DARWIN® 9.2 released for production use in June 2019
New Fatigue Crack Growth Models
In previous versions of DARWIN, treatment of small cracks was limited to the NASGRO fatigue crack growth model. The NASGRO model uses a small crack correction parameter to account for small crack effects on the driving force of fatigue crack growth analyses.
DARWIN 9.2 includes a new El Haddad small crack correction parameter to account for small crack effects. It can be applied to all DARWIN fatigue crack growth models except the NASGRO model, which already provides treatment for small cracks. This is an optional feature that is independent of temperature or stress ratio.
DARWIN 9.2 also includes a new Dhondt cyclic crack growth model (a modified version of the DARWIN Walker crack growth model) that includes dependency of threshold on stress ratio. The GUI provides a display of Dhondt model parameters and a plot of the associated crack growth rate curves.
New Surface Autozoning Capability
In previous versions of DARWIN, autozoning was not supported for risk assessment of surface damage anomalies. Manual zoning was the only option for this anomaly type. In DARWIN 9.2, the optimal autozoning framework that was previously developed for risk assessment of inherent anomalies was extended to support surface damage anomalies. It enables users to assign properties (e.g., anomaly distributions, materials, inspections) to finite element surfaces and identifies the optimal placement of these surfaces into surface zones. Similar to the autozoning capability for inherent anomalies, this new feature provides life and risk contours as well as zone risk convergence values.
Speed & Robustness Improvements
The computational efficiency of many DARWIN analyses is directly related to the efficiency of the DARWIN autoplate algorithm. In previous versions of DARWIN, the autoplate algorithm was limited to use on a single computer processor even when multiple processors were available.
In Version 9.2, DARWIN was enhanced to perform autoplate computations in parallel utilizing multiple processors. Initial results indicate that the DARWIN 9.2 autoplate algorithm runs 10-20 times faster than the previous 9.1 version when life computations are executed in parallel mode using 4 processors. Autoplate was also enhanced to skip nodes of the finite element model at which FCG computations cannot be completed. This enhancement enables users to view skipped nodes (if any) after the analysis is complete, substantially improving the robustness of the computation.
DARWIN 9.2 was also enhanced to provide improved responsiveness for GUI node selection operations. Initial benchmark test results indicate that DARWIN 9.2 GUI response times are significantly faster than previous versions. In addition, the GUI was enhanced to perform internal computations directly using the HDF5 file format. This enhancement significantly improves GUI speed for large project files.
Direct Import of ANSYS and ABAQUS Load Step Times
In previous versions of DARWIN, load step times associated with imported ANSYS or ABAQUS FE results files were defined manually by the user. DARWIN 9.2 includes an enhancement to directly import load step time values contained in ANSYS and ABAQUS FE results files into DARWIN projects. When an FE results file containing load step times is imported into a time-dependent DARWIN project, the load step time input fields are populated automatically.
Zoneless Deterministic Analysis
DARWIN was originally designed to assess the fracture risk of components containing rare material anomalies. The original DARWIN GUI workflow (i.e., content and sequence of GUI menus and features) was intended to support the zone-based risk assessment methodology described in FAA Advisory Circulars 33.14-1 and 33.70-2. However, many analysts also use DARWIN to assess deterministic fatigue crack growth (FCG) life as described in AC 33.70-1. Previous versions of DARWIN enabled users to perform deterministic fatigue crack growth analysis, but the risk assessment-based GUI workflow required additional input that was not required for deterministic analysis. For example, users had to define a zone for each initial crack location in a deterministic FCG life analysis, but much of the zone information was not used in the analysis.
DARWIN 9.1 enables users to perform deterministic life assessments without zones. A new GUI workflow was developed specifically for deterministic assessments in which a new “life assessment” mode is defined in the first (configuration) GUI menu. This enables the GUI to display the analysis configuration settings that apply only to deterministic life assessment. When performing a deterministic life assessment, the user no longer needs to specify the analysis mode and the zone information. The GUI Optional Features menu was enhanced as well to display only the information that is applicable to deterministic life or probabilistic risk assessments.
