Intelligent detection and analysis system and method for semiconductor chip manufacturing based on machine vision
By using a machine vision-based intelligent inspection and analysis system, which combines multiple inspection modules and deep learning algorithms, the problem of low integration in semiconductor chip manufacturing inspection and analysis systems has been solved, achieving efficient and accurate quality control and improved production efficiency.
Patent Information
- Application Number
- CN202411234514.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing semiconductor chip manufacturing testing and analysis systems mostly perform testing on individual manufacturing steps, resulting in low integration and an inability to effectively guarantee stable chip quality.
An intelligent inspection and analysis system based on machine vision is adopted, including a surface defect detection module, a component positioning and assembly inspection module, other inspection modules, a production testing module, and a data processing module. It combines image acquisition, feature extraction, deep learning algorithms, and multiple inspection methods to achieve multi-step inspection and evaluation.
It improves the quality stability and production efficiency of semiconductor chips, reduces costs, achieves efficient and accurate quality control, ensures the precision of component positioning and assembly, and provides comprehensive testing and analysis.
Smart Images

Figure CN119000694A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of detection and analysis, and in particular to a semiconductor chip manufacturing intelligent detection and analysis system and method based on machine vision. Background Art
[0002] Semiconductor chip, also known as semiconductor integrated circuit, refers to a micro electronic component that integrates multiple electronic components, such as transistors, resistors, capacitors, etc., on a silicon base or other semiconductor materials. It is a key component for miniaturization and micro-miniaturization of circuits and can complete specific circuit or system functions;
[0003] During the manufacturing process of semiconductor chips, detection and analysis are required to ensure the stable production of semiconductor chips. During the semiconductor chip manufacturing detection and analysis process, a semiconductor chip manufacturing detection and analysis system and method will be used.
[0004] Existing semiconductor chip manufacturing detection and analysis systems are mostly for the detection of a single manufacturing step, with a low degree of integration, and cannot better guarantee the stable quality of the manufactured semiconductor chips, which has a certain impact on the use of the semiconductor chip manufacturing detection and analysis system. Therefore, a semiconductor chip manufacturing intelligent detection and analysis system and method based on machine vision are proposed. Summary of the invention
[0005] The technical problem to be solved by the present invention is: how to solve the problem that the existing semiconductor chip manufacturing detection and analysis systems are mostly for the detection of a single manufacturing step, have a low degree of integration, cannot better guarantee the stable quality of the manufactured semiconductor chips, and have a certain impact on the use of the semiconductor chip manufacturing detection and analysis system. Provided is a semiconductor chip manufacturing intelligent detection and analysis system and method based on machine vision.
[0006] The present invention solves the above technical problems through the following technical solutions, which include a surface defect detection module, a component positioning and assembly detection module, other detection modules, a production test module, a data processing module and an information sending module;
[0007] The surface defect detection module is used to detect surface defects of chip products and obtain surface defect detection information;
[0008] The component positioning and assembly detection module is used to detect the component positioning of chip products and the assembly of chip products, and obtain component positioning and assembly detection information;
[0009] The other test modules are used to test the chip product by other testing methods to obtain other testing information;
[0010] The production test module is used to test the chip product after it is produced and obtain production test information;
[0011] The data processing module is used to perform visual processing on surface defect detection information, component positioning and assembly detection information and other detection information to obtain detection display information;
[0012] The data processing module is also used to process the production test information to obtain production evaluation information;
[0013] The information sending module is used to send the detection display information and the production evaluation information to a preset receiving terminal.
[0014] Furthermore, the specific process of the surface defect detection module detecting the surface defect of the chip product and obtaining the surface defect detection information is as follows:
[0015] The image information of the chip product is collected by the set image collection device according to the preset collection rules, that is, the real-time product image information, and the real-time product image information includes the first image, the second image and the third image;
[0016] Then, the feature extraction algorithm is used to extract the real-time product image information to obtain three real-time chip product features;
[0017] Process the real-time product features through a deep learning algorithm to obtain three real-time chip product features, process the three real-time information products to obtain the first identification information, the second identification information and the third identification information;
[0018] The process of obtaining the first identification information, the second identification information and the third identification information is as follows: extract three real-time chip product features, extract any one of them and perform feature classification through a classifier, the classifier includes:
[0019] Support Vector Machine (SVM): It works by finding a hyperplane that maximizes the margin between different classes.
[0020] Decision Trees and Random Forests: Classification is performed by building a decision tree or ensembling multiple decision trees.
[0021] Naive Bayes classifier: Classification based on Bayes theorem and feature conditional independence assumption;
[0022] After the feature classification is completed, the defect identification is performed on the classified feature classification through deep learning algorithm;
[0023] The deep learning algorithms used in the defect recognition process include: Convolutional neural networks: particularly suitable for image processing tasks and can automatically learn hierarchical features in images.
