Artificial intelligence system for quality detection

Abstract

The invention relates to the technical field of quality detection, in particular to an artificial intelligence system for quality detection. The system comprises a conversion processing unit, a calculation screening unit, a model detection unit and a quality detection unit. The model detection unit calculates the gradient direction of the change of the weak edge points according to the horizontal gradient and the vertical gradient, and then detects whether the weak edge points of the overall appearance of the plastic mold are burr points or not according to the second derivative of the weak edge points and the number of the edge points; according to the irregular burr area formed by connecting the burr points and the gradient direction of the burr point change, the curvature change of the edge points can be effectively captured through the second derivative, the normal edge points are prevented from being misjudged as the burr points, the detection reliability is improved, meanwhile, the gradient direction of the weak burr point change is analyzed under the condition that the burr points are determined, and the detection accuracy is improved. Irregular burr areas can be accurately connected, the quality of the plastic mold can be rapidly detected, and the quality detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of quality inspection, and in particular to an artificial intelligence system for quality inspection. Background Art

[0002] Plastic molds play a vital role in industrial production, especially in the processes of injection molding, compression molding and blow molding. The quality of the mold directly affects the precision, appearance and performance of the final product. However, the traditional manual inspection method has the problems of low efficiency, strong subjectivity and easy error, which makes it difficult to meet the modern industry's demand for high quality and high efficiency. Artificial intelligence for plastic mold quality inspection is a kind of integration of artificial intelligence technology, such as machine learning, deep learning, and support vector machine, which can automatically detect, analyze and evaluate the quality of plastic molds. When a plastic mold is produced, a small amount or slight burrs will be generated on the edge of the plastic mold. It is impossible to effectively capture the curvature change of the edge points of the plastic mold to detect whether there are burrs on the edge of the plastic mold. Normal plastic mold edge points are often misjudged as burrs, which increases the waste of resources in the production of plastic molds and reduces the reliability of detection. When the burrs are determined, the burrs cannot be accurately connected according to the gradient direction of the burrs, which makes the generated burrs inaccurate, resulting in the subsequent omission of the burrs, affecting the efficiency of detecting the quality of plastic molds, thereby reducing the efficiency of quality inspection. Therefore, we provide an artificial intelligence system for quality inspection. Summary of the invention

[0003] The purpose of the present invention is to provide an artificial intelligence system for quality inspection to solve the problems raised in the above background technology.

[0004] To achieve the above-mentioned object, the present invention provides an artificial intelligence system for quality detection, comprising a conversion processing unit, a calculation and screening unit, a model detection unit and a quality detection unit;

[0005] The conversion processing unit acquires the overall appearance data of the production plastic mold through artificial intelligence to establish a machine learning model, and the machine learning model extracts pixel coordinates from the overall appearance data of the plastic mold for smoothing;

[0006] The calculation and screening unit is used to receive the smoothed appearance data and pixel coordinates in the conversion processing unit. The machine learning model calculates the horizontal gradient and the vertical gradient according to the smoothed appearance data, and then calculates the gradient amplitude of the pixel change. The machine learning model uses the gradient amplitude of the pixel change to screen whether the pixel of the overall appearance of the plastic mold is a strong or weak edge point. When the pixel of the overall appearance of the plastic mold is screened to be a weak edge point, the pixel coordinates are regarded as weak edge point coordinates;

[0007] The model detection unit is used to receive a command from the calculation and screening unit to screen out pixel points of the overall appearance of the plastic mold as weak edge points. The model detection unit obtains the horizontal gradient, the vertical gradient and the coordinates of the weak edge points from the calculation and screening unit to calculate the gradient direction of the change of the weak edge points and the second-order derivative of the coordinates of the weak edge points. The machine learning model detects whether the weak edge points of the overall appearance of the plastic mold are burr points according to the second-order derivatives of the weak edge points. When the weak edge points of the overall appearance of the plastic mold are detected to be burr points, the coordinate points of the burr points are recorded. The machine learning model connects the burr points into irregular burr areas.

[0008] The quality detection unit is used to receive the irregular burr area in the model detection unit. The machine learning model calculates the area of ​​the irregular burr area according to the coordinate points of the edge burr points, and uses the support vector machine algorithm in machine learning to detect whether the overall quality of the plastic mold is qualified based on the calculated irregular burr area.

