters in the database are screened, wherein: a radial basis function neural network is used to screen the parameters in the database.

[0016] Optionally, the input casting process parameters include: the casting process parameters include at least one of alloy liquid casting temperature, metal liquid composition, mold shell temperature, mold shell type, casting system diameter, casting system length and mold shell alloy liquid heat exchange coefficient.

[0017] Optionally, after calculating the porosity according to the simulation results, the method further includes: calculating the casting yield according to the simulation results.

[0018] Optionally, the optimizing the pouring system model includes: adjusting an input independent variable influencing factor.

[0019] Furthermore, the optimization of the pouring system model also includes: adjusting the solidification sequence of various parts of the casting.

[0020] Furthermore, after optimizing the pouring system model, the method further includes: if the target pouring system is not obtained, reselecting a pouring system from the database for optimization.

[0021] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0022] This application reduces the complexity of the gating system design verification by establishing a database and determining the main component influencing factors, and optimizes the gating system design in the anti-gravity pressure-regulated precision casting process by combining artificial intelligence methods. For complex parts of different shapes, the optimal gating system is given to make the filling and solidification process of the molten metal smoother, improve the filling capacity of the pouring process, greatly reduce or even eliminate casting defects such as cold shut and insufficient pouring, and improve the forming quality and finished product rate of the casting. At the same time, it is beneficial to reduce the reaction between the molten metal and the mold shell, reduce the number of flash, burrs, etc., and greatly improve the surface quality of the casting. In addition, the research and design of the gating system through machine learning can solve the problem that the trial and error method of experimental nature consumes a lot of time and causes waste of energy and materials, which is greatly beneficial to the development and application of new technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0024] Figure 1 Schematic diagram of the flow of a method for designing a pressure-regulated precision casting system based on machine learning in one embodiment of the present invention. DETAILED DESCRIPTION

[0025] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several variations and improvements may be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0026] Reference Figure 1 As shown in the flow chart, a method for designing a pressure-regulated precision casting system based on machine learning according to an embodiment of the present invention comprises the following steps:

[0027] S1. Collect pressure-regulated precision casting data and form a database;

[0028] S2. Screen the parameters in the database to obtain the main component influencing factors;

[0029] S3, inputting casting process parameters, specifically, including casting process parameters such as casting alloy liquid composition, temperature, shell composition, and the role of the casting part, and outputting the casting system model based on the main component influencing factors and the pressure holding condition of the solidification process;

[0030] S4. Based on the pouring system model, simulation software such as ProCast software is used to simulate the anti-gravity pressure-regulated casting filling, and the porosity is calculated according to the simulation results;

[0031] S5. Compare the pouring system model with the pouring system in the database. If the porosity is the minimum value, the pouring system model is the target pouring system and the whole process is exited. Otherwise, the pouring system model is optimized again.

[0032] In the embodiment of the present invention, in step S1, the collected data comes from the actual casting production process, including various parameters such as alloy liquid pouring temperature, shell material, shell temperature, holding pressure, etc. For pressure-regulated precision casting, the design of the pouring system needs to take into account the influencing factors in the two processes of filling and solidification. Therefore, the independent variables in the database include data in the two processes of filling and solidification of pressure-regulated precision casting, where the independent variables refer to parameters such as material composition, membrane shell material pouring temperature, etc. that require pressure-regulated casting pouring system design.

[0033] In some embodiments, for the filling process, the independent variables include at least one of the pressure difference between the upper and lower cavities of the mold shell, the filling speed, the alloy liquid temperature, the mold shell temperature, the pouring system diameter, the pouring system length, and the heat exchange coefficient of the mold shell and alloy liquid.

[0034] In some embodiments, for the solidification process, the independent variables are the holding pressure and / or the holding time, etc.

[0035] The above-mentioned embodiment of the present invention forms a database of the anti-gravity pressure-regulated precision casting pouring system by collecting data based on existing empirical data.

[0036] Pressure-regulated casting is the same as the existing anti-gravity casting principle, but the process is different, so the original influencing parameters of anti-gravity casting have different influences in the pressure-regulated casting process, so it is necessary to screen a large number of influencing factors to achieve a degree that can be calculated and simulated. In some embodiments, in step S2, a radial basis function neural network (RBF) is used to screen the parameters in the database to obtain the principal component influencing factors. The training of the radial basis function neural network is usually carried out in an unsupervised learning manner. First, it is necessary to determine the number of hidden layer neurons, which is usually achieved by techniques such as grid search. Then, the parameters of the network are trained in an unsupervised learning manner, such as minimizing the error between the output layer and the true value. After the training is completed, the trained radial basis function neural network can be used to predict new data. The design of the anti-gravity pressure-regulated precision casting pouring system is optimized using RBF, which has the advantages of high efficiency, flexibility and controllability.

