Manufacturing process optimization system based on artificial intelligence deep mining
The AI-based manufacturing process optimization system solves the problem of inaccurate production parameter adjustments in traditional manufacturing. By using multiple linear regression and principal component analysis, it achieves scientific accuracy in production parameters, thereby improving production efficiency and product quality stability.
Patent Information
- Application Number
- CN202510770414.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-10
- Publication Date
- 2025-11-07
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In traditional manufacturing processes, the adjustment of production parameters lacks scientific rigor and accuracy, making it difficult to cope with complex and ever-changing production conditions. Simple statistical analysis is also insufficient to handle the complex linear and nonlinear relationships between production parameters, resulting in poor quality control and production optimization.
The manufacturing process optimization system, based on deep mining of artificial intelligence, includes a regression model building module, a linear relationship judgment and processing module, a quality error analysis module, and a production parameter importance calculation module. It builds the model using multiple linear regression, adjusts the linear relationship using principal component analysis, locates quality problems using error analysis, calculates the importance of production parameters, and provides precise production parameter adjustment strategies.
It enables precise positioning of production parameters under complex production conditions, improves production efficiency and product quality stability, reduces resource waste, optimizes production processes, and enhances the scientific nature and consistency of production management.
Smart Images
Figure CN120911646A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of industrial big data, in particular to a manufacturing process optimization system based on artificial intelligence deep mining. BACKGROUND
[0002] Under the background of the vigorous development of modern manufacturing, product quality and production efficiency have become the key elements of enterprise core competitiveness. Manufacturing process involves many complex production parameters, such as speed, pressure and temperature in stamping process. The production parameters are interrelated and have an important influence on product quality.
[0003] In traditional manufacturing process, the adjustment of production parameters often depends on the experience of workers and simple statistical analysis. However, this method has many limitations: on the one hand, experience judgment lacks scientificity and accuracy, and it is difficult to deal with complex and changeable production conditions. For example, when product quality fluctuates in stamping production, workers may not be able to accurately locate which production parameter is the problem, resulting in untimely or inaccurate adjustment, and thus affecting product quality and production efficiency.
[0004] On the other hand, simple statistical analysis is difficult to deal with the complex linear and nonlinear relationships between production parameters. In actual production, multiple collinearity may exist between multiple production parameters, which will interfere with the accuracy of parameter estimation in the regression model, making the effect of quality control and production optimization based on traditional statistical analysis greatly discounted. For example, in a multiple linear regression model, if there is a strong linear relationship between the independent variables (production parameters), the estimated value of the regression coefficient will be unstable, the variance will increase, and the real impact of each parameter on product quality cannot be accurately reflected, so it is difficult to develop effective production parameter adjustment strategies.
[0005] In view of this, we propose a manufacturing process optimization system based on artificial intelligence deep mining. SUMMARY
[0006] The purpose of the present application is to solve the problem that in the face of complex production conditions, there is a lack of scientificity and accuracy, it is difficult to accurately locate the production parameters, and it affects product quality and production efficiency; simple statistical analysis is difficult to deal with the complex linear and nonlinear relationships between production parameters, multiple collinearity interferes with the parameter estimation of the regression model, resulting in poor quality control and production optimization effect.
[0007] To achieve the above purpose, the present application provides a manufacturing process optimization system based on artificial intelligence deep mining, which comprises a regression model establishment module, a linear relationship judgment and processing module, a quality error analysis module and a production parameter importance calculation module, wherein:
[0008] The regression model establishing module adopts a multiple linear regression method to establish a multiple linear regression model between the stamping parameters and the stamping quality; the linear relationship judging and processing module is configured to judge whether there is a linear relationship between the stamping parameters, and if there is a linear relationship between the stamping parameters, the principal component analysis method is adopted to adjust the regression coefficients of the stamping parameters with the linear relationship in the multiple linear regression model;
[0009] The quality error analysis module is configured to perceive the predicted stamping quality and the stamping quality standard interval, and if the predicted stamping quality is not in the stamping quality standard interval, the error analysis method is adopted to analyze the production parameters to be adjusted and the adjustment parameter values, and then output to the production staff; the production parameter importance calculation module is configured to perceive the adjustment parameter values corresponding to the plurality of production parameters, the regression coefficients corresponding to the production parameters, calculate the importance of the adjusted production parameters, and sort the importance in a descending order, and output the adjustment parameter values to the production staff.