[0009] As a further improvement of the technical solution, the conversion processing unit includes establishing a conversion module and a model processing module;

[0010] The conversion module is established to obtain the overall appearance data of the production plastic mold through artificial intelligence, and a machine learning model is established according to the obtained overall data of the plastic mold. The machine learning model stores the obtained overall appearance data of the plastic mold as a three-dimensional matrix, and ensures that the obtained overall data of the plastic mold is in RGB format, extracts pixel coordinates from the obtained overall appearance data of the plastic mold, and extracts corresponding red channel values, green channel values, and blue channel values ​​from the three-dimensional matrix according to the pixel coordinates. The machine learning model converts the obtained overall appearance data of the plastic mold into overall grayscale appearance data of the plastic mold according to the red channel values, green channel values, and blue channel values;

[0011] The model processing module is used to receive the grayscale appearance and pixel coordinates in the conversion module, the machine learning model performs Gaussian kernel coverage on the acquired overall appearance data of the plastic mold, and extracts the offsets of the rows and columns around the pixel coordinates, calculates the Gaussian kernel based on the offsets, and the machine learning model extracts the adjacent pixel points of the pixel coordinates from the grayscale appearance, and smoothes the grayscale appearance based on the adjacent pixel points and the Gaussian kernel.

[0012] As a further improvement of the technical solution, the calculation screening unit includes a calculation gradient module and a model screening module;

[0013] The calculation gradient module is used to receive the smoothed appearance data in the model processing module and establish the pixel coordinates in the conversion module. The machine learning model uses a multivariate function to calculate the pixel coordinates based on the smoothed appearance data. Small changes in the axis direction and Small changes in the axis direction ,according to Small changes in the axis direction and Small changes in the axis direction Calculate horizontal and vertical gradients;

[0014] The model screening module is used to receive the horizontal gradient and vertical gradient in the gradient calculation module. The machine learning model calculates the gradient amplitude of the pixel point change based on the horizontal gradient and the vertical gradient. The dual threshold detection algorithm is used to screen the gradient amplitude of the pixel point change and the set standard high gradient amplitude and the set standard high gradient amplitude to screen out whether the pixel points of the overall appearance of the plastic mold are strong or weak edge points, and record the number of edge points.

[0015] As a further improvement of the technical solution, the specific screening conditions in the model screening module include:

[0016] Case ①: When the gradient amplitude of the pixel point change is greater than or equal to the set standard high gradient amplitude, the pixel point of the overall appearance of the plastic mold is screened out as a strong edge point;

[0017] Case ②: When the set standard high gradient amplitude is less than or equal to the gradient amplitude of the pixel point change, and the gradient amplitude of the pixel point change is less than the set standard high gradient amplitude, the pixel point of the overall appearance of the plastic mold is screened out as a weak edge point, and the pixel point coordinates are regarded as weak edge point coordinates;

[0018] Case ③: When the gradient amplitude of the pixel point change is less than the set standard high gradient amplitude, the pixel points of the overall appearance of the plastic mold are screened out as not edge points.

[0019] As a further improvement of the present technical solution, the model detection unit is used to receive a command from the model screening module to screen out pixel points of the overall appearance of the plastic mold as weak edge points. The model detection unit obtains the horizontal gradient, vertical gradient and the coordinates and number of edge points of the weak edge points in the model screening module from the gradient calculation module. The machine learning model then calculates the gradient direction of the change of the weak edge point based on the horizontal gradient and the vertical gradient, and calculates the second-order derivative of the coordinates of the weak edge point based on the coordinates of the weak edge point. The machine learning model uses a smoothness algorithm to detect whether the weak edge points of the overall appearance of the plastic mold are burr points based on the second-order derivative of the weak edge point and the number of edge points. When it is detected that the weak edge points of the overall appearance of the plastic mold are burr points, the machine learning model connects the burr points and the gradient direction of the change of the burr points into irregular burr areas, and records the number of irregular burr areas.

[0020] As a further improvement of the technical solution, the machine learning model in the model detection unit uses a smoothness algorithm to detect burr points to implement the following steps:

[0021] Step ①: Collect the second-order derivatives of weak edge points and the number of edge points , the machine learning model is based on the second-order derivative of the weak edge point and the number of edge points Calculate the smoothness of the weak edge points of the overall appearance of the plastic mold and obtain the calculated smoothness , the specific algorithm formula is:

[0022] ;

[0023] Step 2: Machine learning model uses calculated smoothness The weak edge points of the overall appearance of the plastic mold are detected to see if they are burrs with the set standard smoothness threshold. The specific detection conditions include:

[0024] Case 1: When the calculated smoothness When it is less than the set standard smoothness threshold, the calculated smoothness is small, indicating that the curvature of the overall appearance contour of the plastic mold is small, the overall appearance edge of the plastic mold is relatively smooth, and the weak edge point detected on the overall appearance edge of the plastic mold is not a burr point;

[0025] Case 2: When the calculated smoothness When it is greater than the set standard smoothness threshold, the calculated smoothness is detected to be large, indicating that the curvature of the overall appearance contour of the plastic mold is large, and there are irregularities on the edge of the overall appearance of the plastic mold. The weak edge points of the overall appearance edge of the plastic mold are detected to be burr points, and the coordinate points and number of burr points are recorded. At this time, the gradient direction of the change of the weak edge points is regarded as the gradient direction of the change of the burr points.

