Squeeze casting process parameter design method based on radar map

Abstract

The invention provides a squeeze casting process parameter design method based on a radar map, and belongs to the field of squeeze casting process parameter designing.The method comprises the steps that a squeeze casting process data set is constructed, then influence factors of squeeze casting process parameters are converted into the radar map, and a similar case set is obtained through similar characterization of the map; and predicting the extrusion casting process parameters of the new casting by using the K-nearest neighbor algorithm weighted regression. According to the method, characteristic visualization is achieved, material component composition and numerical value difference and functionality are fully considered, the physical significance of the method is enriched, and compared with traditional methods such as Euclidean distance, a more accurate result can be obtained, so that design errors of extrusion casting process parameters are smaller. Meanwhile, on the basis of fully utilizing the existing historical process data, the design result flow is visual and reliable, the process parameter design time and components are saved, and a new thought is provided for mining the relation between the process influence factors and the casting process parameters.

Description

Technical Field

[0001] The invention relates to the field of extrusion casting process parameter design, and in particular to an extrusion casting process parameter design method based on a radar chart. Background Art

[0002] Squeeze casting is a process that combines the process characteristics of metal casting and die forging. It applies high mechanical pressure to the liquid metal poured into the mold at a low speed to solidify it into a near-net shape. It has been widely used in the preparation of high-performance materials and the production of parts in the fields of automobiles, home appliances, high-speed rail, aviation, etc., and has good development prospects. The process parameters of squeeze casting (including pouring temperature, extrusion pressure, mold preheating temperature, etc.) are the key to affecting the quality and performance of squeeze castings. Determining the process parameters of squeeze casting is a basic link in squeeze casting production, and its design efficiency, cost, and quality directly affect the efficiency, cost, and quality of squeeze casting product development. To this end, a large number of studies have been conducted at home and abroad on the process parameters of squeeze casting, and several design methods for squeeze casting process parameters have been formed, but an efficient and unified design method has not yet been formed. In the existing production process of squeeze castings, physical experiments or numerical simulation tests are often used to obtain better process parameters. However, the experimental method is not only affected by the analysis and optimization capabilities of designers, but also because the process parameters are directly selected from discrete experimental results, it is impossible to guarantee the global optimality of the obtained process parameters. Therefore, in recent years, some studies have constructed fitting models based on experimental data to optimize the process parameters of squeeze casting, which cannot reflect the influence of material composition and geometric shape on the process parameters and cannot adapt well to the production of new castings.

[0003] With the maturity of big data technology, data has become a new production resource, and data-driven has become an important direction for the new scientific research paradigm and intelligent manufacturing. Digitalization, networking, and intelligence have become the inevitable path for the survival and development of foundry enterprises. As a material preparation process and a method for manufacturing high-performance castings, it is necessary to adapt to this trend and improve the efficiency and intelligence of process parameter design. A large number of existing studies on extrusion casting processes have accumulated a large number of design cases, but for the design of extrusion casting process parameters, only a very small number of data samples for specific casting experiments have been applied. Case reasoning has been widely used in parameter design in the fields of injection molding, machining, mechanical product design, etc., but there are few reports on its application in the field of extrusion casting. To this end, this paper constructs an extrusion casting case library based on the collected extrusion casting process parameter data, and proposes an extrusion casting process parameter design method based on case reasoning technology. Summary of the invention

[0004] The purpose of the present invention is to provide a method for designing process parameters of squeeze casting based on radar chart to solve the technical problems existing in the background technology.

