A mobile phone hinge injection molding parameter optimization method and system

Abstract

The present application relates to the field of data processing, more particularly, the present application relates to a kind of mobile phone hinge injection molding parameter optimization method and system, method includes: obtaining the influence factor dataset of mobile phone hinge injection molding;Obtain the sampling quantity obtained in advance;Randomly sample in influence factor dataset based on sampling quantity, and based on the sample extracted, construct random forest, calculate fitting stability;According to the moderate of fitting stability, calculate appropriate degree;In response to appropriate degree less than preset appropriate degree threshold, according to fitting stability, adjust sampling quantity, and according to the sample of adjusted sampling quantity, construct random forest again;In response to appropriate degree not less than preset appropriate degree threshold, the random forest corresponding to sampling quantity is regarded as target random forest, to realize mobile phone hinge injection molding processing parameter setting. Improve the accuracy of mobile phone hinge injection molding processing parameter setting.

Description

Technical Field

[0001] This invention relates to the field of data processing. More specifically, this invention relates to a method and system for optimizing injection molding parameters of mobile phone hinges. Background Technology

[0002] With the rapid expansion of the foldable phone market, the quality and performance of the phone hinge, as a core component, have attracted much attention. The hinge structure is precise and complex, requiring extremely high precision in injection molding processes. Accurate injection molding parameters are crucial for ensuring hinge quality, affecting not only dimensional accuracy and mechanical properties but also production efficiency and cost control.

[0003] Existing methods for determining injection molding process parameters for mobile phone hinges have many limitations. Traditional methods rely primarily on manual experience and repeated trial and error, lacking systematicity and precision. Process engineers initially set parameters based on past experience, then adjust them through multiple trial moldings—a cumbersome, time-consuming, and costly process. This approach is ill-suited to the complex and ever-changing design requirements of mobile phone hinges, especially for new materials and complex structures, where the number of trial and error increases exponentially, resulting in extremely low efficiency.

[0004] In recent years, machine learning technology has rapidly emerged, providing new ideas and methods for solving the aforementioned problems. Random forest, as an efficient machine learning algorithm, has been widely applied in many fields due to its excellent performance in handling complex multivariate problems. However, whether the random forest algorithm can construct a high-precision random forest model depends on the appropriateness of the sampling size. Setting the sampling size too large can easily lead to an excessive number of samples used to build the random tree, resulting in overfitting; setting the sampling size too small can easily lead to insufficient samples used to build the random tree, resulting in insufficient fitting ability. Therefore, how to set an appropriate sampling size is the focus of this invention. Summary of the Invention

[0005] To address the issue of setting an appropriate sampling quantity, this invention proposes a method and system for optimizing injection molding parameters of mobile phone hinges.

[0006] In a first aspect, the present invention provides a method for optimizing injection molding parameters of a mobile phone hinge, comprising:

[0007] Obtain the dataset of influencing factors in mobile phone hinge injection molding and the setting process parameters for each influencing factor;

[0008] Obtain the pre-determined sampling quantity; randomly select samples from the influencing factor dataset based on the sampling quantity, and construct a random forest based on the selected samples. Input each sample into each random tree to obtain the output results, obtain the outlier degree of each sample and the corresponding set processing parameters, and calculate the fitting stability. The fitting stability is negatively correlated with the difference between the set processing parameters and the output results, and positively correlated with the outlier degree; calculate the suitability based on the moderateness of the fitting stability.

[0009] If the suitability is less than the preset suitability threshold, the sampling quantity is adjusted according to the fitting stability, and a random forest is constructed based on the samples re-sampled with the adjusted sampling quantity; if the suitability is not less than the preset suitability threshold, the random forest corresponding to the sampling quantity is used as the target random forest to realize the setting of the injection molding process parameters for mobile phone hinges.

[0010] This invention improves the accuracy of random forest model fitting by adaptively setting the sampling quantity, thereby providing more suitable processing parameters for mobile phone hinges. Furthermore, it analyzes the fitting performance of the random forest model to determine the appropriateness of the sampling quantity setting, and adjusts the sampling quantity accordingly to adaptively obtain an accurate sampling quantity. Further, when analyzing the fitting performance of the random forest model, it introduces fitting stability to reflect the suitability of the model's fit, providing a basis for adjusting the sampling quantity. Furthermore, when analyzing fitting stability, it introduces outlier degree and the difference between the set processing parameters and the output results to reflect the fitting performance of the random forest model, accurately measuring underfitting and overfitting, and providing a basis for adjusting the sampling quantity.

