로고 로고

SITE MAP

REFERENCE AI·Big Data

REFERENCE

Based on over 20 years of experience in building smart factories,
DL Information Technology Co. Ltd. is creating manufacturing big data analysis
and AI application cases in various industrial fields

AI/Big Data Construction Case_Data Inquiry Analysis Processing and Wafer Chip Defect Prediction Application
for Efficiency Improvement

In order to improve wafer test result analysis efficiency, DL Information Technology Co., Ltd. provides data inquiry analysis processing and wafer chip defect prediction to semiconductor post-processing companies

Raw data Analysis environment/Technology Data collection/Loading Visualization

Raw Data

Test results for each test site showed slightly different formats
Semi-structured text data
1.Wafer information 2.Defect status by chip 3.Wafer defect status
It can be divided into three parts, and the wafer consists
of 25 sheets

Analysis environment / technology

  • AutoEncoder model structure

Applying the AutoEncoder deep learning model to predict defects for each wafer chip location
- Encoder: Converts data input into internal representation
- Decoder: Converts internal representation into output data
Learn to find latent variables of input data through encoding and decoding of data for learning
Enter new data to the learned model to predict defects

Data Collection / Loading

Upload test result files from the test company to the FTP server
(Extension: .P1, .P2, .P3, .PM)
Download semi-structured test result files from the FTP server
Convert each file into 3 structured datasets
Save the datasets in each DB

Visualization

  • Defect status for each wafer chip

  • Fail Bin Chart

Paint Point

  • Difficult to manage
    and analyze data

  • High probability
    of errors

  • Data inaccuracy

  • Lower reliability
    of results

Absence of quality data management
Difficult to manage and analyze data due to different formats and storage forms of quality test result data
High probability of human errors due to manual classification/determination statistical work of quality data
by workers
Lower reliability of data analysis results due to data inaccuracy
Introduction effects
  • Improving data analysis efficiency through automatic integration of wafer test result data
    for each site
    Converting the contents of wafer test result files in different formats for each site and semi-structured (text) formats into structured data
    Integrating the converted data and then loading it into a database for management
    You can quickly and easily search for desired data from file management to database management, and easily analyze data through data visualization
  • Predicting Wafer chip defect using AutoEncoder
    AI model
    Improving prediction accuracy by applying prediction models
    for each number of wafer dies
    Hyperparameter custom model learning available
    Improved productivity through wafer chip defect prediction
Client company