[0024] Recurrent Neural Networks and Long Short-Term Memory Networks: Suitable for processing sequence data, such as sequences of frames in a video.
[0025] Generative adversarial networks: used to generate new image data, and can also be used for image restoration and defect detection;
[0026] Finally, the first identification information, the second identification information and the third identification information are processed to obtain surface defect detection information;
[0027] The specific process of processing the first identification information, the second identification information and the third identification information to obtain the surface defect detection information is as follows: the identification information includes whether there is a defect or not;
[0028] When the first identification information, the second identification information and the third identification information are all defect-free, the surface defect detection information indicates that the chip product is defect-free;
[0029] When the first identification information, the second identification information and the third identification information are all defect-free, there is a defect, and the surface defect detection information indicates that the chip product has a defect;
[0030] When two of the first identification information, the second identification information and the third identification information are free of defects, but the other one is defective, the first identification information, the second identification information and the third identification information are collected again.
[0031] Furthermore, the specific process of extracting features from real-time product image information using a feature extraction algorithm to obtain three real-time chip product features is as follows:
[0032] Extracting a first image, a second image, and a third image from real-time product image information;
[0033] The first image is processed simultaneously using an edge-based feature extraction method, a region-based feature extraction method, and a transformation-based feature extraction method to obtain three first preliminary feature information, and then the three first preliminary feature information are compared. When the features in the three first preliminary feature information are the same, they are exported as a real-time information product feature information. When the features in the three first preliminary feature information are different, feature collection is performed again.
[0034] The second image and the third image are processed in the same manner as the first image to obtain two more real-time information product feature information.
[0035] Furthermore, the image acquisition device needs to acquire status information of the image acquisition device during the process of acquiring image information, and processes the status information of the image acquisition device to obtain image device evaluation information, wherein the image device evaluation information includes whether the image device is normal or abnormal;
[0036] The process of processing the status information of the image acquisition device to obtain the image acquisition device evaluation information is as follows: extracting the status information of the image acquisition device, the status information of the image acquisition device includes the real-time position information of the image acquisition device, the real-time temperature information of the image acquisition device and the clarity information of the images acquired by the image acquisition device three times in the past;
[0037] The real-time position information of the image acquisition device is compared with the set standard position information to obtain the position deviation. When the position deviation is greater than the preset value, an image device abnormality is generated; when the position deviation is less than the preset value, an image device abnormality is generated;
[0038] When the real-time temperature information of the image acquisition device is greater than the preset value a1 for more than the preset time, an image device abnormality is generated;
[0039] When the real-time temperature information of the image acquisition device is greater than the preset value a2, an image device abnormality is generated;
[0040] When the real-time temperature information of the image acquisition device is less than a1, the image generation device is normal;
[0041] When the fluctuation range of the image clarity information collected three times in the past is greater than the preset value, an image device abnormality is generated;
[0042] When the fluctuation range of the image clarity information collected three times in the past is less than the preset value, and the image clarity information collected three times is greater than the preset clarity, the image generation device is normal.
[0043] Furthermore, the specific process of the component positioning and assembly detection information is as follows: first, positioning is performed, an image of the installed component is captured by an image acquisition device, and image recognition features of the installed component are extracted by an image processing algorithm, and then the real-time position of the recognition feature is extracted, and the position is compared with the preset standard feature position to obtain a position deviation value. When the position deviation value is greater than the preset value, the component positioning and assembly detection information is generated, and at this time, the component positioning and assembly detection information is abnormal component positioning;
[0044] Then, the image information of the semiconductor chip during assembly is collected by an image acquisition device, and assembly features are extracted from the image information of the semiconductor chip during assembly to obtain assembly features. The assembly features are distributed on assembly components of each assembled semiconductor chip. Then, the distance information of each assembly feature is obtained from the image information of the semiconductor chip during assembly to obtain a real-time evaluation distance.
[0045] The difference between the real-time evaluation distance and the distance of the preset standard assembly feature corresponding to the feature is calculated to obtain the feature evaluation difference. When any one of the feature evaluation differences is greater than the preset value, the component positioning and assembly detection information is generated. At this time, the component positioning and assembly detection information is an assembly abnormality.
[0046] Furthermore, the other test modules detect the chip product by other detection methods, and the specific process of obtaining other detection information is as follows:
[0047] Other detection methods include AOI detection and AVI detection. Through AOI detection and AVI detection, other detection information can be obtained by detecting chip products. Other detection information includes AOI detection information and AVI detection information.
