Pouring liquid filter design method and pouring liquid filter device

Abstract

Disclosed is a filter design method based on pouring data, comprising: obtaining pouring variables and configuration variables in a plurality of filter samples, and determining filtering results of the plurality of filter samples on pouring liquid; screening the plurality of filter samples based on the filtering results, and performing first processing on the pouring variables of the screened filter samples to obtain a plurality of sample characteristics, and constructing a characteristic matrix of the filter samples; constructing a prediction model by using the characteristic matrix and the configuration variables; inputting the required pouring variables after calculation and preprocessing into the prediction model, and designing the filter by using the value interval of the configuration variables output by the prediction model, so that the pouring liquid can flow through the filter faster under the premise of ensuring high impurity filtration rate, thereby improving the filtration efficiency and being conducive to the manufacturing cost control of the filtering device.

Description

Case Analysis

[0001] This case is a divisional application of the application filed on August 14, 2025, entitled "Design Method and Apparatus for Casting Liquid Filter Components" and with application number 202511149013X. Technical Field

[0002] This application belongs to the field of casting technology, specifically relating to the design method of casting liquid filter element and casting liquid filtration device. Background Technology

[0003] The gating system for castings is usually designed based on the casting material, casting size, and production equipment. The molten metal used for casting may contain non-metallic impurities. During the casting process, these molten metal materials may also produce inclusions after contact with air or the casting. For example, when casting aluminum alloys, increasing the casting time may cause some of the molten metal to transform into alumina.

[0004] Therefore, in a gating system, filtration devices (such as filters of various shapes) are usually installed upstream or downstream of the casting liquid to remove impurities, ensuring high purity of the casting liquid flowing into the casting and that the purity of the manufactured casting meets product requirements. For example, CN104128599A discloses a molten metal filtration device with multiple filters arranged in parallel between the inlet and outlet chambers of the device. A buffer plate is installed between the filters and the inlet to effectively prevent the filters from deforming or breaking due to the impact of the injection liquid. However, this filtration configuration design is only suitable for large casting gating systems, where the components are all large, making it difficult to extend this configuration to smaller gating systems.

[0005] Furthermore, CN221983843U discloses a casting system for a wind turbine bearing housing casting. In this system, a filter channel is formed by a first filter section and a second filter section, and a foam filter element is provided between the first filter section and the second filter section. The periphery of the two filter sections that are in contact with the foam filter element is recessed inward by 8-12 mm. By providing sufficient molding sand support, the filter element is effectively prevented from being crushed. However, this system does not mention the optimization of the filtration performance of the filter element itself, so its filtration efficiency cannot be determined.

[0006] Therefore, a filter element design method and a filter device based on the filter element designed by the method are needed, so that the filter device can not only provide good impurity filtration, but also ensure stable flow of the casting liquid and avoid fluid blockage. Summary of the Invention

[0007] To address the problems existing in the prior art, the purpose of this application is to provide a design method for a casting fluid filter element. The design method constructs a predictive model based on the configuration variables of multiple filter element samples and the casting variables of the casting system in which they are located. The model is used to predict the preferred filtration length, preferred filtration pore size range, and / or preferred length / pore size range ratio provided by the filter element itself in different casting system models, so as to optimize the balance between impurity interception and casting fluid flow of the filter element, thereby making the filter element and the entire filtration device more practical.

[0008] Specifically, this application relates to the following aspects: According to one aspect of this application, a filter element design method based on casting data is provided, comprising: information acquisition, obtaining casting variables and configuration variables for each filter element sample among multiple filter element samples, and determining the filtration result of each filter element sample on the casting liquid; feature calculation, screening multiple filter element samples based on the filtration results and performing a first processing on the casting variables of each filter element sample among the screened multiple filter element samples to obtain multiple sample features, and constructing a feature matrix for each filter element sample among multiple filter element samples; parameter prediction, constructing a prediction model using the feature matrix and configuration variables of each filter element sample among multiple filter element samples; and device construction, inputting the required casting variables after calculation and preprocessing into the prediction model, and designing the filter element based on the value range of the configuration variables output by the prediction model; wherein, the filtration results include: slag inclusions and porosity on the casting surface and inside the casting, casting tensile strength, and casting pass rate; casting variables include casting weight, runner filtration volume, casting temperature, casting time and / or casting method, and configuration variables include filtration length and / or pore size range.

[0009] According to some embodiments of this application, the first process calculates multiple sample features using the pouring variables, including: performing exponential, logarithmic, and power series transformations on the pouring weight, runner filtration volume, pouring time, and / or pouring temperature to obtain nonlinear features of the pouring variables; performing pairwise multiplication and / or pairwise division on the pouring weight, runner filtration volume, and / or pouring temperature to obtain cross features of the pouring variables; and combining the pouring weight, runner filtration volume, and / or pouring temperature with the pouring time to obtain a time-domain signal of the pouring variables changing with time, and performing frequency domain processing on the time-domain signal to obtain time-frequency features.

[0010] According to some embodiments of this application, the first process of calculating multiple sample features using casting variables further includes: encoding the casting method to obtain category features; normalizing nonlinear features, cross features and / or time-frequency features and concatenating them with the category features respectively to obtain a feature matrix.

[0011] According to some embodiments of this application, before inputting the configuration variable into the prediction model, the method further includes: multiplying or dividing the filter length and pore size range in pairs, so that the configuration variable is updated to the product of the filter length and the pore size range or the ratio of the filter length to the pore size range.

[0012] According to some embodiments of this application, constructing a prediction model includes: determining the kernel function of the prediction model to calculate the similarity relationship between the feature matrix and the configuration variables; training the prediction model using multiple filter samples; and iterating the kernel parameters of the kernel function and the inter-variable covariance matrix of the configuration variables by comparing the prediction results of the prediction model with the actual results, so that the prediction model reaches a preset index or reaches the maximum number of iterations.

[0013] According to some embodiments of this application, constructing a prediction model includes: determining the kernel function of the prediction model to calculate the similarity relationship between the feature matrix and the configuration variables; training the prediction model using multiple filter samples; and iterating the kernel parameters of the kernel function by comparing the prediction results of the prediction model with the actual results, so that the prediction model reaches a preset index or reaches the maximum number of iterations.

[0014] According to some embodiments of this application, constructing a casting fluid filtration device using configuration variables output by a prediction model includes: calculating and preprocessing the required casting variables to obtain a feature matrix, inputting the feature matrix into a prediction model; the prediction model outputs the predicted mean and predicted standard deviation of the configuration variables, calculates the a% confidence interval of the configuration variables, and selects the values ​​of the configuration variables within the a% confidence interval as the values ​​of the configuration variables of the casting fluid filtration device.

