Session: 12-12-01: Modeling of the Fracture, Failure, and Fatigue in Solids

Paper Number: 149829

149829 - A Multi-Fidelity Modeling Approach to Predict Tool Fatigue Fracture in Friction Stir Welding 

Friction Stir Welding (FSW) enables multi-material joining while minimizing mass and energy usage, which contributes to its potential cost-effectiveness. An additional criterion for cost in high-volume production environments, such as the automotive industry, is the weld length achievable before tool replacement is necessary. Although tool wear can also influence the achievable weld length, this study focuses exclusively on fatigue fracture. Unlike the gradual tool wear process, fatigue fracture represents a sudden failure mode that could result in costly work stoppages. This work aims to develop an approach to predict the fatigue fracture of a given tool under different processing conditions and enable optimizing the tool design for longer life.

Most studies on friction stir weld tool life have been primarily empirical, involving time-consuming welding carried out until tool failure. Such an approach would become prohibitively expensive when considering multiple process parameters and iterating numerous tool designs. The other option would be using physics-based models. All such models rely on determining the stresses at key locations on the tool for a single rotation. With this cyclic stress data, the tool's fatigue life was estimated using the stress-life (S-N) approach. The issue is that these models are generally analytical representations, relying on simplified approximations of force distribution and tool geometry. Although these models identify some key factors affecting tool life, their predictions often deviate significantly from observed results. For the issue of determining force distributions, a reliable approach would be to use process models such as Smoothed Particle Hydrodynamics (SPH); however, these models are complex, expensive, and have yet to be fully realized. In contrast, the issue of improving tool geometry models can be addressed more easily.

In this work, we present a modeling approach that utilizes models of two different levels of fidelity to determine tool life. The force distribution on the tool is essentially based on analytical models, i.e., a low-fidelity approach. However, this represents an improvement over previous work because the analytical models were calibrated to the net forces and torque recorded by the gantry. The model is high-fidelity in terms of tool geometry. Every detail of the tool—including the threads, the flat region, and other features—was accurately modeled using Finite Element (FE) analysis in Abaqus. In the FE model, the force distribution was applied for a single rotation to determine the cyclic stress at all locations on the tool. Key locations were then identified, and the cyclic stress was extracted. The cyclic stress state was multiaxial and not fully reversed. Therefore, to compare it to fully reversed rotation-bending experimental data for H13 steel, Fatemi-Socie parameter and the Goodman diagram were employed. The comparison with selected experimental data was favorable, so the model was validated. Additional exploratory simulations consider possible design, material, and process parameter changes that could improve tool life.

Presenting Author: Kranthi Balusu Pacific Northwest National Lab

Presenting Author Biography: Kranthi Balusu specializes in computational modeling and mechanics of materials. Kranthi is experienced with a wide variety of multiscale & multiphysics computational modeling methods in solid mechanics. In addition, his work involves the formulation & development of these computational methods. His modeling expertise has been applied to a wide variety of materials such as biological tissues, metals at the scale of dislocations and crystals, and composite materials at a microstructural level.

Kranthi obtained his bachelor’s in aerospace engineering and master’s in applied mechanics from the Indian Institute of Technology Madras (IITMadras). He has earned his Ph.D. in aerospace engineering from the University of Texas at Arlington, with a dissertation in microscale plasticity.

Kranthi has been a postdoc at PNNL since 2021. Here, his modeling work on projects from the DOE’s vehicle technology office (VTO) assists in the efficient development and deployment of advanced manufacturing techniques. He is always interested in opportunities that would allow him to develop and utilize a broader variety of computational methods and apply them to a broader variety of materials, material processes, and behaviors.

Authors:

Kranthi Balusu Pacific Northwest National Lab
Hrishikesh Das Pacific Northwest National Lab
Shivakant Shukla Pacific Northwest National Lab
Ayoub Soulami pacific northwest national lab
Piyush Upadhyay Pacific Northwest National Lab