Quality factor analysis method, manufacturing method, quality factor analysis device, and manufacturing system

Abstract

To improve accuracy of factor analysis for quality of an object.SOLUTION: A method according to one embodiment of the present invention is a method executed by a quality factor analysis device, including the steps of: acquiring an image of a cosmetic appearance of a manufactured object; calculating a value that quantifies the image; acquiring manufacturing conditions of the object; and performing machine learning by using the manufacturing conditions of the object as input data and the quantified value of the image as output data, and calculating a contribution degree of each variable included in the manufacturing conditions of the object to the quantified value of the image.SELECTED DRAWING: Figure 12

Description

[Technical field]

[0001] The present invention relates to a quality factor analysis method, a manufacturing method, a quality factor analysis device, and a manufacturing system. [Background technology]

[0002] It is known that in the manufacture of products, factors that deteriorate the quality of the products are analyzed. In such a factor analysis of the quality of the products, the quality of the manufactured products may be judged by a sensory evaluation. Summary of the Invention [Problem to be solved by the invention]

[0003] However, sensory evaluation involves subjective elements, making it difficult to make a quantitative judgment, and it is not always possible to make a consistent judgment. When multiple people make the judgment, there are individual differences. As a result, the accuracy of the factor analysis of the quality of the product decreases.

[0004] Therefore, an object of the present invention is to improve the accuracy of factor analysis of product quality. [Means for solving the problem]

[0005] A method according to one embodiment of the present invention is a method executed by a quality factor analysis device, and includes the steps of acquiring an image of the appearance of a manufactured object, calculating a quantified value of the image, acquiring manufacturing conditions of the object, and performing machine learning using the manufacturing conditions of the object as input data and the quantified value of the image as output data, and calculating the contribution of each variable included in the manufacturing conditions of the object to the quantified value of the image. Effect of the Invention

[0006] According to the present invention, it is possible to improve the accuracy of the factor analysis of the quality of an object. [Brief description of the drawings]

[0007] [Figure 1] 1 is an overall configuration diagram according to an embodiment of the present invention; [Diagram 2] 1 is a hardware configuration diagram of a quality factor analysis apparatus according to an embodiment of the present invention. [Diagram 3] 1 is a functional block diagram of a quality factor analysis device according to an embodiment of the present invention. [Figure 4] 2 is an example of a manufacturing apparatus according to an embodiment of the present invention. [Diagram 5] 1 is an example of an inspection device according to an embodiment of the present invention. [Figure 6] FIG. 1 is a diagram for explaining quantification of an image according to an embodiment of the present invention. [Figure 7] FIG. 1 is a diagram for explaining quantification of an image according to an embodiment of the present invention. [Figure 8] FIG. 13 is a diagram showing the relationship between a quantified value of an image according to one embodiment of the present invention and the result of a sensory evaluation of defects in the appearance of an object. [Figure 9] FIG. 2 is a diagram for explaining an overall processing flow according to an embodiment of the present invention. [Figure 10] 1 is a diagram showing an example of variables and contributions included in manufacturing conditions according to an embodiment of the present invention. [Figure 11] 11 is an example of quantified values ​​of images when manufacturing conditions are changed according to an embodiment of the present invention. [Figure 12] 10 is a flowchart showing a process for calculating the contribution of each variable included in the manufacturing conditions of an object to a quantified value of an image of the appearance of the object according to one embodiment of the present invention. [Figure 13] 11 is a sequence diagram showing a process of calculating the contribution of each variable included in the manufacturing conditions of an object to a quantified value of an image of the appearance of the object according to an embodiment of the present invention. FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0009] <Overall composition> 1 is an overall configuration diagram according to an embodiment of the present invention. The manufacturing system 1 includes a quality factor analysis device 10, an inspection device 21, and a manufacturing device 22. Note that in this specification, the quality factor analysis device 10, the inspection device 21, and the manufacturing device 22 are described as separate devices, but two or more of the quality factor analysis device 10, the inspection device 21, and the manufacturing device 22 may be implemented in a single device.

