Workflow to determine Impington Mill's biotope
Workflow to determine Impington Mill's
biotope by Victor Reijs
is licensed under CC BY-NC-SA 4.0
Introduction
On this webpage the wind biotope of Impington Mill is being evaluated.
First the making of the
CAD-model is described and after that the results
of simulations.
At the end an
evaluation is given.
Making a CAD-model with SketchUp
SketchUp CAD-program is used to make
the CAD-model. SketchUp Free version
is well equipped for the windmill environment. In this program one
can easily make 3D forms, such as: boxes, roofed boxes, cylinders,
capped cones, trees, etc, etc.
Buildings that are around a windmill, say with a radius of 250m,
can be simulated by a box (with its height: the wall height plus
half the roof height); or otherwise a more precise model (box with
hipped roof, etc.). A windmill can be a simple standing
cylinder/capped cone. Don't go into too much details as that will
not have much influence on the CFD results.
Furthermore, buildings below the ground/bailey height are
certainly not in need of precise dimensions. As a guideline
include anything that is higher than the DHM-rule
minus 2m.
The following steps for making the CAD-model can be seen as a
guide line:
- Define in SketchUp the unit of the dimensions (CFD uses
m[etres], so do the same in SketchUp): Model Info (i) →
Length Units → Meter; Model Info (i) →
Display precission → 0.00 m; and Model Info
(i) → Length Snapping → OFF
- One can add a piece map (500 by 500m with the mill in the
middle) from of Google Map (so when get an idea where buildings
are), use: ≡ → Add Location → Select Region →
Import
- Having imported this, one can now trace the contours of the
building-plans (Line-tool). Make sure you have zoomed in enough
as that is easier to trace. Make also sure that your buildings
all go down to a height of 0m. One can also disable Linear
inferences (OFF by using the Alt-button).
There is no real need to be very accurate (say within 0.25m).
- After tracing the contours of the building plans, one can
increase their height by extruding them (Push-Pull tool). At the
right bottom of the window (Measurements) one can see the
pull height.
- For UK there is the Defra Data Service which provides a
DSM (Digital Surface Model) for 2019 and 2022 (which uses British National Grid
and see also OS Maps).
- The GeoTIFF (DSM) file has been changed into a csv file by
using a small program in R. This csv file has been imported in
Excel for presentation and easy viewing.
- Here is the 2022 DSM (Digital Surface Model: green lowest;
red highest) for an area of 500*500m with Impington Mill in
the middle (top is North; bottom is South; right is East; and
left is West):
- Using the DHM-rule
on this 2022 DSM, one gets the following picture (top is
North; bottom is South; right is East; and left is West):
The colouring gives a height referenced
from the DHM-rule: green from -2 to 0m; orange from 0 to
4m; brown from 4 to 8m; red from 8 to 12m; black: above
12m
- For simplicity one simple use a box for a roofed
building: with its height: height of walls + half of height of
roof.
There is no real need to be very accurate (say within 0.25m).
Remark: Some other people
recommend to use the ridge height of the roof. Need to
investigate what is better.
- The mill is made by using a circle (Circle-tool), pull up
(Pull tool) to form a cylinder (or capped cone): height of cap
is 16.2m, and windshaft height is around 15m.
- Make sure an object does not end very close to 0m; otherwise 'mesh'
problems can happen during the simulation.
- Beside all buildings within a radius of 100m from the mill;
the buildings that have at least a green colouring (so which are
higher than DHM-rule minus 2m) are included in the CAD-model.
- Here is a bird eye's view of the resulting CAD-model (without
trees) (top is North; bottom is South; right is East; and left
is West):
Simulation without trees
using SIMSCALE
- General guidelines how to use CFD in an urban environment can
be found in this guideline
[Franke, 2007].
- As much as possible default SIMSCALE options have been used.
- Open space on: the left side (25H with H=5m); right side (10H
with H=10m); inflow side (7H with H=10m) and outflow side (10H
with H=10m).
- Turbulence model: k-omega SST and residuals < 0.01
- This model (without trees) has been simulated in SIMSCALE (nl)
using an ABL for an Eastern wind of 7.1m/sec (4.6Bft) @ 10m
above ground level mill and an z0 = 1m.
CAD-model including the trees
- Including trees in SketchUp Free is not easy if using Google
Map, as one can't see the trees in Street map. A better way is
to Import the DSM height Image, which includes buildings and
trees.
- The trees are including by using the stacked-cylinder object.
- Starting with a standard tree (15.3m high and 15.3m diameter)
in SketchUp Free.