[0048] Among them, AOI detection information includes AOI detection qualified and AOI detection unqualified;
[0049] AVI test information includes AVI test pass and AVI test fail.
[0050] Furthermore, the AOI detection information and the AVI detection information are processed to generate production abnormality warning information;
[0051] The specific process of producing abnormal warning information is as follows: collect the product quantity information of semiconductor chips manufactured within a preset time, mark it as G, then extract the number of AOI detection failures in G AOI detection information, and obtain the first other detection evaluation parameter;
[0052] Extract the number of AVI test failures from the G AVI test information, and obtain the second other test evaluation parameter;
[0053] A standard first other detection parameter and a second standard other detection parameter are set;
[0054] When the first other detection evaluation parameter is greater than the standard first other detection parameter or the second other detection evaluation parameter is greater than the standard second other detection parameter, production abnormality warning information is generated.
[0055] Furthermore, the production test module tests the chip product after it is produced, and the specific process of obtaining the production test information is as follows:
[0056] Setting up image acquisition equipment to acquire image information of the test chip product in real time, processing the image information of the test chip product, and obtaining the real-time position of the test chip;
[0057] Then, the chip product is tested for working voltage to obtain working voltage parameters, working temperature is collected to obtain working temperature, and power consumption is collected to obtain power consumption information;
[0058] In the process of collecting working voltage parameters, working temperature and power consumption information, it is necessary to analyze the changes in the real-time position of the test chip in real time;
[0059] When the change of the real-time position of the test chip during the collection of the working voltage parameter, working temperature and power consumption information is greater than the preset value, it is necessary to collect the working voltage parameter, working temperature and power consumption information again;
[0060] Operating voltage parameters, operating temperature and power consumption information constitute production test information.
[0061] Furthermore, the production evaluation information includes normal chips and abnormal chip products. The specific process of the production evaluation information is as follows: extracting the collected production test information, and extracting the operating voltage parameters, operating temperature and power consumption information from the production test information;
[0062] When any of the operating voltage parameter exceeds the preset range, the operating temperature is greater than the preset temperature, and the power consumption information is greater than the preset power consumption, the production assessment information indicates that the chip product is abnormal;
[0063] When the operating voltage parameters are within the preset range, the operating temperature is lower than the preset temperature, and the power consumption information is within the standard range, the production evaluation information indicates that the chip is normal.
[0064] A semiconductor chip manufacturing intelligent detection and analysis method based on machine vision, the method comprising the following steps:
[0065] Step 1: Detect surface defects of the test chip product and obtain surface defect detection information;
[0066] Step 2: Component positioning and assembly detection information of the chip product are obtained by detecting the component positioning and assembly detection information of the chip product;
[0067] Step 3: Test the chip product through other testing methods to obtain other testing information;
[0068] Step 4: After the chip product is produced, test it and obtain production test information;
[0069] Step 5: Visualize the surface defect detection information, component positioning and assembly detection information and other detection information to obtain detection display information;
[0070] Step 6: Process the production test information to obtain production evaluation information;
[0071] Step 7: Send the detection display information and production evaluation information to the preset receiving terminal.
[0072] Compared with the prior art, the present invention has the following advantages: the intelligent detection and analysis system and method for semiconductor chip manufacturing based on machine vision is provided with a variety of different detection methods. In the semiconductor chip manufacturing process, intelligent detection is performed, which provides strong support for improving the quality of semiconductor products, reducing costs and improving production efficiency. By performing surface defect detection, efficient, accurate and reliable semiconductor product production quality control is achieved, and defect detection is faster and more accurate. At the same time, component positioning and assembly detection are performed. It really plays a vital role in the automatic positioning and assembly of parts. It accurately identifies and locates parts through image processing, thereby significantly improving production efficiency and product quality, and finally performs overall testing of finished chip products, thereby allowing the system to achieve more comprehensive intelligent detection and analysis of semiconductor chip manufacturing, so that the quality of the final output chip is better, and the system and method are more intelligent and comprehensive. BRIEF DESCRIPTION OF THE DRAWINGS
[0073] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION
[0074] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiment.
[0075] like Figure 1 As shown, this embodiment provides a technical solution: a semiconductor chip manufacturing intelligent detection and analysis system based on machine vision, including a surface defect detection module, a component positioning and assembly detection module, other detection modules, a production test module, a data processing module and an information sending module;
[0076] The surface defect detection module is used to detect surface defects of chip products and obtain surface defect detection information;
[0077] The component positioning and assembly detection module is used to detect the component positioning of chip products and the assembly of chip products, and obtain component positioning and assembly detection information;
[0078] The other test modules are used to test the chip product by other testing methods to obtain other testing information;
[0079] The production test module is used to test the chip product after it is produced and obtain production test information;
[0080] The data processing module is used to perform visual processing on surface defect detection information, component positioning and assembly detection information and other detection information to obtain detection display information;
[0081] The data processing module is also used to process the production test information to obtain production evaluation information;
[0082] The information sending module is used to send the detection display information and the production evaluation information to a preset receiving terminal.