[0010] <<Quality factor analysis device>> The quality factor analysis device 10 analyzes factors of the quality of an object manufactured by the manufacturing device 22. Specifically, the quality factor analysis device 10 acquires an image of the appearance of the object manufactured by the manufacturing device 22, calculates a value that quantifies the image, and acquires the manufacturing conditions of the object. The quality factor analysis device 10 performs machine learning using the manufacturing conditions of the object as input data and the quantified value of the image as output data, and calculates the contribution of each variable included in the manufacturing conditions of the object to the quantified value of the image. In addition, the quality factor analysis device 10 can determine optimal manufacturing conditions based on the contribution of each variable included in the manufacturing conditions of the object to the quantified value of the image.

[0011] The devices described in the examples are merely one of a number of computing environments for implementing the embodiments disclosed herein. In one embodiment, the quality factor analysis apparatus 10 includes a number of computing devices, such as a server cluster. The computing devices are configured to communicate with each other via any type of communication link, including a network, shared memory, and the like, and perform the processes disclosed herein.

[0012] <<Inspection equipment>> The inspection device 21 is a device that captures an image of the appearance of a manufactured product. Specifically, the inspection device 21 captures an image of the appearance of the product manufactured by the manufacturing device 22 and transmits the image of the appearance of the product to the quality factor analysis device 10.

[0013] The quality factor analysis device 10 may obtain the image of the appearance of the object captured by the inspection device 21 from a device other than the inspection device 21.

[0014] <<Manufacturing equipment>> The manufacturing apparatus 22 is an apparatus that manufactures an object. Specifically, the manufacturing apparatus 22 transmits manufacturing conditions for manufacturing the object to the quality factor analysis apparatus 10, and acquires the optimal manufacturing conditions determined by the quality factor analysis apparatus 10 from the quality factor analysis apparatus 10.

[0015] The quality factor analysis device 10 may acquire manufacturing conditions for manufacturing an object that are input to the quality factor analysis device 10 by an operator of the quality factor analysis device 10. The quality factor analysis device 10 may display the optimal manufacturing conditions determined by the quality factor analysis device 10 on a screen and present them to the operator of the quality factor analysis device 10.

[0016] <Hardware configuration> Fig. 2 is a hardware configuration diagram of a quality factor analysis apparatus 10 according to an embodiment of the present invention. As shown in Fig. 2, the quality factor analysis apparatus 10 is constructed by a computer, and includes a CPU 1001, a ROM 1002, a RAM 1003, an HD 1004, an HDD (Hard Disk Drive) controller 1005, a display 1006, an external device connection I / F (Interface) 1007, a network I / F 1008, a data bus 1009, a keyboard 1010, a pointing device 1011, a DVD-RW (Digital Versatile Disk Rewritable) drive 1013, and a media I / F 1015.

[0017] Among these, the CPU 1001 controls the operation of the entire quality factor analysis device 10. The ROM 1002 stores programs such as IPL used to drive the CPU 1001. The RAM 1003 is used as a work area for the CPU 1001. The HD 1004 stores various data such as programs. The HDD controller 1005 controls reading or writing of various data from the HD 1004 under the control of the CPU 1001. The display 1006 displays various information such as a cursor, a menu, a window, a character, or an image. The external device connection I / F 1007 is an interface for connecting various external devices. In this case, the external device is, for example, a USB (Universal Serial Bus) memory or a printer. The network I / F 1008 is an interface for data communication using a communication network. The bus line 1009 is an address bus, a data bus, or the like for electrically connecting each component such as the CPU 1001 shown in FIG. 2.