[0005] Case-based reasoning technology is a branch of analogical reasoning in artificial intelligence today. It lays the foundation for model reasoning based on similarity reasoning. It is an artificial intelligence reasoning paradigm that uses past experience to quickly solve new problems. The basic process of case-based reasoning is: case retrieval, case reuse, case correction and case preservation. Case retrieval is the central link of case-based reasoning, and determining case retrieval technology is the key to the successful application of case-based reasoning technology. In traditional case-based reasoning technology, the accuracy of case retrieval depends on case feature selection and weight allocation. Different from this, for the key attributes (material composition) of extrusion casting cases, this paper proposes a case matching method without feature reduction, which can achieve more accurate retrieval results and complete the design of extrusion casting process parameters.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A method for designing process parameters of squeeze casting based on radar chart, the method comprising the following steps:

[0008] Step 1: Build a structured case;

[0009] Step 2: Graphicalize the case;

[0010] Step 3: Retrieve casting cases based on graph similarity, build a similarity algorithm based on the radar graph, and obtain a reference group of casting cases similar to the designed casting;

[0011] Step 4: Based on the similar case group and the K nearest neighbor idea, refer to the k most similar cases to construct a new casting process parameter design based on similar casting cases.

[0012] Furthermore, the specific process of step 1 is:

[0013] Firstly, the squeeze casting process parameters and their influencing factors are represented as cases, and the cases are digitized to form structured cases. The squeeze casting process parameter samples are divided into process influencing factors X o , where o = 1, 2, ..., l, l represents the number of process influencing factors, of which the material composition is the first θ, and the process parameter is Y j , where j = 1, 2, ..., n, n represents the number of process parameters, and each data sample Q in the case library i , i=1, 2, ..., m, where m represents the number of data samples.

[0014] Furthermore, the specific process of step 2 is:

[0015] In case-based reasoning, accurately finding similar cases is the key to the accuracy of case-based reasoning. The acquisition of similarity cases in case-based reasoning is mainly based on the mathematical calculation of distance functions such as Euclidean distance and their improved functions, which cannot accurately reflect the differences in the materials and geometric shapes of extrusion castings. Materials are used as the key attributes of extrusion casting process parameters, and the materials are transformed into radar charts. The similarity of the cases is measured by the similarity of the charts.

[0016] Further, in step 2, the process influencing factor X o Standardize the process and visualize the cases. The specific process is as follows:

[0017] Step 2.1, Factors affecting the process X o After standardization, the formula is as follows:

[0018]

[0019] Where o = 1, 2, ..., θ, X o ′ represents the normalized data, X o Represents the original value of each attribute in the DB;

[0020] Step 2.2: Treat each component type as an evaluation index, and distribute each index on the radar chart at a uniform angle, with X o ′ is used as the evaluation index value to construct a radar chart, o = 1, 2, ..., θ;

[0021] Step 2.3, using the statistical method to first count the frequency of occurrence of various components in the case library, and then generate the order of component indicators in the radar chart in descending order of frequency, and then draw the components of each sample in a counterclockwise uniform arrangement, so as to obtain the material or casting radar chart;

[0022] Step 2.4, with the final effect of parameter design as the goal, determine the size of the material radar chart to be 100×100 pixels.

[0023] Furthermore, in step 3, the casting case retrieval based on graph similarity and the method of constructing a similarity algorithm based on radar graph are as follows:

[0024] Step 3.1: Grayscale the generated radar image based on OpenCV, denoted as I (i) , the formula is as follows:

[0025] I (i) (px,py)=0.299×R(px,py)+0.587×G(px,py)+0.114×B(px,py)

[0026] Where (px,py) represents the pixel position, R(px,py), G(px,py), and B(px,py) represent the pixel values ​​of the red channel, green channel, and blue channel at the position (px,py), respectively;

[0027] Step 3.2, use Otsu's adaptive threshold method to binarize the image, denoted as Here’s how:

[0028] Calculate the normalized histogram of a grayscale image:

[0029] h(i′)=n i′ / N′

[0030] Where i′ represents the gray level 0, 1, 2, ..., L-1, L represents the gray level, the default is 256; n i′ is the number of pixels with gray value i′ in the image, N′ is the total number of pixels, that is, N′ is equal to N×N;

[0031] The mean grayscale value of the image foreground is u1, and the background is u2:

