Tutorial 5 - FOSWEC
This tutorial models the floating oscillating surge wave energy converter (FOSWEC) and examines adding top-mounted ballast to the FOSWEC flaps. The FOSWEC has been tested in multiple experimental campaigns at the Oregon State University (OSU) O.H. Hinsdale Wave Research Laboratory (HWRL) Directional Wave Basin (DWB) and a WEC-Sim numerical model has been previously developed and validated against experimental data.
The FOSWEC is a coupled multibody system consisting of two flaps pitching relative to a floating platform. Efficiently modeling this system in WecOptTool requires a reduced-coordinate formulation. In this tutorial, that reduction is performed using Multibody for Everybody (M4E), a package developed by the National Laboratory of the Rockies (NLR).
While previous WecOptTool tutorials focus on a single degree of freedom (DOF) or a small set of coupled DOFs, many WEC archetypes have more complex and coupled dynamics that would otherwise require manually defining kinematic constraints or reduced-coordinate equations of motion. M4E automates this process by transforming the hydrodynamic, hydrostatic, kinematic, and external force terms from global to joint coordinates. For the FOSWEC, this reduces the planar model from 9 global coordinates to 5 joint coordinates, preserving important platform-flap coupling while reducing the number of optimization state variables significantly. M4E eliminates the need for manually deriving reduced-coordinate equations, which can be time-consuming and error prone, and greatly improves the computational efficiency, since the WecOptTool optimization cost roughly scales with the square of the number of optimization variables.
Using Multibody for Everybody with WecOptTool requires the following main steps:
Create an M4E input file. This defines the body properties, joint locations, and joint types and needs to be imported into the current notebook.
After running BEM for all rigid body degrees of freedom, call
wot.utilities.setup_from_M4E()to assemble the reduced impedance matrix, reduced excitation coefficients, and coordinate transformation matrix.Convert PTO kinematics and external force matrices to reduced coordinates using
wot.utilities.reduce_PTO_kinematics_M4E()andwot.utilities.reduce_damping_stiffness_M4E().Set up the WEC object using the
from_impedancemethod and solve the optimization problem.
This tutorial consists of three parts. First, the model is set up and converted to reduced coordinates using M4E. Next, the unoptimized response is compared to an equivalent WEC-Sim model for code-to-code verification. Then, the impact of adding top-mounted ballast to the flaps is analyzed.
[1]:
import capytaine as cpy
from capytaine.io.meshio import load_from_meshio
import numpy as np
import jax.numpy as jnp
import matplotlib.pyplot as plt
from matplotlib.lines import Line2D
import xarray as xr
import gmsh, pygmsh
import wecopttool as wot
## set colorblind-friendly colormap for plots
plt.style.use('tableau-colorblind10')
1. Model setup
1.1 Wave Conditions
For the initial code-to-code verification, three difference regular wave amplitudes are tested. The first wave has a very small amplitude to ensure the response remains within the linear range of operation. The intermediate wave amplitude matches a condition previously tested in the wave tank, while the largest wave height is included to further evaluate the effect of larger-amplitude motion on the linearized M4E formulation.
Since we are using only regular waves and linear proportional control gains in this tutorial, the system is entirely linear. Thus, only 2 frequencies are needed, enabling a computationally efficient optimization. For irregular waves or any control introducing nonlinearities (e.g., unstructured), additional frequencies would be necessary to properly model the dynamics.
[2]:
# Frequency object
wavefreq = 1/2.61 # Hz
f1 = wavefreq
nfreq = 2
freq = wot.frequency(f1, nfreq, False) # False -> no zero frequency
# Define wave cases
wave_heights = [0.001, 0.136, 0.272] # m
wave_amplitudes = [height / 2 for height in wave_heights]
phase = 0 # degrees
wavedir = 0 # degrees
waves = [
wot.waves.regular_wave(f1, nfreq, wavefreq, amplitude, phase, wavedir)
for amplitude in wave_amplitudes
]
1.2 WEC Geometry
The FOSWEC consists of two flaps pitching relative to a floating platform. The geometry is defined below including each of the three meshes, their hydrostatic properties, and the degrees of freedom (DOFs) for solving the boundary element method (BEM) problem.
Usually, only the relevant DOFs are defined for each floating body in Capytaine. However, since M4E is handling the transformation from global to reduced coordinates, we add all 6 rigid body DOFs defined about the center of gravity using Capytaine.
