Revision History

Version 18.3.3, Build 4240 (April 25, 2022)

• A parameter definition-substitution capability is now available. Parameters and their values are de-
fined with the *parameter command, e.g., *parameter ymod = ’88000’, yield = ’485’. Param-
eter names in WARP3D commands are referenced with @, e.g., properties ... e @ymod sig yld
@yield .... Parameter definitions may appear anywhere in the input. They often are gathered into
a separate file and included with *input from file ... at the beginning of the WARP3D input
file. This capability may simplify the organinzation of sensitivity analyses where a small number of
model values change between analyses.
• Automatic assignment of element blocks to to domains for MPI+threads execution. The assignment
algorithm now uses a simple round-robin scheme.
• The domain integral processor has long had a very convenient capability to generate the lists of
elements and nodes automatically. But the algorithms do not detect when/if a domain reaches a
free surface in a model. The new command print domain extents displays the min-max global
coordinates X-Y -Z that often enable simple confirmation of the domain by the user.
• Comment lines may now be included in the (possibly large) data files of crystal orientations in the
CP material model
• Now use newest versions of the Intel Fortran Compiler Classic (Windows, Linux, MacOS) and GNU-
Fortran (Linux, MacOS). The Intel compiler system and Math Kernel Library (w/ Pardiso solver)
are available for no charge. The MKL-Pardsio is used in the GNU-Fortran builds.
• Added new full input files for plane-sided C(T) specimen and a shear fracture test specimen in the
standard fracture specimens directory.

Manual Sections updated: TOC, 2.6 (Blocking and Domains), 2.15 (Utility * Commands), 2.17
(*parameter), 3.12.7.2 (Crystal orientations), 4.4.8.1 (Automatic Domain Scheme).

Version 18.3.2, Build 4220 (January 25, 2022)

• Expanded and simplified definition of domains for J-integral computation. Any number of nodes
may be defined in a node set at each front location. This enables the 1st domain to extend well away
from the front. Automatically generated remote domains then may have fewer ill-shaped elements
and possibly avoid mesh transitions that degrade domain (path) independence. This new capability
simplifies the definition of remote domains needed for large-displacement (finite-strain) analyses and
crack-growth analyses. The definition of domains to obtain a single, average J-value for the front is
now much more convenient.
• New options to improve performance of crack initiation/growth by element deletion. Addresses the
challenging issue of severe distortions that cause solutions to be stopped. A (optional) distortion
metric parameter, the so-called Oddy parameter, can be tracked over the solution and trigger element
deletion before distortions become too severe. Separate from the element deletion criteria (damage),
the accumulated plastic strain, ε̄p , may have also have a limiting (maximum) value to trigger element
deletion.
• Crack initiation/growth by SMCS criterion now supports multiple options for post-initiation soft-
ening that gradually remove elements from the solution. This process helps reduce element mesh
dependency of the solutions and often improves numerical convergence.
• New output options to produce files with key information as elements are removed with SMCS
controlled crack growth.

Revision History RH-i Updated: April 23, 2022 at 7:55am

 • New option for crack growth by node release (CTOA controlled growth) stops the analysis if any of
a list of provided nodes is released. This is a convenient option when a target amount of growth is
known but the number of load(time) steps required is unknown due to various adaptive and overshoot
modifications.
Manual Sections updated: TOC, 4.4 (Commands for J-Integral Computations), 5.2 (Growth by
Element Deletion, full re-write), 5.3 (Crack Growth by Node Release).