[0032]

[0033] Where t is the threshold, w1(t) and w2(t) are the weights of the foreground and background categories, respectively, and are defined as follows:

[0034]

[0035] Calculate the inter-class variance of the image foreground and background, that is, the square of the difference between the means of the two categories:

[0036] σ 2 (t)=w1(t)·w2(t)v(μ1(t)-μ2(t)) 2

[0037] Select the optimal threshold, among all possible gray levels i′, so that the between-class variance σ 2 (t) The maximum gray level is taken as the optimal threshold t:

[0038] t=argmax(σ 2 (t))

[0039] Use the optimal threshold t to transform the grayscale image I (i) Perform binarization to obtain a binary image

[0040]

[0041] in, Represents the pixel value at the position (px,py) in the binary image, I(i) (px,py) represents the pixel value at the corresponding position in the original grayscale image, and t is the selected threshold;

[0042] Step 3.3, calculate the final difference between the two material radar images, assuming that any two binary material radar images are The difference is calculated as:

[0043]

[0044] Furthermore, in step 4, based on the similar case group and the K nearest neighbor idea, the k most similar cases are referred to to construct a new casting process parameter design based on similar casting cases. The specific process is as follows:

[0045] Step 4.1, let the material of the new casting be P, then the difference between the material radar image of the i-th historical case and the new casting is

[0046] Step 4.2, according to the distance The size of k cases that are most similar to the new casting material is taken as Q′={Q1′,Q2′,...,Q k '};

[0047] Step 4.3, weight each cluster case according to the distance between the cluster case and the new casting, and the case weight is calculated as:

[0048]

[0049] Step 4.4, calculate the design value of the process parameters of the new casting as:

[0050]

[0051] The present invention has the following beneficial effects due to the adoption of the above technical solution:

[0052] The present invention directly uses the existing squeeze casting process parameter data to design the process parameters of new castings. There is no need to conduct experiments, which is low-cost and high-efficiency. It provides a new data-driven design method for squeeze casting process parameter design. It is not bound to a specific single casting, can adapt to different castings at the same time, has universality, and can promote the application and development of squeeze casting. Based on this framework, the reuse of historical process data (here refers to material composition data) is realized, which saves the time for casting process parameter design, and provides a method for exploring and mining the relationship between materials and casting process parameters, which is conducive to exploring more material compositions for squeeze casting technology. At the same time, this method realizes the visualization of case features through radar charts, which not only enriches the physical meaning of the method, but also avoids subjective operations such as setting case feature weights and feature reduction in the process of similar case retrieval. Compared with traditional methods such as Euclidean distance, more accurate results can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is an overall flow chart of the method of the present invention.

[0054] Figure 2 It is the frequency distribution diagram of the aluminum alloy elements of the present invention.

[0055] Figure 3 This is a case diagram of drawing the material radar diagram principle of the present invention.

[0056] Figure 4 It is a schematic diagram of the material radar chart similarity calculation principle of the present invention.

[0057] Figure 5 It is a radar chart of the target casting case for the retrieval experiment comparison of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and preferred embodiments. However, it should be noted that many details listed in the specification are only for the purpose of enabling the reader to have a thorough understanding of one or more aspects of the present invention, and these aspects of the present invention can be implemented even without these specific details.

[0059] like Figure 1-5 As shown, a method for designing process parameters of squeeze casting based on radar chart, the method comprising the following steps:

[0060] Step 1: Construct structured cases. First, the squeeze casting process parameters and their influencing factors are represented as cases. At the same time, they are digitized to form structured cases. The squeeze casting process parameter samples are divided into process influencing factors X o (o = 1, 2, ..., l, l represents the number of process influencing factors, of which material composition is the first θ), and process parameters Y j (j=1, 2, ..., n, n represents the number of process parameters), each data sample Q in the case library i (i=1, 2, ..., m, where m represents the number of data samples).