[0011] Preferably, the calculation of fitting stability includes:

[0012] ;

[0013] in, This represents the difference between the set processing parameters of the i-th sample and the output result of the i-th sample in the j-th random tree. This represents the outlier degree of the i-th sample and the set processing parameters. This indicates the preset first zero-prevention coefficient. This indicates a preset second zero-prevention coefficient. This represents the output result of the i-th sample in the j-th random tree. Let |i| represent the mean of the output results of the i-th sample across all random trees, |i| represent the magnitude of the vector, and M and N represent the number of random trees and samples, respectively.

[0014] In analyzing the stability of the fit, this invention introduces the variance of the random number output to reflect the diversity of the output results of different random trees, thereby accurately measuring underfitting and overfitting, and providing a basis for the sampling quantity.

[0015] Preferably, obtaining the pre-determined sample size includes:

[0016] The data on influencing factors with the same processing parameters are grouped into one category. Boundary data between each two categories are obtained. Based on the boundary data, the boundary clarity and boundary pattern complexity of each two categories are obtained. The quotient of boundary pattern complexity and boundary clarity is used as the distinguishing difficulty between each two categories. The average distinguishing difficulty of all two categories is used as the overall distinguishing difficulty. The sampling size is set according to the overall distinguishing difficulty, and the sampling size is positively correlated with the overall distinguishing difficulty.

[0017] Preferably, the method for obtaining the boundary sharpness includes:

[0018] Obtain the geometric center of the influencing factor data for each category. Draw a straight line through the geometric center of the influencing factor data for the two categories. Project the influencing factor data of the two categories onto the straight line to obtain projection points. Arrange the category label values ​​of all influencing factor data of the two categories according to the order of the projection points to obtain the category label value sequence. The reciprocal of the permutation entropy of the category label value sequence is recorded as the boundary clarity between the two categories.

[0019] This invention accurately reflects the boundary between two types of influencing factor data by using the permutation entropy of the distribution of two types of data, providing a basis for accurately setting the sampling size.

[0020] Preferably, the method for obtaining the category flag value includes:

[0021] Set the category flag value of one category of influencing factor data to 1, and set the category flag value of another category of influencing factor data to 0.

[0022] Preferably, the method for obtaining the complexity of the boundary regularity includes:

[0023] Set the number of items variable By fitting a polynomial with T terms using boundary points, T is sequentially assigned different values ​​in ascending order, and the fitting error of the polynomial corresponding to each value of T is obtained. The polynomial whose fitting error is first less than a preset error threshold is denoted as the target polynomial.

[0024] The highest degree with non-zero coefficients in the objective polynomial is denoted as the objective degree; the boundary condition complexity is obtained by multiplying the objective degree by the number of terms in the objective polynomial.

[0025] This invention uses the number of terms and degree of a polynomial to accurately reflect the complex situation of the changing pattern of the boundary between two categories, thereby providing a basis for accurately setting the sampling size.

[0026] Preferably, the method for obtaining the outlier degree includes:

[0027] Each sample is concatenated with the set processing parameters to obtain the analysis data. The distribution density of each analysis data is obtained. The outlier degree of the analysis data is obtained by dividing the distribution density of each analysis data by the mean distribution density of all neighboring data. This outlier degree of the sample and the corresponding set processing parameters is recorded as the outlier degree of the sample and the corresponding set processing parameters.

[0028] Preferably, the calculation of suitability based on the moderateness of fitting stability includes:

[0029] The fitting stability is normalized by taking the reciprocal of the absolute value of the difference between the normalized fitting stability and the preset center value, and then normalizing again to obtain the appropriateness of the sampling size.

[0030] Preferably, adjusting the sampling size based on the stability of the fit includes:

[0031] If the normalized fitting stability is less than the preset center value, the preset adjustment step size is multiplied by the difference between the normalized fitting stability and the preset adjustment step size, and then rounded to obtain the adjustment quantity. The sampling quantity is then added to the adjustment quantity to obtain the adjusted sampling quantity.