[0010] As a further improvement of the technical solution, the working principle of the multiple linear regression method in the regression model establishing module is that: a plurality of historical production data are perceived, and each historical data set includes stamping parameters and stamping quality; and then a multiple linear regression model is established:
[0011] Wherein, y is the stamping quality, x1, x2, x n are the stamping parameters, β0 is the intercept term, specifically the value of y when all the stamping parameters are 0, β1, β2, β n are the regression coefficients, and ∈ is the random error term.
[0012] As a further improvement of the technical solution, the regression coefficients β1, β2, β n in the multiple linear regression model are specifically calculated by the least square method to minimize the sum of squares of errors between the predicted stamping quality and the actual value.
[0013] The beneficial effects of the above further solution are that the relationship between the stamping parameters (such as stamping speed, pressure, temperature, etc.) and the stamping quality is quantified by the established multiple linear regression model, and the direction and degree of the influence of each stamping parameter on the stamping quality are determined by the regression coefficients; for example, if the regression coefficient of the stamping speed is positive, it means that when the speed increases, the stamping quality has an upward trend under the model framework; if it is negative, the quality shows a downward trend;
[0014] In modern manufacturing industry, neither technical personnel nor production management personnel need to have high and deep professional knowledge, and they can understand the relationship between the stamping parameters and the quality expressed by the model, as well as the meaning of the regression coefficients, so as to facilitate the application of the model to the actual production process.
[0015] Based on the technical solution, the application can be further improved as follows.
[0016] As a further improvement of the technical solution, the working principle of the linear relationship judging and processing module for judging whether there is a linear relationship between the stamping parameters is as follows:
[0017]
[0018] Wherein r ij is the correlation coefficient of the stamping parameter x i and the stamping parameter x j , n is the sample quantity of the stamping parameter x i and the stamping parameter x j , x ik is the value of the stamping parameter x i in the kth sample, is the mean value of the stamping parameter x i ; x jk is the value of the stamping parameter x j in the kth sample, is the mean value of the stamping parameter x j .
[0019] A correlation threshold is set, if the absolute value of the correlation coefficient |r ij | is greater than the correlation threshold, it is judged that there is a linear relationship between the stamping parameter x i and the stamping parameter x j .
[0020] As a further improvement of the technical solution, the principal component analysis method in the linear relationship judging and processing module converts the multiple stamping parameters with linear relationship into a group of main components without linear relationship. The main components are linear combinations of the multiple stamping parameters, and are independent of each other.
[0021] The beneficial effect of the above further solution is that, after adjusting the regression coefficient by using the principal component analysis method, the linear relationship between the stamping parameters is eliminated, so that the multiple linear regression model is more stable and accurate. Therefore, when the stamping quality problem occurs, the stamping parameters related to the quality problem can be accurately found based on the accurate model and regression coefficient, and it is clear which parameters change or are related to each other to cause the quality fluctuation, thereby providing an accurate direction for solving the quality problem.
[0022] Based on the technical solution, the application can be further improved as follows.