[0061] Table 1 Squeeze casting process parameter data example

[0062]

[0063] Step 2: Case graphics. In case reasoning, accurately finding similar cases is the key to the accuracy of case reasoning. The acquisition of similarity cases in existing case reasoning is mainly based on the mathematical calculation of distance functions such as Euclidean distance and its improved functions, which cannot accurately reflect the differences in the materials, geometric shapes and other components of extrusion castings. To this end, this paper takes materials as the key attributes of extrusion casting process parameters, transforms materials into radar charts, and measures the similarity of cases by the similarity of the charts.

[0064] Step 2.1, Factors affecting the process X o After standardization, the formula is as follows:

[0065]

[0066] Among them, X o ′ represents the normalized data, X o Represents the original value of each attribute in the DB;

[0067] Step 2.2: Treat each component type as an evaluation index, and distribute each index on the radar chart at a uniform angle, with X o ′(o=1,2,...,θ) is used as the evaluation index value to construct a radar chart.

[0068] Step 2.3, use the statistical method to first count the frequency of occurrence of various components in the case library, then generate the order of component indicators in the radar chart in descending order of frequency, and then draw the components of each sample in a counterclockwise even arrangement to obtain the material (casting) radar chart.

[0069] Step 2.4, with the final effect of parameter design as the goal, determine the size of the material radar chart to be 100×100 pixels.

[0070] Take the case data from the case library as shown in Table 2, and draw the irregular material polygon as shown in Figure 3 (a). Then, the redundant information on the graph is removed to obtain the final material polygon, as shown in Figure 3 (b) as shown.

[0071] Table 2 Squeeze casting process parameter data example

[0072]

[0073] It can be seen that converting the (casting) material composition into a radar chart realizes the visualization of the casting material characteristics, integrates the changes in the two dimensions of material composition type and composition, provides physical meaning for evaluating the composition of materials, and can more intuitively discover the changes and similarities of materials. When the type of material changes, the enclosed area changes, which indicates a change in the material properties. When the material type is the same but the content changes, it will cause the image to change in shape, especially for trace components, even small changes can be clearly reflected. Material 2 in Table 2 is obtained based on material 1. Material 2 only adjusts the Si element content of material 1 (increase by 0.95%), that is, 0.95% of the decimal point of the mass fraction of the main component AI is increased to the Si component. Its material radar chart ( Figure 3 (c)), compared with the graph of Material 1 ( Figure 3 (b)) has obvious changes, which is more intuitive than numerical calculation and judgment. Since the total content of the material composition remains unchanged (100%), the increase of a certain component will inevitably lead to the decrease of some components, which will cause the generated material radar chart to change significantly. In this case, Figure 3 (c) The image shrinks toward Al and expands toward Si, so this method can accurately reflect and reveal the changes and differences in material composition. Conversely, when the material case properties are similar, their material radar graphs must be similar. Therefore, it is easy to reflect the differences and similarities of materials based on the material radar graph, and establish a new channel for building a graph-based method to find similar extrusion castings.

[0074] In step 3, the casting case retrieval based on graph similarity and the method of constructing a similarity algorithm based on radar graph are as follows:

[0075] Step 3.1: Grayscale the generated radar image based on OpenCV, denoted as I (i) , the formula is as follows:

[0076] I (i) (px,py)=0.299×R(px,py)+0.587×G(px,py)+0.114×B(px,py)

[0077] Among them, (px,py) represents the pixel position, R(px,py), G(px,py), and B(px,py) represent the pixel values ​​of the red channel, green channel, and blue channel at the position (px,py), respectively.

[0078] Step 3.2, use Otsu's adaptive threshold method to binarize the image, denoted as Here’s how:

[0079] Calculate the normalized histogram of a grayscale image:

[0080] h(i′)=n i′ / N′

[0081] Where i′ represents the gray level 0, 1, 2, ..., L-1, L represents the gray level (the default is 256); n i′ is the number of pixels with gray value i′ in the image, and N′ is the total number of pixels, that is, N′ is equal to N×N.