[0015] According to some embodiments of this application, the a% confidence interval of the configuration variable is a 95% confidence interval.

[0016] According to another aspect of this application, a filtering device is also provided, comprising: a filter element fixed in and partially protruding from a runner, including a plurality of filter layers, each of the plurality of filter layers having a plurality of filter holes; a runner for receiving molten metal flowing therein, including a receiving portion for fixing the filter element and at least partially receiving the filter element; wherein the ratio of the aperture range of the plurality of filter holes to the thickness of the filter element is between a first threshold and a second threshold.

[0017] According to some embodiments of this application, the ratio of the pore size range of a plurality of filter holes to the thickness of the filter element being located between a first threshold and a second threshold includes: setting the ratio of the pore size range of a plurality of filter holes to the thickness of the filter element to be the ratio of the larger of the values ​​to the smaller of the values, such that the first threshold is 1 ppi / mm or 1 mm / ppi and the second threshold is 4 ppi / mm or 4 mm / ppi.

[0018] According to some embodiments of this application, the first threshold to the second threshold is the range of values ​​for the configuration variable predicted by the aforementioned filter design method based on casting data.

[0019] According to some embodiments of this application, the pore size of the plurality of filter holes ranges from 5 ppi to 40 ppi; the thickness of the filter element is from 10 mm to 40 mm.

[0020] According to some embodiments of this application, the length of the filter element protruding from the gating system ranges from 2 mm to 15 mm; the cross-sectional shape of the filter element along the direction perpendicular to the flow velocity of the poured molten metal is one of a circle, an ellipse, a triangle, or an n-sided polygon; wherein n ≥ 4.

[0021] According to some embodiments of this application, in the gating system, the receiving portion has a first buffer area and a second buffer area on both sides; the edge of each buffer area extends from the edge of the gating system at a distance of a third threshold from the filter element to each side of the filter element at a certain angle to the axial direction of the gating system.

[0022] According to some embodiments of this application, the range of the third threshold is 20 mm to 300 mm.

[0023] According to some embodiments of this application, when the length or diameter of the filter element is not greater than 60 mm, the thickness of the filter element is 10 mm to 25 mm; when the length or diameter of the filter element is greater than 60 mm, the thickness of the filter element is 15 mm to 40 mm.

[0024] According to some embodiments of this application, the filter element is fixed in the receiving part in the gating channel, perpendicular or parallel to the direction of the flow rate of the molten metal. Attached Figure Description

[0025] Figure 1 The illustration shows a flowchart of a filter design method based on casting data according to an embodiment of this application.

[0026] Figure 2A The illustration shows a first schematic diagram of casting variables and configuration variables according to an embodiment of this application.

[0027] Figure 2B The illustration shows a second schematic diagram of casting variables and configuration variables according to an embodiment of this application.

[0028] Figure 3A The figure shows a schematic diagram of the structure of a filter according to an embodiment of this application.

[0029] Figure 3B The figure shows a cross-sectional schematic diagram of a filtering device according to an embodiment of the present application.

[0030] Figure 4A The illustration shows a first schematic diagram of the use of a filter element and a gating system in accordance with an embodiment of this application.

[0031] Figure 4B The illustration shows a second schematic diagram of the use of a filter element and a gating system in accordance with an embodiment of this application.

[0032] Figure 5A The illustration shows a third schematic diagram of the use of a filter element and a gating system in accordance with an embodiment of this application.

[0033] Figure 5B The illustration shows a fourth schematic diagram of the use of a filter element and a gating system in accordance with an embodiment of this application.

[0034] Figure Labels 1-Filter element; 2-Gating system; 10-Filter layer; 11-Filter holes 20-Containing section 21-First gating 22-Second gating 200 - First buffer zone; 210 - Second buffer zone Detailed Implementation

[0035] The present application is further illustrated below with reference to embodiments. It should be understood that the embodiments are only used to further illustrate and explain the present application and are not intended to limit the present application.

[0036] Unless otherwise defined, technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art. While similar or identical methods and materials may be applied in experimental or practical applications, materials and methods are described herein. In case of conflict, the definitions included herein shall prevail. Furthermore, materials, methods, and examples are for illustrative purposes only and are not intended to be limiting. The present application is further described below with reference to specific embodiments, but is not intended to limit the scope of the application.

[0037] Application Overview As mentioned above, there are many filtration devices or filter elements with various configurations available on the market that are suitable for non-pouring environments. However, their applicability is relatively limited. A filtration system used in a certain casting operation may become unusable after the material, amount, and specified pouring time of the pouring liquid are changed. This may be because its filtration structure is too rigid, which greatly reduces the efficiency of the normal pouring liquid flow. It may also be because it cannot be effectively used with a new pouring system, resulting in a deterioration in the filtration effect. A typical example is that the same filter element is difficult to use universally in pouring systems with different pouring methods (such as when the position of the filter element or the shape of the gating channel needs to be changed).

[0038] The limitations of these existing filters often necessitate a reassessment of their usage when altering at least a portion of the gating system's configuration to determine whether iterative design changes are necessary, such as optimizing their pore size range. Without such assessment, quality issues in the finished castings may emerge, either proactively or reactively, after multiple casting operations, leading to dissatisfaction among downstream users and damage to the filter manufacturer's reputation. There is a need in the art for a design method that can evaluate and predict suitable filters for a specific casting operation scenario based on parameters, thereby providing manufacturers with optimal manufacturing solutions for the filters while meeting casting requirements. Furthermore, this design method can yield filters with optimized configuration parameters within a certain applicable range, and filtration devices constructed based on these filters.

[0039] Based on the above considerations, this application proposes a filter design method based on casting data. The aim is to determine how to design a good filter under given working conditions by constructing a framework of in-depth relationships between various inherent variables of the casting system with known filtration capacity and variables that interact with the casting liquid during the casting process. In particular, it aims to determine the optimal filtration thickness and pore size, thereby guiding those skilled in the art to manufacture filters most suitable for given working conditions. From a methodological perspective, this avoids the situation where the filters manufactured at the production end are unsuitable or not durable, so as to achieve efficiency improvement and cost reduction in casting.

[0040] Specifically, the pore size range (in ppi) mentioned in this application refers to the number of pores within a specific diameter (or length) range. Unless otherwise specified, this specific diameter (or length) is typically 25.4 mm, or one inch. Industrially, a line segment of length A (in mm) is drawn on the filter element, and the number of pores B that the line passes through is observed and recorded. The pore size X (in ppi) is then calculated using the formula X = A / B * 25.4. Since the pores are small and their number is difficult to control precisely during manufacturing, the pore size X can be approximate (e.g., close to an integer) to facilitate filter element production. It can be considered that a larger pore size range in ppi indicates a higher density and smaller gaps in the pores; conversely, a smaller pore size range in ppi indicates a lower density and larger gaps in the pores. The purpose of considering the pore size range is that, since the size of a single filter pore is typically very small, it is extremely difficult to precisely control the size and position of a single filter pore during the manufacturing process of the filter element. Therefore, using predictive models to predict and optimize the diameter or length (or even the specific shape) of a single filter pore is not significant in the actual production of the filter element. Therefore, this application instead optimizes the overall pore size range of the filter element based on its working environment to determine the optimal overall size and density of multiple filter pores in the filter element under different pouring liquid volumes, temperatures, and filter runner sizes.