A new “deterministic mode” option was added to the configuration menu that is available only for deterministic life assessments. It enables users to perform individual FCG analyses at user-specified locations (User-Defined Cracks option) and/or to construct a series of FCG life contours (Crack Growth Contours option). Users may select one or both deterministic mode options. The User-Defined Cracks option enables users to place cracks at multiple locations in an FE model and to perform a deterministic FCG life analysis at these locations. Users provide the crack location, crack type, crack growth plane, and fracture mechanics plate (via the Autoplate algorithm). The crack growth contours option enables users to define one or more crack growth contour models via a new “Crack Growth Contours” preprocessing screen. When this option is selected, a deterministic life assessment is performed for every node within the crack growth contour regions specified by the user. This option enables users to view fatigue crack growth life contours associated with specified initial crack sizes. It also provides contours for other fatigue crack growth properties such as Kmax, crack depth, and crack length at the critical crack state.
Critical Initial Crack Size
DARWIN 9.1 extends the new zoneless deterministic life capability to compute deterministic critical initial crack sizes at selected nodes. The critical initial crack size refers to the crack size above which a crack will grow to fracture at some point during its service life. For example, a critical initial crack size of 0.010” means that all cracks of 0.010” and larger will grow to fracture for the given service life (e.g., 20,000 missions), whereas cracks smaller than 0.010” will not grow to fracture for the given service life. The critical initial crack size varies throughout the component and depends on the loading, geometry, and material.
DARWIN 9.1. computes the critical initial crack size at selected nodes in the geometry using the same framework established for deterministic crack growth analyses. Users provide properties at these locations (e.g,. an initial crack shape) and the mission history of interest. Users also provide the service life for the component. DARWIN then computes the critical initial crack sizes at these locations. This information is shown as contours on the finite element geometry that enables users to locate hotspots.
Autoplate for 3D Finite Element Models
The DARWIN Autoplate algorithm identifies the size and orientation of fracture mechanics models (i.e., rectangular plates) based on the geometry, temperature, and stresses at specified locations in a finite element (FE) model. The DARWIN autozoning algorithms use the Autoplate algorithm to create zones. In previous versions of DARWIN, users could also invoke the Autoplate algorithm when manually creating zones via the DARWIN GUI, but this capability was limited to 2D FE models. For manual zone creation using 3D FE models, users were required to determine the size and orientation of fracture mechanics models using engineering judgment. Furthermore, for 3D models the initial crack locations were limited to nodes on the surfaces of FE models.
DARWIN 9.1 now enables users to create zones manually for 3D FE models using the Autoplate algorithm. When the user selects an initial crack location on the 3D FE model, the slice plane and fracture model are created simultaneously using a single mouse click based on the Autoplate algorithm. The GUI also enables users to specify the initial crack location anywhere on the surface of a 3D FE model. These enhancements have resulted in a common interface for manually creating zones in both 2D and 3D FE model geometries.
The Autoplate algorithm supports fracture models for surface, embedded, and corner crack types (SC30, EC05, and CC11, respectively). For all other crack types, the user must provide the fracture model parameters (i.e., fracture model dimensions and orientation). In previous versions of DARWIN, it was often difficult for users to determine whether a fracture model had been created by a user or via the Autoplate algorithm. This information was only provided in a table. In DARWIN 9.1, the GUI was enhanced to enable users to distinguish among models created by users and via Autoplate. Autoplate-defined plates are represented with a green border, and user-defined plates are represented with a white border.
Improved Support for Large Finite Element Models
Previous versions of DARWIN did not provide adequate support for large FE models with large numbers of load steps. The time required to import and display large FE models in the GUI could be measured in double digit minutes or even hours. The time required to process these FE models in the risk assessment code (RAC) was extensive, and often terminated because the memory required to process the analysis exceeded available computer random access memory. The memory limitation was due to the use of a text-based file format for FE models. DARWIN 9.1 introduces a new binary file format called HSIESTA to store finite element results data. HSIESTA replaces an earlier file format (SIESTA) used by DARWIN to store the same information. HSIESTA stores information in a binary format that is more easily accessible to DARWIN.