[0018] The keyboard 1010 is a type of input means having a plurality of keys for inputting characters, numbers, various instructions, etc. The pointing device 1011 is a type of input means for selecting and executing various instructions, selecting a processing target, moving a cursor, etc. The DVD-RW drive 1013 controls reading and writing of various data from a DVD-RW 1012 as an example of a removable recording medium. Note that the medium is not limited to a DVD-RW, and may be a DVD-R, etc. The media I / F 1015 controls reading and writing (storing) of data from a recording medium 1014 such as a flash memory.

[0019] <Functional configuration> 3 is a functional block diagram of a quality factor analysis device 10 according to an embodiment of the present invention. The quality factor analysis device 10 includes an image acquisition unit 101, an image quantification unit 102, a manufacturing condition acquisition unit 103, a machine learning unit 104, an optimal manufacturing condition determination unit 105, and a manufacturing condition storage unit 106. The quality factor analysis device 10 functions as the image acquisition unit 101, the image quantification unit 102, the manufacturing condition acquisition unit 103, the machine learning unit 104, and the optimal manufacturing condition determination unit 105 by executing a program.

[0020] The image acquiring unit 101 acquires an image of the appearance of the manufactured product. For example, the image acquiring unit 101 acquires an image of the product manufactured by the manufacturing apparatus 22, captured by the inspection apparatus 21, from the inspection apparatus 21. For example, the image of the appearance of the product is an image of the surface of a water-foamed silicone rubber tube.

[0021] The image quantification unit 102 calculates a quantified value of the image acquired by the image acquisition unit 101 (i.e., a quantified value of the appearance of an object in the image). For example, the image quantification unit 102 calculates a quantified value of a defect in the appearance of an object in the image. For example, the quantified value of a defect in the appearance of an object in the image is a quantified value of uneven color of the object in the image (for example, the total number of pixels in an area where uneven color occurs (for example, an area where a shading difference equal to or greater than a threshold occurs)).

[0022] The manufacturing condition acquisition unit 103 acquires the manufacturing conditions of the product (e.g., the raw materials and processing conditions of the product (note that the processing conditions are conditions related to manufacturing and include a plurality of variables)). For example, the manufacturing conditions are the raw materials of the water-foamed silicone rubber tube and the processing conditions of the emulsification and casting process (e.g., the temperature of the tank, the rotation speed of the mixer, etc.).

[0023] The machine learning unit 104 performs machine learning using the manufacturing conditions of the product acquired by the manufacturing condition acquisition unit 103 as input data and the quantified value of the image calculated by the image quantification unit 102 as output data, and calculates the contribution of each variable included in the manufacturing conditions of the product to the quantified value of the image (also called feature importance or variable importance. For example, the contribution may be calculated using Permutation Importance or SHAP (SHapley Additive exPlanations)).

[0024] The optimum manufacturing condition determination unit 105 determines optimum manufacturing conditions based on the contribution degree (i.e., the contribution degree of each variable included in the manufacturing conditions of the product to the quantified value of the image) calculated by the machine learning unit 104. The optimum manufacturing condition determination unit 105 determines the manufacturing conditions in which the variable with the high contribution degree is changed as the optimum manufacturing conditions.

[0025] The manufacturing condition storage unit 106 stores the manufacturing conditions and the contribution degree calculated by the machine learning unit 104 (i.e., the contribution degree of each variable included in the manufacturing conditions of the product to the quantified value of the image).

[0026] The quality factor analysis device 10 can analyze the factors of product quality for each product lot. In this case, the image quantification unit 102 statistically processes the quantified value of the image of the product's appearance for each product lot. The manufacturing condition acquisition unit 103 statistically processes the manufacturing conditions of the product for each product lot. The machine learning unit 104 performs machine learning using the quantified value of the image of the product's appearance for each statistically processed lot and the manufacturing conditions of the product for each statistically processed lot, and calculates the contribution of each variable included in the manufacturing conditions to the quantified value of the image of the appearance.

[0027] FIG. 4 shows an example of a manufacturing apparatus 22 according to an embodiment of the present invention (a foam molding apparatus for manufacturing a water-foamed silicone rubber tube).