[0032] If the normalized fitting stability is greater than the preset center value, then the preset adjustment step size is multiplied by the normalized fitting stability and then rounded to obtain the adjustment quantity. The sampling quantity is then subtracted from the adjustment quantity to obtain the adjusted sampling quantity.

[0033] This invention adjusts the sampling size by fitting stability, thereby obtaining a more accurate sampling size.

[0034] Secondly, the present invention provides a mobile phone hinge injection molding parameter optimization system, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned mobile phone hinge injection molding parameter optimization method is implemented.

[0035] By adopting the above technical solution, a computer program is generated from the above-mentioned method for optimizing injection molding parameters of a mobile phone hinge, and stored in a memory for loading and execution by a processor. This allows for the production of terminal devices based on the memory and processor, making them convenient to use.

[0036] The present invention has the following beneficial effects:

[0037] This invention improves the accuracy of random forest model fitting by adaptively setting the sampling number, thereby providing more suitable processing parameters for mobile phone hinges;

[0038] Furthermore, by analyzing the fit of the random forest model, we can determine whether the sampling size setting is appropriate, and then adjust the sampling size according to the appropriateness of the sampling size setting to obtain an accurate sampling size adaptively.

[0039] Furthermore, when analyzing the fit of the random forest model, the fit stability is introduced to reflect the suitability of the random forest model, thereby providing a basis for adjusting the sampling size.

[0040] Furthermore, when analyzing the stability of the fit, the degree of outlier and the difference between the processing parameters and the output results are introduced to reflect the fit of the random forest model, so as to accurately measure the underfit and overfit, and provide a basis for adjusting the sampling quantity.

[0041] Furthermore, when analyzing the stability of the fit, the variance of the random number output is introduced to reflect the diversity of the output results of different random trees, thereby accurately measuring underfitting and overfitting, and providing a basis for the sampling quantity. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the steps of a method for optimizing injection molding parameters of a mobile phone hinge according to an embodiment of the present invention. Detailed Implementation

[0043] Please see Figure 1 The diagram illustrates a flowchart of a method for optimizing injection molding parameters of a mobile phone hinge according to an embodiment of the present invention. The method includes the following steps:

[0044] S1: Obtain the dataset of influencing factors for injection molding of mobile phone hinges and the setting process parameters for each influencing factor.

[0045] Specifically, data on influencing factors for injection molding of each type of mobile phone hinge is obtained, and the set of data on influencing factors for injection molding of all types of mobile phone hinges is denoted as the influencing factor dataset. The influencing factor data for injection molding of mobile phone hinges consists of relevant data that need to be considered when setting the injection molding processing parameters of mobile phone hinges. The influencing factor data includes, but is not limited to, the following aspects: structural point location data of the mobile phone hinge, thermal performance data of the injection material, flowability data of the injection material, and the location of the injection gate, etc.

[0046] The processing parameters for injection molding of each type of mobile phone hinge are obtained from historical data and recorded as the set processing parameters for injection molding of mobile phone hinges.

[0047] S2: Obtain the pre-determined sample size.

[0048] It's important to note that the Random Forest algorithm uses different samples to construct different random trees in order to increase the diversity of results and prevent overfitting. This is typically done by randomly selecting a subset of samples from the total dataset to construct each random tree. The sampling size determines the number of samples used to build the random trees. Setting the sampling size too large leads to excessive overlap between samples from different random trees, resulting in smaller differences between the results and thus excessive stability of the output. Conversely, a large sampling size can also lead to overly complex random trees, resulting in overfitting. Conversely, setting the sampling size too small results in less overlap between samples from different random trees, leading to larger differences between the results and thus less stability of the output. Furthermore, a small sampling size can result in overly simple random trees, leading to underfitting. Therefore, the appropriateness of the sampling size significantly impacts the prediction accuracy of the Random Forest algorithm.

[0049] Optional, as an example, obtaining the pre-determined sample size includes:

[0050] Use the preset value as the sampling quantity.