[0041] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0042] Exemplary methods Figure 1 The illustration shows a filter design method based on casting data according to an embodiment of this application.

[0043] like Figure 1 As shown, the filter design method based on casting data according to an embodiment of this application includes the following steps: Step S110: Obtain the casting variable and configuration variable of each filter element sample among the multiple filter element samples, and determine the filtration result of each filter element sample on the casting liquid among the multiple filter element samples; Step S120: Based on the filtering results, filter the plurality of filter element samples and perform a first processing on the casting variable of each filter element sample among the filtered plurality of filter element samples to obtain multiple sample features, and construct a feature matrix for each filter element sample among the plurality of filter element samples. Step S130: Construct a prediction model using the feature matrix and configuration variables of each of the multiple filter samples; Step S140: The required casting variables are calculated and preprocessed and then input into the prediction model. The configuration variables output by the prediction model are used to construct the casting fluid filtration device.

[0044] In step S110, the multiple filter sample samples are samples of gating systems for castings made of stainless steel, carbon steel, and low alloy steel at temperatures of 1500℃ and above, collected by the applicant from production lines and customers, and samples of multiple filters used therein. These samples include pouring variables of the pouring liquid used in the gating system, including pouring volume, filtration volume of the filtration device (including filters for removing solid impurities and runners for restricting the flow of the pouring liquid), and the pouring temperature, pouring time, and pouring method (divided into ladle pouring or bottom pouring) required for the pouring operation set by the gating system. Furthermore, the parameters of the filters used in the gating system included in the multiple samples also include configuration variables, including the filtration length and pore size range of the filter itself. The filtration length refers to the path length of the pouring liquid flowing through the filter, which can be understood as equivalent to the thickness of the filter itself. Additionally, these samples also include finished castings manufactured by the gating system under the above parameters.

[0045] For each of the multiple samples, the applicant also collected its filtration results, typically categorized as good or bad. For the gating system, there are many indicators to measure the quality of the cast parts. Among these, indicators reflecting the impurity filtration level of the filtration device mainly include: inclusions on and inside the casting, porosity, tensile strength, and casting pass rate. Inclusions visually indicate the remaining impurities; high porosity, such as a high porosity scrap rate, indicates excessive impurities leading to uneven flow of the casting liquid and numerous pores; mechanical properties such as tensile strength indirectly reflect impurity content, with lower impurity content resulting in mechanical properties closer to theoretical levels; finally, the casting pass rate, such as the process yield rate, is a statistical result from the production end, reflecting the overall casting capacity and stability of the gating system.

[0046] The testing process for the above indicators is as follows: For the scrap rate of castings with inclusions and porosity, the applicant uses liquid penetrant testing (PT) to test, requiring that any linear indices on the good sample be no larger than 2 mm, the size of a single circular indices be no larger than 4 mm, and dense indices be no larger than 2 mm in any 100 mm * 100 mm area. For the casting yield (process output rate), the applicant uses volumetric radiography (RT) to test, requiring that the average internal quality yield of the good sample measured by this method reach 95% or higher, which is much higher than the current average yield of 71.2% for castings without filters. For the mechanical properties of castings, the applicant uses a universal testing machine to test the tensile strength (Rm), requiring that the tensile strength of the good sample be no less than 612 MPa.

[0047] Furthermore, samples where all the above indicators are above the average level are designated as good samples, and the remaining samples as bad samples. This ensures that the samples included in the model training do not contain any parameters that would lead to poor filtering performance. The resulting prediction model has higher accuracy and is more practical in guiding filter component production. Therefore, this application further selects samples with good filtering results and uses these samples as valid data to construct the prediction model for the configuration variables.

[0048] In one example, the variable values ​​of the multiple filter element samples obtained based on the filtering results are shown in Table 1 below. These filter element samples have good filtering performance in their gating system. Therefore, the complex relationship between the gating variables and configuration variables in the samples can be further explored to determine how the values ​​of the gating variables affect the design of the filter element itself.

[0049] Table 1 shows the parameter values ​​for multiple filter samples.

[0050] Specifically, the overall gating weight is usually considered together with the overall volume provided by the gating system's runner, i.e., the ratio of gating weight to runner filtration volume is treated as a gating variable to reflect the amount of gating liquid required to be filtered per unit volume under different tasks in different gating systems. However, those skilled in the art can also consider both as independent parameters for training the prediction model to highlight their independent contributions to the filter element, or consider their product as a polynomial expansion of the gating variable to derive information, using this derived information as a sample feature to train the prediction model. When the nonlinear relationship between the gating variable and the configuration variable is strong (which is indeed the case, as traditional linear regression methods are not good at processing gating system data), it is preferable to include the product of the gating weight and the runner filtration volume in the sample features to help the prediction model learn the synergistic effect between the two and improve the linearity between variables.

[0051] In the method described in this application, pouring temperature and time are two other pouring variables that need to be considered as having a significant impact on the filter element. It is understood that higher pouring temperatures pose a greater challenge to the structural stability of the filter element, while longer pouring times increase the likelihood of impurity formation and filter element clogging. Pouring methods are generally divided into two types: shaker (teapot pourer) and bottom pourer. The gating system under a shaker has a shape similar to a teapot, and the flow of the pouring liquid from the gate to the mold is controlled by tilting or shaking the gating system. Bottom pourers, on the other hand, typically use larger capacities and more stable gating structures, making them suitable for pouring large castings. Therefore, the fluid state of the pouring liquid in the gating system and filter element differs significantly between the two pouring methods, and their impact on the filtration of impurities needs to be considered separately.

[0052] In step S120, the first processing may include preprocessing and feature construction of the multiple casting variables obtained in step S110 to obtain sample features containing deep correlation information of the configuration variables of the filter element. In one example, one or more casting variables may be subjected to exponential, logarithmic, and / or power series transformations to obtain their respective nonlinear independent features, thereby reducing the data unevenness of the left-skewed or right-skewed distribution of the original casting variables. In another example, polynomial cross features such as the ratio of casting weight to runner filter volume and / or the product of casting weight to runner filter volume, as mentioned above, may be constructed. Since the casting variables selected in this application are all macroscopic variables of the casting system with significant differences and relatively easy-to-obtain parameters, it is necessary to consider the synergistic effect between casting variables one by one. Such cross features may also be the product and / or ratio of casting time to casting weight, the product and / or ratio of runner filter volume to casting temperature, etc. This application will not elaborate on specific examples of the pairwise products / ratios of all casting variables here.