[0023] As a further improvement of the technical solution, the working principle of the linear relationship judging and processing module for judging whether there is a linear relationship between the stamping parameters is as follows:
[0024] Step one: sense multiple stamping parameters with linear relationship, construct a matrix: X=(x ij ) n×p , where n is the sample number of stamping parameters x i with linear relationship, p is the number of stamping parameters with linear relationship, x ij is the data of the jth stamping parameter of the ith sample; and standardize the data, and the standardized matrix is defined as Z=(z ij ) n×p ;
[0025] Step two: calculate the covariance matrix of the standardized matrix Z The covariance matrix is a p×p matrix, and the main diagonal elements of the covariance matrix are the variances of the stamping parameters, and the non-main diagonal elements are the covariances between the stamping parameters;
[0026] Step three: solve the eigenvalue equation |λI-Σ| of the covariance matrix Σ, get and eigenvectors e1, e2, e p , and the eigenvectors satisfy Σe i =λ i e i ,
[0027] As a further improvement of the technical solution, the principal components F i in the linear relationship judgment and processing module are composed of linear combinations of stamping parameters, and the expression is where e ij is the jth component of the ith eigenvector, and z j is the jth standardized stamping parameter; select the principal components of the first k eigenvalues, so that the cumulative variance contribution rate where A is the contribution threshold.
[0028] As a further improvement of the technical solution, the working principle of the error analysis method in the quality error analysis module is as follows: calculate the prediction error of the predicted stamping quality and the stamping quality standard interval, call out the regression coefficients of each production parameter in the multiple linear regression model, determine the contribution degree of each production parameter to the prediction error, and then further adjust the production parameter, calculate the production parameter adjustment value, so that the predicted stamping quality is within the stamping quality standard interval.
[0029] The beneficial effects of the above further scheme are that when there is a difference between the predicted stamping quality and the stamping quality standard interval in the manufacturing process, the error analysis method in the quality error analysis module can accurately find out the specific production parameter that causes the quality deviation by calculating the prediction error, and determine the contribution degree of each parameter to the prediction error according to the regression coefficients of each production parameter in the multiple linear regression model:
[0030] For example, if the absolute value of the regression coefficient of the stamping pressure is large and the prediction error is positive, it means that the stamping pressure is one of the key factors causing the quality to exceed the standard range. This allows for precise identification of the root cause of the problem, avoiding blind troubleshooting. It also prevents production staff from adjusting production parameters based on experience or blind attempts, saving a lot of time and resources. Production staff can quickly and accurately adjust parameters based on the results of error analysis, reducing stagnation and waste in the production process, improving production efficiency, and enabling the company to produce more qualified products in a shorter time, thereby reducing production costs.
[0031] Based on the above technical solution, the present invention can be further improved as follows.
[0032] As a further improvement to this technical solution, the error analysis method calculation formula in the quality error analysis module is as follows: Calculate the prediction error between the predicted stamping quality and the standard range of stamping quality: Where e is the prediction error The predicted stamping quality is calculated using a multiple linear regression model. The midpoint of the stamping quality standard interval is used as a reference point in the calculation of prediction error. 标准max y 标准min These represent the maximum and minimum values within the standard range for stamping quality, respectively.
[0033] Retrieve the regression coefficients β1, β2, and β3 of each production parameter in the multiple linear regression model. β n To determine the contribution of each production parameter to the prediction error: if the absolute value of the regression coefficient |β i The larger the value, the more significant the impact of production parameters on the predicted stamping quality.
[0034] The prediction error e is decomposed into the contributions of each parameter: Where Δx i For production parameter x i Adjustment values;
[0035] If the prediction error e > 0, the production parameter x i Corresponding regression coefficient β i >0, then according to Δx i Reduce production parameters x i ;
[0036] If the prediction error e > 0, the production parameter x i Corresponding regression coefficient β i <0, then according to Δx i Improve production parameters x i ;
[0037] Conversely, if the prediction error e < 0, then the production parameter x is adjusted.i The directions are opposite.
[0038] As a further improvement to this technical solution, the working principle of the production parameter importance calculation module for calculating the importance of adjusting production parameters is as follows:
[0039] The production parameter for the linear relationship judgment and processing module that does not have a linear relationship is the absolute value of the regression coefficient |β| in the regression model building module. i |As a measure of its importance I i , that is I i =|β i |;
[0040] The linear relationship judgment and processing module detects production parameters with linear relationships. It adds the regression coefficient to another production parameter using the correlation coefficient to increase its importance. It compares the absolute values of the two regression coefficients, determining the one with the larger coefficient. The regression coefficient of the smaller coefficient is then added to the corresponding regression coefficient of the other production parameter through the correlation coefficient, thus determining its importance. The specific expression is as follows: I m =|β m |+|r mn |×|β n |;
[0041] Where |β m |≥|β n |,r mn For production parameter X m and X n Correlation coefficient, β m and β n X m and X n The regression coefficients.