[0082] The mean grayscale value of the image foreground is u1, and the background is u2:

[0083]

[0084] Where t is the threshold, w1(t) and w2(t) are the weights of the foreground and background categories, respectively, and are defined as follows:

[0085]

[0086] Calculate the inter-class variance of the image foreground and background, that is, the square of the difference between the means of the two categories:

[0087] σ 2 (t)=w1(t)·w2(t)·(μ1(t)-μ2(t)) 2

[0088] Select the optimal threshold, among all possible gray levels i′, so that the between-class variance σ 2 (t) The maximum gray level is taken as the optimal threshold t.

[0089] t=argmax(σ 2 (t))

[0090] Use the optimal threshold t to transform the grayscale image I (i) Perform binarization to obtain a binary image

[0091]

[0092] in, Represents the pixel value at the position (px,py) in the binary image, I (i) (px,py) represents the pixel value at the corresponding position in the original grayscale image, and t is the selected threshold.

[0093] Step 3.3, calculate the final difference between the two material radar images. Suppose any two binary material radar images are The difference is calculated as:

[0094]

[0095] The smaller the difference value, the more similar the cases are. The similarity calculation process is as follows: Figure 4 Hint.

[0096] Some material cases are given to verify the retrieval accuracy, as shown in Table 3. The target case is drawn as follows Figure 5 The search results are as follows:

[0097] Table 3 Material composition of some cases

[0098]

[0099] Table 4 Comparison of search results

[0100]

[0101] It can be seen that this paper converts the material composition into a radar chart to provide a visual evaluation method for evaluating the retrieval results of different measurement methods. The similarity gap between different material compositions not only has numerical differences but also visual differences, which makes it possible to make a more intuitive judgment on whether different materials are similar.

[0102] From the search results, we can see that the first two case search results obtained by several methods are the same, and differences begin to appear in the third one. Analysis of the data shows that the material composition and value of the case with index 38 are very close to the target sample. Although case 88 has more Si (0.07%) and Fe (0.09%) elements, the maximum proportion of Si element in the case library is 13.50%, and the maximum proportion of Fe element is 1.3%. After normalization, the values ​​of these two components are 0.0052 and 0.07. Therefore, even if converted into a graph, these two elements are almost not expressed, so case 88 ranks second. This shows that when the difference is very small, several methods can get accurate results, and it also shows that the method in this paper has the same effect as Euclidean distance, etc. Therefore, the similarity calculation based on radar chart and its algorithm proposed in this paper can replace algorithms such as Euclidean distance. However, in the search results ranked 3 to 5, the results obtained by traditional methods based on Euclidean distance, cosine similarity, and Chebyshev distance similarity are significantly different from the target casting samples, especially the calculation results of Chebyshev distance are more different. From the radar chart, it can be seen that the case results obtained by Euclidean distance and cosine similarity obviously lack the lower left corner (i.e., Ti material composition), while the cases ranked 8th (77th) and 9th (62nd) are more similar; Case 57, which has a significant gap with the target sample, is ranked in the front position by the three methods. Case 57 has 5 more elements than the target case, namely Si, Mg, Fe, Zn, and Ni, among which the normalized values ​​of Mg and Fe are 0.20 and 0.23 respectively; The results obtained by the proposed method are significantly better than those of the other three methods. Cases 77, 62, and 83 have the same material composition as the target case, and only the Ti element among the trace elements has a significant gap with the target sample, so they should be more similar to the target sample in terms of perception. This shows that the retrieval effect of the proposed method is more stable, and the material composition of the retrieval results is generally closer to the target sample. It also shows that the similarity method proposed in this paper is more accurate and effective, and is more suitable for case-based reasoning design of extrusion castings.