Scale this tree by using:
Scale → Uniform Scale About Opposite Point →
enter xyz scaling
Scale → Blue Scale About Opposite Point →
enter z scaling
- At first, only trees that will affect the winds between South
and West are included:
- De CAD model changes for every wind sector (steps of 15deg) in
SIMSCALE CAD mode:
- Rotated with a certain value (Transform →→
Rotate Z-axis → Rotation angle →
<value> → right click → Assign All →
Apply)
- Flow volume → input X min, X max, Y min, Y max, Z
min, Z max → Excluded parts → Select trees →
Apply
- Flow region → Hide → Select Trees →
Invert selection → Delete → Apply →
Flow region Unhide → Save as Copy
Simulation with trees using SIMSCALE
- General guidelines how to use CFD in an urban environment can
be found in this guideline
[Franke, 2007].
- As much as possible default SIMSCALE options have been used.
- Open space on: the left side (25H with H=5m); right side (10H
with H=10m); inflow side (7H with H=10m) and outflow side (10H
with H=10m).
- Turbulence model: at start k-omega SST and at the end Realized
k-epsilon and residuals < 0.01
- In SIMSCALE we use a Forchheimer
coefficient of f=0.45 [1/m] for leafed/summer trees (using
the Darcy-Forchheimer
medium).
- Simulations were done for wind directions from 180 to 270deg
in steps of 15deg using the project: DSM2022 Impington.
- Input ABL wind speed is 7.1m/sec @ 10m and z0=0.5.
A lower z0 has been used (compared to earlier this
page) as the wind speed at the mill is compared to wind speed at
the Mildenhall (which is in quite open terrain).
- When measuring the calculated speed at 75%
of the sail span (75% of 9m) in a vertical sail-plane in
front of the mill body (2.7m from the centre of the mill body)
with wind shaft at 15m height, the following wind speeds
(relative to Mildenhall ABL wind speeds: u(ABL)) were
calculated:
- When repeating the measurement of the calculated wind
velocities a standard deviation of some 0.2m/sec is found.
- When a sail is at its lowest point (bottom, around 8.5m),
the lowest wind speeds were calculated (due to upwind effect
of the mill body).
- The wind speeds at the highest point (top, around 21.5m) are
for most wind direction close to the Mildenhall ABL wind speed
for such heights.
- The wind speeds at the low locations (over2,
bottom, over3) is effected by buildings and trees
in the neighbourhood.
- For Impington Mill: Trees (up to 19m) have the most effect
compared to buildings (up to 8m).
- The influence of the terrain (at 15m height) within100m
upwind is as follows for each azimuth:
180
|
195
|
210
|
225
|
240
|
255
|
270
|
open
|
nearby
tree
|
open
|
far
trees
|
open
|
open
|
open
|
- For wind directions of 195 and 225deg, the wind speed at all
locations is reduced; due to respectively a nearby high tree
(around 15.5 high at a distance of some 20m) and a line of
four high trees (between 12 and 15m high at a distance between
20 and 200m).
- A few possible locations were also simulated (as these could
be indicators of rotational energy or positions of sensors).
- The locations are:
- 0.4m above the mill cap: yellow (here Impington
Mill has a anemometer)
- avg around the
circumference of 75% of sail length
(8 averaged): green (an indicator of the mill's rotational
energy)
The dotted green line is a trendline through the calculated
values.
- avg of right and left locations of 75% of sail
length (2 averaged): blue (a simpler indicator of the mill's
rotational energy)
- in front of wind shaft: black (an alternative
location for the anemometer)
- Impington: red (measured by
Temple [2024])
Remark: Is Impington's
speed at same height as the Mildenhall speed?
- The wind speed above the mill cap (yellow) is high compared
to ABL (due to turbulence?).
- It is expected that the average
wind speed over the circumference (green) is the best proxy for
the rotational energy available to the mill. Within the
azimuth of 180 to 270deg (the general wind direction is around
240deg), the wind speed at Impington Mill is between 60 and
100% of Mildenhall's.
- The average wind speed over the circumference (green) is
slightly higher than the average wind speed over only left and
right position (blue: at wind shaft height).
- The wind speed in front of the wind shaft (black) is low
(due to upwind effects of mill cap and/or obstructions).
- In general, all the locations have similar calculated
behaviour (like due to the high trees).
- .
Sensitivity analysis around
number/placement of trees and building height
Several scenarios have been simulated to see how the (magnitude of
the) average wind speed (from SW direction, with ABL's z0=0.5m)
was influenced. Eight things have been varied: the placement and
number of trees; the tree form (stacked cylinder [sc], cylinder [c]
and ellipsoid [e]) the height of trees; the height of building
roofs; the roughness of the ground; the height of the flow region;
turbulence model (k-omega SST or realized k-epsilon);
parameter initialisations and boundary layers in mesh.