[0051] It's important to note that the prediction difficulty varies across different types of data. Some data exhibits complex patterns, making prediction more difficult, while others have simpler patterns, resulting in lower prediction difficulty. Data with high prediction difficulty requires less data to fit the pattern, while data with low prediction difficulty requires more data. Therefore, the sampling size needs to be set based on the prediction difficulty of the data. According to the basic principle of the Random Forest algorithm, which relies on classification for prediction, the accuracy of the Random Forest algorithm is highly dependent on the accuracy of its classification. Since the classification difficulty varies across different types of data, the sampling size can be set by analyzing the classification difficulty of the data.

[0052] Preferably, as an example, obtaining the pre-determined sample size includes:

[0053] The data on influencing factors with the same processing parameters are grouped into one category. Boundary data between each two categories are obtained. Based on the boundary data, the boundary clarity and boundary pattern complexity of each two categories are obtained. The quotient of boundary pattern complexity and boundary clarity is used as the distinguishing difficulty between each two categories. The average distinguishing difficulty of all two categories is used as the overall distinguishing difficulty. The normalized value of the overall distinguishing difficulty is multiplied by the preset baseline quantity and then rounded up to obtain the sampling quantity.

[0054] Understandably, boundary clarity reflects the boundary between two categories; a higher value indicates a clearer boundary, making it easier to distinguish between the two categories. Boundary pattern complexity reflects the complexity of the patterns along the boundary lines / surfaces between the two categories; greater complexity indicates that the boundary lines / surfaces are harder to fit, making it more difficult to distinguish between the two categories.

[0055] The above embodiments involve boundary data, boundary clarity, and boundary pattern complexity. The methods for determining boundary data, boundary clarity, and boundary pattern complexity will be explained below.

[0056] First, let's introduce the boundary data.

[0057] Preferably, as an example, the method for obtaining boundary data includes:

[0058] Obtain adjacent data for each influencing factor. If adjacent data contains both data of the same type and data of another type, then record the influencing factor data as boundary data.

[0059] Then, we will discuss boundary clarity.

[0060] It should be noted that when the classification boundary between two categories is clear, data from one category is distributed on one side of the boundary, and data from the other category is distributed on the other side. This distribution pattern results in a relatively low permutation entropy.

[0061] Preferably, as an example, methods for obtaining boundary sharpness include:

[0062] Set the category flag value of one category of influencing factor data to 1, and set the category flag value of another category of influencing factor data to 0.

[0063] Obtain the geometric center of the influencing factor data for each category. Draw a straight line through the geometric center of the influencing factor data for the two categories. Project the influencing factor data of the two categories onto the straight line to obtain projection points. Arrange the category label values ​​of all influencing factor data of the two categories according to the order of the projection points to obtain the category label value sequence. The reciprocal of the permutation entropy of the category label value sequence is recorded as the boundary clarity between the two categories.

[0064] It is understandable that the distribution pattern of the two categories of data is reflected by the sequence of category label values, and the disorder of the distribution of the two categories is reflected by analyzing the permutation entropy of the sequence of category label values. The larger the permutation entropy, the more disordered the distribution of the two categories of data and the smaller the clarity of the boundary between the two categories.

[0065] Finally, the complexity of boundary patterns is introduced.

[0066] It should be noted that when the data is complex, a large number of terms and a high degree are required to fit the regularity using a polynomial. Therefore, the complexity of the boundary regularity can be assessed by analyzing the number of terms and degree of the polynomial when the regularity is accurately fitted.

[0067] Preferably, as an example, methods for obtaining the complexity of boundary patterns include:

[0068] Set the number of items variable Based on boundary points, a polynomial with T terms is fitted using the least squares method. T is taken in different values ​​in ascending order, and the fitting error of the polynomial corresponding to each value of T is obtained. The polynomial whose fitting error is first less than the preset error threshold is denoted as the target polynomial.

[0069] The highest degree with non-zero coefficients in the objective polynomial is denoted as the objective degree; the boundary condition complexity is obtained by multiplying the objective degree by the number of terms in the objective polynomial.

[0070] Understandably, when the polynomial fitting error is less than a preset error threshold, it indicates that the polynomial can fit the boundary patterns relatively accurately. The more terms and the higher the degree of the polynomial that accurately fits the boundary patterns, the more complex the boundary variation patterns.