[0053] For optional, more detailed feature mining, the method described in this application also considers a construction approach that combines the pouring weight, gating filter volume, and / or pouring temperature with the pouring time to obtain time-domain signals of other pouring variables changing with the pouring time. These time-domain signals are then transformed into time-frequency features through frequency-domain processing. For example, the values ​​of pouring temperature and / or pouring weight are continuously recorded during the pouring time (requiring additional temperature and / or mechanical sensors and proper insulation) to obtain time-varying curves of pouring temperature and / or pouring weight. The endpoints of the curves represent the start and end points of the pouring time. Then, a Fourier transform is performed on this signal, and low, medium, and high-frequency amplitudes are extracted as new sample features. These new time-frequency features contain more and more complex mechanical amplitude-frequency response characteristics of the pouring system itself, enabling the prediction model to learn more deeply about the influence of structural changes in the casting process of the pouring system on the filter. However, this significantly increases the economic and time costs of obtaining filter samples, and therefore is only considered as an optional solution.

[0054] Specifically, before inputting the multiple sample features obtained through the above method into the prediction model, it may also include calculating the Pearson correlation coefficient between the multiple sample features to consider the correlation between the features and the configuration variables, and eliminating features unrelated to the configuration variables to avoid model overfitting caused by the explosion of sample feature dimensions. For the example above in this application, considering that the number of features is relatively large compared to the number of samples, this operation can be preferably implemented.

[0055] That is, according to the first processing of the embodiments of this application, the plurality of sample features are calculated using the pouring variables, including: performing exponential, logarithmic, and power series transformations on the pouring weight, runner filtration volume, pouring time, and / or pouring temperature respectively to obtain the nonlinear features of the pouring variables; performing pairwise multiplication and / or pairwise division on the pouring weight, runner filtration volume, and / or pouring temperature respectively to obtain the cross features of the pouring variables; and combining the pouring weight, runner filtration volume, and / or pouring temperature with the pouring time respectively to obtain the time-domain signal of the pouring variables changing with time, and performing frequency domain processing on the time-domain signal to obtain time-frequency features.

[0056] Considering that the above feature construction method lacks the inclusion of the casting method in the casting variables, the sample features can be preprocessed (including outlier removal, missing value imputation, and normalization) and then concatenated with the casting method to incorporate the casting method into the sample features. Specifically, in one example, the applicant uses the above feature construction method (i.e., the first processing) to obtain the casting volume / sprue filter volume, casting temperature, casting time, and the logarithm, square, and square root values ​​of casting temperature and casting time, as well as the casting temperature x casting time and casting temperature / casting time values, through the aforementioned 55 collected and screened samples (see Table 1, some samples have missing variables) and performs Z-score normalization to eliminate the differences in the dimensions and orders of magnitude among these sample features; then, the casting method is binary encoded (the jacking method is [0, 1], and the bottom jacking method is [1, 0]); finally, for each filter sample, the normalized sample features and the encoded casting method are concatenated into a feature vector for input into the prediction model. Thus, for each filter sample, multiple sample features are constructed and obtained from different data characteristics based on the collected casting variables, so that the prediction model can learn the relationship between these features and the target variable value (i.e., the configuration variable with unknown values ​​in different casting systems). The selection of the prediction model and the optimization of the model parameters will be described in detail below.

[0057] In particular, the configuration variables of the filter samples are also preprocessed before being input into the prediction model, including outlier removal, interpolation, and normalization. For example, the two configuration variables are Z-score normalized to maintain consistency with the sample features in terms of format.

[0058] That is, the first processing according to the embodiments of this application, which calculates the multiple sample features using the casting variables, further includes: encoding the casting method to obtain category features; normalizing the nonlinear features, cross features and / or time-frequency features and concatenating them with the category features to obtain a feature matrix.

[0059] In step S130, a suitable kernel for the prediction model is selected based on the sample features themselves and the relationship between the sample features and configuration variables. First, the linear relationship between the model input (sample features) and the model prediction output (configuration variables) can be determined by the value relationship between each sample feature and the configuration variable. For example, the linearity can be directly observed through a scatter plot of the sample features versus configuration variables, or the Pearson correlation coefficient between each pair of sample features and configuration variables can be calculated to determine the linearity; the higher the absolute value of the correlation coefficient, the stronger the linear relationship. Further, the linearity of the linear combination of configuration variables and multiple sample features can be calculated. For example, the covariance matrix between all sample features and each configuration variable can be calculated, and the column vector (eigenvalue) decomposition of the covariance matrix can be used to determine whether multiple sample features have an approximate linear dependence on the configuration variables.

[0060] If the linear relationship between multiple sample features and configuration variables is low, a nonlinear model should be selected for prediction. For example... Figure 2A and Figure 2B As shown in Table 1, the collinearity between the casting time and filter length of the 52 filter samples (considering practical application, all non-integer points are approximated to the closest integer points or half points, more closely resembling actual operation control rather than system recording) is low, making linear fitting difficult; the collinearity between the casting time and aperture range of the 48 filter samples in Table 1 is also low. Based on this, it can be understood that the linearity between the nonlinear features and cross features constructed based on casting time and the aperture range is also low. Therefore, this example is more suitable for selecting a nonlinear model to predict configuration variables. Since the number of sample features is not significantly higher than the number of filter samples, and the differences in dimensions and sizes among the sample features are large, it is also unsuitable to choose methods such as random trees or LASSO regression to predict the required configuration variables using small samples and low-dimensional features. Therefore, in this example, a Gaussian process regression model, which has better ability to handle nonlinear data on small samples, is selected.

[0061] Specifically, considering the relatively uniform sample size (with only slightly fewer samples for pouring weight / runner filtration volume), Gaussian noise can be omitted for data augmentation of imbalanced samples. The 12-dimensional feature matrix (1 pouring method code + 11 remaining numerical sample features) obtained in the previous example, along with the two normalized configuration variables, is input into the spatial kernel of the Gaussian process regression, and its kernel function is constructed as follows: (Formula 1) in and These represent the variances of the sample eigenvalues ​​and noise values ​​in the feature matrix, respectively. l The length scale of the kernel function.i and j Let be a feature value of a sample from different filter samples, and n be the number of samples. The kernel function exp can be an RBF kernel with strong non-linear processing capabilities. For noise, white noise or pink noise can be selected to simulate vibration noise or fluid noise in the gating system. Since the prediction model outputs multiple (two) values ​​in this example, a composite kernel function needs to be constructed based on the covariance matrix of the output configuration variables and the aforementioned kernel function. This allows for the explicit determination of the correlation between the filter length and aperture range in the spatial kernel of the prediction model, thereby accurately outputting both predicted values. Specifically, the covariance matrix between configuration variables is introduced into the spatial kernel, and the composite kernel function is optimized by iteratively solving this covariance matrix through model training. , and l These are the indicators. Therefore, the covariance matrix of the composite kernel function is shown in the following equation: (Formula 2) Where B is the covariance matrix among configuration variables of size 2x2, K input To remove noise from the kernel function of Gaussian process regression, it can be set as an identity matrix before iterating the prediction model. As the prediction model continuously optimizes the composite kernel function based on the results of the test set samples, the covariance value in B will get closer and closer to the value that makes the prediction result optimal.