[0026] As a further improvement of the present technical solution, the quality inspection unit is used to receive irregular burr areas, the number of irregular burr areas, the number of edge burr points, and the coordinate points of edge burr points in the model inspection unit. The machine learning model uses a polygon area algorithm to calculate the area of ​​the irregular burr area based on the coordinate points of the edge burr points and the number of edge burr points. The support vector machine algorithm in machine learning is used to detect whether the overall quality of the plastic mold is qualified based on the calculated irregular burr area and the number of irregular burr areas.

[0027] As a further improvement of the technical solution, the machine learning model in the quality detection unit uses a polygon area algorithm to calculate the area of ​​the irregular burr area to achieve the following principle:

[0028] Machine learning model collects edge burr point coordinates , Coordinate points of adjacent edge burrs and the number of edge burr points Calculate the area of ​​the irregular burr area and obtain the calculated area of ​​the irregular burr area , the specific algorithm formula is:

[0029] ;

[0030] Here, 1 and 2 are constants.

[0031] As a further improvement of the technical solution, the quality detection unit uses a support vector machine algorithm to detect whether the overall quality of the plastic mold is qualified to implement the following steps:

[0032] Step 1: Collect and calculate the area of ​​irregular burr areas and the number of irregular burr areas And input into the machine learning model, the machine learning model learns the input data and outputs the corresponding weight coefficient , ;

[0033] Step 2: According to the calculated irregular burr area , the number of irregular burr areas and weight coefficient , Calculate the overall quality index of the plastic mold and obtain the calculated quality index , the specific algorithm formula is:

[0034] ;

[0035] in, refers to the partial derivative;

[0036] Step 3: Use calculated quality indicators Check whether the overall quality of the plastic mold is qualified according to the set standard quality indicators. The specific inspection conditions include:

[0037] Case 1: When the quality index is calculated When it is greater than the set standard quality index, there are a lot of burrs on the edge of the overall appearance of the plastic mold, and the overall quality of the plastic mold is detected to be unqualified, and the plastic mold is destroyed and reprocessed using artificial intelligence equipment;

[0038] Case 2: When the quality index is calculated When it is less than the set standard quality index, there is a small amount of burrs on the edge of the overall appearance of the plastic mold, and the overall quality of the plastic mold is detected to be qualified. The artificial intelligence equipment is used to process the small amount of burrs on the overall edge of the plastic mold according to the irregular burr area.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. In the artificial intelligence system for quality inspection, the model detection unit obtains the horizontal gradient, vertical gradient and weak edge point coordinates and the number of edge points in the model screening module from the calculation gradient module, calculates the gradient direction of the change of the weak edge point according to the horizontal gradient and the vertical gradient, and then calculates the second-order derivative of the weak edge point coordinate according to the weak edge point coordinate, detects whether the weak edge point of the overall appearance of the plastic mold is a burr point according to the second-order derivative of the weak edge point and the number of edge points, and when the weak edge point of the overall appearance of the plastic mold is detected to be a burr point, the machine learning model connects the burr point and the gradient direction of the change of the burr point into an irregular burr area, and can effectively capture the curvature change of the edge point through the second-order derivative, avoid misjudging normal edge points as burr points, and improve the reliability of detection, and at the same time, analyze the gradient direction of the change of the weak burr point when the burr point is determined, so as to more accurately connect the burr points and generate irregular burr areas, avoid subsequent omission of burr areas and quickly detect the quality of plastic molds, and improve the efficiency of quality inspection.

[0041] 2. In the artificial intelligence system for quality inspection, the quality inspection unit is used to receive irregular burr areas, the number of irregular burr areas, the number of edge burr points, and the coordinate points of edge burr points in the model inspection unit. The machine learning model uses a polygon area algorithm to calculate the area of ​​the irregular burr area according to the coordinate points of the edge burr points and the number of edge burr points. The support vector machine algorithm in the machine learning is used to detect whether the overall quality of the plastic mold is qualified based on the calculated area of ​​the irregular burr area and the number of irregular burr areas. By detecting the overall quality of the plastic mold, further processing of unqualified plastic molds is stopped in time, and subsequent grinding and polishing of unqualified plastic molds are reduced, thereby reducing energy consumption and reducing subsequent processing of unqualified plastic molds. Ineffective processing can be avoided, production line downtime caused by quality problems can be reduced, and production efficiency can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A unit block diagram of the present invention;

[0043] Figure 2 It is a block diagram of the module unit of the present invention.