[0071] It should be noted that in this embodiment, T starts from 3 and increases upwards. Other embodiments may set other starting points, and this embodiment does not impose specific restrictions.

[0072] S3: Randomly select samples from the influencing factor dataset based on the sampling quantity, construct a random forest based on the selected samples, calculate the fitting stability, and calculate the suitability based on the moderateness of the fitting stability.

[0073] S30: Randomly select samples from the influencing factor dataset based on the sampling quantity, and construct a random forest based on the selected samples.

[0074] Preferably, as an example, samples are randomly drawn from the influencing factor dataset based on the sampling size, and a random forest is constructed based on the drawn samples, including:

[0075] Randomly select a number of samples from the influencing factor dataset as the construction sample set for each random tree; obtain the construction sample set for each random tree.

[0076] Based on the sample set constructed for each random tree, random trees are constructed using the random forest algorithm, and all random trees are combined to form a random forest.

[0077] It should be noted that using the random forest algorithm to construct random trees is an existing technology, and will not be elaborated on here.

[0078] S31: Calculate the stability of the fit.

[0079] It should be noted that, in order to determine whether the sampling size setting is appropriate, the fitting performance of the random forest constructed based on the sampling size needs to be evaluated. When the random forest is overfitted, its fitting stability will be too high; when it is underfitted, its fitting stability will be too low. Therefore, the fitting stability can be analyzed to determine the fitting accuracy of the random forest constructed under the given sampling size.

[0080] Preferably, as an example, calculating the fit stability includes:

[0081]

[0082] The process involves inputting each sample into a random tree to obtain the output result. This represents the difference between the set processing parameters of the i-th sample and the output result of the i-th sample in the j-th random tree. This represents the outlier degree of the i-th sample and the set processing parameters. This indicates the preset first zero-prevention coefficient. This indicates a preset second zero-prevention coefficient. This represents the output result of the i-th sample in the j-th random tree. Let M represent the mean of the output results of the i-th sample across all random trees, and M and N represent the number of random trees and samples, respectively. Let || represent the magnitude of the vector.

[0083] Understandable This value reflects how well the random forest fits data with varying degrees of outliers. A larger value indicates that the random forest also fits the patterns of data with a high degree of outliers. In other words, the random forest also fits the patterns of some abnormal or noisy data, thus the random forest is more likely to overfit. A smaller value indicates that the random forest fits the patterns of non-abnormal or noisy data poorly, thus the random forest is more likely to underfit. This value reflects the similarity of the fitting results of different random trees. The larger the value, the greater the similarity of the fitting results of different random trees, and therefore the poorer the fitting diversity of the random forest. Thus, the random forest has a higher probability of overfitting. The smaller the value, the smaller the similarity of the fitting results of different random trees, and therefore the better the fitting diversity of the random forest. However, the random forest has poorer stability of its fitting results, and thus the random forest has a higher probability of underfitting.

[0084] It should be added that the methods for obtaining outlier degree include:

[0085] Each sample is concatenated with the set processing parameters to obtain analytical data. The distribution density of each analytical data point is then obtained. The outlier degree of the analytical data is obtained by dividing the distribution density of each analytical data point by the mean distribution density of all neighboring data points, and this outlier degree is recorded as the outlier degree of the sample and the set processing parameters. In this embodiment, the analytical data within a preset radius centered on the analytical data point is considered as the neighboring data point. Other embodiments may use other methods to set the neighboring data point, and this embodiment does not impose specific limitations. In addition, this embodiment uses the derivative of the mean distance between the analytical data point and the neighboring data point as the distribution density of the analytical data point. Other embodiments may use other methods to set the distribution density, and this embodiment does not impose specific limitations.

[0086] It is necessary to further supplement the method for obtaining the difference between the set processing parameters of the i-th sample and the output result of the i-th sample in the j-th random tree, including:

[0087] The magnitude of the difference between the set processing parameters of the i-th sample and the output result of the i-th sample in the j-th random tree is taken as the difference between the set processing parameters of the i-th sample and the output result of the i-th sample in the j-th random tree.

[0088] S32: Calculate the suitability based on the moderateness of the fitting stability.

[0089] Optionally, as an example, the suitability can be calculated based on the moderateness of fit stability, including:

[0090] If the fitting stability falls within the preset range, set the suitability level to 1; if the fitting stability does not fall within the preset range, set the suitability level to 0.