[0062] Next, for Gaussian process regression, the marginal maximum likelihood method can be used to optimize its kernel function in each iteration of the prediction model, and this method can also be applied to the composite kernel function constructed in the example above: (Formula 3) in For the predicted value vector, θ This is the set of all kernel parameters, including those mentioned above. , , l The values ​​of the four covariance matrices for the two configuration variables in B are also included. For the training iteration process of the prediction model, a maximum number of iterations can be set. The calculation stops after the maximum marginal likelihood is reached, and the current kernel parameters are used as the final parameters of the prediction model. Alternatively, preset indicators (such as mean squared error, mean absolute error, coefficient of determination, etc.) can be set for the prediction model. The iteration stops and the kernel parameters are finally determined after the predicted values ​​for the configuration variables meet the prediction indicator requirements. In the example of this application, the prediction model is trained by setting the maximum number of iterations to 300.

[0063] Thus, in this example, steps S110-S130 achieve the acquisition of relevant parameters of the filter element, feature construction, and the prediction model for the filter element's filtration length and pore size range. In step S140, for the prediction model based on the Gaussian process regression kernel function, after inputting the feature matrix of the new sample, the prediction model can provide the predicted values ​​for the filtration length and pore size range of the new sample. (Formula 4) (Formula 5) in To predict the mean, To predict the covariance matrix, and These are the covariance matrix between the training and test sets and the covariance matrix of the test set samples themselves, respectively. It's important to note that both are normalized values ​​and can be obtained using known inverse normalization methods. , The true predicted mean and true covariance of the filter length and aperture range in their respective dimensions are calculated using the off-diagonal terms, the mean and standard deviation of the filter length and aperture range of the filter sample used to train the prediction model.

[0064] That is, the construction of the prediction model according to the embodiments of this application includes: determining the kernel function of the prediction model to calculate the similarity relationship between the feature matrix and the configuration variables; training the prediction model using the multiple filter samples; and iterating the kernel parameters of the kernel function and the inter-variable covariance matrix of the configuration variables by comparing the prediction results of the prediction model with the actual results, so that the prediction model reaches a preset index or reaches the maximum number of iterations.

[0065] Specifically, the prediction result output by the method described in the example above according to this application is preferably an interval, such as outputting the prediction result as a confidence interval with a certain level of confidence. For example, the 90% confidence interval of the prediction result is the inversely normalized interval. ± 1.645 1 / 2 The 95% confidence interval for the prediction results is the inverse normalized interval. ±1.960 1 / 2This provides an understanding of the uncertainty and fluctuation range of the prediction results. Preferably, a 95% confidence interval of the prediction results is used as the final prediction results for the filter length and pore size range according to the method described in this application. Filter elements with specific thicknesses and pore size ranges are then manufactured using values ​​falling within these prediction results to achieve a balance between impurity filtration and flow of the casting process in the filter unit containing the filter element, thereby improving the overall efficiency of the casting system.

[0066] That is, constructing a casting fluid filtration device according to the configuration variables output by the prediction model according to the embodiments of this application includes: calculating and preprocessing the required casting variables to obtain the feature matrix, inputting the feature matrix into the prediction model; the prediction model outputs the predicted mean and predicted standard deviation of the configuration variables, calculates the a% confidence interval of the configuration variables, and selects the values ​​of the configuration variables within the a% confidence interval as the values ​​of the configuration variables of the casting fluid filtration device.

[0067] Furthermore, the a% confidence interval of the configuration variable is a 95% confidence interval.

[0068] Furthermore, it is also possible to consider transforming two configuration variables into one configuration variable, thus changing the prediction model from multivariate regression prediction to univariate regression prediction. This avoids the problems of covariance structure assumptions and inaccurate iterations, thereby improving the prediction accuracy of the prediction model. Therefore, the solution provided by the method described in this application is to first determine the relationship between the filter length (i.e., its thickness) and the pore size range of the filter element, and then artificially define the relationship between the two and transform them into a single configuration variable for prediction. In this way, a single-output regression prediction model can be trained using this single configuration variable. The performance of the single-output model is usually better than that of the multi-output model, with a narrower confidence interval and greater stability, ensuring that the provided prediction configuration variable is more accurate (which can be understood as making the accuracy of each predicted sample higher, avoiding providing erroneous prediction results, and thus reducing the trial and error costs for filter element manufacturers).

[0069] Specifically, considering the relationship between the two configuration variables of the filter element itself, the smaller the pore size, the more sensitive it is to impurity size and the better the overall filtration effect. Furthermore, the thicker the filter element, the longer the path of the molten metal through the mesh within the filter element, and the better the filtration effect. However, the filtration effect is often positively correlated with the resistance encountered by the molten metal flow. When the filter element reduces the flow rate of the molten metal to allow a large amount to remain upstream, it will cause problems such as molten metal oxidation due to temperature loss, turbulence, and porosity, and the filter element itself will also experience greater pressure. In actual casting production, it is often necessary to balance the filtration effect of the filter element and the molten metal flow effect. This requires maintaining a certain relationship between the two key parameters reflecting the filtration and blockage effects of the molten metal: the pore size range and the thickness. Furthermore, based on the 55 filter element samples in the above example, the constraints are specified: the pore size range of the filter element is 5 ppi ~ 40 ppi, and the thickness is 10 mm ~ 40 mm, to avoid extreme values ​​of the two configuration variables and the ratio falling within the predicted range.

[0070] In this way, it is feasible to use the ratio of the filter length to the pore size range of the filter element as the only configuration variable. This can provide the original relationship between the filter length and the pore size range to the prediction model, while also taking into account the limitations of the variable itself and reducing the loss of the independent characteristics of the filter length and the pore size range.