Version 18.3.0, Build 4200 (May 15, 2021)

• This release includes a new, quite general capability to reduce the effects of meshing and element
sizes in simulations to predict ductile crack extension via element deletion. Various forms of the
Stress Modified Critical Strain Model (SMCS) set the conditions to start the process to remove
an element from the solution. The new mesh regularization schemes provide options to reduce the
element stresses and stiffness to zero using a damage parameter d that increases from 0 at the
onset of deletion to 1.0 when stresses and stiffness become zero. The recommended procedure adopts
a cohesive, traction separation concept where the user specifies a fracture energy (Gf ) and the
functional form for evolution of d(G/Gf ) from 0 → 1. New verification suite case test88 exercises
this capability.
• Updated Intel compiler system from 2019 to 2021 (oneAPI) on Linux and Mac OSX. The oneAPI
compiler system is now available without charge from Intel.
Manual Sections updated: TOC, Executive Summary, 5.1, 5.2 (Crack Growth by Element Deletion).

Version 18.2.4, Build 4160 (February 20, 2021)

• Corrections to off-diagonal terms of the Q matrix for NLGEOM analyses. The default is to omit this
contribution in forming the element tangent stiffness.
• Updated to a new pivoting strategy offered in the MKL Sparse Solver.
• Added code to suppress MKL warning messages about use of deprecated OMP features. Intel is
supposed to fix this issue in coming MKL releases.
• Increased the maximum number of user-definable lists from 100 to 500. There is no limit on the
number of entries in each list. Updated Intel compiler and MKL from 2019 to 2021 on Linux and
OSX. Windows remains at 2019.
Manual Sections updated: 1.9 (Finite-Strain Plasticity), 2.10.14 (Consistent Q Matrix).

Version 18.2.4, Build: 4150 (January 1, 2021)

• The Stress-Modified Critical Strain (SMCS) approach to initiate/continue crack extension via ele-
ment deletion has new options to define the critical plastic strain: the Modified Mohr-Coulomb form
from Wierzbicki’s MIT group that includes the effects of stress triaxiality and Lode angle. This form
offers a fourth option to define the critical plastic strain. A fifth option provides the Wellman/Sandia
tearing parameter as a somewhat simpler criterion for the onset of element deletion.
• Several new features support solutions with crack growth by element deletion. The user may provide
a list of elements such than when any of the listed elements is deleted, the analysis is stopped and
a restart file written. This option may be used to detect when the crack extends beyond a region of
regularly shaped elements intended for crack growth. Element deletion governed by critical porosity
or critical plastic strain may employ this option.
• For element deletion by SMCS, the user may request that at file be written during solution to list
the state (plastic strain, triaxiality, Lode angle, ...) of each element as it is deleted. This text file
may prove convenient in post-processing of crack growth analyses.
• For element deletion by SMCS, the user may request generation of special element states result files
for all load steps or for a list of load steps. These flat files are written in stream or text (ascii)

Revision History RH-ii Updated: April 23, 2022 at 7:55am

 format and follow the concepts of flat output files described in Section 2.12 [Flat Text & Stream
(Binary) Result Files]. These files provide SMCS related quantities (plastic strain, current estimate of
critical plastic strain, triaxiality, Lode angle,...) for elements useful in post-processing with ParaView
for example. The warp3d2exii (Python) program recognizes these files and includes the results in
creating a .exo file.
Manual Sections updated: 5.2 (Crack Growth by Element Deletion).

Version 18.2.3 (September 15, 2020)