DARWIN 9.1 now reads, displays, and utilizes data from HSIESTA. These enhancements have significantly reduced the time required to read and display the stresses and temperatures associated with large FE models. For example, consider a finite element model with approximately 100,000 nodes and 1,000 load cases. In the previous version of DARWIN (Version 9.0), the GUI required nearly ten minutes to import and display the stresses and temperatures associated with a single load case. The time required to display the stresses and temperatures associated with another load case in the same model exceeded ten minutes and eventually timed out. Using the HSIESTA capability implemented DARWIN 9.1, the same FE model was imported and displayed in approximately ten seconds, or roughly 60 times faster than the previous DARWIN version that used SIESTA files. The GUI was also able to display the stresses and temperatures associated with other load cases in the file in roughly 8 seconds for each load case.
FE2NEU (DARWIN FE results file translator) was enhanced to translate FE models from commercial FE software (e.g., ANSYS, ABAQUS) to the HSIESTA format. It was further enhanced to convert legacy SIESTA-formatted files to the HSIESTA format. This will enable users to convert FE results from legacy files to the new HSIESTA format for use in DARWIN 9.1.
The DARWIN computational engine was also enhanced to read and process data from HSIESTA. The RAC was enhanced with new random access API functions that enable it to read data from specific regions of the HSIESTA file rather than importing the entire file into memory. Preliminary results indicate reduced memory usage for large finite element models when the number of zones is much smaller than the number of elements/nodes.
Optimal Gaussian Process Pre-zoning
DARWIN 8.2 expanded the exhaustive and optimal autozoning capabilities that were previously available for 2D axisymmetric geometries to 3D geometries with inherent anomalies. These techniques automatically build DARWIN zones (i.e., set material volume and locate cracks) based on user-defined property regions such as anomaly distributions, material parameters, and other factors. Autozoning leads to consistent results between users and reduces user intervention needed to define zones. For example, risk limiting locations defined by experienced and inexperienced users may differ drastically, leading to drastically differing risk values. Autozoning eliminates this issue by determining the expected risk limiting location without human intervention. However, the computational cost of autozoning increases dramatically for increasingly complex finite element models.
DARWIN 9.0 introduces a new optimal Gaussian Process (GP) pre-zoning capability to create DARWIN zones automatically and efficiently. Pre-zones are elements grouped by stress range, distance-to-surface, and temperature that have similar risk values. DARWIN 9.0 samples points within these pre-zones to build an approximate risk surface on local pre-zone domains using a response surface defined by training points. An iterative scheme determines the approximate risk limiting location within each pre-zone. DARWIN then employs the previously developed optimal autozoning methodology to determine the optimal zone break-up of the model. Pre-zones, instead of individual elements, define the minimum sizes for zones. For large models where the number of pre-zones is significantly smaller than the number of finite elements, the optimal pre-zoning capability can significantly reduce the computation time associated with risk assessment. Initial studies indicate that the new pre-zoning algorithm is up to three orders of magnitude faster than the previous optimal autozoning algorithm. The new pre-zoning algorithm is up to five orders of magnitude faster than the previous exhaustive algorithm. Furthermore, the pre-zoning method requires less memory than either the exhaustive or optimal methods. This feature enables the pre-zoning method to solve much larger models than either previous method. Note that the performance of the new algorithm may vary among different FE models.
3D Sector Models
Rotating engine components often have 3D geometric features that repeat cyclically around an axis. For example, a rotor might have eight “sectors” where the geometry remains identical between any two rays originating from the axis of rotation if these rays are separated by 45 degrees. Previously, DARWIN only supported full 3D models or axisymmetric geometries where the cross section was constant about the axis of rotation.
DARWIN 9.0 includes a new capability that enables users to import 3D sector models directly into DARWIN and to assess the life and fracture risk based on the complete component geometry. The finite element results file translator FE2NEU was enhanced to provide information regarding the number of sectors and the axis of rotation. When the sector model option is selected, DARWIN displays the original sector model and external boundaries based on cyclically repeating sectors. For life and risk assessments, cracks and associated fracture plates are based on the full model geometry. For example, a crack location on a sector boundary would be treated correctly as an embedded crack rather than a surface crack. This capability is available for both manually and automatically zoned models.
SIF Solution for Angled Corner Cracks
DARWIN 9.0 features a new stress intensity factor solution (CC18) for a quarter-elliptical surface crack centered on an angled corner. In a previous solution (CC12), DARWIN required cracks at chamfers to span the enter chamfer edge. Large chamfers required large cracks as a result. The new solution (CC18) is designated for cracks that originate at a clipped end of a 45˚ chamfered corner. CC18 is a bivariant SIF solution that supports stress gradients that vary strongly in more than one direction. The CC18 SIF solution is available for life assessment analyses using 2D axisymmetric models that are manually zoned.