[0044] The meaning of each number in the figure is:

[0045] 1. Conversion processing unit; 11. Establishing conversion module; 12. Model processing module;

[0046] 2. Calculation and screening unit; 21. Calculation gradient module; 22. Model screening module;

[0047] 3. Model detection unit; 4. Quality detection unit. DETAILED DESCRIPTION

[0048] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The present invention provides an artificial intelligence system for quality inspection, please refer to Figure 1 , including a conversion processing unit 1, a calculation and screening unit 2, a model detection unit 3 and a quality detection unit 4;

[0050] The conversion processing unit 1 acquires the overall appearance data of the production plastic mold through artificial intelligence to establish a machine learning model, and the machine learning model extracts pixel coordinates from the overall appearance data of the plastic mold for smoothing;

[0051] The calculation and screening unit 2 is used to receive the appearance data and pixel coordinates smoothed in the conversion processing unit 1. The machine learning model calculates the horizontal gradient and the vertical gradient according to the smoothed appearance data, and then calculates the gradient amplitude of the pixel change. The machine learning model uses the gradient amplitude of the pixel change to screen out whether the pixel of the overall appearance of the plastic mold is a strong or weak edge point. When the pixel of the overall appearance of the plastic mold is screened out as a weak edge point, the pixel coordinates are regarded as weak edge point coordinates;

[0052] The model detection unit 3 is used to receive a command from the calculation and screening unit 2 that the pixel points of the overall appearance of the plastic mold are weak edge points. The model detection unit 3 obtains the horizontal gradient, the vertical gradient and the coordinates of the weak edge points from the calculation and screening unit 2 to calculate the gradient direction of the change of the weak edge points and the second-order derivative of the coordinates of the weak edge points. The machine learning model detects whether the weak edge points of the overall appearance of the plastic mold are burr points according to the second-order derivatives of the weak edge points. When the weak edge points of the overall appearance of the plastic mold are detected to be burr points, the coordinate points of the burr points are recorded. The machine learning model connects the burr points into irregular burr areas.

[0053] The quality detection unit 4 is used to receive the irregular burr area in the model detection unit 3. The machine learning model calculates the area of ​​the irregular burr area according to the coordinate points of the edge burr points, and uses the support vector machine algorithm in machine learning to detect whether the overall quality of the plastic mold is qualified based on the calculated irregular burr area.

[0054] The following is a refinement of the above units, see Figure 2 ;

[0055] The conversion processing unit 1 includes a conversion establishment module 11 and a model processing module 12;

[0056] Establish a conversion module 11 to obtain the overall appearance data of the production plastic mold through artificial intelligence, establish a machine learning model based on the obtained overall data of the plastic mold, the machine learning model stores the obtained overall appearance data of the plastic mold as a three-dimensional matrix, and ensures that the obtained overall data of the plastic mold is in RGB format (the channel order is red, green, and blue), and the machine learning model extracts pixel coordinates from the obtained overall appearance data of the plastic mold , according to the pixel coordinates Extract the corresponding red channel value from the three-dimensional matrix , Green channel value , blue channel value , the machine learning model is based on the red channel value , Green channel value , blue channel value The acquired overall appearance data of the plastic mold is converted into the overall grayscale appearance data of the plastic mold to obtain the grayscale appearance ;

[0057] Grayscale appearance algorithm formula:

[0058] ;

[0059] Among them, 0.299 refers to the red channel weight, which plays a relatively small role because the human eye is not as sensitive to red as to green. 0.587 refers to the green channel weight, which has the largest weight because the human eye is most sensitive to green light. The value of the green channel plays a major role in determining the brightness of the grayscale appearance. 0.114 refers to the blue channel weight, which has the smallest weight because the human eye is relatively less sensitive to blue light. This formula is used to convert the acquired overall data of the plastic mold into the overall grayscale appearance data of the plastic mold.

[0060] The model processing module 12 is used to receive the grayscale appearance of the conversion module 11. and pixel coordinates The machine learning model performs Gaussian kernel coverage on the acquired plastic mold overall appearance data (Gaussian kernel coverage is to use the Gaussian kernel function to perform convolution operations on each pixel and its neighborhood in the appearance) and extracts the pixel coordinates Offsets of surrounding rows and columns , the machine learning model is based on the offset Calculate Gaussian kernel , the machine learning model is derived from the grayscale appearance Extract pixel coordinates from The neighboring pixels , the machine learning model is based on the neighboring pixels and Gaussian kernel Grayscale appearance Perform smoothing to obtain smoothed appearance data ;

[0061] The implementation principle of calculating the Gaussian kernel:

[0062] Machine Learning Model Collection Offsets , the machine learning model is based on the offset Calculate Gaussian kernel , the specific algorithm formula is:

[0063] ;

[0064] in, refers to the normalization coefficient, refers to the integral result of the two-dimensional Gaussian function, refers to the index part, is the standard deviation of the Gaussian kernel. This formula is used to calculate the Gaussian kernel for subsequent smoothing.