[0103] From Table 5, we can see that for the Euclidean distance, case 34 in the search results and the last three cases in the Chebyshev distance search results are all quite different from the target case material composition, and their own data are quite different. This is because these three methods cannot characterize the physical meaning of material composition. (1) Euclidean distance is widely used when the case attribute characteristics are balanced. However, due to the objective condition of feature imbalance in the material case library, Euclidean distance cannot measure the similarity of material composition well, and can only measure the sum of the differences in the numerical values ​​of material composition. Since it only considers numerical results, it cannot take into account the impact of differences in component types. For materials, a small amount of component difference, especially trace elements, may lead to significant differences in material properties. (2) Although cosine distance can measure the change of material properties, it ignores the difference in attribute values, resulting in the similarity values ​​of the search results being very close, and the differences between different cases cannot be well displayed. It is more suitable for similarity measurement of non-numerical feature vectors. (3) Chebyshev distance only measures the attribute with the largest difference, so it is more sensitive to cases with unbalanced features, and it is easier to obtain material composition that is significantly different from the target case. It is more suitable for measuring the stability of time series data. In summary, the traditional measurement method used in material case retrieval has certain defects and cannot accurately and effectively identify similar materials. The method proposed in this paper can better measure the changes in material composition and their numerical differences, thereby obtaining more accurate historical cases.

[0104] Step 4: Based on the similar case group and the K nearest neighbor idea, refer to the k most similar cases to construct a new casting process parameter design based on similar casting cases.

[0105] The specific steps are as follows:

[0106] Step 4.1, let the material of the new casting be P, then the difference between the material radar image of the i-th historical case and the new casting is

[0107] Step 4.2, according to the distance The size of k cases that are most similar to the new casting material is taken as Q′={Q1′,Q2′,...,Q k ′}.

[0108] Step 4.3, weight each cluster case according to the distance between the cluster case and the new casting. The case weight is calculated as:

[0109]

[0110] Step 4.4, calculate the design value of the process parameters of the new casting as:

[0111]

[0112] Based on the squeeze casting process data in Table 1, 80% of the data is divided into training samples and 20% is divided into test samples to design the actual process parameters, mainly focusing on the design of pouring temperature. The relative mean error (MRE) and mean square error (MSE) are used as design performance evaluation indicators, as shown in the following formula:

[0113]

[0114] Where: n is the number of test samples; y ij is the predicted value of the process parameters of the i-th test sample; y ij is the true value of the process parameters of the i-th test sample.

[0115] Through experimental verification, the average error of the pouring temperature designed by this method is only 1.06%, the mean square error is 97.78, and the maximum error of the pouring temperature is controlled within 20°C. The error is very small and meets the use requirements of the squeeze casting process.

[0116] Matters not covered by the present invention are known technologies.

[0117] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for designing process parameters of squeeze casting based on radar chart, characterized in that: The method comprises the following steps: Step 1: Build a structured case; Step 2: Graphicalize the case; Step 3: Retrieve casting cases based on graph similarity, build a similarity algorithm based on the radar graph, and obtain a reference group of casting cases similar to the designed casting; Step 4: Based on the similar case group and the K nearest neighbor idea, refer to the k most similar cases to construct a new casting process parameter design based on similar casting cases.

2. The method for designing process parameters of squeeze casting based on radar chart according to claim 1, characterized in that: The specific process of step 1 is: Firstly, the squeeze casting process parameters and their influencing factors are represented as cases, and the cases are digitized to form structured cases. The squeeze casting process parameter samples are divided into process influencing factors X o , where o = 1, 2, ..., l, l represents the number of process influencing factors, of which the material composition is the first θ, and the process parameter is Y j , where j = 1, 2, ..., n, n represents the number of process parameters, and each data sample Q in the case library i , i=1, 2, ..., m, where m represents the number of data samples.

