中国·金年金字招牌诚信至上(股份)有限公司

INF-WPA

An Intelligent and Efficient Platform for Wafer Pattern Analysis

Overview

INF-WPA is a product for wafer map pattern analysis. It leverages a variety of advanced machine learning techniques to extract features from abnormal patterns based on the distribution of data on the wafer, identifies product quality issues, and assists engineers in subsequent analyses. With INF-WPA, engineers can quickly pinpoint anomalies in production equipment, confirm process fluctuations, and analyze process windows and weaknesses. INF-WPA helps reduce product yield loss, lower production costs, and enhance production efficiency.

Advantages

INF-WPA helps fab engineers conduct pattern summary analysis and associate analysis results with the yield management system. This enables the long-term monitoring of various patterns. For instance, engineers can perform tool commonality analysis based on pattern classification results and employ machine learning algorithms to identify wafers with similar user-defined patterns.

One-click analysis correlated with other modules
Display of map pattern

User-friendly interface
Data visualization

Accurate and fast classification models
Easy and efficient pattern clustering

Features

Wafer Pattern Classification

INF-WPA offers a high classification accuracy of wafer map patterns that exceeds 98% and supports customized categorization of wafers based on the patterns. Wafer map patterns are strongly correlated with their underlying root causes. By classifying wafer map patterns with high accuracy, INF-WPA helps users categorize and analyze low-yield wafers, which reduces root cause troubleshooting time and improves engineers' work efficiency. Besides, wafer patterns and fab production anomalies, along with wafer testing issues, mutually corroborate and complement each other.

Wafer Pattern Match

INF-WPA enables users to customize fail patterns based on defective wafers and specific bins of interest. It calculates the similarity between all wafers and the customized patterns through pattern matching. By correlating production and testing process data, it pinpoints commonalities, thereby shortening root cause investigation time, providing robust data support to clients, and mitigating losses. The vast wafer pattern data on the platform establishes a powerful database for users, enabling the retrospective analysis of historical issues.

Wafer Pattern Cluster

INF-WPA employs unsupervised deep learning methods to rapidly cluster all parameters based on the distribution of failed maps. It enables users to analyze defects, identify their causes, and resolve the issues so as to accelerate the yield improvement speed.

Wafer Pattern Measure

Through AI-based feature parameter extraction, INF-WPA confirms the central cluster range, scratch radius, angle, and other feature dimensions. By comparing with the equipment dimension feature parameter database and integrating WIP data, INF-WPA helps pinpoint abnormal equipment in the production process.

Wafer Ink

INF-WPA enables users to ink dies within pattern paths that present reliability risks based on pattern identification and specific rules, which prevents problematic chips from entering subsequent processes and causing significant quality issues. This reduces costs and enhances the safety of final products as well.