[0028] The foam molding apparatus 22 includes a material tank A 2001A, a material tank B 2001B, a material tank C 2001C, a liquid supply pump A 2002A, a liquid supply pump B 2002B, a liquid supply pump C 2002C, a liquid supply pipe A 2003A, a liquid supply pipe B 2003B, a liquid supply pipe C 2003C, a mixer 2004, and a molding die 2005.

[0029] Raw materials are put into each of the material tanks A to C. The raw materials in the material tank A 2001A are sent to the mixer 2004 through the liquid feed pump A 2002A and the liquid feed piping A 2003A. Similarly, the raw materials in the material tank B 2001B are sent to the mixer 2004 through the liquid feed pump B 2002B and the liquid feed piping B 2003B. Similarly, the raw materials in the material tank C 2001C are sent to the mixer 2004 through the liquid feed pump C 2002C and the liquid feed piping C 2003C. The raw materials are mixed in the mixer 2004 and poured into a molding die 2005 to manufacture a water-foamed silicone rubber tube.

[0030] In water-foamed silicone rubber tubes, if there are any areas where the foaming is poor, those areas (areas where the foaming is poor) will appear as uneven coloring. Conventionally, when inspecting the quality of uneven coloring, uneven coloring was divided into five levels, from least uneven to most uneven, and each level was ranked from 1 to 5 (these levels are called the uneven coloring ranks), and a sensory evaluation was conducted to determine which level the tube fell into.

[0031] 5 is an example of an inspection device 21 according to an embodiment of the present invention. The inspection device 21 images the surface of a water-foamed silicone rubber tube. Specifically, a water-foamed silicone rubber tube 2014 is placed on a rotation mechanism 2013, rotated, and an image is taken of it. Using a light source (LED bar) 2012 makes it possible to image at a constant brightness. By capturing an image with a line camera 2011, it is possible to obtain a single image data of the entire circumference of the water-foamed silicone rubber tube as shown in FIG. 6.

[0032] Fig. 6 is a diagram for explaining quantification of an image according to an embodiment of the present invention. [Before quantification] in Fig. 6 is an image captured by the method described in Fig. 5, and [after quantification] in Fig. 6 is an image in which color unevenness has been quantified by the method described in Fig. 7.

[0033] 7 is a diagram for explaining quantification of an image according to one embodiment of the present invention. In one embodiment of the present invention, a captured image is processed into an image in which color unevenness can be quantified.

[0034] For example, a method for quantifying an image (i.e., the appearance of objects in an image) involves convolution processing of the image, applying a filter to extract features and remove noise. For example, an algorithm called ADAPTIVE THRESH MEAN is used for the required range of the image to perform binarization processing that removes luminance unevenness and clarify areas of color unevenness. After that, noise other than areas where color unevenness occurs and noise below a certain size are removed. From the completed binarized image, mask processing is used to extract only black areas, and an image with luminance distribution is formed. Based on the formed image, the number of pixels where differences in shading occur is measured to quantify color unevenness. A detailed explanation is given below.

[0035] Step 1 (S1) shows a captured image. Unnecessary parts of the image in S1 (e.g., both sides) are cut off.

[0036] Step 2 (S2) shows an image with unnecessary parts cut off. The image of S2 is binarized by adaptive threshold processing. For example, the median value of the neighboring region may be used as the threshold, or the weighted average value of the neighboring region (the weight value is calculated so as to have a Gaussian distribution) may be used as the threshold.

[0037] Step 3 (S3) shows the binarized image. Morphological processing fills small holes and fine gaps in the image in S3.

