Reliability and Prognostics Study for Thermal-mechanical Failures of Memory Product

Samsung Company

Motivation & Problem Description

Lifetime is a critical metric for evaluating design options during the development of semiconductor memory, where ensuring long-term reliability is essential to reduce failure risks and manage lifecycle costs. However, traditional lifetime prediction methods rely heavily on limited experimental testing and empirical models, which often fail to capture the multiphysics degradation behaviors of modern devices. This project aims at developing a domain-specific prognostics framework designed to overcome these limitations in semiconductor applications. 

Objective

 To develop a novel FEA-based prognostics framework aimed at enhancing early design- stage lifetime prediction for memory product. 

Solution

The developed novel FEA-based prognostics framework achieves two primary objectives: (1) it leverages finite element analysis to systematically generate multimodal multiphysics degradation data under combined external stresses, enabling the adaptation of existing prognostic methods from other domains or the development of new ones tailored to memory products; and (2) it establishes a complete end-to-end pipeline that integrates this rich simulation data into downstream lifetime prediction and uncertainty quantification. 

Impact

A case study on solder ball interconnects in memory modules demonstrates that the framework can match industry-standard lifetime predictions while reducing simulation time and costs by approximately one-third, and providing insights into prediction confidence. Beyond introducing a new predictive methodology, this work presents a scalable platform for data-driven prognostics in semiconductor memory, enabling earlier and more informed design decisions.

Reliability and Prognostics Study for Thermal-mechanical Failures of Memory Product
Figure 1. The overview of the proposed framework. As illustrated in Figure 1, this framework spans two critical phases of the product lifecycle: (i) the development phase, where physical prototypes do not yet exist and simulations are the primary source of data, and (ii) the qualification phase, where prototypes become available for real-world testing and validation of simulation results.
References:
  1. Zhang, Z., Gorbunova, A., Rhew, K., & Shi, J. (2025). A Review of Prognostics Methods for Electronic Packages: From a Structure-Aware System-Level Perspective. IEEE Transactions on Reliability, vol. 74, no. 3, pp. 4116-4130.
  2. Zhang, Z., Rhew, K., Hwang, Y., Chun, K. C., & Shi, J. (2025). Early design-stage prognostics of semiconductor memory using finite element analysis. IISE Transactions, 58(9), 1093–1110.

Project PI

Jianjun Shi

Jianjun Shi

Carolyn J. Stewart Chair
Professor

ISyE ProfilePersonal Website