• Input of initial stresses for the model at time t = 0 may now be accomplished via an Abaqus
compatible user-routine SIGINI. See Section 2.9 for inout commands and Appendix I for details on
the implementation.
• Thermal expansion coefficients can be temperature dependent using the secant definition of the
coefficients. The code now supports input definition of a reference (or zero) temperature used in
measuring the secant values. This material property values is alpha zero. See discussion in Section
2.2 and new material property alpha zero for bilinear, mises, gurson material models in Chapter 3.
• The verification directory contains a large collection of problems to exercise most aspects of the
code. With this release, the small programs that supervise execution of these models and checking
of solution values for comparison with known correct values has been changed from a Bash shell
to Python 3. The codes run tests.bash and run J tests.bash are replaced by run tests.py and
run J tests.py. This change is especially important for Windows – it removes the need to install
the Cygwin software that provides a Bash shell. The installation of Python 3 on Windows and use
inside a standard command shell is straightforward. The QuickStart document for each platform and
the document Workflow run WARP3D verification provide complete details to install Python3 and
to run the verifications test cases.
• WARP3D can output files of computed results with a simple format (flat files, Section 2.12). These
files are used by the Python code warp3d2exii provided in the distribution. Other user-developed
programs and spreadsheets (Excel) can readily read the flat files. The warp3d2exii program reads
flat files and creates a single file with the model definition and results using the Exodus II1 format
(a .exo file). The free and open-source ParaView2 program reads the Exodus II files and provides
outstanding capabilities for graphical post-processing of results. The Exodus II format is widely used
by other post-processing codes, e.g., Visit3 from LLNL.
With this release, warp3d2exii has been converted from Python 2.7.3 to be fully compatible Python
3.5.x or newer. Python 2.7.3 is no longer supported.
• The Quickstart file for each platform has been updated and improved.
• Usual bug fixes and minor improvements.
Manual Sections updated: 2.2 (Material definition), 2.9 (Initial Conditions), 2.16 (User Named Lists),
3.5 (bilinear material model, 3.6 mises material model, 3.7 gurson material model, Appendix I (User
Routines). All Quickstart documents. Workflow documents: Workflow run WARP3D verification, Work-
flow WARP3D to Paraview

Version 18.2.0 (February 13, 2020)

• Crack initiation/growth by the smcs framework with element deletion. Computational procedures
are revised that introduce three types (criteria) for the critical plastic strain with a dependence
on stress triaxiality and Lode angle parameters. To accommodate (approximately) non-proportional
loading histories, the critical plastic strain computation uses a plastic-strain weighted value of the
triaxiality and Lode angle parameters. The printed output for smcs growth in each load(time) step
1 The Exodus II format specification is part of the SECAS system developed/supported by Sandia national Laboratory
2 paraview.org
3 wci.llnl.gov/simulation/computer-codes/visit/

Revision History RH-iii Updated: April 23, 2022 at 7:55am

 includes this triaxiality and a plastic-strain weighted value of the Lode angle parameter for reference.
See Chapter 5 on Crack Growth Procedures for details.
• New commands to support crack growth by element deletion are now available to (1) print a list of
killable elements ranked by their current nearness to reaching the element deletion criterion, (2) a
command to stop the solution with if the total number of killed (deleted) elements reaches a specified
number (prevents runaway simulations). This option writes a restart file.
• The element deletion processor using smcs can now output element states files of key values after
specified load(time) steps. Values include the accumulated plastic strain, current estimate of the
critical plastic strain, current mean and mises stress, and plastic-strain weighted values of the tri-
axiality and Lode parameters. These states files have simple text or stream (binary) storage, have
a simple flat format, and follow the same construction and naming scheme as the flat states files
described in Chapter 2 section on Output. These files are particularly useful in post-processing (e.g.,
via ParaView) to understand the evolution of crack initiation/growth.
• All of Chapter 5 on Crack Growth Procedures has been updated with Sections 5.1 and 5.2 rewritten.
• New verification tests are included in the usual locations of the distribution to exercise the revised
smcs criteria.
• Binary packets output for types 22, 23 to support element deletion by the smcs criterion are updated
to include additional terms. See revised Appendix F and the example packet reader.f code included
in the distribution.
• The manual section for the deformation material model has been re-written to clarify features of the
computations and definition of the scalar plastic strain value included in stress output.
• A new option to define simple multipoint constraints is now available: rigid. This enforces equality
of displacements u, v, w at two nodes.
• The release constraints command is extended to support removal of multipoint constraints that
enforce equality of displacement components at two nodes.
• Certain tractions applied to crack faces may not contribute to the interaction integral I(s) for
computation of the T -stress. The new Section 4.5.6.4 discusses the traction integral. New Section
4.5.6.5 describes the use of superposition to compute T -stresses using solutions for the uncracked
body and tractions applied to the crack faces. New verification test cases are included for computation
of T -stresses.
• Additional verification test cases are now include for the PPR option of the cohesive material model.
• Prior versions had built-in limits on the number terms that could be stored in lists of integers, e.g.,
<integer list>. Models with millions of nodes/elements have occasionally encountered this limitation.
All integer lists are now dynamically sized to required number of terms, thereby removing the prior
limitations.
• Users of WARP3D frequently modify the source code and rebuild the executables. With release 18.0
and newer, GNU gfortran may now be used to rebuild executables on Linux and with experimental
support on Mac OSX. Run-time performance decreases with gfortran but not to unacceptable levels.
The necessary, freely available files from the Intel MKL library (e.g., Pardiso solver) for Linux and
Mac OSX are included in the WARP3D distribution to support rebuilding with gfortran. The Intel
Fortran compiler remains required to rebuild on Windows and is the strongly recommended build
system for Linux and Mac OSX. The support of two compiler systems required a significant number
of mostly cosmetic changes in the source code to resolve incompatibilities and in the scripts/makefiles
to drive the rebuilding process.
• Intel has changed the procedures to set the number of threads used in the Pardiso solver and other
MKL routines (LAPACK, LINPACK, ...). WARP3D now explicitly sets the number of threads for
Pardiso with a call to an MKL utility routine. The default number of threads is the same as set with
the OMP NUM THREADS environment variable. The number of threads just for Pardiso may now be set
with the new solver threads ... command described in Section 2.10, 7.2, 7.3.

