Board-Level Reliability (BLR) traditionally refers to the durability and lifespan of a package's solder interconnections after it has been mounted onto a printed circuit board (PCB). Because solder joints typically represent the weakest link in the assembly, they have historically defined the reliability of the entire board. Our lab aims to expand this definition — moving beyond solder-joint assessment toward the diagnosis and prognosis of the full printed circuit board, drawing on newly proposed failure theories and component-level lifetime models that account for system-wide interactions. 

• Failure Theories: We develop novel failure theories that do not assume the independent reliability of each component. This involves deriving complex, nonlinear mathematical and physical degradation relationships among the components that comprise the board, capturing how the deterioration of one component can influence the reliability of its neighbors. 

• Lifetime Model Developments: Board-level reliability emerges from the individual yet interdependent reliabilities of its constituent components. We develop robust mathematical models to predict the Remaining Useful Life (RUL) of electronic components across varying boundary and loading conditions, building the foundation needed to assess overall board-level reliability as an integrated system. 

• Machine Learning: A key challenge in board-level reliability is the computational cost of simulating interdependent degradation across numerous components using traditional finite element methods, which require re-meshing and re-solving for each new set of loading conditions. Machine learning offers a more scalable alternative: once trained on sufficient data, these models can generalize across complex, high-dimensional degradation patterns at a fraction of the computational cost. 

Prognostics and Health Management (PHM) frameworks shift maintenance from a reactive "run-to-failure" approach to a predictive, proactive strategy. By drawing on operational data and statistical lifetime models, PHM tools assess current system health and forecast future performance - enabling continuous monitoring of critical systems, preventing catastrophic field failures, extending operational lifespans, and optimizing maintenance schedules.

Our core research areas in PHM include:

• Condition Monitoring & Feature Extraction: We identify measurable precursors to failure before macroscopic damage develops. This involves tracking subtle changes in electrical and thermal signatures under real-world operating stresses — such as shifts in thermal impedance, increases in parasitic resistance, or rising leakage currents in degrading power components. 

• Predictive Lifetime Modeling: A core objective of our PHM research is to accurately predict the Remaining Useful Life (RUL) of electronic components. We develop robust mathematical models that estimate RUL across a range of component types, boundary conditions, and loading profiles. 

• Physics-Informed Learning: Effective field-level decision-making demands models grounded in physical reasoning, not purely data-driven pattern recognition. Rather than relying solely on black-box machine learning, we fuse AI techniques with our deep Physics-of-Failure (PoF) expertise. Conventional finite element methods remain the gold standard for mechanical accuracy, but their computational cost makes them impractical for field applications that require responses in seconds. By embedding experimentally observed degradation behavior and physics-of-failure knowledge into our models, we develop lightweight yet reliable tools for both prognosis and diagnosis. 

Design for Reliability (DfR) is the proactive integration of Physics-of-Failure (PoF) insights directly into the early stages of the product development cycle. Traditionally, industry has relied on a reactive "build-test-fix" approach — waiting for physical prototypes to fail in the lab before iterating on the design. Our DfR methodology replaces this costly and time-consuming cycle with predictive, simulation-driven engineering. 

Our DfR processes include:

• Virtual Prototyping: We use Finite Element Method (FEM) simulations to create digital twins of PCB layouts and power modules, then subject them to realistic operational environments encompassing thermal, mechanical, and electrical stresses. 

• Design Optimization: We conduct systematic parametric studies informed by our simulation data and damage accumulation models. By iteratively varying key design parameters — such as interconnect geometries, substrate thicknesses, and material properties — we map how each variable influences stress and strain distributions across the assembly. 

• Reliability Assessment: Finally, we quantify the expected operational lifespan of a design under specific mission profiles, verifying that it meets the required reliability targets before physical prototyping begins. 

Physics-of-Failure (PoF) methodology investigates the physical, chemical, and mechanical root causes of degradation to accurately predict component lifespan. Finite Element Method (FEM) analysis serves as a critical bridge between microscopic stress mechanisms and macroscopic failure outcomes. 

Our core research activities include:

• Multi-Physics FEM Simulations: We use FEM to model the complex, real-world stressors that components experience, including thermal cycling and dynamic vibration. These simulations resolve stress, strain, and temperature distributions across intricate board-level geometries and packaging architectures, providing physically grounded explanations for observed failure behavior. 

• Degradation Analysis: We investigate the specific failure mechanisms driven by localized stresses. Field failure data are often scarce and costly to obtain, limiting the reach of purely statistical approaches. Where failure statistics are insufficient, our PoF methodology fills this gap by identifying the most physically consistent failure mechanisms from known geometries, boundary conditions, and loading profiles.