[0038] Morphological processing is a process of image processing in which image processing is performed based on the shape of the image. For example, there are the following types of morphological processing: Inflate: Increase the object size by expanding the object's boundaries Erosion: Reduce the object size by shrinking the object's boundaries Opening: Erosion followed by dilation removes small objects and reduces noise Closing: By performing a contraction operation after an expansion operation, small holes and fine gaps are filled. Edge detection: Detects the boundaries of objects by subtracting the result of the erosion operation from the result of the dilation operation.

[0039] Step 4 (S4) shows the image that has been subjected to morphological processing. The image of S4 is subjected to noise removal.

[0040] Step 5 (S5) shows the image after noise removal. Masking of the image in S5 is performed.

[0041] Step 6 (S6) shows the image after the mask process.

[0042] FIG. 8 is a diagram showing the relationship between the quantified value of an image according to one embodiment of the present invention and the result of a sensory evaluation of defects in the appearance of an object. The "color unevenness rank (image area sensory evaluation)" on the horizontal axis is the color unevenness rank ("1" to "5") that is the result of the sensory evaluation, and the "total number of pixels" on the vertical axis indicates the quantified value of the image (the total number of pixels of color unevenness derived from the image). As shown in FIG. 8, there was a correlation between the quantified value of the image and the result of the sensory evaluation. Furthermore, by showing such a relationship, it was possible to make the boundary between the color unevenness ranks clearer.

[0043] FIG. 9 is a diagram for explaining the overall processing flow according to one embodiment of the present invention.

[0044] The quality factor analysis device 10 acquires raw material data (step 111), and executes data processing (data cleansing) of the raw material data (step 112).

[0045] The quality factor analysis device 10 acquires data on processing conditions (step 121), and executes data processing (data cleansing) of the data on the processing conditions (step 122).

[0046] The quality factor analysis device 10 acquires an image of a manufactured product (step 131), executes image processing to quantify the image (step 132), and calculates a value that quantifies the image (step 133).

[0047] Next, the quality factor analysis device 10 performs machine learning using the raw material data of S112, the processing condition data of S122, and the quantified value of the image of S133 to calculate the contribution degree (step 11). Next, the quality factor analysis device 10 extracts important variables (variables with high contribution degrees and easy to adjust) based on the contribution degree of S11 (step 12). Next, the quality factor analysis device 10 changes the values ​​of the important variables to calculate optimal manufacturing conditions (i.e., operation amounts in the process) (step 13).

[0048] In one embodiment of the present invention, the machine learning method can be Logistic Regression, Random Forest, or Multivariate Statistical Process Control (MSPC). Each of these methods will be described below.

[0049] [Logistic Regression] Model Type: Linear Model Troublesome problems: Binary classification problems Learning method: Use maximum likelihood estimation for parameter estimation

[0050] [Random Forest] Model type: A model created by combining randomly generated decision trees Problems that can be handled: Classification, regression, clustering Training method: Randomly create decision trees and train each one using bootstrap sampling.

[0051] [MSPC (Statistical Method)] Method: Statistical method. Used for anomaly detection. Tacklable problems: Process anomaly detection and process control Learning method: Learn only normal data and be able to detect abnormalities when they occur

[0052] <Extraction of variables with high contribution> 10 shows an example of variables and their contributions included in the manufacturing conditions according to one embodiment of the present invention. Specifically, the variables with high contributions are shown for each machine learning method (logistic regression, random forest, and multivariate statistical process control (MSPC)).

[0053] In logistic regression, Tank A pressure had the highest contribution, followed by Agitation Speed, Apparent Viscosity, Filling Pressure, and Mean Flow A. In random forest, Agitation Speed ​​had the highest contribution, followed by Material Lot_A, Pump Torque_A, Material Lot Code_A, and Pump Torque_B. In multivariate statistical process control, Mean Flow A had the highest contribution, followed by Agitation Speed, Filling Pressure, Material Lot_A, and Tank A Pressure.

