SHF: Small: End-To-End Test Data Analytics For Automotive Chip Production Lines

University Of California-Santa Barbara, Santa Barbara CA

Investigators

Abstract

Using automotive industry as the research driver, this project aims to develop novel data mining solutions to address the challenge of reliable low-power, high quality chip production by improving the effectiveness and reducing the cost of testing them. While research in the data mining community focuses more on developing generic approaches, this project focuses on developing dedicated approaches, optimized for mining test measurement data. The techniques developed in this project can complement and bring synergies to existing data mining research. The technologies developed through this project will be transferred to the industry, providing solutions to overcome the challenges in the automotive chip sector. Furthermore, the research will be integrated with educational activities to produce publications, curriculum materials, tutorials, and software tools for broader impacts to the semiconductor industry. In production, automotive chip products go through a rather comprehensive test process to assure their quality. End-to-end test data refers to all data collected in this process which comprises multiple stages, starting from product manufacturing all the way to evaluation in an electronic system. Analytics refers to the discovery, interpretation, and utilization of knowledge extracted from the data. This research aims to enable effective and robust analytics in order to improve product quality and reduce production test cost. For effectiveness, novel software tools and methodologies will be designed to automatically incorporate domain knowledge in the analytics. For robustness, new approaches will be developed to determine the meaningfulness of data mining results. From a practical perspective, solutions developed through the research will benefit the semiconductor industry by facilitating the effective use of large-scale data mining for test process optimization. From a scientific perspective, this research will provide a deeper understanding of the limitations with test data analytics and enable its robust implementation. Overall, this project aims to develop the next-generation test data analytics software, thereby enabling application of the research to diverse scenarios encountered in semiconductor chip production test environments.