[0065] How the machine learning model smooths the grayscale appearance:

[0066] The machine learning model is based on the adjacent pixels and Gaussian kernel Grayscale appearance Perform smoothing to obtain smoothed appearance data , the specific algorithm formula is:

[0067] ;

[0068] in, Refers to the size of the Gaussian kernel, ranging from ,This formula is used to smooth the grayscale appearance, and through the weighted average of adjacent edge pixels, it reduces the random noise in the mold appearance and makes the mold appearance clearer.

[0069] The calculation and screening unit 2 includes a calculation gradient module 21 and a model screening module 22;

[0070] The calculation gradient module 21 is used to receive the appearance data smoothed by the model processing module 12. And establish the pixel coordinates in the conversion module 11 , the machine learning model uses a multivariate function to smooth the appearance data Calculate pixel coordinates of Small changes in the axis direction and Small changes in the axis direction ,according to Small changes in the axis direction and smoothed appearance data Calculate horizontal gradient ,in, Refers to the partial derivative, and then according to Small changes in the axis direction and smoothed appearance data Calculate vertical gradient , refers to the partial derivative;

[0071] The model screening module 22 is used to receive the horizontal gradient in the calculation gradient module 21 and vertical gradient , the machine learning model is based on the horizontal gradient and vertical gradient Calculate the gradient amplitude of pixel changes The machine learning model uses a dual threshold detection algorithm based on the gradient amplitude of pixel changes. With the standard high gradient amplitude set and the standard high gradient amplitude set Screen to see whether the pixels of the overall appearance of the plastic mold are strong or weak edge points, and record the number of edge points ,The gradient amplitude can effectively reflect the edge strength of the pixel point. By setting a standard high gradient amplitude, it can accurately distinguish strong edge points from weak edge points, avoid misjudging noise or non-edge points as edge points, and improve the accuracy of detection;

[0072] The specific screening conditions in the model screening module 22 include:

[0073] Case ①: When the gradient amplitude of the pixel changes Greater than or equal to the set standard high gradient amplitude When , the pixel points of the overall appearance of the plastic mold are screened out as strong edge points;

[0074] Case ②: When the standard high gradient amplitude is set Less than or equal to the gradient amplitude of the pixel change , and the gradient amplitude of the pixel change Smaller than the set standard high gradient amplitude When , the pixel points of the overall appearance of the plastic mold are screened out as weak edge points, and the pixel coordinates are Coordinates of weak edge points ;

[0075] Case ③: When the gradient amplitude of the pixel changes Smaller than the set standard high gradient amplitude When , the pixel points of the overall appearance of the plastic mold are screened out as not edge points;

[0076] The model detection unit 3 is used to receive the command from the model screening module 22 that the pixel points of the overall appearance of the plastic mold are weak edge points, and the model detection unit 3 obtains the horizontal gradient from the calculation gradient module 21 , vertical gradient and the coordinates of weak edge points in the model screening module 22 , the number of edge points , the machine learning model then uses the horizontal gradient and vertical gradient Calculate the gradient direction of the weak edge point change , where the gradient direction is usually quantized into four directions: 0°, 45°, 90°, and 135°. Calculate the second-order derivative of the weak edge point coordinates to obtain the second-order derivative of the weak edge point ,in, is the derivative, Refers to the The machine learning model uses the smoothness algorithm to calculate the coordinates of weak edge points according to the second-order derivatives of the weak edge points. and the number of edge points Detect whether the weak edge points of the overall appearance of the plastic mold are burr points. When the weak edge points of the overall appearance of the plastic mold are burr points, the machine learning model is based on the burr points and the gradient direction of the burr point changes. Connect into irregular burr areas and record the number of irregular burr areas The second-order derivative can effectively capture the curvature change of edge points, avoid misjudging normal edge points as burr points, and improve the reliability of detection. At the same time, when the burr points are determined, the gradient direction of the change of weak burr points can be analyzed, which can more accurately connect the burr points and generate irregular burr areas, avoiding the subsequent omission of burr areas and quickly detecting the quality of plastic molds, thereby improving the efficiency of quality inspection;

[0077] The machine learning model in model detection unit 3 uses the smoothness algorithm to detect burrs to implement the following steps:

[0078] Step ①: Collect the second-order derivatives of weak edge points and the number of edge points , the machine learning model is based on the second-order derivative of the weak edge point and the number of edge points Calculate the smoothness of the weak edge points of the overall appearance of the plastic mold and obtain the calculated smoothness , the specific algorithm formula is:

[0079] ;