Revision History RH-iv Updated: April 23, 2022 at 7:55am

 • The symmetric version of the Pardiso direct solver in the Intel MKL library detects the occurrence
of negative eigenvalues during factorization indicating an indefinite set of linearized equations in
the global Newton’s method. WARP3D now outputs a count of the negative eigenvalues when they
occur. Section 2.10 on the Consistent [Q] Matrix discusses possible reasons for the development of
negative eigenvalues.
Manual Sections updated: 2.7 (Constraints), 2.10 (Nonlinear Solution Parameter), 3.4 (Deformation
material model), 3.8 (Cohesive material model), 3.10 (Cyclic plasticity model - typos fixed in some
equations) 4.5 (Interaction Integrals), Chapter 5 (Crack Growth Procedures), 7.2-7.3 on setup for parallel
execution, Appendix F (packets output file).

Version 17.9.3 (December 2, 2018)

• No fixed limits on model sizes. The few remaining data arrays with built-in, fixed sizes have been
removed from the code on all platforms. All previously fixed-size data structures are now dynamically
allocated and resized during execution as needed, including the number of nodes, elements, blocks,
load steps, etc. This removes the need some users faced for multiple executables of WARP3D having
different limits on model nodes and elements. The memory footprint at start of execution is now less
than 10 MB. Build 3500 on all platforms. Models with millions of nodes and elements are routinely
processed with the code.
• The domain integral procedures are verified for simulations that employ the crystal plasticity material
models.
Manual Sections updated: 2.1 (Model sizes).

Version 17.9.2 (September 18, 2018)