[0042] The beneficial effect of the above-mentioned further scheme is that, for production parameters that do not have a linear relationship, the importance is directly measured by the absolute value of the regression coefficient through the production parameter importance calculation module; for production parameters that have a linear relationship, the importance is determined by reasonably adding and integrating the regression coefficient based on the correlation coefficient.
[0043] To avoid confusion or delays in adjustments due to production staff's lack of understanding of the importance of parameters, optimize resource allocation so that production staff can focus on handling important parameters, thereby improving the scientific and orderly nature of production management;
[0044] According to the parameter importance ranking, the production staff can quickly lock the key parameters and make targeted adjustment, reduce invalid operation on the secondary parameters, save production time, improve production efficiency, make the production process more smooth and efficient, effectively reduce the product quality problems caused by the out-of-control key parameters, improve the product quality stability and consistency, reduce the defective rate, and enhance the product market competitiveness.
[0045] Based on the above technical solutions, the application can also be improved as follows.
[0046] In addition to the purposes, features and advantages described above, the application has other purposes, features and advantages. The application will be further described in detail below with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS
[0047] Figure 1 The overall module schematic diagram of the application.
[0048] The meanings of the various reference numbers in the drawings are as follows:
[0049] 100, regression model establishment module; 200, linear relationship judgment and processing module; 300, quality error analysis module; 400, production parameter importance calculation module. DETAILED DESCRIPTION
[0050] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the application. Obviously, the described embodiments are only some of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.
[0051] The manufacturing process optimization system based on artificial intelligence deep mining includes a regression model establishment module 100, a linear relationship judgment and processing module 200, a quality error analysis module 300, and a production parameter importance calculation module 400, wherein:
[0052] Embodiment one, referring to Figure 1 As shown: the regression model establishment module 100 adopts a multiple linear regression method to establish a multiple linear regression model between the stamping parameters and the stamping quality; thereby through a mapping relationship, the relationship between the stamping parameters such as stamping speed, pressure, and temperature and the stamping quality indicators such as stamping part size error is quantitatively presented by a specific linear equation; the fuzzy way of empirically judging the stamping quality is changed, and the influence direction and degree of each parameter on the stamping quality are clear and explicit; for example, through the regression coefficient, it can be accurately known that the average change amount of the stamping part size error is one unit when the stamping speed increases by one unit, so that the stamping part quality is more stable, and the quality fluctuation is reduced.
[0053] The working principle of the multiple linear regression method in the regression model building module 100 is as follows: Multiple historical production data are collected, each including stamping parameters (stamping speed, pressure, temperature, etc.) and stamping quality (dimensional errors of stamped parts); then, a multiple linear regression model is established.
[0054] Where y is the stamping mass, x1, x2, x n Here are the stamping parameters, β0 is the intercept term, specifically the value of y when all stamping parameters are 0, β1, β2, ... β n The regression coefficients measure the influence of each stamping parameter on stamping quality, and ∈ represents the random error term with a mean of 0 and a variance of σ. 2 The normal distribution, i.e., ∈ ~ N(0, σ 2 );
[0055] In the above multiple linear regression model, the regression coefficients β1, β2, and β3 are... β n Specifically, the least squares method is used to minimize the sum of squared errors between the predicted stamping quality and the actual value for calculation: There are m historical production data points x in total. i1 x i2 , x in , Predicting stamping quality for:
[0056]
[0057] The formula for calculating the sum of squared errors (SSE) is:
[0058] To find the minimum β value of SSE, we consider SSE with respect to β1, β2, and β3 respectively. β n Find the partial derivatives and set them equal to 0: Solving the above system of equations yields an estimate of β. That is, the regression coefficients β1, β2, ... in the multiple linear regression model β n .