[0071] In this case, the aforementioned spatial kernel of Gaussian process regression can still be used, and the parameters of the kernel function of the prediction model can be optimized by repeatedly calculating the marginal maximum likelihood. Unlike the case with two configuration variables, the covariance matrix of the kernel function in this case does not need to include the covariance matrix between the configuration variables. Only the covariance matrix of the Gaussian process regression kernel function needs to be iterated, thus avoiding the introduction of additional parameter terms that would increase the error rate and fuzz the confidence interval, while also simplifying the prediction model and improving its training efficiency. Similarly, the 95% confidence interval for the ratio of filter length to aperture range can be calculated using existing inverse normalization methods. In this way, the method described in this application fully considers the inherent relationship between configuration variables from the user's perspective, simplifies the construction efficiency of the prediction model, and improves the prediction effect of the prediction model.

[0072] Specifically, the method described in this application constructs a minimum casting variable simulation sample (all casting variables are set to minimum) and a maximum casting variable simulation sample (all casting variables are set to maximum) by taking the minimum and maximum values ​​of each casting variable for the above 55 filter sample examples. The prediction model constructed based on the above examples is then used to predict the ratio of filter length to pore size range for the minimum and maximum casting variable simulation samples, respectively. For the minimum casting variable simulation sample, the prediction model gives a ratio of filter length to pore size range in the interval [0.74, 3.27] with a variance of 0.41; for the maximum casting variable simulation sample, the prediction model gives a ratio of filter length to pore size range in the interval [2.21, 4.28] with a variance of 0.28. Considering that the casting of actual filter components may involve more complex situations, such as high temperature and wide runner, high casting volume and short casting time, the confidence interval may be wider. Therefore, the intersection of two confidence intervals is taken to obtain a candidate preferred interval, ensuring that the preferred interval for the ratio of filter length and pore size range provided in this application is accurate in most cases. The candidate preferred interval is [0.74, 4.28]. Considering that the pore size range is still difficult to control precisely in actual production, the interval is directly rounded to allow technicians to adjust the pore size range. Thus, the final preferred interval is determined to be [1, 4].

[0073] In other words, based on the casting and configuration variables of the filter sample collected by the applicant, a preferred ratio of thickness to pore size range for the filter element in a casting system suitable for most casting operations can be obtained. This ratio, rounded to 1-4, can be used to guide the production and manufacturing of the filter element, thereby achieving high-efficiency casting operations with the help of the filter element. It is understood that the method described in this application is not limited to processing data within the casting variable range of the aforementioned filter sample; the above example is merely an exemplary operational method provided by the applicant based on the collected sample data.

[0074] By utilizing a preferred range of the ratio of this thickness value to the pore size range, this application provides a filter device including a filter element with a preferred configuration variable, which will be described in detail below.

[0075] Exemplary device The filtration device according to an embodiment of this application includes the filter element 1 and the gating channel 2. Figure 3A The illustration shows an example shape of the filter element 1, with reference to... Figure 3A According to the embodiments of this application, the filter element 1 is cylindrical in shape and has a smooth curved surface on its outer periphery. It is usually fixed in the receiving portion 20 of the gating 2 and partially protrudes from the gating 2, which is a redundancy in the structure of the gating 2.

[0076] Specifically, the filter element 1 includes multiple filter layers, each of which has multiple filter holes 11. The product of the pore size range of the multiple filter holes 11 and the thickness of the filter element 1 is between a first threshold and a second threshold. When the pore size range of the multiple filter holes 11 is large, the flow rate of the molten metal flowing through the multiple filter holes 11 within a certain volume per unit time will increase, thereby improving the filter element 1's ability to remove impurities. Therefore, the required filtration volume of the filter element 1 can be appropriately reduced, thereby reducing the manufacturing cost, volume occupied, and filtration time of the filter element 1, and facilitating the fit and fixation of the filter element 1 with the gating system 2.

[0077] Therefore, it is preferable to maintain a certain relationship between the pore size range of the filter holes 11 and the thickness of the filter element 1, so that when the requirements of manufacturing process, cost, and application casting system necessitate setting one of the pore size range or the filter element thickness, the preferred range of the other can be determined based on this relationship, thereby ensuring high filtration performance and cost control of the filter element 1. That is, the product of the pore size range of each filter hole 11 and the thickness of the filter element 1 is set between a first threshold and a second threshold.

[0078] The applicable range of the pore size of the plurality of filter holes 11 in the filter device according to the embodiments of this application is 5 ppi to 40 ppi, and the applicable range of the thickness of the filter element 1 is 10 mm to 40 mm. It is suitable for placement in a direct gating system (e.g., Figure 4A and Figure 4B (The vertical portion of the sprue 2 shown); preferably, the filter element 1 is adapted to be placed at the bottom of the sprue, such as... Figure 4B As shown, when the molten metal reaches the filter element 1 after being buffered by the sprue, the impact force is reduced, which can reduce the risk of breakage of the filter element 1. At the same time, the molten metal can maintain a certain flow rate after passing through the filter element, which can be used to effectively pour the downstream horizontal sprue, especially the multi-branch horizontal sprue.

[0079] When the filter element 1 is within the aforementioned size range, it can also be placed in the horizontal runner (e.g. Figure 5A and Figure 5BAs shown in the horizontal section of the gating system 2, the impact force of the molten metal in the horizontal gating is relatively low. Compared to being placed in the sprue, the filter element 1 in the horizontal gating can preferably be considered for its impurity filtration capacity and improved molten metal purity. Therefore, the pore size and / or thickness of the filter element 1 can be appropriately adjusted to maximize filtration capacity while ensuring normal flow of molten metal in the horizontal gating. Furthermore, the filter element 1 can also be placed in the pouring cup. This application does not limit the position of the filter element in the gating system; it merely provides a placement scheme that is effective in a specific scenario through the example described above.

[0080] Based on the range of configuration variables obtained from existing filter element samples in the "Exemplary Method" section, the preferred range of the size parameters of the filter element 1 is further determined. Specifically, the ratio of the pore diameter of the filter hole 11 to the thickness of the filter element 1 is set to the ratio of the larger of the two values ​​to the smaller of the two values. This ratio is used as the molten metal filtration / clogging efficiency parameter of the filter element 1. If this parameter is too small, the impurity filtration effect is poor; if it is too large, the molten metal clogging degree is high. Therefore, under the specified available pore diameter and thickness values, the endpoints of the range of thickness to pore diameter ratios are determined: a first threshold of 1 ppi / mm or 1 mm / ppi, and a second threshold of 4 ppi / mm or 4 mm / ppi, which are the ranges of configuration variables obtained according to the filter element design method based on casting data. In this way, the number of filter holes 11 can limit the filter element 1 from becoming thinner or thicker when there are more or fewer holes, so as to balance its filtration performance and clogging level.