[0080] Among them, this formula is used to calculate the smoothness of the weak edge points of the overall appearance of the plastic mold. The second-order derivative can reflect the curvature change of the weak edge points. Combined with the number of edge points, the smoothness of the plastic mold surface can be more accurately evaluated;

[0081] Step 2: Machine learning model uses calculated smoothness The weak edge points of the overall appearance of the plastic mold are detected to see if they are burrs with the set standard smoothness threshold. The specific detection conditions include:

[0082] Case 1: When the calculated smoothness When it is less than the set standard smoothness threshold, the calculated smoothness is small, indicating that the curvature of the overall appearance contour of the plastic mold is small, the overall appearance edge of the plastic mold is relatively smooth, and the weak edge point detected on the overall appearance edge of the plastic mold is not a burr point;

[0083] Case 2: When the calculated smoothness When it is greater than the set standard smoothness threshold, the calculated smoothness is detected to be large, indicating that the curvature of the overall appearance contour of the plastic mold is large, and there are irregularities on the edge of the overall appearance of the plastic mold. The weak edge points of the overall appearance edge of the plastic mold are detected to be burrs, and the coordinates of the burrs are recorded. and the number of edge burr points ,in, Refers to the The coordinates of the edge burr points, at this time, the gradient direction of the weak edge point change is regarded as the gradient direction of the burr point change;

[0084] The quality detection unit 4 is used to receive the irregular burr area and the number of irregular burr areas from the model detection unit 3. 、Number of edge burr points 、Edge burr point coordinate point , get the edge burr point coordinates from the irregular burr area The coordinates of the adjacent points of the edge burr point ,in, Refers to the The machine learning model uses the polygon area algorithm to calculate the coordinate points of the edge burrs. , Coordinate points of adjacent edge burrs and the number of edge burr points Calculate the area of ​​the irregular burr area and use the support vector machine algorithm in machine learning to calculate the area of ​​the irregular burr area and the number of irregular burr areas Check whether the overall quality of the plastic mold is qualified. By checking the overall quality of the plastic mold, stop further processing of unqualified plastic molds in time, reduce subsequent grinding and polishing of unqualified plastic molds, reduce energy consumption, and reduce subsequent processing of unqualified plastic molds, which can avoid invalid processing, reduce production line downtime caused by quality problems, and improve production efficiency;

[0085] The machine learning model in quality inspection unit 4 uses the polygon area algorithm to calculate the area of ​​irregular burr areas:

[0086] Machine learning model collects edge burr point coordinates , Coordinate points of adjacent edge burrs and the number of edge burr points Calculate the area of ​​the irregular burr area and obtain the calculated area of ​​the irregular burr area , the specific algorithm formula is:

[0087] ;

[0088] Among them, 1 and 2 refer to constants. This formula is used to calculate the area of ​​irregular burrs. By calculating the area of ​​burrs, the burrs on the edge of the overall appearance of the plastic mold can be discovered in time to avoid affecting the quality of the final product and reduce the generation of burrs.

[0089] The steps to use the support vector machine algorithm to detect whether the overall quality of the plastic mold is qualified are as follows:

[0090] Step 1: Collect and calculate the area of ​​irregular burr areas and the number of irregular burr areas And input into the machine learning model, the machine learning model learns the input data and outputs the corresponding weight coefficient , ;

[0091] Step 2: According to the calculated irregular burr area , the number of irregular burr areas and weight coefficient , Calculate the overall quality index of the plastic mold and obtain the calculated quality index , the specific algorithm formula is:

[0092] ;

[0093] in, Refers to partial derivative. This formula is used to calculate the overall quality index of plastic molds. By combining multiple data, a comprehensive quality index is formed, which can fully reflect the overall quality status of the mold.

[0094] Step 3: Use calculated quality indicators Check whether the overall quality of the plastic mold is qualified according to the set standard quality indicators. The specific inspection conditions include:

[0095] Case 1: When the quality index is calculated When it is greater than the set standard quality index, there are a lot of burrs on the edge of the overall appearance of the plastic mold, and the overall quality of the plastic mold is detected to be unqualified, and the plastic mold is destroyed and reprocessed using artificial intelligence equipment;

[0096] Case 2: When the quality index is calculated When it is less than the set standard quality index, there is a small amount of burrs on the edge of the overall appearance of the plastic mold, and the overall quality of the plastic mold is detected to be qualified. The artificial intelligence equipment is used to process the small amount of burrs on the overall edge of the plastic mold according to the irregular burr area.