[0059] Before stamping production, the stamping parameters input by the production staff are fed into a multiple linear regression model to predict stamping quality in advance and identify potential quality risks in a timely manner. For example, if the dimensional error is predicted to exceed the allowable range, it can prevent the actual production of defective products, reduce the scrap rate, and reduce the cost waste caused by quality problems, including raw material, labor and time costs.
[0060] If multiple stamping parameters have strong linear relationship, the estimated value of regression coefficient in multiple linear regression model will be unstable, and the variance will increase. For example, when stamping speed and pressure have linear correlation, multiple linear regression model is difficult to accurately distinguish the influence of stamping speed and pressure on stamping quality, resulting in deviation of regression coefficient calculated by least square method, and affecting the reliability of multiple linear regression model for stamping quality prediction. Therefore, the linear relationship judgment and processing module 200 is used to judge whether there is linear relationship between stamping parameters: Wherein r ij is the correlation coefficient of stamping parameter x i and stamping parameter x j , n is the sample number of stamping parameter x i and stamping parameter x j , x ik is the value of stamping parameter x i in the kth sample, is the mean value of stamping parameter x i ; x jk is the value of stamping parameter x j in the kth sample, is the mean value of stamping parameter x j ;
[0061] Set the correlation threshold, if the absolute value of the correlation coefficient |r ij |> correlation threshold, it is judged that there is linear relationship between stamping parameter x i and stamping parameter x j ;
[0062] Call out the stamping parameters with linear relationship, and adjust the regression coefficient of the stamping parameters with linear relationship by principal component analysis method. Principal component analysis method is used to convert multiple stamping parameters with linear relationship into a group of main components without linear relationship. The main components are linear combinations of multiple stamping parameters, which are independent of each other. The specific working steps are as follows:
[0063] Step one: perceive multiple stamping parameters with linear relationship, and construct matrix X=(x ij ) n×p , wherein n is the sample number of stamping parameters x i with linear relationship, p is the number of stamping parameters with linear relationship, x ij is the data of the jth stamping parameter in the ith sample; in order to eliminate the influence of variable dimension, the data is standardized: Wherein is the mean value of the jth stamping parameter, is the standard deviation of the jth stamping parameter, and the standardized matrix is defined as Z=(z ij ) n×p ;
[0064] Step two: calculate the covariance matrix of the normalized matrix Z The covariance matrix is a p x p matrix, the main diagonal elements of the covariance matrix are the variances of the stamping parameters, reflecting the fluctuation degree of the stamping parameters themselves, and the non-diagonal elements are the covariances between the stamping parameters, which are used to measure the linear correlation between multiple stamping parameters;
[0065] Step three: solve the characteristic equation |λI-Σ| of the covariance matrix Σ, so as to obtain and the eigenvectors e1, e2, e p , and the eigenvectors satisfy Σe i =λ i e i ,
[0066] Principal component F in principal component analysis i It is composed of a linear combination of stamping parameters, and the expression is Where e ij is the jth component of the ith eigenvector, and z j is the jth normalized stamping parameter; Select the first k (k < p) principal components, so that the cumulative variance contribution rate Where A is the contribution threshold, which is used to judge whether the cumulative variance contribution rate of the selected principal components meets the standard, where the cumulative variance contribution rate In order to make the k principal components retain most of the information of the original data, and realize the reduction of the number of stamping parameters (dimension reduction) while not losing too much key content of the original data.
[0067] Thus, the regression coefficients of the stamping parameters with linear relationship are adjusted, the stamping parameters with linear relationship are converted into independent principal components, the regression coefficients in the multiple linear regression model are adjusted, the key factors of the stamping parameters are highlighted, and the quality error analysis module 300 can be more accurate when analyzing the production parameters and adjustment parameter values that need to be adjusted.