[0081] In other words, the filter element 1 can ensure high flow efficiency of the molten metal while avoiding poor filtration of impurities (especially when the filter element 1 is placed in a sprue), or prevent the molten metal from accumulating upstream while maintaining a high impurity filtration effect (especially when the filter element 1 is placed in a runner). Considering that the filter element cannot be replaced during the casting process, the configuration of the filter element 1 in the filtration device provided in this application is more effective, and the practicality of the filtration device is significantly improved.

[0082] In addition to the main filter component 1, refer to Figure 3B The filtering device according to an embodiment of this application further includes a gating channel 2 for pouring molten metal into which it flows, and the gating channel 2 includes a receiving portion 20 for fixing the filter element 1 and at least partially accommodating the filter element 1. Specifically, the gating channel 2 includes a first gating channel 21 and a second gating channel 22, which are respectively fixed to both sides of the filter element 1 to clamp the filter element 1. The clamping power can usually be provided directly by gravity. It should be noted that the gating channel 2 also includes other parts (such as... Figures 4A-5BThe part shown is not the part where filter element 1 is placed. The following mainly describes the part that cooperates with filter element 1 and is used to place filter element 1.

[0083] The plurality of filter layers are staggered between each pair of adjacent filter layers 10 along the direction of the pouring molten metal flow, forming a three-dimensional lattice configuration. Each pair of adjacent filter layers 10 is connected by multiple connecting lines, which additionally form multiple filter holes 11. That is, the plurality of filter layers can form a three-dimensional mesh-like interconnected filter channel, which has a stronger adsorption capacity for large particle inclusions than a simple multi-layer filter. This structure allows the filter element 1 to effectively mechanically intercept and capture large particle impurities (such as oxides with a particle size of not less than 50 μm) when high-temperature molten metal passes through. Furthermore, this interconnected mesh structure allows for convenient coating of active coatings onto the mesh surface in various ways to enhance its chemical adsorption and remove small impurities (such as small Al2O3 clusters).

[0084] In particular, compared to multi-layer filters, the filter element 1 has filter holes 11 between each filter layer 10. Therefore, the viscous resistance encountered by the molten metal flowing through each filter layer is lower. That is, when the molten metal passes through the filter holes 11 of the filter layer, it is guided by the new filter holes 11 connected to it downstream. Therefore, this low flow resistance design can better ensure the smooth flow of molten metal, thereby improving the filtration efficiency and improving the situation where the molten metal is blocked at the filter element 1 in the runner.

[0085] When the filter element 1 is fixed in place with the first gating channel 21 and the second gating channel 22 on both sides, the shape of its contact cross-section with the gating channel can be determined according to the cross-sectional shape of the gating channel along the direction perpendicular to the flow velocity of the molten metal. Specifically, for example, when a high flow rate is required, the shape of the gating channel 2 is usually cylindrical. In this case, the shape of the contact surface of the filter element 1 with the gating channel 2 is circular and the entire filter element remains cylindrical, so that it can also cooperate with the gating channel 2 to quickly filter the molten metal with the maximum flow radius. Similarly, when the gating channel 2 is of other shapes, the filter element 1 is also shaped accordingly, such as ellipse, triangle, polygon, etc., to achieve the effects of directional flow control, turbulence suppression, and reduction of molten metal oxidation required by the entire casting device.

[0086] That is, the cross-sectional shape of the filter element 1 along the direction perpendicular to the flow velocity of the molten metal is one of a circle, an ellipse, a triangle, or an n-sided polygon; wherein n ≥ 4. In particular, it is preferred that the cross-sectional shape of the filter element 1 along the direction perpendicular to the flow velocity of the molten metal is a circle or a regular n-sided polygon, with n ≥ 4. For example, the filter element 1 can be circular or rectangular, allowing the molten metal to flow in the filter element 1 at a higher flow rate, achieving rapid filtration. It is understood that the above shapes are merely exemplary shapes selected to optimize the filtration performance of the filter element 1, and this application does not limit the filter element 1 to other shapes; it can be any shape adapted to fit with the gating system.

[0087] refer to Figure 3B In the gating system 2, the receiving portion 20 has a first buffer region 200 and a second buffer region 210 on both sides. Since the filter element 1 needs to quickly filter impurities from the pouring liquid, and considering that the flow velocity of the filter element 1 in the entire pouring device is lower than the flow velocity upstream, the cross-sectional area of ​​the filter element 1 along the direction perpendicular to the flow velocity of the pouring molten metal needs to be larger than the cross-sectional area of ​​the gating system 2. This increases the flow area of ​​the pouring liquid to compensate for the decrease in flow velocity and match the flow rate of the gating system 2, ensuring that the upstream pouring liquid does not accumulate at the filter element 1. It is understood that the cross-sectional area of ​​the filter element 1 typically needs to be several times (e.g., 3 to 5 times) larger than the cross-sectional area of ​​the gating system 2 to achieve flow matching; for example, the length of the filter element 1 protruding from the gating system 2 ranges from 2 mm to 15 mm, so that its flow cross-sectional area is significantly larger than that of the gating system 2.

[0088] Therefore, buffer areas are provided at both ends of the gating 2 near the filter element 1. Specifically, the first gating 21 is located upstream of the filter element 1, and the first buffer area 200 is provided near the filter element 1; the second gating 21 is located downstream of the filter element 1, and the second buffer area 210 is provided near the filter element 1. The edge of each buffer area extends from the edge of the gating 2 at a third threshold distance from the filter element 1, and extends to its respective side of the filter element 1 at a certain angle to the axial direction of the gating 2; for example, as... Figure 3B The first buffer region 200 and the second buffer region 210 shown are trapezoidal cross sections parallel to the direction of the pouring fluid flow. The gating channel 2 extends outward from a certain angle relative to its axial direction at a distance from the third threshold of the filter element 1, so that the volume of the gating channel 2 in the extended portion increases.

[0089] Thus, when the casting liquid flows through the filter element 1, the reduced flow rate inevitably leads to liquid accumulation. However, by providing buffer zones upstream and downstream (especially upstream) of the filter element 1, the flow area between the casting liquid and the filter element 1 can be increased. Furthermore, when the filter element 1 becomes clogged with impurities and accumulates, it can accommodate any unpassed casting liquid for a certain period, reducing potential safety hazards during casting. Therefore, the problem of casting liquid retention in the filter element 1 can be effectively solved by the buffer zones.

[0090] Specifically, continue to refer to Figure 3B The cross-sectional dimension of the filter element 1 needs to be larger than the dimension of the buffer area to ensure that the casting liquid does not seep out from the outer edge of the buffer area. The cross-sectional dimension of the filter element 1 along the direction of the casting liquid flow can exceed the cross-sectional dimension of the first buffer area 200 along the direction of the casting liquid flow by 3 mm to 5 mm, and exceed the cross-sectional dimension of the second buffer area 210 along the direction of the casting liquid flow by 5 mm to 12 mm, in order to ensure the buffering effect of the casting liquid and avoid seepage. It can be seen that the cross-sectional area of ​​the first buffer area 200 upstream of the filter element 1 is larger, which can effectively increase the flow cross-sectional area of ​​the casting liquid when it reaches the filter element 1 to avoid accumulation.

