Data Structures and Workflows for ICME
29
is the storage of the parameter settings used in each function of the workflow. Since
the constructed workflow may be complex, and itself hierarchical, a sufficiently
flexible format must be used for serialization. Ideally, this format would also be
human-readable and should be standardized for broad access outside of materialsspecific tools.
4.3 Data Access and Metadata Labeling Requirements
After a successful execution of a workflow, the computed data should be serializable
into an accessible data format. Since the implemented data structure is highly
generic and flexible, the chosen data format must also be equally flexible. Similar to
the format for workflow serialization, the data format must also belong to an open
standard, allowing access outside of materials toolsets. Ideally, this format should
also be size and speed efficient, allowing for fast reading and writing, an enabler
for the large datasets inherent in many ICME workflows. Stored information may
be either heavy, the dense data comprising the bulk of the content in terms of size,
or light, consisting of metadata such as material name, component dimensions, data
type, time step, etc. The format must therefore be capable of storing either heavy
or light data. It is desirable to store the processing history of a set of data along
with the data itself; thus, the history remains innately coupled with the data, which
allows for reproducibility and transparency. Therefore, the data format must enable
the workflow to be stored alongside the data and additionally allow for further
functions and their parameter settings to be appended to the workflow should future
processing be necessary.
5 SIMPL and DREAM.3D: Enabling ICME Workflows
To satisfy the above requirements for an ICME workflow tool, the Air Force
Research Laboratory in partnership with BlueQuartz Software developed a software
framework known as the Digital Representation Environment for the Analysis of
Microstructure in 3D (DREAM.3D). DREAM.3D is an open-source software tool
explicitly designed to enable the creation of generic materials analytics workflows
that are adaptable to any kind of input, regardless of geometric topology or
data type [33]. Specific capabilities include 2D and 3D EBSD reconstruction and
analysis, n-D image processing, feature identification and quantification, surface
meshing, texture analysis, and synthetic microstructure generation. DREAM.3D’s
unique capabilities stem from an underlying data structure and management library:
The Spatial Information Management Protocol Library (SIMPL). SIMPL is an
open-source C++ library that implements an abstract data structure, including a
well-defined application programming interface [50]. Additionally, SIMPL defines
a functional interface for interacting with the data structure. This interface is
29
is the storage of the parameter settings used in each function of the workflow. Since
the constructed workflow may be complex, and itself hierarchical, a sufficiently
flexible format must be used for serialization. Ideally, this format would also be
human-readable and should be standardized for broad access outside of materialsspecific tools.
4.3 Data Access and Metadata Labeling Requirements
After a successful execution of a workflow, the computed data should be serializable
into an accessible data format. Since the implemented data structure is highly
generic and flexible, the chosen data format must also be equally flexible. Similar to
the format for workflow serialization, the data format must also belong to an open
standard, allowing access outside of materials toolsets. Ideally, this format should
also be size and speed efficient, allowing for fast reading and writing, an enabler
for the large datasets inherent in many ICME workflows. Stored information may
be either heavy, the dense data comprising the bulk of the content in terms of size,
or light, consisting of metadata such as material name, component dimensions, data
type, time step, etc. The format must therefore be capable of storing either heavy
or light data. It is desirable to store the processing history of a set of data along
with the data itself; thus, the history remains innately coupled with the data, which
allows for reproducibility and transparency. Therefore, the data format must enable
the workflow to be stored alongside the data and additionally allow for further
functions and their parameter settings to be appended to the workflow should future
processing be necessary.
5 SIMPL and DREAM.3D: Enabling ICME Workflows
To satisfy the above requirements for an ICME workflow tool, the Air Force
Research Laboratory in partnership with BlueQuartz Software developed a software
framework known as the Digital Representation Environment for the Analysis of
Microstructure in 3D (DREAM.3D). DREAM.3D is an open-source software tool
explicitly designed to enable the creation of generic materials analytics workflows
that are adaptable to any kind of input, regardless of geometric topology or
data type [33]. Specific capabilities include 2D and 3D EBSD reconstruction and
analysis, n-D image processing, feature identification and quantification, surface
meshing, texture analysis, and synthetic microstructure generation. DREAM.3D’s
unique capabilities stem from an underlying data structure and management library:
The Spatial Information Management Protocol Library (SIMPL). SIMPL is an
open-source C++ library that implements an abstract data structure, including a
well-defined application programming interface [50]. Additionally, SIMPL defines
a functional interface for interacting with the data structure. This interface is