[0068] Example two: In order to adjust the production parameters in time when the predicted stamping quality is problematic, therefore, the difference between this embodiment and the above-mentioned embodiment one is that: the quality error analysis module 300 perceives the predicted stamping quality and the stamping quality standard interval, if the predicted stamping quality is not in the stamping quality standard interval, then the error analysis method is used to analyze the production parameters and adjustment parameter values that need to be adjusted, and then output to the production staff.
[0069] The working principle of the error analysis method in the quality error analysis module 300 is as follows: calculate the prediction error between the predicted stamping quality and the standard range of stamping quality, retrieve the regression coefficients of each production parameter in the multiple linear regression model, determine the contribution of each production parameter to the prediction error (the larger the absolute value of the regression coefficient, the more significant the influence of the production parameter on the predicted stamping quality), and then further adjust the production parameters and calculate the adjustment values of the production parameters so that the predicted stamping quality is within the standard range of stamping quality.
[0070] The error analysis method in the quality error analysis module 300 calculates the following formula: Calculate the prediction error between the predicted stamping quality and the standard range of stamping quality: Where e is the prediction error The predicted stamping quality is calculated using a multiple linear regression model in module 100 for the regression model establishment. The midpoint of the stamping quality standard interval is used as a reference point in the prediction error calculation. 标准max y 标准min These are the maximum and minimum values of the stamping quality standard range, respectively. Since the stamping quality standard range defines the acceptable range of product quality, the midpoint of the stamping quality standard range is the center position of the stamping quality standard range. Using the midpoint as a reference, the degree of deviation of the predicted quality from the standard range can be intuitively reflected, and the deviation of the predicted stamping quality from the ideal state can be reflected.
[0071] Retrieve the regression coefficients β1, β2, and β3 of each production parameter in the multiple linear regression model. β n To determine the contribution of each production parameter to the prediction error: if the absolute value of the regression coefficient |β i The larger the value, the more significant the impact of production parameters on the predicted stamping quality.
[0072] The prediction error e is then decomposed into the contributions of each parameter: Where Δx i For production parameter x i Adjustment values;
[0073] If the prediction error e > 0, the production parameter x i Corresponding regression coefficient β i >0, then according to Δx i Reduce production parameters x i ;
[0074] If the prediction error e > 0, the production parameter x i Corresponding regression coefficient β i <0, then according to Δx i Improve production parameters x i ;
[0075] Conversely, if the prediction error e < 0, the production parameter x is adjusted i The regression coefficients in the multiple linear regression model reflect the relationship between the production parameters and the stamping quality in the opposite direction. The prediction error is calculated by the quality error analysis module 300 and combined with the regression coefficients to determine the contribution of each production parameter to the prediction error, thereby accurately locating the key factors that cause quality deviation and avoiding blind adjustment, so that the stamping quality is maintained within the standard interval, the quality fluctuation in the production process is reduced, the production process is more stable and controllable, and it is conducive to maintaining stable production order.
[0076] In order to avoid the situation that the production staff cannot know the importance of the production parameter when obtaining the production parameter that needs to be adjusted and the adjustment parameter value, and the adjustment is not timely, the embodiment is different from the above-mentioned embodiment one. The production parameter importance calculation module 400 perceives the adjustment parameter values corresponding to the plurality of production parameters, the regression coefficients corresponding to the production parameters, calculates the importance of adjusting the production parameters, and sorts the importance in descending order. The production staff is outputted by the quality error analysis module 300. For the production parameters that exist linear relationship, the importance is calculated combined with the correlation coefficient and the regression coefficient, which can comprehensively consider the correlation between parameters and the influence of each parameter on the result. Compared with the judgment of the regression coefficient, the evaluation result is more in line with the actual situation, which provides more reliable basis for production decision. At the same time, the accurate importance sorting can let the production staff know which parameter has greater influence on the production result, so as to preferentially invest the limited resources (such as time, manpower and material resources) into the adjustment and monitoring of the key parameters, improve the resource utilization efficiency, and optimize the production process.
