# 4. Recap

In this workshop we added a geometry to an atmospheric boundary layer simulation, set up
its refinements and mesh, exported the pressure data needed to compute the $c_p$ of the
geometry, and finally processed that data in ParaView.

## What we did

```{eval-rst}
.. list-table::
   :class: ghost
   :widths: auto
   :header-rows: 1

   * - Block
     - What was done
   * - 1
     - Uploaded the building STL, checked its triangulation and attached it to a
       duplicate of the ABL case with the TBL wall model.
   * - 2
     - Added the buffer and body refinements, updated the simulation time from the
       building's convective time scale and set up the two pressure exports.
   * - 3
     - Downloaded the exports, computed :math:`c_p` from the body and reference
       pressures and visualized it in ParaView.
```

## The settings we ended with

```{eval-rst}
.. list-table::
   :class: ghost
   :widths: auto
   :header-rows: 1

   * - Item
     - Value
   * - Body
     - ``ar1_gable``, :math:`20 \times 20\,\text{m}` base,
       :math:`25.75\,\text{m}` at the ridge, TBL wall model
   * - Buffer refinement
     - target :math:`0.8\,\text{m}`, maximum :math:`1.2\,\text{m}`, resolving to
       :math:`0.63\,\text{m}` (level 4)
   * - Body refinement
     - target :math:`0.3\,\text{m}`, maximum :math:`0.5\,\text{m}`, resolving to
       :math:`0.31\,\text{m}` (level 5)
   * - Mesh
     - ~41.55M nodes
   * - Time window
     - :math:`-600\,\text{s}` to :math:`1300\,\text{s}` (500 CTS of acquisition at
       :math:`2.66\,\text{s}`)
   * - Exports
     - body pressure with :math:`0.62\,\text{m}` normal offset and a reference point at
       :math:`z = 150\,\text{m}`, both at :math:`0.1\,\text{s}`
   * - Runtime
     - ~3h13 at the default 500 MLUPS
```

## Things worth carrying to the next case

* **The body's triangulation is part of the setup.** It generates the IBM points and it
  carries the exported data, so its shape, size and normals decide both how the wall
  model behaves and the resolution of the results.
* **Refine in steps.** The buffer box between the ABL refinement and the body refinement
  is what keeps the transition smooth instead of jumping several levels at once.
* **The body sets the acquisition time.** Once there is a body of interest, the
  convective time scale is built on its smallest side, not on the fetch.
* **Both pressure exports must share the sampling period.** The $c_p$ is a per-instant
  subtraction, so a mismatch in timesteps makes the two files impossible to combine.
* **The probes have to sit outside the diffusive layer.** A normal offset of 2 cells is
  what separates a surface pressure from an IBM artefact.

## Where to go from here

The pressure time series we ended with is the input for the rest of the work: surface
force integration, structural loads and the comparison against code values.

```{seealso}
* {doc}`Pressure Coefficient Measurement <../../guidelines/pressure-coefficient-measurement>` -
  the $c_p$ definition and the script used here.
* {doc}`Structural Load from Simulation <../../guidelines/structural-load-from-simulation>` -
  turning the surface pressures into forces on the structure.
* {doc}`Geometry Representation with IBM <../../reference/body-discretization>` - how the
  triangles become Lagrangian points and what the diffusive layer around them is.
```