[0083] The specific process of the surface defect detection module detecting the surface defect of the chip product and obtaining the surface defect detection information is as follows:
[0084] The image information of the chip product is collected by the set image collection device according to the preset collection rules, that is, the real-time product image information, and the real-time product image information includes the first image, the second image and the third image;
[0085] Then, the feature extraction algorithm is used to extract the real-time product image information to obtain three real-time chip product features;
[0086] Process the real-time product features through a deep learning algorithm to obtain three real-time chip product features, process the three real-time information products to obtain the first identification information, the second identification information and the third identification information;
[0087] The preset acquisition rules are set as follows: three image acquisition devices are set at three different positions, and image information of chip products is collected from different angles at the same time. Multiple images of the same position are collected at the same time, and the image with the highest definition is selected to be finally exported as the first image, the second image, or the third image;
[0088] The process of obtaining the first identification information, the second identification information and the third identification information is as follows: extract three real-time chip product features, extract any one of them and perform feature classification through a classifier, the classifier includes:
[0089] Support Vector Machine: It works by finding a hyperplane that maximizes the margin between different classes.
[0090] Decision Trees and Random Forests: Classification is performed by building a decision tree or ensembling multiple decision trees.
[0091] Naive Bayes classifier: Classification based on Bayes theorem and feature conditional independence assumption;
[0092] After the feature classification is completed, the defect identification is performed on the classified feature classification through deep learning algorithm;
[0093] The deep learning algorithms used in the defect recognition process include: Convolutional neural networks: particularly suitable for image processing tasks and can automatically learn hierarchical features in images.
[0094] Recurrent Neural Networks and Long Short-Term Memory Networks: Suitable for processing sequence data, such as sequences of frames in a video.
[0095] Generative adversarial networks: used to generate new image data, and can also be used for image restoration and defect detection;
[0096] Finally, the first identification information, the second identification information and the third identification information are processed to obtain surface defect detection information;
[0097] The specific process of processing the first identification information, the second identification information and the third identification information to obtain the surface defect detection information is as follows: the identification information includes whether there is a defect or not;
[0098] When the first identification information, the second identification information and the third identification information are all defect-free, the surface defect detection information indicates that the chip product is defect-free;
[0099] When the first identification information, the second identification information and the third identification information are all defect-free, there is a defect, and the surface defect detection information indicates that the chip product has a defect;
[0100] When two of the first identification information, the second identification information and the third identification information are free of defects, but the other one is defective, the first identification information, the second identification information and the third identification information are collected again;
[0101] Through the above process, more accurate identification information can be collected, thereby ensuring the accuracy of the final surface detection results.
[0102] The specific process of extracting features from real-time product image information using a feature extraction algorithm to obtain three real-time chip product features is as follows:
[0103] Extracting a first image, a second image, and a third image from real-time product image information;
[0104] The first image is processed simultaneously using an edge-based feature extraction method, a region-based feature extraction method, and a transformation-based feature extraction method to obtain three first preliminary feature information, and then the three first preliminary feature information are compared. When the features in the three first preliminary feature information are the same, they are exported as a real-time information product feature information. When the features in the three first preliminary feature information are different, feature collection is performed again.
[0105] The second image and the third image are processed in the same manner as the first image to obtain two more pieces of real-time information product feature information;
[0106] Through the above process, more accurate product information can be obtained, thereby providing data support for the final identification of product surface defects.
[0107] In the process of collecting image information, the image acquisition device needs to collect status information of the image acquisition device, and process the status information of the image acquisition device to obtain image device evaluation information, where the image device evaluation information includes whether the image device is normal or abnormal;
[0108] The process of processing the status information of the image acquisition device to obtain the image acquisition device evaluation information is as follows: extracting the status information of the image acquisition device, the status information of the image acquisition device includes the real-time position information of the image acquisition device, the real-time temperature information of the image acquisition device and the clarity information of the images acquired by the image acquisition device three times in the past;
[0109] The real-time position information of the image acquisition device is compared with the set standard position information to obtain the position deviation. When the position deviation is greater than the preset value, an image device abnormality is generated; when the position deviation is less than the preset value, an image device abnormality is generated;
[0110] When the real-time temperature information of the image acquisition device is greater than the preset value a1 for more than the preset time, an image device abnormality is generated;
[0111] When the real-time temperature information of the image acquisition device is greater than the preset value a2, an image device abnormality is generated;
[0112] When the real-time temperature information of the image acquisition device is less than a1, the image acquisition device is normal, a1<a2;
[0113] When the fluctuation range of the image clarity information collected three times in the past is greater than the preset value, an image device abnormality is generated;
[0114] When the fluctuation range of the image clarity information collected three times in the past is less than the preset value, and the image clarity information collected three times is greater than the preset clarity, the image generation device is normal;
[0115] Through the above process, the status of the equipment for image acquisition is analyzed, which greatly reduces the occurrence of errors in the final surface defect identification caused by large deviations between the collected image information and the actual situation due to abnormalities in the image acquisition equipment.