[0097] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An artificial intelligence system for quality inspection, characterized by: It includes a conversion processing unit (1), a calculation and screening unit (2), a model detection unit (3) and a quality detection unit (4); The conversion processing unit (1) acquires the overall appearance data of the production plastic mold through artificial intelligence to establish a machine learning model, and the machine learning model extracts pixel point coordinates from the overall appearance data of the plastic mold for smoothing; The calculation and screening unit (2) is used to receive the appearance data and pixel coordinates smoothed in the conversion processing unit (1), the machine learning model calculates the horizontal gradient and the vertical gradient based on the smoothed appearance data, and then calculates the gradient amplitude of the pixel change. The machine learning model uses the gradient amplitude of the pixel change to screen whether the pixel of the overall appearance of the plastic mold is a strong or weak edge point. When the pixel of the overall appearance of the plastic mold is screened to be a weak edge point, the pixel coordinates are regarded as the weak edge point coordinates; The model detection unit (3) is used to receive a command from the calculation and screening unit (2) that the pixel points of the overall appearance of the plastic mold are weak edge points. The model detection unit (3) obtains the horizontal gradient, the vertical gradient and the coordinates of the weak edge points from the calculation and screening unit (2) to calculate the gradient direction of the change of the weak edge points and the second-order derivative of the coordinates of the weak edge points. The machine learning model detects whether the weak edge points of the overall appearance of the plastic mold are burr points based on the second-order derivatives of the weak edge points. When the weak edge points of the overall appearance of the plastic mold are detected to be burr points, the coordinates of the burr points are recorded. The machine learning model connects the burr points into irregular burr areas. The quality detection unit (4) is used to receive the irregular burr area in the model detection unit (3), the machine learning model calculates the area of ​​the irregular burr area according to the coordinate points of the edge burr points, and uses the support vector machine algorithm in the machine learning to detect whether the overall quality of the plastic mold is qualified based on the calculated area of ​​the irregular burr area.

2. The artificial intelligence system for quality inspection according to claim 1, characterized in that: The conversion processing unit (1) comprises a conversion establishment module (11) and a model processing module (12); The conversion module (11) is used to obtain the overall appearance data of the plastic mold produced by artificial intelligence, and a machine learning model is established based on the obtained overall data of the plastic mold. The machine learning model stores the obtained overall appearance data of the plastic mold as a three-dimensional matrix, and ensures that the obtained overall data of the plastic mold is in RGB format, extracts pixel coordinates from the obtained overall appearance data of the plastic mold, and extracts corresponding red channel values, green channel values, and blue channel values ​​from the three-dimensional matrix based on the pixel coordinates. The machine learning model converts the obtained overall appearance data of the plastic mold into overall grayscale appearance data of the plastic mold based on the red channel values, green channel values, and blue channel values; The model processing module (12) is used to receive the grayscale appearance and pixel coordinates in the conversion module (11), the machine learning model performs Gaussian kernel coverage on the acquired overall appearance data of the plastic mold, and extracts the offset of the rows and columns around the pixel coordinates, calculates the Gaussian kernel based on the offset, and the machine learning model extracts the adjacent pixel points of the pixel coordinates from the grayscale appearance, and smoothes the grayscale appearance based on the adjacent pixel points and the Gaussian kernel.

3. The artificial intelligence system for quality inspection according to claim 2, characterized in that: The calculation screening unit (2) comprises a calculation gradient module (21) and a model screening module (22); The gradient calculation module (21) is used to receive the appearance data smoothed in the model processing module (12) and to establish the pixel coordinates in the conversion module (11). The machine learning model uses a multivariate function to calculate the pixel coordinates based on the smoothed appearance data. Small changes in the axis direction and Small changes in the axis direction ,according to Small changes in the axis direction and Small changes in the axis direction Calculate horizontal and vertical gradients; The model screening module (22) is used to receive the horizontal gradient and the vertical gradient in the gradient calculation module (21), and the machine learning model calculates the gradient amplitude of the pixel point change based on the horizontal gradient and the vertical gradient, and uses a dual threshold detection algorithm to screen the pixel point change gradient amplitude and the set standard high gradient amplitude and the set standard high gradient amplitude to screen out whether the pixel point of the overall appearance of the plastic mold is a strong or weak edge point, and record the number of edge points.

4. The artificial intelligence system for quality inspection according to claim 3, characterized in that: The specific screening conditions in the model screening module (22) include: Case ①: When the gradient amplitude of the pixel point change is greater than or equal to the set standard high gradient amplitude, the pixel point of the overall appearance of the plastic mold is screened out as a strong edge point; Case ②: When the set standard high gradient amplitude is less than or equal to the gradient amplitude of the pixel point change, and the gradient amplitude of the pixel point change is less than the set standard high gradient amplitude, the pixel point of the overall appearance of the plastic mold is screened out as a weak edge point, and the pixel point coordinates are regarded as weak edge point coordinates; Case ③: When the gradient amplitude of the pixel point change is less than the set standard high gradient amplitude, the pixel points of the overall appearance of the plastic mold are screened out as not edge points.