[0037] In the above-mentioned embodiment of the present invention, the principal component analysis method is used to reduce the dimension of the influencing factors. Different component materials and different membrane shell casting processes have different influencing factors. The principal component influencing factors are the correlation degree of the input variables, which are the most critical factors affecting the pressure regulating casting process. These factors are counted and their respective influence degrees on the pressure regulating casting are obtained. After the input sample composition, size and other parameters interact with the principal component influencing factors, the most reasonable casting system can be obtained.

[0038] The above-mentioned embodiment of the present invention forms a database by collecting a large amount of data and determines the main component influencing factors of the pressure-regulated precision casting pouring system, which can reduce the amount of calculation and facilitate the establishment of a machine learning model. The most critical factors are obtained through the principal component analysis method; by adjusting the degree of influence of different factors and combining the input pouring process parameters, a pouring model is obtained, and a filling model is made using PROCAT software, etc., to detect the filling condition and defect condition, thereby guiding the design of the pouring system.

[0039] In some embodiments, in step S3, the pouring process parameters include parameters related to the properties of the molten metal and the shell before the molten metal enters the shell, illustratively including at least one of the alloy liquid pouring temperature, molten metal composition, shell temperature, shell type, pouring system diameter, pouring system length, and heat exchange coefficient of the shell alloy liquid. In this step, the casting model formed by a large amount of casting production process data collected in the previous step is input with parameters such as the composition of the poured sample. After the parameters are input, interactive calculations are performed based on the existing data model, the main component influencing factors, and the pressure holding conditions of the solidification process. Specifically, pressure-regulated casting includes two links, namely, low-pressure filling and pressure-holding solidification. Low-pressure filling can ensure a smooth filling process without splashing, and pressure-holding solidification is not only conducive to the stability of the thin-walled structure, but also makes the entire workpiece structure evenly distributed through a certain pressure, improving performance, thereby obtaining a reasonable pouring system design.

[0040] In step S4, after the porosity is counted according to the simulation results, it also includes: counting the casting yield rate according to the simulation results. By counting the casting yield rate, it is determined whether the workpiece after molding, that is, the workpiece taken out from the pouring system, has large-scale defects, that is, insufficient pouring in a large area, to ensure that the structural requirements of the workpiece can be met. In general casting processes, the performance needs to be further enhanced by heat treatment after molding. Therefore, it is necessary to consider whether its strength performance meets the service performance requirements after heat treatment and strengthening processes. If the above requirements are met, it is considered that the pouring system design is reasonable and usable, and can be used to guide the subsequent pouring system design; if the above requirements are not met, it is necessary to search for other pouring systems in the database for replacement.

[0041] In step S5, the porosity corresponding to the gating system in the database is obtained during the initial data collection process. If it is a gating system of an existing material, it is compared in the database; otherwise, it is necessary to compare the casting results of other types of gating systems of the same material.

[0042] In some optional embodiments, in step S5, the pouring system model is optimized, including: adjusting the influencing factors of the input independent variables. In step S2, the most critical factors are obtained through principal component analysis, but for different pouring scenarios, the influence of factors such as alloy liquid temperature, shell type, shell temperature, shell composition, etc. is different. Therefore, the influence of each factor needs to be considered according to the actual situation, and the influence factor is flexibly adjusted, that is, the influence of different components in the main component is adjusted. Specifically, the influence of each main component influencing factor is first adjusted. If a reasonable pouring system cannot be obtained, the model is adjusted, that is, the influence factors corresponding to other main components are used.

[0043] If the target molded sample cannot be obtained after all the above-mentioned principal component factors have been adjusted, it is necessary to consider whether the solidification sequence is reasonable. In some optional embodiments, in step S5, the pouring system model is optimized, and also includes: adjusting the solidification sequence of each part of the casting. It should be noted that the solidification sequence has a certain influence on casting. Considering the counter-gravity filling process, the counter-gravity filling is generally from bottom to top, and the influence of the solidification sequence is small, and the control of the solidification sequence is more complicated, and it is necessary to comprehensively consider the design of various aspects of the pouring system. Therefore, the solidification sequence is adjusted after adjusting the input independent variable influencing factors.

[0044] In a further embodiment, after optimizing the gating system model, the method further includes: if the target gating system is not obtained, reselecting the gating system model from the database.

[0045] When collecting a large amount of casting data, the design model of the existing pouring system will also be collected, but whether it is the most reasonable model remains to be confirmed. In the casting process, the most important defects are shrinkage and shrinkage holes. Both types of defects can be characterized by statistical porosity. Other types of defects such as insufficient pouring and cold shut are relatively obvious pouring failures. Therefore, when the porosity is the minimum value, it can be considered that the optimal pouring system design is obtained. If the porosity is not the minimum value, the various ratios of the main component influencing factors are changed. If the porosity still cannot be reduced, consider adjusting the solidification sequence. Reselecting the pouring system is when a suitable pouring system design cannot be obtained after comprehensively considering the main component influencing factors and the solidification sequence. It is necessary to reconsider whether this type of pouring system is suitable for the casting process of the workpiece. Therefore, other existing pouring systems in the database can be tried to detect whether a suitable runner design can be output. The above embodiment of the present invention compares the pouring system models obtained by different methods to obtain the most suitable model for guiding the subsequent pouring system model design.