Here is an overview of these scenarios and their (magnitude of the)
averaged windspeed (relative to the ABL speed at the same height)
over the wing rotation surface:
Scenario
|
1
|
2
|
3
|
4
|
5
|
6
|
7
|
8
|
9
|
10
|
11
|
12
|
13
|
Roof
height
|
midway
|
midway
|
midway
|
ridge
|
ridge |
ridge |
ridge |
ridge |
ridge
|
ridge |
ridge |
ridge
|
ridge |
Trees,
placement
|
14sc, hand
|
21sc, DSM2019
|
25sc, DSM2022
|
25sc, DSM2022 |
25sc, DSM2022 |
25sc, DSM2022 |
25sc, DSM2022 |
25sc, DSM2022 |
25sc, DSM2022 |
25sc, DSM2022 |
25sc, DSM2022 |
25c, DSM2022 |
25e, DSM2022 |
Turbulence
model
|
k-omega SST
|
k-omega SST |
k-omega SST |
k-omega SST |
k-omega SST |
k-omega SST |
realized k-epsilon
|
realized k-epsilon |
realized k-epsilon
initialised
|
realized k-epsilon
initialised, boundary layer
|
realized k-epsilon
initialised, boundary layer |
realized k-epsilon
initialised, boundary layer |
realized k-epsilon
initialised, boundary layer |
flow
region
height [m]
|
40
|
40 |
40 |
40 |
40 |
40 |
40 |
100
|
100 |
100 |
100 |
100
|
100
|
Ground
roughness
height [m]
|
small
|
small
|
small
|
small
|
0.5
|
10
|
10
|
1
|
1
|
1
|
10
|
10
|
10
|
View
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Relative
speed
|
Relatieve
energy
|
|
|
The scenarios are:
- The trees (modelled as stacked cylinders; sc) were positioned
by hand, say within 10m of their actual position (the ground
plane from Google Map does not incude trees).
- Now DSM2019 was included as ground plane and thus trees are
positioned more accurately. Some trees were missing, perhaps
because it was a DSM from 2019.
- The DSM2020 was included as ground plane and the height of the
trees was adjusted (first echo, so they were on average some 2m
higher than in scenario 2).
- The height of buildings is up to ridge roof height; on average
some 30% higher then when using the average roof height.
- The ground layer has a roughness height of 0.5m.
- The ground layer has a roughness height of 10m (=20*z0
[Dong, 2015, page 11004]; Rasaders).
- The turbulence model realized k-epsilon, which should
be better for turbulence in this mill environment [pers. comm.
fvgool, 2024].
- The flow region height is 100m. In the earlier scenarios this
flow region height was too small, as it should be at least 6*H
(and H=16m in this case). Ground Roughness heights of 3, 5 and
10m did not work (Gauge pressure field started diverging).
Adding the necessary boundary
layer at ground level (and perhaps general Initialisation) looks
to solve this issue (scenario
11).
- Several parameters were initialised (gauge pressure, velocity,
k, epsilon, and potential flow initialisation enabled).
- The boundary layer in the mesh was included for the ground
level (this is essential for a good mesh)
- Increased Roughness height of ground
level to 10m
- Changed stacked cylinder (cs) trees to one cylinder (c) trees,
plus the buildings were sometimes merged to improve the meshing.
- Changed one cylinder (c) trees to one
ellipsoid (e) trees.
The ellpsoid trees do not yet calculate in SIMSCALE, but it is
expected that the results will be close to scenario 11.
Remark: A question has been
asked in SIMSCALE Froum about this.
- Depending on the tree height, change the f. This will be
calculated when scenario 13 is working.
Remark: at this moment all
trees have the same f, this could be further improved by
making it depending on the height.
Other observations the following can be deducted:
- The 8 averaged wind speed (green) is calculated by
using every 45 degrees of the wing rotation surface; while the 2
averaged (blue) is only done for the left and right
position of the sails (this
is the 'recommended' wind speed position). There is some
14% difference between the two. And if we do not measure at left
and right but 1m above left and right; both averages
are close. So, it might be good to put anemometer(s) at 1m above
the left and/or right wing position.
So it is proposed to change the definition of this
'recommended' position.
- Increasing the height of the buildings, deceased the wind
velocity at shaft height (scenario 3 and 4).
This is expected as the higher buildings will provide more
blockage. It is not yet known if one should use ridge height or
average roof height.
Remark: investigate what height of a
roofed building is most representative.
- A round roughness height of 10m (=20*z0
[Dong, 2015, page 11004]; Rasaders), this
increased the (magnitude of the) wind speed (scenarios 4, 5 and
6 & scenarios 9 and 10).
Roughness height increases the turbulence and this will also
increase the velocity.
- Increasing the height of the flow region to 100m, this
decreased the wind velocity at shaft height (scenario 7 and 8)
This is expected as the lower flow region will return the wind
downwards and this increasing the wind at shaft height. One
needs to align with normal CFD modelling rules: height flow
region > 6Hmax (Franke, 2007, section 5.1.1). In
our case Hmax = Hmill = 16m.