5. The artificial intelligence system for quality inspection according to claim 3, characterized in that: The model detection unit (3) is used to receive a command from the model screening module (22) to screen out pixel points of the overall appearance of the plastic mold as weak edge points. The model detection unit (3) obtains the horizontal gradient, the vertical gradient and the coordinates of the weak edge points and the number of edge points in the model screening module (22) from the calculation gradient module (21). The machine learning model then calculates the gradient direction of the change of the weak edge points based on the horizontal gradient and the vertical gradient, and calculates the second-order derivative of the coordinates of the weak edge points based on the coordinates of the weak edge points. The machine learning model uses a smoothness algorithm to detect whether the weak edge points of the overall appearance of the plastic mold are burr points based on the second-order derivative of the weak edge points and the number of edge points. When the weak edge points of the overall appearance of the plastic mold are detected to be burr points, the machine learning model connects the burr points and the gradient direction of the change of the burr points into irregular burr areas, and records the number of irregular burr areas.

6. The artificial intelligence system for quality inspection according to claim 5, characterized in that: The machine learning model in the model detection unit (3) uses a smoothness algorithm to detect burr points to implement the following steps: Step ①: Collect the second-order derivatives of weak edge points and the number of edge points , the machine learning model is based on the second-order derivative of the weak edge point and the number of edge points Calculate the smoothness of the weak edge points of the overall appearance of the plastic mold and obtain the calculated smoothness , the specific algorithm formula is: ； Step 2: Machine learning model uses calculated smoothness The weak edge points of the overall appearance of the plastic mold are detected to see if they are burrs with the set standard smoothness threshold. The specific detection conditions include: Case 1: When the calculated smoothness When it is less than the set standard smoothness threshold, the calculated smoothness is small, indicating that the curvature of the overall appearance contour of the plastic mold is small, the overall appearance edge of the plastic mold is relatively smooth, and the weak edge point detected on the overall appearance edge of the plastic mold is not a burr point; Case 2: When the calculated smoothness When it is greater than the set standard smoothness threshold, the calculated smoothness is detected to be large, indicating that the curvature of the overall appearance contour of the plastic mold is large, and there are irregularities on the edge of the overall appearance of the plastic mold. The weak edge points of the overall appearance edge of the plastic mold are detected to be burr points, and the coordinate points and number of burr points are recorded. At this time, the gradient direction of the change of the weak edge points is regarded as the gradient direction of the change of the burr points.

7. The artificial intelligence system for quality inspection according to claim 5, characterized in that: The quality detection unit (4) is used to receive the irregular burr area, the number of irregular burr areas, the number of edge burr points, and the coordinate points of the edge burr points from the model detection unit (3); the machine learning model uses a polygon area algorithm to calculate the area of ​​the irregular burr area according to the coordinate points of the edge burr points and the number of edge burr points; and uses a support vector machine algorithm in machine learning to detect whether the overall quality of the plastic mold is qualified according to the calculated area of ​​the irregular burr area and the number of irregular burr areas.

8. The artificial intelligence system for quality inspection according to claim 7, characterized in that: The machine learning model in the quality detection unit (4) uses a polygon area algorithm to calculate the area of ​​the irregular burr area to achieve the following principle: Machine learning model collects edge burr point coordinates , Coordinate points of adjacent edge burrs and the number of edge burr points Calculate the area of ​​the irregular burr area and obtain the calculated area of ​​the irregular burr area , the specific algorithm formula is: ； Here, 1 and 2 are constants.

9. The artificial intelligence system for quality inspection according to claim 7, characterized in that: The quality detection unit (4) uses the support vector machine algorithm to detect whether the overall quality of the plastic mold is qualified. The implementation steps are: Step 1: Collect and calculate the area of ​​irregular burr areas and the number of irregular burr areas And input into the machine learning model, the machine learning model learns the input data and outputs the corresponding weight coefficient , ; Step 2: According to the calculated irregular burr area , the number of irregular burr areas and weight coefficient , Calculate the overall quality index of the plastic mold and obtain the calculated quality index , the specific algorithm formula is: ； in, refers to the partial derivative; Step 3: Use calculated quality indicators Check whether the overall quality of the plastic mold is qualified according to the set standard quality indicators. The specific inspection conditions include: Case 1: When the quality index is calculated When it is greater than the set standard quality index, there are a lot of burrs on the edge of the overall appearance of the plastic mold, and the overall quality of the plastic mold is detected to be unqualified, and the plastic mold is destroyed and reprocessed using artificial intelligence equipment; Case 2: When the quality index is calculated When it is less than the set standard quality index, there is a small amount of burrs on the edge of the overall appearance of the plastic mold, and the overall quality of the plastic mold is detected to be qualified. The artificial intelligence equipment is used to process the small amount of burrs on the overall edge of the plastic mold according to the irregular burr area.

Images