[0046] The above embodiment of the present invention optimizes the gating system design in the process of anti-gravity pressure-regulated precision casting by establishing a database and combining the artificial intelligence method, calculates the most critical factors by using the principal component analysis method, and combines the input process parameters with the model formed in the database to calculate a new gating system. After obtaining the machine learning model by combining calculation and simulation, each input can promote the model's ability to analyze and dissect complex casting components, and in the process of establishing the model, the complexity of the components will also be considered in the collected casting data, so that for complex parts of different shapes, the optimal gating system is given to improve the filling capacity of the casting process. The design of the gating system is less dependent on the "master", and the optimized gating system will make the filling and solidification process of the molten metal smoother, greatly reduce or even eliminate casting defects such as cold shut and insufficient pouring, and improve the forming quality and finished product rate of the casting. At the same time, it is conducive to reducing the reaction between the molten metal and the mold shell, reducing the number of flash and burrs, and greatly improving the surface quality of the casting. In addition, anti-gravity precision casting is a new casting process proposed in recent years. Different from traditional casting methods, not only the properties of the alloy liquid and the mold shell itself have a great influence on the filling and solidification process, but also the pressure difference between the upper and lower chambers during the filling process and the size of the holding pressure during the solidification process have a great influence on the final yield of the casting. The trial and error method of experimental nature will not only consume a lot of time, but also easily cause waste of energy and materials. The research and design of the pouring system through machine learning can avoid repeated experiments, which can effectively solve the above problems and greatly benefit the development and application of new technologies.

[0047] For anti-gravity pressure-regulated precision casting, the method in the above embodiment of the present invention can promote the development and application of new casting processes. It should be noted that the method in the above embodiment is not limited to pressure-regulated precision casting. By expanding the database, the above method can be used for all casting process gating system designs.

[0048] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various modifications or variations within the scope of the claims, which does not affect the essence of the present invention. The above preferred features can be used in any combination without conflicting with each other.

Claims

1. A method for designing a pressure-regulated precision casting gating system based on machine learning, characterized in that: include: Collect pressure-regulated precision casting data and form a database; Screening the parameters in the database to obtain the main component influencing factors; Input the pouring process independent variables, and output the pouring system model based on the main component influencing factors and the pressure holding condition of the solidification process; Based on the pouring system model, simulation software is used to perform anti-gravity pressure-regulated casting filling simulation, and the porosity is calculated according to the simulation results; The pouring system model is compared with the pouring system in the database. If the porosity is the minimum value, the pouring system model is the target pouring system and the whole process is exited. Otherwise, the pouring system model is optimized.

2. The method for designing a pressure-regulated precision casting gating system based on machine learning according to claim 1, characterized in that: The pressure-regulated precision casting data is collected to form a database, wherein the independent variables in the database include data from the two processes of filling and solidification of the pressure-regulated precision casting.

3. The method for designing a pressure-regulated precision casting gating system based on machine learning according to claim 2 is characterized in that: For the filling process, the independent variables include at least one of the pressure difference between the upper and lower cavities of the mold shell, the filling speed, the alloy liquid temperature, the mold shell temperature, the pouring system diameter, the pouring system length and the heat exchange coefficient of the mold shell alloy liquid.

4. The method for designing a pressure-regulated precision casting system based on machine learning according to claim 2, characterized in that: For the solidification process, the independent variables are the holding pressure and / or the holding time.

5. The method for designing a pressure-regulated precision casting gating system based on machine learning according to claim 1, characterized in that: The parameters in the database are screened, wherein: a radial basis function neural network is used to screen the parameters in the database.

6. The method for designing a pressure-regulated precision casting gating system based on machine learning according to claim 1, characterized in that: The input casting process parameters include at least one of alloy liquid casting temperature, metal liquid composition, mold shell temperature, mold shell type, casting system diameter, casting system length and heat exchange coefficient of mold shell alloy liquid.

7. The method for designing a pressure-regulated precision casting gating system based on machine learning according to claim 1, characterized in that: After calculating the porosity according to the simulation results, the method further includes: calculating the casting yield rate according to the simulation results.

8. The method for designing a pressure-regulated precision casting system based on machine learning according to claim 1, characterized in that: The optimization of the pouring system model includes: adjusting the input independent variable influencing factor.

9. The method for designing a pressure-regulated precision casting pouring system based on machine learning according to claim 8, characterized in that: The optimization of the pouring system model also includes: adjusting the solidification sequence of various parts of the casting.

10. The method for designing a pressure-regulated precision casting gating system based on machine learning according to claim 1, characterized in that: After the pouring system model is optimized, the method further includes: if the target pouring system is not obtained, reselecting a pouring system from the database for optimization.

Images