Proposal to develop a postprocessor for design of reinforced concrete shell structures.

This proposal is based on the work we are doing on the White Rose CGS.

I would be happy to share my insight which is based on working with such tools developed in Norway the last 30 years. In addition I have been instrumental in the newest development of OO’s postprocessor, which is capable of doing simple nonlinear analysis by feeding back the stiffness to the FEM software and doing interactions.

The development should be made in several steps:

1. A stand-alone program with user input only. This have been developed on the White Rose project and is called ShellD. ShellD is written in Mathlab. The first step of a development would be to write the same code, or a code doing the same, using a more appropriate platform. I have done a Fortran code doing the same as ShellD. The speed is in the order of 10 to 40 times faster than the Matlab program. The post-processing is a heavy computational effort, so the speed is important. My recommendation is that the post-processor is written on a platform fit for heavy computations, such as C++ or Fortran.
   o For offshore concrete project the postprocessor should be able to do calculations for:
    ULS
    SLS
    ALS
    FLS
   o It would be appropriate to investigate the needs of other sectors in Arup and have a discussion on what type of calculations the post-processor should be capable of doing.


2. Link the post-processor to one or more FEM packages. This has proven to increase the efficiency of the design process.
   
   o The FEM packages should be selected. GSA and LS-Dyna are obvious candidates. Sesam is probably a good option for offshore application:
    GSA is well known by most structural engineers in Arup. However, the question is whether GSA is powerfull enough for large structures, such as the White Rose CGS. We have experienced issues with GSA on the project. This seems to be related to the drilling DOF of shell elements in GSA. Further, the preprocessor (modelling and load application) in GSA is not effective, especially of curved shells. Further, GSA do not have solid elements.
   • Well known in Arup
   • Technically “weak”, has limitations (pore performance of the shell elements)
   • Pore post-processor
   • Do not have solid elements
    LS-Dyna is an advanced FEM package. It could be convenient to have a link to the postprocessor.
   • Well known in Arup
   • Technically strong/proven
   • Probably to advance for the regular structural engineer
   • Have its own non-linear capabilities
    Sesam is not known to many engineers in Arup. However, it is well suited for offshore structures as it has a relatively good pre-processor and are well proven on large FEA with many DOFs. Further, it has most of the capabilities one need to design an offshore structure, such as wave loading, stochastic wave calculations, floating stability, etc.
   • Not well known in Arup
   • Technically proven
   • Good pre-processor
   • Well suited for offshore structure
   • Not aimed at on-shore structures
   • Do have solid elements (have a good choice of element types)
   • Expensive
    Based on the points above it may be the best option to make the necessary improvements to GSA?
   o Development of a link between the post-processor and the FEA:
    Make a link between a grid of design points in the post-processor and the FEM. We should have an internal discussion how this link should be. The “Norwegian” method has been to make a grid out of each Gaussian point in the FEM. There are 4 Gaussian points in a 8-nodes shell element. The reason for using the Gaussian point is that the FEA method tends to predict the stresses most accurate in the Gaussian points. I could be merits in developing an alternative method:
   • Gaussian Pont method
   o Well proven
   o Does require a regular mesh. It can be discussed whether this is good or bad as a regular mesh gives better stress predictions but sometimes the geometry does not allow a regular mesh.
   o Require all input to be changes if the mesh I changed
   o Extra so-called extrapolation points has to be defined on the edge of a structural element
   • Alternative method
   o A grid is defined independently of the element mesh
   o A general routine should be developed, to interpolate/extrapolate stress resultants in any grid points
   o The postprocessor is independent of the element mesh. This means that the input files do not change if the element mesh have to change. This would be a big improvement.
    Definition of materials
   • Concrete
   • Reinforcing steel
   • Post-tensioning steel
    Definition of reinforcement in the design points
   • Reinforcement types
   o Longitudinal reinforcement
   o Transverse reinforcement (shear links)
   o Post-tensioning reinforcement
   • Definition of reinforcement types for the design points, individually or areas of design points
   • A table of reinforcement can be evaluated for the design points, individual or by area. This means the post-processor is starting with minimum reinforcement and are increasing the reinforcement according to a user defined table of reinforcement types until a utilization of 1.0 or less is reach.
    Load combination feature
   • It is recommended to develop a load combination feature. A hierarchic system of load combinations should be developed.
   • Develop a method of combining load actions and load reactions (soil reactions) so the reactions is in equilibrium with the actions. This has proven effective on large offshore structures
   • Load combination with complex wave loads (Real and imaginary load case), making it possible to step through a number of wave phases and/or to define single wave load phases.
   • Define temperature and deformation loads
    Define code checks
   • ULS design load cases
   • ALS design load cases
   • SLS design load cases
   • FLS design load cases
    Develop presentation of results
   • Tabulated results
   • 3D plots (2D plots?)


3. Develop nonlinear analysis based on stiffnesses calculated by the post-processor
   o The post-processor should be able to start/stop the FEA
   o Calculate the stiffness properties to feed back to each individual element in the FEA
   o Feedback the stiffnesses into the FEM
   o Start the FEA with updated stiffnesses
   o Read results from the updated FEA and evaluate convergence criteria
   o If necessary update the stiffness again. The stiffness calculation should be based on the strain in the FEA (one of the main lesson learnd)
   o Repeat the interactions until the convergence criteria is met, or the maximum number of iterations are made.