Semiconductor Manufacturing International has patented a computer-implemented detection method using machine learning to improve defect analysis accuracy in semiconductor manufacturing. By comparing layout and scan graphics, extracting patterns, and training a neural network, the method accelerates technology development and enhances production efficiency. GlobalData’s report on Semiconductor Manufacturing International gives a 360-degree view of the company including its patenting strategy. Buy the report here.
According to GlobalData’s company profile on Semiconductor Manufacturing International, Quantum dot devices was a key innovation area identified from patents. Semiconductor Manufacturing International's grant share as of May 2024 was 70%. Grant share is based on the ratio of number of grants to total number of patents.
Defect detection method using machine learning for semiconductor devices

A recently granted patent (Publication Number: US11995814B2) outlines a computer-implemented detection method that involves superimposing and comparing a layout graphic of a sample device with a scan graphic of the same device. The method further includes extracting a sample non-overlapping pattern, encoding it to form sample coded data, and training a machine learning model based on a convolutional neural network using this data. The trained model is then used to detect defects in a to-be-detected device by inputting its layout graphic and scan graphic, ultimately outputting the defect detection result. The patent also details the use of an electron microscope to obtain scan graphics, as well as specific processing rules for extracting non-overlapping patterns effectively.
Additionally, the patent describes a detection apparatus and electronic device that implement the detection method outlined in the claims. The apparatus includes modules for obtaining layout and scan graphics, extracting non-overlapping patterns, encoding data, and executing the detection model to identify defects in a to-be-detected device. The electronic device, on the other hand, comprises memory to store instructions and processor circuitry to execute the detection process. Furthermore, a non-transitory machine-readable media is detailed, containing instructions that, when executed, enable a processor to carry out the defect detection process using a convolutional neural network-based machine learning model. Overall, the patent provides a comprehensive framework for efficiently detecting defects in devices using advanced computational methods and machine learning techniques.
To know more about GlobalData’s detailed insights on Semiconductor Manufacturing International, buy the report here.
Data Insights
From
The gold standard of business intelligence.
Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.