[0116] The specific process of component positioning and assembly detection information is as follows: first, positioning is performed, the image of the installed component is captured by an image acquisition device, and the image recognition feature of the installed component is extracted by an image processing algorithm, and then the real-time position of the recognition feature is extracted, and the position is compared with the preset standard feature position to obtain the position deviation value. When the position deviation value is greater than the preset value, the component positioning and assembly detection information is generated. At this time, the component positioning and assembly detection information is component positioning abnormality. After the component positioning abnormality is generated, a warning message will be sent to the preset receiving terminal to alert the management personnel;
[0117] Then, the image information of the semiconductor chip during assembly is collected by an image acquisition device, and assembly features are extracted from the image information of the semiconductor chip during assembly to obtain assembly features. The assembly features are distributed on assembly components of each assembled semiconductor chip. Then, the distance information of each assembly feature is obtained from the image information of the semiconductor chip during assembly to obtain a real-time evaluation distance.
[0118] Calculate the difference between the real-time evaluation distance and the distance of the preset standard assembly feature, and obtain the feature evaluation difference. When any one of the feature evaluation differences is greater than the preset value, the component positioning and assembly detection information is generated. At this time, the component positioning and assembly detection information is an assembly anomaly. After the assembly anomaly is generated, a warning message is sent to the preset receiving terminal to alert the management personnel;
[0119] Through the above process, relevant inspections are carried out on the component position and assembly in the production process of semiconductor chips, so as to timely discover whether the component positioning and assembly are abnormal, and timely warnings are given when abnormalities are found to ensure the quality of the semiconductor products produced.
[0120] The other test modules detect the chip product by other detection methods, and the specific process of obtaining other detection information is as follows:
[0121] Other detection methods include AOI detection and AVI detection. Through AOI detection and AVI detection, other detection information can be obtained by detecting chip products. Other detection information includes AOI detection information and AVI detection information.
[0122] Among them, AOI detection information includes AOI detection qualified and AOI detection unqualified;
[0123] AVI test information includes AVI test qualified and AVI test unqualified;
[0124] The specific process of AOI inspection is as follows: Image acquisition: Use a high-resolution camera to capture images of the components to be inspected. These cameras are usually equipped with specific light sources and lighting systems to ensure the quality and clarity of the image.
[0125] Image preprocessing: Preprocess the collected images, including denoising, contrast enhancement, binarization and other operations, for subsequent analysis and recognition.
[0126] Defect detection: Analyze pre-processed images through image processing algorithms and pattern recognition technology to identify defects on the surface of components, such as scratches, stains, and missing parts.
[0127] These algorithms can be customized and optimized for different types of defects to improve detection accuracy and efficiency.
[0128] Dimension measurement: Use the pixel information in the image and the known scaling ratio to calculate the dimensional parameters of the component, such as length, width, spacing, etc.
[0129] These dimensional parameters can be compared with preset specifications to determine whether the component is acceptable.
[0130] Welding quality inspection: For welding points, AOI technology can detect whether the welding is complete, whether there are cold joints, short circuits and other problems.
[0131] The quality of the weld can be assessed by analyzing characteristics such as shape, color and brightness of the weld.
[0132] Data output and decision-making: Output the test results in the form of data or images for subsequent processing by operators or automated systems.
[0133] Based on the test results, components can be automatically classified as qualified or unqualified, and corresponding treatment measures can be taken, such as reprocessing, scrapping, etc.
[0134] AOI technology has broad application prospects and important practical value in the field of semiconductor manufacturing, providing strong support for improving the quality of semiconductor products, reducing costs and improving production efficiency;
[0135] The specific process of AVI detection is as follows: Image acquisition: AVI detection technology first captures high-definition images of the conductor components to be inspected in real time through a high-precision camera.
[0136] During image acquisition, special attention was paid to light uniformity and stability to ensure clear, high-quality images were captured.
[0137] Image preprocessing: The collected images will go through a series of preprocessing algorithms, such as grayscale, smoothing, edge detection, morphological processing, etc., to extract key feature information in the image.
