Root cause analysis methods, devices, electronic equipment, media and products

By combining machine learning models with engineer experience in semiconductor manufacturing, features are extracted from multi-dimensional data, solving the problem of low accuracy in existing traceability analysis and achieving more accurate yield traceability and production optimization.

CN122310068APending Publication Date: 2026-06-30SHANGHAI INTEGRATED CIRCUIT RESEARCH & DEVELOPMENT CENTER CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411954460.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-27
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing traceability analysis has low accuracy in semiconductor manufacturing and cannot fully reflect the complexity of yield issues, leading to low analysis efficiency and misleading optimization.

Method used

By employing a pre-trained machine learning model, feature data is extracted from multi-dimensional data through multiple feature extraction layers. Combined with engineers' experience and prior information, the source tracing results are comprehensively analyzed to ensure the representativeness and interpretability of the results.

Benefits of technology

It improves the accuracy and efficiency of yield traceability analysis, enabling the identification of the root causes of yield anomalies caused by multiple factors, reducing misleading optimization, and improving production efficiency and product quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122310068A_ABST
    Figure CN122310068A_ABST
Patent Text Reader

Abstract

This application provides a root cause analysis method, apparatus, electronic device, medium, and product, relating to the field of data mining technology. The method includes: acquiring data to be processed, which includes various types of parameter values ​​related to yield; inputting the data into a trained machine learning model to obtain at least one first root cause analysis result; the model includes multiple feature extraction layers, each with corresponding weight coefficients; determining at least one second root cause analysis result based on the second feature data output by each feature extraction layer, prior information, and weight coefficients; and determining a target root cause analysis result based on the first and second root cause analysis results. Because the model can extract feature data from multiple dimensions, thereby identifying key features, and combining prior information with the output of the feature extraction layers, it identifies features with significant impact. By comprehensively analyzing the root cause analysis results obtained from both methods, the accuracy of root cause analysis is improved.
Need to check novelty before this filing date? Find Prior Art