So important to have the height of the flow region at least 6Hmax
(Franke, 2007, section 5.1.1).
- The scenarios the velocities (@ 2 averaged) are
relatively insensitivity to: number of tree, tree form, the
height of trees/buildings; ground roughness, turbulence model
and mesh.
- With height flow region of 40m (scenarios 1 to 7) have a
standard deviation of ~6%
- With height flow region of 100m (scenarios 8 to
12) have a standard deviation of ~3%
IMHO the thirteenth
scenario is closest to reality. The wind speed at 15m (@ 2
averaged) is some 55% of the ABL speed at 15m.
Anemometer measurements at
Impington Mill and Mildenhall
Average windspeeds (m/sec) around Impington
Looking at the average windspeed, the airport locations of
Cambridge, Mildenhall and Lakenheath all could be ok for Impington
Mill.
Simulated speeds
Below is an overview of the simulated speeds (wind direction of
225deg) at several locations for scenarios 8 to 12:
Using Temple's measurements
The anemometer was at
around 17m and behind the sails and in front of middle of mill
cap (ano old loc)
The measured averaged wind speed (m/sec) of Impington Mill (black:
Anemometer) and meteorological station Mildenhall (grey: Open
Fields) are provided for summer time (May to September) in below
figure (Temple, 2024, @ 21:36). All speeds in below graph are
referenced to ???m.
The Impington Mill wind speeds are on average some 40% of the
Mildenhall values (for azimuths of 180 to 270deg), see also the
red line the earlier figure. This relative value (40%) is
much lower than simulated
relative wind speeds in the CDF (between 50 and 90%), but
the behaviour
of the lines could be seen as slightly comparable (a dip
around 195deg): the dotted green line (trendline of avg. of
circumference) matches the form of Impington line.
Using smartmolen data
At present (April 2024) the Impington Mill weather station is at the
back of the fantail frame, at more or less the same level as the top
of the mill cap: 16m (ano new loc).
Here is average wind speed data from Impington Mill weather station (black),
meteorological stations (Mildenhall: orange, Lakenheath: grey and
Wittering: green); and the ratio between Impington Mill and
Mildenhall (dotted orange) over a (more or less leafless) period of
2.5 Months (last week Febr, March to first week May). All speeds are
referenced to 16m.
The Wittering curve (green) is only based on half the datapoints, so
it is not yet comparable with the other curves.
It is understood that meteorological station Cambridge speeds are
not fully correct (seem to be interpolation of a few data points),
so for now they are not useable.
Evaluation
At this moment the difference between measurements and CDF are
very large. This needs more study! This will be done in the coming
weeks.
Aspects to look at are:
- How are the anemometer measurements at Impington Mill and
Mildenhall calibrated?
It looks, that regardless of the wind direction, that
the windspeed over Impington is on average reduced to around 40%
compared to Mildenhall. How is the measured wind speed converted
to wind speed at wind shaft level?
- Is the CAD-model correct?
Are all trees incorporated at the correct height. The first
return DSM 2022 is used from the Defra Data Service
- What happens with the wind speed for relatively tree free wind
direction?
In the wind direction where not many trees exist (270deg), the
calculated wind speed is quite close (98%) to ABL. In the
anemometer measurements it is regardless of wind directions on
average around 40%.
- Is the porosity of the trees properly included in the CFD?
To reduce the wind speed in the CFD to the levels of the anemometer
measurements; the porosity of the trees should be reduced
considerable. Would that still be realistic trees? The
derivation of f (Darcy-Forchheimer's measure of porosity)
is not yet 100% understood; the simulated empirical way has been
used for now.
- Even if the CDF is not 100% correct in absolute sense, it can
be utilised in a relative sense for comparing different
scenarios. Most simulations (if they are in general correctly
setup, aka dependencies are properly configured) have this
property.
Points 1 and 2 might be important to check (keeping in mind
points 3 and 4).
When the above points are understood, the wind rose sector
between 270 and 360deg will also be simulated.
References
Dong, Zhibao et al.: Aerodynamic roughness of
fixed sandy beds. In: Journal of Geophysical Research: Solid
Earth 106 (2015), issue February, pp. 11000-11011.
Franke, Jörg et al. COST Action 732: Best
practice guideline for the CFD simulation of flows in the urban
environment. Brussels, COST Office 2007.
Temple, Stephen: Wind report. In: Mill
news (2020), issue January, pp. 6-10.
Temple, Stephen: A modern molenbiotope.
https://www.youtube.com/watch?v=rNxU6xD3Iuc, 2024.
Acknowledgements
I would like to thank people, such as Justin Coombs,
fvgool, Stephen Temple, SIMSCALE
support team, and others for their help,
encouragement and/or constructive feedback. Any remaining errors
in methodology or results are my responsibility of course!!! If
you want to provide constructive feedback, please let me know.
Major content related
changes: March 8, 2024