New Features
• Domain integrals for fracture mechanics. The entire code base has been re-designed and implemented
to include an initial state framework required to obtain path independent J-values in the presence
of (1) user-defined residual (initial) stresses, (2) eigenstrains, and/or (3) residual stresses/inelastic
strains from thermo-plastic deformation during load(time) steps in the analysis prior to introduc-
tion of cracks in the model. The new capabilities include full support for finite deformations, func-
tionally graded materials, anisotropic thermal expansion coefficients, all inertia contributions to
the J-integrals, crack-face tractions, thermal effects. J-integral computations are performed using
thread-parallel and thread+MPI parallel.
• Chapter 4 (Domain Integrals) has been re-written and much expanded to include extensive, first
principle discussions using finite deformations throughout of: local energy flux into the crack tip
in a 2D setting; the somewhat subtle transition to energy release rates in 2D for non-steady and
quasi-static advance of the crack tip; Rice’s J-integral in a 2D setting and the limitations imposed
for path independence; local energy flux into a 3D crack front and the transition to an energy
release rate; extensions of Rice’s J-integral for 3D (the area contribution); concepts of the domain
integral in 2D/3D for numerical computation of the energy release rate in a finite element framework;
incorporation of thermal loading, inhomogeneous material properties, stresses and inelastic strains
present prior to introduction of cracks; details of various numerical schemes required to evaluate the
integrals.
• Initial-state framework for J computations described in Chapter 2, section on Solution Parameters.
The user indicates before the solution begins for load (time) step one that an initial state will be
declared at some time during a simulation that runs over potentially thousands of load(time) steps
and multiple restarts before a crack is introduced into the model. When the simulation evolves to
a point/time where the initial state is to be defined for J computations, the load step number for
the initial state is given under the Solution Parameters command (see details in Chapter 2). In the
simplest case, the user-defines residual stresses at time zero using the initial conditions command.

Revision History RH-v Updated: April 23, 2022 at 7:55am

 The initial-state step is declared to be step one that represents a so-called equilibrium step, followed
by constraint releases to introduce one or more cracks.
• Initial stresses in elements may now be specified as model Initial Conditions to incorporate welding
residual stresses, for example, and other effects that generate equilibrium or non-equilibrium initial
stresses at tn=0 .
• The new Appendix K provides a review of methods to model residual strains-stresses in WARP3D
and to compute J-values.
• A parallel assembly process is available for use with the hypre solver that executes with MPI. The
hypre solver is suitable for very large models containing millions of nodes. This option distributes
the work (and memory) of equation assembly across MPI ranks.
Minor Updates/Fixes
• J-integral verification. A comprehensive suite of automated testing/verification problems for the J-
integral capabilities is now included in the verification directory. The same setup is followed for the
other, existing verification suites. A Bash shell script executes one or more simulations sets at the
user’s direction and compares results to known good values.
• The fixed maximum limit on the number of definable load (time) steps has been removed. Related
data structures are resized as needed.
• The proximity option in the constraints command to locate nodes using geometric tolerances now
works properly. Optional words are now included in constraints commands for readability.
• Data for nodal coordinates and element incidences may now be entered using sequential stream and
(Fortran) binary files rather than the usual text-based input form. This new feature is especially
useful for very large models where the parsing of text-based values can require significant processing
time, is executed in serial mode, and only on rank 0 (for MPI-based execution).
• The keyword nlgeom may now be used to request the finite displacement, finite strain element formu-
lations in the analysis (has the same effect as use of the keyword in Abaqus). Element formulations
default to linear small-displacement, small-strain formulations; the keyword nonlinear continues to
be synonym for nlgeom.
• Many formatting issues fixed for node and element output in models with more than 1M nodes-
elements.
• The executables for WARP3D included in the distribution now require Intel/AMD processors that
support the AVX instruction set (for vector operations). These processors first became available in
2011-2012, with even more capable processors now found in all but the oldest machines. It is our
intent next year to require processors that support the AVX2 instruction set (more vector registers).
• Recent testing of WARP3D at the Ohio Supercomputer Center showed excellent performance using
64 compute nodes with 1328 cores for models with several millions nodes and elements. Test cases
employed both Intel Cluster Pardiso (direct) solver and the hypre iterative solver
Manual Sections updated: Table of Contents, 1.6 (include initial stresses in formulation), 2.3 (Ele-
ment Types and Properties), 2.4 (coordinates), 2.5 (incidences), 2.7 (constraints), 2.9 (commands to
define initial conditions for model), 2.10 (solution parameters – command to define initial-state for
J computations), all of Chapter 4 (Domain Integrals). New Appendix K on step-by-step methods to
model residual strains-stresses and to compute JK-values.