[0138] These preprocessing steps help remove noise from images, enhance key features of images, and provide a basis for subsequent analysis and judgment.
[0139] Defect identification and positioning: Using computer vision technology and deep learning algorithms, the AVI inspection machine can intelligently identify various defects on the surface of conductor components, such as cracks, burrs, bubbles, stains, depressions, etc.
[0140] By analyzing and comparing the feature information in the image, the AVI inspection machine can accurately locate the location and type of defects.
[0141] Threshold and algorithm setting: According to the specific requirements and defect characteristics of the conductor components, the AVI detection machine will set reasonable thresholds and algorithms.
[0142] These thresholds and algorithms are used to differentiate between normal and abnormal conditions of conductor components to identify and locate any non-compliant defects.
[0143] Data analysis and decision-making: AVI detection machine has powerful data integration capabilities and can perform statistics and analysis on the detected data.
[0144] Based on these data, AVI inspection machines can assist in production line process optimization and intelligent quality control, providing data support for companies to achieve continuous improvement in yield.
[0145] Test result output: AVI test machine will output the test results in the form of images, data reports, etc.
[0146] These output results can help operators quickly understand the quality status of conductor components and make corresponding treatments and improvements;
[0147] During the inspection process of semiconductor chips, VI technology achieves efficient and accurate inspection of conductor components through high-precision image acquisition, image preprocessing, defect identification and positioning, threshold and algorithm setting, data analysis and decision-making, self-learning and optimization.
[0148] Process AOI and AVI inspection information to generate production abnormality warning information;
[0149] The specific process of producing abnormal warning information is as follows: collect the product quantity information of semiconductor chips manufactured within a preset time, mark it as G, then extract the number of AOI detection failures in G AOI detection information, and obtain the first other detection evaluation parameter;
[0150] Extract the number of AVI test failures from the G AVI test information, and obtain the second other test evaluation parameter;
[0151] A standard first other detection parameter and a second standard other detection parameter are set;
[0152] When the first other detection evaluation parameter is greater than the standard first other detection parameter or the second other detection evaluation parameter is greater than the standard second other detection parameter, a production abnormality warning message is generated;
[0153] Through the above process, it is possible to understand the production quality information of semiconductor chip products over a period of time, issue timely warnings, and thus avoid the occurrence of a decline in yield rate.
[0154] The production test module tests the chip product after it is produced. The specific process of obtaining the production test information is as follows:
[0155] Setting up image acquisition equipment to acquire image information of the test chip product in real time, processing the image information of the test chip product, and obtaining the real-time position of the test chip;
[0156] Then, the chip product is tested for working voltage to obtain working voltage parameters, working temperature is collected to obtain working temperature, and power consumption is collected to obtain power consumption information;
[0157] In the process of collecting working voltage parameters, working temperature and power consumption information, it is necessary to analyze the changes in the real-time position of the test chip in real time;
[0158] When the change of the real-time position of the test chip during the collection of the working voltage parameter, working temperature and power consumption information is greater than the preset value, it is necessary to collect the working voltage parameter, working temperature and power consumption information again;
[0159] Operating voltage parameters, operating temperature and power consumption information constitute production test information;
[0160] In the above process, the application of machine vision technology in chip parameter detection realizes non-contact detection: the machine vision system can detect chips non-contactly, avoiding potential physical damage to the chips. This is especially important for tiny chips because any tiny physical contact may affect their performance. During the relevant testing process, although machine vision itself does not directly measure electrical properties, it can be combined with electrical performance testing equipment to assist in positioning, aligning and identifying chips.
[0161] Before electrical performance testing, machine vision systems can ensure that chips are correctly placed and connected, thereby improving the accuracy and efficiency of testing.
[0162] The production evaluation information includes normal chips and abnormal chip products. The specific process of production evaluation information is as follows: extract the collected production test information, and extract the operating voltage parameters, operating temperature and power consumption information from the production test information;
[0163] When any of the operating voltage parameter exceeds the preset range, the operating temperature is greater than the preset temperature, and the power consumption information is greater than the preset power consumption, the production assessment information indicates that the chip product is abnormal;
[0164] When the operating voltage parameters are within the preset range, the operating temperature is lower than the preset temperature, and the power consumption information is within the standard range, the production evaluation information indicates that the chip is normal.
[0165] All the image acquisition devices used in the above process are different types of cameras set according to actual acquisition requirements.
