Industrial Data Challenge

Are you ready to tackle a real-world problem and showcase your advanced analytical skills? We invite you to participate in our exciting Industrial Data Challenge! Collaborate with peers, share your insights, and compete for exciting prizes exclusively from Micron Singapore.

This challenge is a fantastic opportunity to:

·         Apply your expertise in advanced analytics, data science, and artificial intelligence.

·         Work with complex, multi-dimensional time series and spatial data.

·         Contribute to advancements in semiconductor manufacturing technology.

Signals to Quality: Unlocking Insights from Process Data

Problem Statement

In semiconductor manufacturing, complex process steps and toolings are involved. In this data challenge, we are particularly interested in one of the critical process steps. Various tools are involved in this step, and multiple sensors capture the signals during the process. At the end of each run, a wafer level measurement is collected which records the output value at each location of the product (indexed by the x,y coordinates). Due to equipment variations and process dynamics, the resulting measurements can exhibit non-uniformity across different locations. Accurately predicting both the average measurement and its variability is essential for optimizing manufacturing processes.

On top of the complex relations between tool setup, process signals, and the output, the tools degrade over time, and have varying lifetimes. As a result, the performance of the tools fluctuates over time. It is of interest to account for the tool degradation in the output prediction. It is noted that certain hardwares inside the tools can be replaced, and their lifetime will be reset after replacement. An illustration of the data structure and relation is shown in Figure 1.

Figure 1. Illustration of the process data and quality data 

Data

Data comes from a processing step from a semiconductor manufacturer. The dataset contains in-situ sensor measurement in each run. The input data are organized in the following format. Essentially, the inputs record time series of various sensors associated with tools during the process runs. The age of the tools are recorded as “ConsumableLife” to consider the impact of tool degradation over the performance.

Objective
The primary objective of the data challenge is to develop a predictive model that utilizes the given signal data and tool usage data in each run to predict the output quality. The model should account for the variations in tool performance across different lifetimes and the impact of hardware changes.

A successful model will be evaluated based on its accuracy in predicting the outcome of the product. The accuracy will be evaluated on a separate set of test data provided by the competition committee. The model should demonstrate high precision and robustness in handling the time series data and the inherent variability in tool performance.

Submission Details

Team of up to 3 participants may participate in this challenge. The submission deadline is May 31, 2025. The submission should consist of a report along with the source code used to generate figures and tables in the report. The report should be limited to 25 pages, excluding references. It should clearly highlight your methodology, assumptions, preprocessing steps, and implementation details. The prediction outcome on the test dataset should also be submitted. Please put all predicted measurement values on the test data in a single CSV file following the data format of the output data. Participants can use any coding language to solve the problem.

Submission will be evaluated by a panel of judges based on (i) the accuracy of the methods; (ii) the suitability and innovation of the used methodology, (ii) the insights derived from the model, and (iii) the clarity, technical correctness and completeness of the report and the presentation. Finalists will present their solutions live at the ICQSR conference workshop. The first place and runner up of the competition will be then announced and recognized at the conference. A cash prize or an equivalent award will be granted.

The competition judges consist of experts from academic and industry. All submissions should contain a single zip file named “ICQSR2025_Data_Teamname.zip” and emailed to the organizers Chen Zhang zhangchen01@tsinghua.edu.cn by May 31 2025, anywhere on earth.

Who Can Participate

Scholars, scientists, researchers, engineers, and students with a passion for solving industrial problems. A team of maximum 3 people can be formed to participate the data challenge. 

If you are interested in participating in the data challenge, please register your interest here by March 31 2025, anywhere on earth. The link to the data will be sent to registered participants.

Registration link: https://docs.google.com/forms/d/1scXiPXGNcsXGCQpV9GpaSrE9LWfojnPsA3m5-Ir_T_I/edit 

Timeline

• Mar 1 2025: Starts of the competition

• Mar 20, 2025:  Q&As and challenge explanations from Micron

• May 31, 2025: Submission deadlines

• June 8, 2025: Notification of the finalist

• Jun 30, 2025: Finalist Presentation and Prize Ceremony

Presentation and Prize

• Four teams as finalists will be invited to present their solution at the conference workshop

• At least one member from the team needs to attend and present the solution

• A panel judge from both academic and Micron will be formed to decide the winner

• Cash prize

Don't miss out on this chance to make a significant impact in the semiconductor industry. Register today and be part of the innovation!