Version 17.8.4 (September 12, 2017)

• New link2 element provides orthogonal, linear-elastic springs connecting two nodes that maybe
initially coincident. When combined with existing multi-point constraints capabilities, the most gen-
eral periodic boundary conditions may be imposed, for example, on representative volume elements
(RVEs). Also useful to model non-rigid supports, elastic foundations and to define additional point
masses in models. (manual section 3.15).

Revision History RH-vi Updated: April 23, 2022 at 7:55am

 • New bar2 element provides an axial force-only behavior, i.e., a spacetruss. Material behavior is linear-
elastic. Small and large-displacement (strains and rotations) formulations are available. (manual
section 3.14).
• New options are available for rigid contact with cylinders and spheres (manual sections 6.2, 6.3).
Contact conditions may now be set for inside/outside of the rigid surfaces.
Manual Sections updated: Table of Contents, 2.1 (Model sizes), 2.7 (Constraints), 2.10 (Solution Pa-
rameters), 3.14 (new bar2 element), 3.15 (new link2 element), re-write of Appendix C (Translation of
Patran-style Neutral Files to WARP3D).

Version 17.8.2 (June 10, 2017)

• The limit on maximum number of model nodes has been removed.
• The crystal plasticity (CP) model now works with MPI + threads execution.
• The memory footprint at WARP3D startup is now much smaller – arrays based on number of model
nodes are now dynamically allocated.

Version 17.8.0 (April 20, 2017)

• Extensive new capabilities have been implemented in the crystal plasticity (CP) material model.
Further, much of CP source code has been re-written/re-structured to dramatically improve computa-
tional efficiency that becomes especially important for large models with multiple crystals/orientations
at integration points. The new modeling capabilities provide:
(1) transversely isotropic elastic behavior,
(2) hcp6 and hcp18 slip systems (in addition to the existing fcc, bcc, bcc48 and single
systems),
(3) the Ma-Roters-Raabe dislocation density based constitutive model for fcc crystals
which adopts a homogeneous dislocation structure and that tracks the temperature de-
pendent dislocation evolution on each slip system, and
(4) the Deka, et al. rate-dependent (temperature independent) hardening model for α
phase Ti polycrystalline alloys that accommodates anisotropy in material behavior.
(5) simple methods to model Norton power-law creep with Voce hardening model
• Use of MPI+threads for solutions has been made markedly simpler for most analyses. (1) The
Cluster Pardiso solver (CPardiso) of the Intel MKL library is now accessible in WARP3D. This is a
direct solver for symmetric and asymmetric systems that leverages the additional resources (cores,
memory, memory subsystems) available in compute clusters and supercomputers; (2) WARP3D will
now automatically assign element blocks to domains in a scheme suitable for use with CPardiso
(greatly simplifies exploring combinations of MPI ranks and threads per rank).
CPardiso is designed for the very largest models with potentially millions of nodes. It is a direct,
sparse solver and is especially efficient for quasi-static loading. The iterative hype solver remains avail-
able for extremely large numbers of nodes and is more efficient for high-rate, impact-type loadings
where the small time steps greatly improve convergence of the conjugate gradient iterations.
• The MPI+threads solutions generate partial results files (on each MPI rank) that must first be
combined before creating an Exodus II (.exo) file for ParaView or for creating Patran-compatible
results files used by a number of post-processing codes. An entirely new code has been created to
combine the partial files and to add support for state variable output in the MPI execution. The new
code is located in the directory combine mpi results files of the WARP3D distribution. See Appendix
E.
• New problems for both threads-only and MPI+threads versions of WARP3D are included in the
growing set of verification tests located in the verification subdirectory of the distribution. Interactive