[0091] Understandably, determining suitability by setting fixed values ​​has poor adaptability and scalability. When dealing with different data, a new preset range needs to be set, and its accuracy is highly dependent on the accuracy of the preset range.

[0092] Preferably, as an example, the suitability is calculated based on the moderateness of the fit stability, including:

[0093] The fitting stability is normalized by taking the reciprocal of the absolute value of the difference between the normalized fitting stability and the preset center value, and then normalizing again to obtain the appropriateness of the sampling size.

[0094] Understandably, the closer the fit stability is to the preset center value, the greater its suitability; the farther it is from the preset center value, the smaller its suitability. This approach offers higher adaptability and scalability.

[0095] S4: In response to the suitability being less than the preset suitability threshold, the sampling quantity is adjusted according to the fitting stability, and a random forest is constructed based on the samples re-sampled according to the adjusted sampling quantity; in response to the suitability being not less than the preset suitability threshold, the random forest corresponding to the sampling quantity is used as the target random forest to realize the setting of the injection molding process parameters for mobile phone hinges.

[0096] S40: In response to the suitability being less than the preset suitability threshold, the sampling quantity is adjusted according to the fitting stability, and a random forest is constructed based on the samples re-sampled according to the adjusted sampling quantity.

[0097] It should be noted that if the suitability score is less than the preset suitability threshold, it indicates that the sampling size is not set accurately and needs to be adjusted. Therefore, the sampling size needs to be adjusted based on the fitting performance of the random forest.

[0098] Preferably, as an example, in response to a suitability score being less than a preset suitability score threshold, the sampling size is adjusted based on the fit stability, and a random forest is constructed based on the samples re-sampled using the adjusted sampling size, including:

[0099] If the normalized fitting stability is less than the preset center value, the preset adjustment step size is multiplied by the difference between the normalized fitting stability and the preset adjustment step size, and then rounded to obtain the adjustment quantity. The sampling quantity is then added to the adjustment quantity to obtain the adjusted sampling quantity.

[0100] If the normalized fitting stability is greater than the preset center value, then the preset adjustment step size is multiplied by the normalized fitting stability and then rounded to obtain the adjustment quantity. The sampling quantity is then subtracted from the adjustment quantity to obtain the adjusted sampling quantity.

[0101] The sampling process is performed based on the adjusted sampling quantity to obtain the construction sample set of each random tree. Based on the construction sample set of the random trees, a random forest is constructed using the random forest algorithm.

[0102] S41: In response to the suitability level being no less than a preset suitability level threshold, the random forest corresponding to the sampling quantity is used as the target random forest.

[0103] Preferably, as an example, in response to a suitability level not being less than a preset suitability level threshold, the random forest corresponding to the sampling size is used as the target random forest, including:

[0104] In response to the suitability level being no less than a preset suitability threshold, the sampling quantity with a suitability level no less than the preset threshold is taken as the target sampling quantity, and the random forest constructed under the target sampling quantity is taken as the target random forest.

[0105] S42: To set the process parameters for injection molding of mobile phone hinges.

[0106] Preferably, as an example, the process parameters for injection molding of mobile phone hinges include:

[0107] The data on factors affecting the production of the mobile phone hinge are input into the target random forest to obtain the processing parameters of the mobile phone hinge to be produced.

[0108] This invention also discloses a mobile phone hinge injection molding parameter optimization system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a mobile phone hinge injection molding parameter optimization method according to the present invention is implemented.

[0109] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0110] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as resistive random access memory (DRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (DRAM), high-bandwidth memory, hybrid memory cube, etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device.

