The mechanical behavior of engineering materials under extreme strain rates — as encountered in impact, crash, and ballistic events — is governed by their dynamic flow stress. For decades, the Johnson-Cook (J-C) equation has been the industry standard for representing this behavior, yet its mathematical foundations have remained unexamined and its calibration from real experimental data has been fraught with ambiguities.
This research establishes a rigorous, data-driven trilogy: first verifying when decoupled constitutive equations are mathematically legitimate, then determining flow stress from varied strain-rate SHPB data, and finally extending the framework to the full three-dimensional (strain, strain-rate, temperature) constitutive surface. The methodology — combining SVD/CP tensor decomposition with artificial neural networks — has broad applicability beyond the J-C equation to any factorized constitutive model.
Background & Motivation
The Johnson-Cook equation assumes that strain hardening, strain-rate sensitivity, and thermal softening can be separated — i.e., the dynamic flow stress is a product (or sum of products) of independent functions of each variable. While this assumption drives nearly all impact simulations globally, its mathematical validity against actual experimental data had never been systematically tested.
Key Contributions
- First rigorous mathematical framework to verify the legitimacy of decoupled (factorized) flow stress equations using matrix/tensor decomposition
- SVD applied to 2D flow stress data arrays reveals when a rank-1 approximation is sufficient — the necessary condition for J-C type equations
- Quantitative criterion established for measuring the validity of decoupled assumptions for any material dataset
- Discrete flow stress representation method developed — avoids the need to assume a functional form
- Problems inherent to the conventional J-C parameter-fitting procedure identified and clarified
- Demonstrated on multiple materials: the legitimacy of decoupling varies by material and must be verified, not assumed
Background & Motivation
Traditional SHPB (Split Hopkinson Pressure Bar) tests require constant strain-rate conditions, which are practically difficult to achieve. Most experimental data contains mixed-strain-rate loading histories. Conventional approaches discard this "impure" data or fit it with simplified models, discarding valuable information and introducing systematic errors.
Key Contributions
- Framework for determining dynamic flow stress directly from varied strain-rate SHPB data — relaxing the conventional constant-strain-rate requirement
- Data qualification criteria developed to screen and validate raw SHPB experimental datasets
- ANN combined with SVD to generate a finely-resolved 2D flow stress matrix from sparse experimental points
- Five inherent uncertainties in conventional flow stress determination methods identified and addressed systematically
- Method demonstrated to produce superior accuracy compared to conventional J-C fitting procedures
- Provides both a discrete flow stress table and an equivalent analytical equation for direct use in FEM simulations
Background & Motivation
Real impact events involve simultaneous evolution of strain, strain-rate, and temperature. Part 2 extends the methodology to the complete three-dimensional constitutive surface — the full dependence of flow stress on all three thermomechanical variables simultaneously. CP (CANDECOMP/PARAFAC) tensor decomposition generalizes the SVD approach to 3D data arrays.
Key Contributions
- Extension of the ANN+decomposition framework from 2D (strain-rate only) to 3D (strain × strain-rate × temperature)
- CP (CANDECOMP/PARAFAC) decomposition applied to three-dimensional flow stress tensors
- ANN used to generate a finely-resolved 3D flow stress tensor from sparse multi-temperature SHPB data
- Thermal softening effect modeled with superior accuracy compared to modified Johnson-Cook approaches
- Demonstrated on C54400 phosphor copper alloy across wide strain-rate and temperature ranges
- Both discrete 3D constitutive tables and equivalent analytical equations derived for simulation use
"This trilogy establishes a complete, mathematically grounded pipeline for dynamic material characterization: from verifying when empirical models are legitimate, to determining constitutive equations from realistic (non-ideal) experimental data, to fully resolving the three-dimensional mechanical response. The same ANN+SVD/CP methodology has since been extended to impact resistance prediction, demonstrating its broad applicability as a data-driven mechanics framework."— Research significance of the constitutive modeling trilogy