[0077] The working principle of the production parameter importance calculation module 400 for calculating the importance of adjusting the production parameters is as follows:
[0078] The production parameters without linear relationship in the linear relationship judgment and processing module 200: the absolute value |β i of the regression coefficient in the regression model establishment module 100 is taken as the importance I i of the production parameter. i That is, I i = |β m |;
[0079] The production parameters with linear relationship in the linear relationship judgment and processing module 200 add the regression coefficient to another production parameter through the correlation coefficient to increase the importance, compare the absolute values of the two regression coefficients, determine the regression coefficient with the larger absolute value, add the regression coefficient with the smaller absolute value to the corresponding regression coefficient of another production parameter through the correlation coefficient, and determine the importance, wherein the specific expression is as follows: I m = |β m |+ |rmn |β n |;
[0080] where |β m |≥|β n |, r mn is the correlation coefficient of production parameters X m and X n , β m and β n are the regression coefficients of X m and X n respectively.
[0081] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited by the above examples, the above examples and descriptions in the specification are only preferred examples of the present application, and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.
Claims
1. A manufacturing process optimization system based on deep mining of artificial intelligence, characterized in that, The method comprises a regression model establishing module (100), a linear relationship judging and processing module (200), a quality error analyzing module (300) and a production parameter importance calculating module (400), wherein: The regression model establishing module (100) adopts a multiple linear regression method to establish a multiple linear regression model between stamping parameters and stamping quality; the linear relationship judging and processing module (200) is used for judging whether there is a linear relationship between stamping parameters, and if there is a linear relationship between stamping parameters, a principal component analysis method is adopted to adjust the regression coefficients of the stamping parameters with linear relationship in the multiple linear regression model; The quality error analyzing module (300) perceives predicted stamping quality and a stamping quality standard interval, and if the predicted stamping quality is not in the stamping quality standard interval, an error analysis method is adopted to analyze the production parameters to be adjusted and the adjustment parameter values, and then output to a production worker; the production parameter importance calculating module (400) perceives the adjustment parameter values corresponding to a plurality of production parameters, the regression coefficients corresponding to the production parameters, calculates the importance of the adjusted production parameters, and sorts the importance from large to small, and outputs the adjustment parameter values to the production worker together with the quality error analyzing module (300). 2.The artificial intelligence deep mining based manufacturing process optimization system according to claim 1, characterized in that: The working principle of the multiple linear regression method in the regression model establishing module (100) is as follows: multiple historical production data are perceived, each historical data set includes stamping parameters and stamping quality; and then a multiple linear regression model is established: where y is the stamping mass, x1, x2, x n are the stamping parameters, β0 is the intercept term, specifically the value of y when all stamping parameters x1, x2, β n are the regression coefficients, and ∈ is the random error term. 3.The artificial intelligence deep mining based manufacturing process optimization system according to claim 1, characterized in that: The regression coefficients β1, β2, β n Specifically, the least square method is used to minimize the sum of squares of errors between the predicted stamping mass and the actual value. 4.The artificial intelligence deep mining based manufacturing process optimization system according to claim 2, characterized in that: The linear relationship judging and processing module (200) judges whether there is a linear relationship between stamping parameters, and the working principle is as follows: wherein r ij is the correlation coefficient of the stamping parameter x i and the stamping parameter x j , n is the sample number of the stamping parameter x i and the stamping parameter x j , x ik is the value of the stamping parameter x i in the kth sample, is the mean value of the stamping parameter x i ; x jk is the value of the stamping parameter x j in the kth sample, is the mean value of the stamping parameter x j ; A correlation threshold is set, and if the absolute value |r ij of the correlation coefficient is greater than the correlation threshold, it is determined that the stamping parameter x i and the stamping parameter x j have a linear relationship. 5.The artificial intelligence deep mining based manufacturing process optimization system according to claim 4, characterized in that: The principal component analysis method in the linear relationship judging and processing module (200) converts a plurality of stamping parameters with linear relationship into a group of main components without linear relationship, and the main components are linear combinations of the plurality of stamping parameters and are independent of each other. 