[0054] When reviewing the manufacturing conditions, variables that have a high contribution to the quantified value of the image (for example, the quantified value of color unevenness) and that are easy to adjust are selected from the variables included in the manufacturing conditions. From the results in Figure 10, "agitation speed" was selected, and Figure 11 shows the results of comparing the quantified values ​​of color unevenness of the completed water-foamed silicone rubber tubes when the "agitation speed" was changed from "low," "medium," and "high."

[0055] <Determining optimal manufacturing conditions> 11 shows an example of quantified values ​​of an image when the manufacturing conditions according to one embodiment of the present invention are changed. Specifically, the figure shows quantified values ​​(total number of pixels of color unevenness) of color unevenness when the variable (stirring speed) included in the manufacturing conditions is "low", "medium", and "high".

[0056] When the stirring speed was "low", the quantified value of color unevenness was "51023", when the stirring speed was "medium", the quantified value of color unevenness was "15438", and when the stirring speed was "high", the quantified value of color unevenness was "92476", showing that the quality was best (less color unevenness) when the stirring speed was "medium". Furthermore, when manufactured under the optimal manufacturing conditions (i.e., stirring speed: "medium"), the non-defective product rate improved from 86.2% to 89.1%.

[0057] In this way, among the variables included in the manufacturing conditions, variables that have a high degree of contribution and are easy to adjust are selected, and the optimal manufacturing conditions can be determined by comparing the quantified values ​​of the images when the variables are changed (for example, the quantified values ​​of color unevenness).

[0058] <Method> The processing method will be described with reference to FIGS.

[0059] FIG. 12 is a flowchart showing a process for calculating the contribution of each variable included in the manufacturing conditions of an object to a quantified value of an image of the appearance of the object according to one embodiment of the present invention.

[0060] In step 101 (S101), the image acquisition unit 101 acquires an image of the appearance of a manufactured product.

[0061] In step 102 (S102), the image quantification unit 102 calculates a quantified value of the image acquired in S101.

[0062] In step 103 (S103), the manufacturing condition acquisition unit 103 acquires manufacturing conditions for the product. Note that S103 may be executed before S102 or simultaneously with S102, or before S101 or simultaneously with S101.

[0063] In step 104 (S104), the machine learning unit 104 performs machine learning using the manufacturing conditions of the object acquired in S103 and the quantified value of the image calculated in S102, and calculates the contribution degree of each variable included in the manufacturing conditions of the object to the quantified value of the image. Specifically, the machine learning unit 104 performs machine learning using the manufacturing conditions of the object as input data and the quantified value of the image as output data, and calculates the contribution degree of each variable included in the manufacturing conditions of the object to the quantified value of the image.

[0064] In step 105 (S105), the optimum manufacturing condition determination unit 105 determines the optimum manufacturing conditions based on the contribution degree calculated in S104.

[0065] FIG. 13 is a sequence diagram showing a process for calculating the contribution of each variable included in the manufacturing conditions of an object to a quantified value of an image of the appearance of the object according to one embodiment of the present invention.

[0066] In step 1001 (S1001), the inspection device 21 captures an image of the appearance of the product manufactured by the manufacturing device 22 and transmits the image of the appearance of the product to the quality factor analysis device 10. After that, the quality factor analysis device 10 acquires the image of the appearance of the product from the inspection device 21.

[0067] In step 1002 (S1002), the quality factor analysis apparatus 10 calculates a quantified value of the image acquired in S101.

[0068] In step 1003 (S1003), the manufacturing apparatus 22 transmits manufacturing conditions when manufacturing the product to the quality factor analysis apparatus 10. After that, the quality factor analysis apparatus 10 acquires from the manufacturing apparatus 22 the manufacturing conditions when manufacturing the product.

[0069] In step 1004 (S1004), the quality factor analysis device 10 performs machine learning using the manufacturing conditions of the product acquired in S1003 and the quantified value of the image calculated in S1002, and calculates the contribution degree of each variable included in the manufacturing conditions of the product to the quantified value of the image. Specifically, the quality factor analysis device 10 performs machine learning using the manufacturing conditions of the product as input data and the quantified value of the image as output data, and calculates the contribution degree of each variable included in the manufacturing conditions of the product to the quantified value of the image.