[0091] Preferably, the range of the third threshold can be 20 mm to 300 mm.

[0092] The strength of the filter element 1 and its fluid resistance to the casting liquid are mainly related to its thickness. After determining the flow cross-sectional area of ​​the filter element 1 based on the length of the flow cross-sectional area of ​​the gating channel 2, an appropriate thickness should be set for it to ensure that the thickness of the filter element 1 is reduced as much as possible under the premise of strength safety and high filtration efficiency. This avoids excessive fluid resistance to the casting liquid caused by excessive thickness of the filter element 1. In this case, even with a buffer area, the speed of the casting liquid passing through the filter element 1 may be too slow, and serious accumulation may still occur. In addition, increasing the thickness significantly will have limited effect on improving the filtration capacity of the filter element 1, but will significantly increase its manufacturing cost.

[0093] Therefore, in this application, when the length or diameter of the filter element 1 is no greater than 60 mm (for small castings), the thickness of the filter element 1 is set to a range of 10 mm to 25 mm; when the length or diameter of the filter element 1 is greater than 60 mm (for large castings), the thickness of the filter element 1 is set to a range of 15 mm to 40 mm. When the thickness does not exceed the given range, severe accumulation of the casting liquid can be avoided, and efficient filtration can be achieved by combining the pore size range of the aforementioned multiple filter holes 11 with the thickness ratio of the filter element 1, without needing to increase the thickness to extend the filtration path and time.

[0094] The material of the filter element 1 may include zirconium oxide, preferably with a zirconium oxide content of more than 50%, so that the strength of the filter element 1 is guaranteed when it is suitable for filtering molten steel (above 1500℃) and high-temperature alloys, avoiding deformation or breakage caused by high temperature or high pouring fluid flow rate, and also avoiding cracking caused by rapid cooling and heating during repeated pouring.

[0095] Thus, the filtration device described in this application efficiently removes impurities from the molten metal by placing its filter element 1 in the gating 2 with a certain pore size / thickness ratio. At the same time, the buffer area of ​​the gating 2 and the redundant design of the molten metal flow velocity cross section being smaller than the cross section of the filter element 1 ensure that the molten metal can pass through the filter element 1 normally without stagnation during one or multiple pours in a short period of time. This allows the filtration device described in this application to be adapted to the manufacturing scenarios of stainless steel, carbon steel, low alloy steel and other high-temperature alloys, effectively avoiding problems such as porosity and secondary oxidation of the molten metal that are prone to occur in existing pouring processes, thereby providing customers with high-quality casting services.

[0096] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0097] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0098] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0099] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0100] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A filtration device, comprising: A filter element, which is fixed in the runner and partially protrudes from the runner, includes multiple filter layers, each of which has multiple filter holes. A runner for receiving molten metal flowing therein, including a receiving portion for fixing the filter element and at least partially receiving the filter element; The ratio of the pore size range of the plurality of filter holes to the thickness of the filter element is located between a first threshold and a second threshold. The pore size range and the thickness are determined as follows: the casting variable and configuration variable of each filter element sample in the plurality of filter element samples are obtained, the filtration result of each filter element sample on the casting liquid is determined, the plurality of filter element samples are screened based on the filtration result, the casting variable of each screened filter element sample is subjected to a first processing to obtain multiple sample features, a feature matrix of each filter element sample is constructed, a prediction model is constructed using the feature matrix and the configuration variable of each filter element sample, the required casting variable is calculated and preprocessed and then input into the prediction model, the filter element is designed based on the value range of the configuration variable output by the prediction model, and the pore size range and the thickness are determined. The filtration results include: slag inclusions and porosity on the casting surface and inside the casting, tensile strength of the casting, and casting qualification rate. The pouring variables include pouring weight, filtration volume of the runner, pouring temperature, pouring time and / or pouring method. The configuration variables include pore size range and filtration length.

2. The filtration device according to claim 1, wherein, The ratio of the pore size range of the plurality of filter pores and the thickness of the filter element to be between a first threshold and a second threshold includes: The ratio of the aperture range of the plurality of filter holes to the thickness of the filter element is set to the ratio of the larger value to the smaller value, such that the first threshold is 1 ppi / mm or 1 mm / ppi and the second threshold is 4 ppi / mm or 4 mm / ppi.

3. The filtration device according to claim 1, wherein, The first threshold to the second threshold is the range of values ​​for the configuration variable predicted by the prediction model.

4. The filtration device according to claim 2, wherein, The aperture range of the plurality of filter pores is 5 ppi to 40 ppi; The thickness of the filter element is 10 mm to 40 mm.

5. The filtration device according to claim 1, wherein, The length of the filter element protruding from the gating channel ranges from 2 mm to 15 mm. The cross-sectional shape of the filter element along the direction perpendicular to the flow velocity of the poured molten metal is one of the following: circular, elliptical, triangular, or n-sided. Where n ≥ 4.

6. The filtration device according to claim 1, wherein, In the gating system, the receiving portion has a first buffer area and a second buffer area on both sides, respectively; The edge of each buffer zone extends from the edge of the gating system at a distance of the third threshold from the filter element, and extends to its respective side of the filter element at a certain angle to the axial direction of the gating system.

7. The filtration device according to claim 6, wherein, The range of the third threshold is 20 mm to 300 mm.

8. The filtration device according to claim 1, wherein, When the length or diameter of the filter element is not greater than 60 mm, the thickness of the filter element is 10 mm to 25 mm. When the length or diameter of the filter element is greater than 60 mm, the thickness of the filter element is 15 mm to 40 mm.

9. The filtration device according to claim 1, wherein, The filter element is fixed in the receiving part in the gating channel, perpendicular or parallel to the direction of the flow rate of the poured molten metal.

10. The filtration device according to claim 1, wherein, The casting variables of each filtered sample are first processed to obtain multiple sample features, including: The nonlinear characteristics of the pouring variables are obtained by performing exponential, logarithmic, and power series transformations on the pouring weight, runner filtration volume, pouring time, and / or pouring temperature, respectively. The cross-characteristics of the pouring variables are obtained by pairwise multiplication and / or pairwise division of the pouring weight, runner filtration volume, and / or pouring temperature; and The pouring weight, runner filtration volume, and / or pouring temperature are combined with the pouring time to obtain the time-domain signal of the pouring variable changing over time. The time-domain signal is then processed in the frequency domain to obtain the time-frequency characteristics.

Images