3. The method for designing process parameters of squeeze casting based on radar chart according to claim 1, characterized in that: The specific process of step 2 is: In case-based reasoning, accurately finding similar cases is the key to the accuracy of case-based reasoning. The acquisition of similarity cases in case-based reasoning is mainly based on the mathematical calculation of distance functions such as Euclidean distance and their improved functions, which cannot accurately reflect the differences in the materials and geometric shapes of extrusion castings. Materials are used as the key attributes of extrusion casting process parameters, and the materials are transformed into radar charts. The similarity of the cases is measured by the similarity of the charts.

4. The method for designing process parameters of squeeze casting based on radar chart according to claim 1, characterized in that: In step 2, the process influencing factor X o Standardize the process and visualize the cases. The specific process is as follows: Step 2.1, Factors affecting the process X o The standardized formula is as follows: Where o = 1, 2, ..., θ, X o ′ represents the normalized data, X o Represents the original value of each attribute in the DB; Step 2.2: Treat each component type as an evaluation index, and distribute each index on the radar chart at a uniform angle, with X o ′ is used as the evaluation index value to construct a radar chart, o = 1, 2, ..., θ; Step 2.3, using the statistical method to first count the frequency of occurrence of various components in the case library, and then generate the order of component indicators in the radar chart in descending order of frequency, and then draw the components of each sample in a counterclockwise uniform arrangement, so as to obtain the material or casting radar chart; Step 2.4, with the final effect of parameter design as the goal, determine the size of the material radar chart to be 100×100 pixels.

5. The method for designing process parameters of squeeze casting based on radar chart according to claim 1, characterized in that: In step 3, the casting case retrieval based on graph similarity and the method of constructing a similarity algorithm based on radar graph are as follows: Step 3.1: Grayscale the generated radar image based on OpenCV, denoted as I (i) , the formula is as follows: I (i) (px,py)=0.299×R(px,py)+0.587×G(px,py)+0.114×B(px,py) Where (px,py) represents the pixel position, R(px,py), G(px,py), and B(px,py) represent the pixel values ​​of the red channel, green channel, and blue channel at the position (px,py), respectively; Step 3.2, use Otsu's adaptive threshold method to binarize the image, denoted as Here’s how: Calculate the normalized histogram of a grayscale image: h(i′)=n i′ / N′ Where i′ represents the gray level 0, 1, 2, ..., L-1, L represents the gray level, the default is 256; n i′ is the number of pixels with gray value i′ in the image, N′ is the total number of pixels, that is, N′ is equal to N×N; The mean grayscale value of the image foreground is u1, and the background is u2: Where t is the threshold, w1(t) and w2(t) are the weights of the foreground and background categories, respectively, and are defined as follows: Calculate the inter-class variance of the image foreground and background, that is, the square of the difference between the means of the two categories: σ 2 (t)=w1(t)·w2(t)·(μ1(t)-μ2(t)) 2 Select the optimal threshold, among all possible gray levels i′, so that the between-class variance σ 2 (t) The maximum gray level is taken as the optimal threshold t: t=argmax(σ 2 (t)) Use the optimal threshold t to transform the grayscale image I (i) Perform binarization to obtain a binary image in, Represents the pixel value at the position (px,py) in the binary image, I (i) (px,py) represents the pixel value at the corresponding position in the original grayscale image, and t is the selected threshold; Step 3.3, calculate the final difference between the two material radar images, assuming that any two binary material radar images are The difference is calculated as:

6. The method for designing process parameters of squeeze casting based on radar chart according to claim 1, characterized in that: In step 4, based on the similar case group and the K nearest neighbor idea, the k most similar cases are referred to to construct a new casting process parameter design based on similar casting cases. The specific process is as follows: Step 4.1, let the material of the new casting be P, then the difference between the material radar image of the i-th historical case and the new casting is Step 4.2, according to the distance The size of k cases that are most similar to the new casting material is taken as Q′={Q1′,Q2′,...,Q k '}; Step 4.3, weight each cluster case according to the distance between the cluster case and the new casting, and the case weight is calculated as: Step 4.4, calculate the design value of the process parameters of the new casting as:

Images