[0166] A semiconductor chip manufacturing intelligent detection and analysis method based on machine vision, the method comprising the following steps:
[0167] Step 1: Detect surface defects of the test chip product and obtain surface defect detection information;
[0168] Step 2: Component positioning and assembly detection information of the chip product are obtained by detecting the component positioning and assembly detection information of the chip product;
[0169] Step 3: Test the chip product through other testing methods to obtain other testing information;
[0170] Step 4: After the chip product is produced, test it and obtain production test information;
[0171] Step 5: Visualize the surface defect detection information, component positioning and assembly detection information and other detection information to obtain detection display information;
[0172] Step 6: Process the production test information to obtain production evaluation information;
[0173] Step 7: Send the detection display information and production evaluation information to the preset receiving terminal.
[0174] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0175] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0176] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.
Claims
1. Intelligent detection and analysis system for semiconductor chip manufacturing based on machine vision, characterized by: It includes surface defect detection module, component positioning and assembly detection module, other detection modules, production test module, data processing module and information sending module; The surface defect detection module is used to detect surface defects of chip products and obtain surface defect detection information; The component positioning and assembly detection module is used to detect the component positioning of chip products and the assembly of chip products, and obtain component positioning and assembly detection information; The other test modules are used to test the chip product by other testing methods to obtain other testing information; The production test module is used to test the chip product after it is produced and obtain production test information; The data processing module is used to perform visual processing on surface defect detection information, component positioning and assembly detection information and other detection information to obtain detection display information; The data processing module is also used to process the production test information to obtain production evaluation information; The information sending module is used to send the detection display information and the production evaluation information to a preset receiving terminal.
2. The semiconductor chip manufacturing intelligent detection and analysis system based on machine vision according to claim 1, characterized in that: The specific process of the surface defect detection module detecting the surface defect of the chip product and obtaining the surface defect detection information is as follows: The image information of the chip product is collected by the set image collection device according to the preset collection rules, that is, the real-time product image information, and the real-time product image information includes the first image, the second image and the third image; Then, the feature extraction algorithm is used to extract the real-time product image information to obtain three real-time chip product features; The real-time product features are processed through a deep learning algorithm to obtain three real-time chip product features, and the three real-time information products are processed to obtain the first identification information, the second identification information and the third identification information.
3. The semiconductor chip manufacturing intelligent detection and analysis system based on machine vision according to claim 2 is characterized by: The specific process of extracting features from real-time product image information using a feature extraction algorithm to obtain three real-time chip product features is as follows: Extracting a first image, a second image, and a third image from real-time product image information; The first image is processed simultaneously using an edge-based feature extraction method, a region-based feature extraction method, and a transformation-based feature extraction method to obtain three first preliminary feature information, and then the three first preliminary feature information are compared. When the features in the three first preliminary feature information are the same, they are exported as a real-time information product feature information. When the features in the three first preliminary feature information are different, feature collection is performed again. The second image and the third image are processed in the same manner as the first image to obtain two more real-time information product feature information.
4. The semiconductor chip manufacturing intelligent detection and analysis system based on machine vision according to claim 2 is characterized by: In the process of collecting image information, the image acquisition device needs to collect status information of the image acquisition device, and process the status information of the image acquisition device to obtain image device evaluation information, where the image device evaluation information includes whether the image device is normal or abnormal; The process of processing the status information of the image acquisition device to obtain the image acquisition device evaluation information is as follows: extracting the status information of the image acquisition device, the status information of the image acquisition device includes the real-time position information of the image acquisition device, the real-time temperature information of the image acquisition device and the clarity information of the images acquired by the image acquisition device three times in the past; The real-time position information of the image acquisition device is compared with the set standard position information to obtain the position deviation. When the position deviation is greater than the preset value, an image device abnormality is generated; when the position deviation is less than the preset value, an image device abnormality is generated; When the real-time temperature information of the image acquisition device is greater than the preset value a1 for more than the preset time, an image device abnormality is generated; When the real-time temperature information of the image acquisition device is greater than the preset value a2, an image device abnormality is generated; When the real-time temperature information of the image acquisition device is less than a1, the image generation device is normal; When the fluctuation range of the image clarity information collected three times in the past is greater than the preset value, an image device abnormality is generated; When the fluctuation range of the image clarity information collected three times in the past is less than the preset value, and the image clarity information collected three times is greater than the preset clarity, the image generation device is normal.
5. The semiconductor chip manufacturing intelligent detection and analysis system based on machine vision according to claim 1, characterized in that: The specific process of component positioning and assembly detection information is as follows: first, positioning is performed, an image of the installed component is captured by an image acquisition device, and image recognition features of the installed component are extracted by an image processing algorithm, and then the real-time position of the recognition feature is extracted, and the position is compared with the preset standard feature position to obtain a position deviation value. When the position deviation value is greater than the preset value, component positioning and assembly detection information is generated, and at this time, the component positioning and assembly detection information is component positioning abnormality; Then, the image information of the semiconductor chip during assembly is collected by an image acquisition device, and assembly features are extracted from the image information of the semiconductor chip during assembly to obtain assembly features. The assembly features are distributed on assembly components of each assembled semiconductor chip. Then, the distance information of each assembly feature is obtained from the image information of the semiconductor chip during assembly to obtain a real-time evaluation distance. The difference between the real-time evaluation distance and the distance of the preset standard assembly feature corresponding to the feature is calculated to obtain the feature evaluation difference. When any one of the feature evaluation differences is greater than the preset value, the component positioning and assembly detection information is generated. At this time, the component positioning and assembly detection information is an assembly abnormality.