[0111] This concludes the embodiment.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing injection molding parameters of a mobile phone hinge, characterized in that, include: Obtain the dataset of influencing factors in mobile phone hinge injection molding and the setting process parameters for each influencing factor; Obtain the pre-determined sample size; Based on the sampling quantity, samples are randomly selected from the influencing factor dataset, and a random forest is constructed based on the selected samples. Each sample is input into each random tree to obtain the output results. The outlier degree of each sample and the corresponding set processing parameters is obtained, including: concatenating each sample with the set processing parameters to obtain the analysis data, obtaining the distribution density of each analysis data, dividing the distribution density of each analysis data by the mean distribution density of all neighborhood data to obtain the outlier degree of the analysis data, which is denoted as the outlier degree of the sample and the corresponding set processing parameters. Calculating fit stability includes: ； in, This represents the difference between the set processing parameters of the i-th sample and the output result of the i-th sample in the j-th random tree. This represents the outlier degree of the i-th sample and the set processing parameters. This indicates the preset first zero-prevention coefficient. This indicates a preset second zero-prevention coefficient. This represents the output result of the i-th sample in the j-th random tree. Let |i| represent the mean of the output results of the i-th sample across all random trees, |i| represent the magnitude of the vector, and M and N represent the number of random trees and samples, respectively. The closer the fitting stability is to the preset center value, the greater its suitability; the farther it is from the preset center value, the smaller its suitability. The suitability is calculated based on the moderateness of the fitting stability, including: The fitting stability is normalized by taking the reciprocal of the absolute value of the difference between the normalized fitting stability and the preset center value and then normalizing it to obtain the appropriateness of the sampling size. In response to a suitability level less than a preset suitability threshold, the sampling quantity is adjusted based on the fitting stability, including: if the normalized fitting stability is less than a preset center value, the preset adjustment step size is multiplied by the difference between the preset adjustment step size and the normalized fitting stability, and then rounded to obtain the adjustment quantity; the sampling quantity is then added to the adjustment quantity to obtain the adjusted sampling quantity. If the normalized fitting stability is greater than the preset center value, the preset adjustment step size is multiplied by the normalized fitting stability, and then rounded to obtain the adjustment quantity; the sampling quantity is then subtracted from the adjustment quantity to obtain the adjusted sampling quantity. Construct a random forest from the samples re-drawn based on the adjusted sampling size; In response to the suitability level being no less than a preset suitability threshold, the random forest corresponding to the sampling quantity is used as the target random forest to achieve the setting of injection molding process parameters for mobile phone hinges.

2. The method for optimizing injection molding parameters of a mobile phone hinge according to claim 1, characterized in that, The process of obtaining the pre-determined sample size includes: The data on influencing factors with the same processing parameters are grouped into one category. Boundary data between each two categories are obtained. Based on the boundary data, the boundary clarity and boundary pattern complexity of each two categories are obtained. The quotient of boundary pattern complexity and boundary clarity is used as the distinguishing difficulty between each two categories. The average distinguishing difficulty of all two categories is used as the overall distinguishing difficulty. The sampling size is set according to the overall distinguishing difficulty, and the sampling size is positively correlated with the overall distinguishing difficulty.

3. The method for optimizing injection molding parameters of a mobile phone hinge according to claim 2, characterized in that, The method for obtaining the boundary sharpness includes: Obtain the geometric center of the influencing factor data for each category. Draw a straight line through the geometric center of the influencing factor data for the two categories. Project the influencing factor data of the two categories onto the straight line to obtain projection points. Arrange the category label values ​​of all influencing factor data of the two categories according to the order of the projection points to obtain the category label value sequence. The reciprocal of the permutation entropy of the category label value sequence is recorded as the boundary clarity between the two categories.

4. The method for optimizing injection molding parameters of a mobile phone hinge according to claim 3, characterized in that, The method for obtaining the category flag value includes: Set the category flag value of one category of influencing factor data to 1, and set the category flag value of another category of influencing factor data to 0.

5. The method for optimizing injection molding parameters of a mobile phone hinge according to claim 2, characterized in that, The method for obtaining the complexity of the boundary rules includes: Set the number of items variable By fitting a polynomial with T terms using boundary points, T is sequentially assigned different values ​​in ascending order, and the fitting error of the polynomial corresponding to each value of T is obtained. The polynomial whose fitting error is first less than a preset error threshold is denoted as the target polynomial. The highest degree with non-zero coefficients in the objective polynomial is denoted as the objective degree; the boundary condition complexity is obtained by multiplying the objective degree by the number of terms in the objective polynomial.

6. A mobile phone hinge injection molding parameter optimization system, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a method for optimizing injection molding parameters of a mobile phone hinge according to any one of claims 1-5.

Images