The DARWIN GUI has been enhanced to support the new crack type. During zone definition, users can select the new crack type (CC18). For circular crack fronts (i.e., ), the crack is placed at the chamfer corner, and the GUI automatically positions the crack tips. Non-circular crack fronts (i.e., ) require more information on the position of the crack front from the user. The GUI positions an arrow in the direction of the ellipse’s major axis. The GUI includes a button that enables users to switch the chamfer leg where the long crack length is located. Similar to CC12, CC18 requires that the user input a chamfer dimension (as is already done for the previous chamfer solution, CC12).
Cracks defined using CC18 may transition to CC12 during the analysis. The transition occurs if the crack grows from one end of the chamfered corner to the opposite edge. A crack modeled with the CC18 SIF solution will transition to the CC12 SIF solution where the crack spans the entire length of the chamfered edge.
Autozoning for 3D components
Increased computational resources enable more complex and higher fidelity models of aerospace and power generation components. Compromises once made to reduce computational demands, such as 2D axisymmetric models where the cross section remains constant around an axis, are no longer necessary in the design process of turbine rotors. Analysis of inherent anomalies (e.g., hard-alpha particles) in earlier versions of DARWIN has been limited to 2D axisymmetric geometries. 3D geometries were only available for the analysis of surface damage.
New capabilities in DARWIN 8.2 support 3D analysis of life predictions and fracture risk assessments for general inherent anomalies. DARWIN 8.2 now enables 3D geometries imported from 3D finite element (FE) models to be analyzed with internal inherent anomalies in addition to the existing 3D surface damage analysis.
DARWIN 8.2 supports inherent anomalies in 3D geometries within the exhaustive autozoning and optimal autozoning frameworks. Graphical-user interface (GUI) enhancements now enable user definition of property region information (i.e., material response, anomaly distributions, inspection schedules, and mission regions) directly on the 3D FE model. Properties may be defined for the entire model, surface and/or internal elements only, or element-by-element. DARWIN employs this information to determine the 3D zones in the model based on using either exhaustive autozoning (one zone for each element) or optimal autozoning (zones are automatically sized to minimize risk based on user-defined restrictions, e.g., the probability of fracture)
Shop visit anomalies
Anomalies may sometimes be introduced on the surfaces of a component that is handled or inspected during routine maintenance. For example, an anomaly might not exist until cycle 5000 when it is introduced during an inspection. Introducing these anomalies at the beginning of the service life may result in an overestimate of the risk of fracture.
Previous versions of DARWIN have been limited to anomalies that are present before the part has entered service. In DARWIN 8.2, users have the option to define shop visit anomalies for improved risk assessments. This feature is available for surface damage investigations of 1D, 2D, and 3D geometries. A new GUI menu has been introduced that enables the user to specify the size distribution and timetables associated with shop visit anomalies. The timetable may be linked to the timetable of inspection schedules if desired. DARWIN tracks risk associated with shop visit anomalies and quantifies its influence on the overall probability of fracture.
Flight profiles in aerospace components are often described in terms of missions that consist of individual load steps. A simple mission could consist of a single large load step to simulate ascent, several smaller steps for maneuvers, and another load step for descent. Previous versions of DARWIN supported the definition of missions and their combinations using mission mix scenarios. DARWIN 8.2 enhances this optional feature with several new capabilities and renames it “Mission Mixing”. Users continue to define missions based on the load steps imported into DARWIN from FE analysis. Users satisfied with a single mission do not need to perform any additional tasks. However, other users will find the new mission mixing capability useful for complex loading profiles. Missions represent the first level grouping of load steps and can be grouped further into mission blocks. Mission blocks occur over some loading duration to define the entire flight profile.
Several options are available through the mission mixer to define the sequence of missions within a mission block. Duration times for missions and mission blocks may be described in terms of flight cycles, hours, and/or TACs. Separate mission blocks may have separate units, and DARWIN handles the conversion automatically. This mission sequence can defined deterministically based on a forward and reverse progression of missions. Alternatively, users can define a random mission sequence set using a mission seed. Users provide either the number of times that a mission repeats or set a mission mix percentage for the mission block.