6. The semiconductor chip manufacturing intelligent detection and analysis system based on machine vision according to claim 1, characterized in that: The other test modules detect the chip product by other detection methods, and the specific process of obtaining other detection information is as follows: Other detection methods include AOI detection and AVI detection. Through AOI detection and AVI detection, other detection information can be obtained by detecting chip products. Other detection information includes AOI detection information and AVI detection information. Among them, AOI detection information includes AOI detection qualified and AOI detection unqualified; AVI test information includes AVI test pass and AVI test fail.
7. The semiconductor chip manufacturing intelligent detection and analysis system based on machine vision according to claim 6 is characterized by: Process AOI and AVI inspection information to generate production abnormality warning information; The specific process of producing abnormal warning information is as follows: collect the product quantity information of semiconductor chips manufactured within a preset time, mark it as G, then extract the number of AOI detection failures in G AOI detection information, and obtain the first other detection evaluation parameter; Extract the number of AVI test failures from the G AVI test information, and obtain the second other test evaluation parameter; A standard first other detection parameter and a second standard other detection parameter are set; When the first other detection evaluation parameter is greater than the standard first other detection parameter or the second other detection evaluation parameter is greater than the standard second other detection parameter, production abnormality warning information is generated.
8. The semiconductor chip manufacturing intelligent detection and analysis system based on machine vision according to claim 1, characterized in that: The production test module tests the chip product after it is produced. The specific process of obtaining the production test information is as follows: Setting up image acquisition equipment to acquire image information of the test chip product in real time, processing the image information of the test chip product, and obtaining the real-time position of the test chip; Then, the chip product is tested for working voltage to obtain working voltage parameters, working temperature is collected to obtain working temperature, and power consumption is collected to obtain power consumption information; In the process of collecting working voltage parameters, working temperature and power consumption information, it is necessary to analyze the changes in the real-time position of the test chip in real time; When the change of the real-time position of the test chip during the collection of the working voltage parameter, working temperature and power consumption information is greater than the preset value, it is necessary to collect the working voltage parameter, working temperature and power consumption information again; Operating voltage parameters, operating temperature and power consumption information constitute production test information.
9. The semiconductor chip manufacturing intelligent detection and analysis system based on machine vision according to claim 8, characterized in that: The production evaluation information includes normal chips and abnormal chip products. The specific process of the production evaluation information is as follows: extract the collected production test information, and extract the operating voltage parameters, operating temperature and power consumption information from the production test information; When any of the operating voltage parameter exceeds the preset range, the operating temperature is greater than the preset temperature, and the power consumption information is greater than the preset power consumption, the production assessment information indicates that the chip product is abnormal; When the operating voltage parameters are within the preset range, the operating temperature is lower than the preset temperature, and the power consumption information is within the standard range, the production evaluation information indicates that the chip is normal.
10. A method for intelligent detection and analysis of semiconductor chip manufacturing based on machine vision, the method being based on the analysis system according to any one of claims 1 to 9, characterized in that: The method comprises the following steps: Step 1: Detect surface defects of the test chip product and obtain surface defect detection information; Step 2: Component positioning and assembly detection information of the chip product are obtained by detecting the component positioning and assembly detection information of the chip product; Step 3: Test the chip product through other testing methods to obtain other testing information; Step 4: After the chip product is produced, test it and obtain production test information; Step 5: Visualize the surface defect detection information, component positioning and assembly detection information and other detection information to obtain detection display information; Step 6: Process the production test information to obtain production evaluation information; Step 7: Send the detection display information and production evaluation information to the preset receiving terminal.
Citation Information
Patent Citations
Online intelligent detection and analysis system for semiconductor chip manufacturing based on machine vision
CN115101431A
Auxiliary intelligent detection system for chip processing industry
CN117038494A
Chip detection method and chip detection system
CN117370755A
Automatic intelligent image inspection quality control management system
CN117575382A
System and method for automatic monitoring and intelligent analysis of flexible circuit board manufacturing process
WO2017084186A1
Cited By
Chip defect detection and optimization system based on knowledge graph reasoning
CN119671776A
AOI detection method based on semiconductor chip consistency defect
CN121961999A