6.The artificial intelligence deep mining based manufacturing process optimization system according to claim 4, characterized in that: The specific working steps of the principal component analysis method in the linear relationship judging and processing module (200) are as follows: Step one: sense multiple stamping parameters with linear relationship, construct a matrix: X=(x ij ) n×p , where n is the sample number of stamping parameters x i with linear relationship, p is the number of stamping parameters with linear relationship, x ij is the data of the jth stamping parameter of the ith sample; and standardize the data, the standardized matrix is defined as Z=(z ij ) n×p ; Step two: calculate the covariance matrix of the standardized matrix Z The covariance matrix is a p x p matrix, the main diagonal elements of the covariance matrix are the variances of the stamping parameters, and the non-diagonal elements are the covariances between the stamping parameters; Step three: Solve the eigen-equation of the covariance matrix Σ, |λI - Σ|, to get and eigenvectors e1, e2, e p and the eigenvectors satisfy Σe i = λ i e i , 7.The artificial intelligence deep mining based manufacturing process optimization system according to claim 6, characterized in that: The principal component F in the principal component analysis method i is composed of linear combination of stamping parameters, and the expression is where e ij is the jth component of the ith eigenvector, z j is the jth normalized stamping parameter; the principal components of the first k eigenvalues are selected, so that the cumulative variance contribution rate is where A is a contribution threshold. 8.The artificial intelligence deep mining based manufacturing process optimization system according to claim 1, wherein: The working principle of the error analysis method in the quality error analyzing module (300) is as follows: the prediction error of the predicted stamping quality and the stamping quality standard interval is calculated, the regression coefficients of each production parameter in the multiple linear regression model are called out, the contribution degree of each production parameter to the prediction error is determined, and then the production parameters are further adjusted and the production parameter adjustment values are calculated, so that the predicted stamping quality is in the stamping quality standard interval. 9.The artificial intelligence deep mining based manufacturing process optimization system according to claim 8, characterized in that: The error analysis formula in the quality error analysis module (300) is as follows: the prediction error of the predicted stamping quality and the standard interval of the stamping quality is calculated. Wherein e is the prediction error, is the predicted stamping quality calculated by the multiple linear regression model, is the midpoint of the standard interval of the stamping quality, which is used as a reference benchmark in the prediction error calculation, y 标准max , y 标准min are the maximum and minimum values of the standard interval of the stamping quality, respectively. The regression coefficients β1, β2 of each production parameter in the multiple linear regression model are called out, β n The contribution of each production parameter to the prediction error is determined: the larger the absolute value |β i | of the regression coefficient, the more significant the impact of the production parameter on the prediction of the stamping quality. The prediction error e is decomposed into contributions of the individual parameters: where Δx i is the adjustment value for the production parameter x i . If the prediction error e > 0, the production parameter x i Corresponding regression coefficient β i >0, then according to Δx i Reduce production parameters x i ; If the prediction error e > 0, the production parameter x i The corresponding regression coefficient β i < 0, then according to Δx i The production parameter x i ; Conversely, if the prediction error e < 0, then the production parameter x is adjusted in the opposite direction. i The direction is opposite. 10.The artificial intelligence deep mining based manufacturing process optimization system according to claim 9, characterized in that: The working principle of calculating the importance of the adjusted production parameters in the production parameter importance calculating module (400) is as follows: sensing the linear relationship determining and processing module (200) to determine the production parameters without linear relationship: the absolute value of the regression coefficient |β i | as a measure of its importance I i , that is, I i = |β i |; The linear relationship judgment and processing module (200) perceives the production parameters with linear relationship, adds the regression coefficient to another production parameter through the correlation coefficient to increase its importance, compares the absolute value of the two regression coefficients, determines the regression coefficient with larger absolute value, adds the regression coefficient with smaller absolute value to the corresponding regression coefficient of another production parameter through the correlation of the correlation coefficient, and determines the importance, and the specific expression is as follows: I m = |β m | + |r mn | x |β n |; where |β m |≥|β n |, r mn is the correlation coefficient of the production parameters X m and X n , β m and β n are the regression coefficients of X m and X n , respectively.