Revision History RH-vii Updated: April 23, 2022 at 7:55am

 shell scripts support running the various test problems and then compare results to known good
solution values.
• Chapter 7 of the User Manual on Parallel Execution has been expanded and sections re-written to
reflect the current concepts of how WARP3D executes in parallel.
• All computationally intensive parts of the code have been updated to leverage the newest vectoriza-
tion capabilities of the Intel compilers and emerging processors with many cores. Some algorithms
have been redesigned for improved performance based on guidance from the code vectorization ef-
forts. In nearly all model tested, there is a notable performance improvement.
This and the CPardiso effort have been supported by the Intel Parallel Computing Center awarded
to the Ohio Supercomputer Center, which focuses on WARP3D.
• Processes to build the WARP3D executable and support programs (patwarp,...) are updated to
Version 17.0.2 of the Intel ifort compiler and software on all three platforms (Linux, Windows,
MacOS). Intel MPI 17.0.2 is used on Linux. Adoption of the newest compilers and toolset is especially
important access the newest performance improvements in Pardiso and CPardiso.
Manual Sections updated: Table of Contents, 1.7 (Linear Equation Solvers), 2.6 (Element Blocking and
Domain Assignments), 2.10 (CPardiso Solver), 3.12 (Crystal Plasticity model), Appendix C (Patran →
WARP3D Translator patwarp, Appendix E (Combing partial MPI Results Files), Appendix H (Sup-
ported Platforms), Appendix I (UMAT and USER Routines), Appendix K (Python Processing of Flat
Model and Flat Result Files).

Version 17.7.4 (September 27, 2016)

• A line search procedure is now available to augment global Newton iterations for simulations with
difficult convergence. This implementation uses a backtracking algorithm to compute a suitable step
length for the corrective displacement vector. We now recommend these options for routine nonlinear
solutions: extrapolation on, line search on, adaptive solution on.
• Chapter 6 on contact analysis procedures has been updated with additional descriptions and recom-
mended procedures to perform contact simulations.
• The solution option linear stiffness on iteration one has been deprecated and replaced by actions of
the (displacement) extrapolation on/off.
• New creep material model. The initial implementation of a creep model is now included in WARP3D.
The model name is creep and is described in Section 3.13. This is a classic Norton-Bailey power-law
model generally applied to characterize secondary creep.
• The WARP3D interactions with UMAT routines (when they are called during the nonlinear solution)
now reflect more closely how Abaqus calls the UMAT for increment=1 of each *STEP and in each
subsequent increment for the *STEP. Updated documentation and code fragments describe the
improvements. The procedures to handle thermal effects, initial strain (e.g., creep) and possibly
initial stresses are much simplified. Existing UMAT routines written to be executed by Abaqus
require very minimal changes to work properly with WARP3D.
• A large-scale overhaul of the internal strategy for the sequencing of computations in global Newton
iterations has substantially simplified the code, eliminated a noticeable fraction of the source code,
and decreased run times. The number of global Newton iterations to reach a converged solution is
often reduced. These modifications are transparent to the user except for the changes made with the
extrapolation on/off nonlinear parameters (we recommend using extrapolation on as the default).
• The crystal plasticity model (CP) has been extended to include a new slip-system type designated
bcc48 {110}<111>, {112}<111>, and {123}<111> [48 slip systems]. Some internal algorithms in
the CP model for stress updating have been revamped to improve computational efficiency and
robustness. The output states capability has been significantly extended to provide ready access
to sets of internal variables of the CP computations, e.g., slip rates on each system, total slip on
each system, hardening variables, rotation tensors, lattice strains, etc. Professor Tim Truster at the

Revision History RH-viii Updated: April 23, 2022 at 7:55am

 University of Tennessee-Knoxville is actively working to extend capabilities of the CP model.
• The UEXTERNALDB routine for Abaqus compatibility has extensions to support inclusion of user
routine data in the WARP3D restart file.
Manual Sections updated: Table of Contents, 1.6.4 (line search, displacement extrapolation), 2.10 (line
search, extrapolation), 3.12 (CP updated for new bcc48 slip systems), 3.13 (new), Chapter 6 (contact)
all sections, Appendix I (all sections).

Revision History RH-ix Updated: April 23, 2022 at 7:55am