[0070] In step 1005 (S1005), the quality factor analysis apparatus 10 determines optimal manufacturing conditions based on the contribution degree calculated in S1004.

[0071] In step 1006 (S1006), the quality factor analysis apparatus 10 transmits the optimal manufacturing conditions determined in S1005 to the quality factor analysis apparatus 10.

[0072] <Effects> In one embodiment of the present invention, when reviewing the manufacturing conditions to improve the quality of the product, the results of a quantitative evaluation of the quality of the manufactured product are used, so that optimal manufacturing conditions can be determined. In other words, instead of using the results of a sensory evaluation for machine learning, the results of quantifying an image are used for machine learning, so that more reliable important variables can be extracted from the variables included in the manufacturing conditions.

[0073] Each function of the above-described embodiments can be realized by one or more processing circuits. Here, the term "processing circuit" in this specification includes a processor programmed to execute each function by software, such as a processor implemented by an electronic circuit, and a device such as an ASIC (Application Specific Integrated Circuit), a DSP (digital signal processor), an FPGA (field programmable gate array), or a conventional circuit module designed to execute each function described above. [Explanation of symbols]

[0074] 1. Manufacturing System 10 Quality factor analysis device 21 Inspection equipment 22 Manufacturing equipment 101 Image acquisition unit 102 Image Quantification Division 103 Manufacturing condition acquisition department 104 Machine Learning Department 105 Optimal Manufacturing Condition Determination Unit 106 Manufacturing condition storage section [Prior art documents] [Patent documents]

[0075] [Patent Document 1] Patent Publication No. 2021-009126

Claims

1. A method performed by a quality factor analysis device, comprising: - acquiring an image of the appearance of the manufactured object; calculating a quantification value of the image; Obtaining manufacturing conditions of the product; A step of performing machine learning using the manufacturing conditions of the object as input data and the quantified value of the image as output data, and calculating the contribution of each variable included in the manufacturing conditions of the object to the quantified value of the image; The method includes:

2. The method of claim 1 , wherein the quantified value of the image is a quantified value of a defect in the appearance of the object.

3. The method of claim 1 or 2, wherein the machine learning is performed using one of the techniques of regression, classification, and multivariate statistical process control.

4. determining optimal manufacturing conditions based on the contribution of each variable included in the manufacturing conditions of the object to a value quantified by the image; The method of claim 1 further comprising:

5. The method according to claim 4 , further comprising determining, as the optimum manufacturing conditions, manufacturing conditions obtained by changing the variables having the high degree of contribution.

6. A method for manufacturing an article by using a manufacturing apparatus under the optimum manufacturing conditions according to claim 4 or 5.

7. an image acquisition unit for acquiring an image of the appearance of the manufactured product; An image quantification unit that calculates a quantified value of the image; A manufacturing condition acquisition unit that acquires manufacturing conditions of the product; a machine learning unit that performs machine learning using the manufacturing conditions of the product as input data and the quantified value of the image as output data, and calculates the contribution of each variable included in the manufacturing conditions of the product to the quantified value of the image; A quality factor analysis device equipped with

8. A manufacturing system including a manufacturing device that manufactures an object, an inspection device that captures an external appearance of the object, and a quality factor analysis device, The quality factor analysis device comprises: an image acquisition unit that acquires an image of the appearance of a manufactured product from the inspection device; An image quantification unit that calculates a quantified value of the image; a manufacturing condition acquisition unit that acquires manufacturing conditions of the object from the manufacturing device; a machine learning unit that performs machine learning using the manufacturing conditions of the product as input data and the quantified value of the image as output data, and calculates the contribution of each variable included in the manufacturing conditions of the product to the quantified value of the image; A manufacturing system equipped with

Images