[3]:
# flap dimensions and properties
flap_thickness_bottom = 0.04
flap_thickness_top = 0.1
flap_center_distance_apart = 1.44 # distance between two flap centers in the direction of wave propagation
flap_width = 0.76
flap_height = 0.58
flap_draft = 0.59 # Fully submerged
cg_height_above_hinge = 0.17
flap1_cg = [-flap_center_distance_apart / 2, 0, -flap_draft + cg_height_above_hinge] # from water surface/origin
flap2_cg = [flap_center_distance_apart / 2, 0, -flap_draft + cg_height_above_hinge] # from water surface/origin
flap_mass = 23.1
flap_inertia = 1.19
# platform dimensions and properties; consists of a frame, a DAQ box, and four columns
platform_frame_length = 1.44
platform_frame_thickness = 0.05
platform_frame_width = 1.06
platform_frame_top_depth = 0.53 + 0.1
platform_cg = [0, 0, -0.8]
platform_mass = 189.8
platform_inertia = 30
cutout_width = platform_frame_width - 2 * platform_frame_thickness
cutout_length = platform_frame_length - 2 * platform_frame_thickness
full_width = 1.63
columns_radius = 0.1
columns_draft = 0.53 + 0.1 + 0.4
columns_xy = [flap_center_distance_apart / 2, full_width / 2 - columns_radius]
daq_length = 0.72
daq_width = 0.6
daq_height = 0.2
daq_top_depth = 0.53 + 0.1 + 0.05
# Define the meshes
with pygmsh.occ.Geometry() as geom:
gmsh.option.setNumber("Mesh.MeshSizeFactor", 0.3)
platform = geom.add_box(
[-platform_frame_length / 2, -platform_frame_width / 2, -platform_frame_top_depth - platform_frame_thickness],
[platform_frame_length, platform_frame_width, platform_frame_thickness],
)
cutout = geom.add_box(
[-cutout_length / 2, -cutout_width / 2, -platform_frame_top_depth - platform_frame_thickness],
[cutout_length, cutout_width, platform_frame_thickness],
)
cyl1 = geom.add_cylinder(
[-columns_xy[0], -columns_xy[1], -columns_draft],
[0, 0, columns_draft + 0.01],
columns_radius,
)
cyl2 = geom.add_cylinder(
[-columns_xy[0], columns_xy[1], -columns_draft],
[0, 0, columns_draft + 0.01],
columns_radius,
)
cyl3 = geom.add_cylinder(
[columns_xy[0], -columns_xy[1], -columns_draft],
[0, 0, columns_draft + 0.01],
columns_radius,
)
cyl4 = geom.add_cylinder(
[columns_xy[0], columns_xy[1], -columns_draft],
[0, 0, columns_draft + 0.01],
columns_radius,
)
daq = geom.add_box(
[-daq_length / 2, -daq_width / 2, -daq_top_depth - daq_height],
[daq_length, daq_width, daq_height],
)
platform_frame = geom.boolean_difference(platform, cutout)
geom.boolean_union([platform_frame, daq, cyl1, cyl2, cyl3, cyl4])
platform_mesh = geom.generate_mesh()
with pygmsh.geo.Geometry() as geom:
flap1 = geom.add_polygon(
[
[-flap_center_distance_apart / 2 - flap_thickness_bottom / 2, -flap_width / 2, -flap_draft],
[-flap_center_distance_apart / 2 + flap_thickness_bottom / 2, -flap_width / 2, -flap_draft],
[-flap_center_distance_apart / 2 + flap_thickness_top / 2, -flap_width / 2, flap_height - flap_draft],
[-flap_center_distance_apart / 2 - flap_thickness_top / 2, -flap_width / 2, flap_height - flap_draft],
],
mesh_size=0.1,
)
geom.extrude(flap1, [0, flap_width, 0])
flap1_mesh = geom.generate_mesh()
with pygmsh.geo.Geometry() as geom:
flap2 = geom.add_polygon(
[
[flap_center_distance_apart / 2 - flap_thickness_bottom / 2, -flap_width / 2, -flap_draft],
[flap_center_distance_apart / 2 + flap_thickness_bottom / 2, -flap_width / 2, -flap_draft],
[flap_center_distance_apart / 2 + flap_thickness_top / 2, -flap_width / 2, flap_height - flap_draft],
[flap_center_distance_apart / 2 - flap_thickness_top / 2, -flap_width / 2, flap_height - flap_draft],
],
mesh_size=0.1,
)
geom.extrude(flap2, [0, flap_width, 0])
flap2_mesh = geom.generate_mesh()
# define functions for creating inertia matrix and floating body
def make_inertia_matrix(body, mass, inertia):
return xr.DataArray(
data=np.diag([mass, mass, mass, inertia, inertia, inertia]),
dims=["influenced_dof", "radiating_dof"],
coords={
"influenced_dof": list(body.dofs),
"radiating_dof": list(body.dofs),
},
name="inertia_matrix",
)
def make_floating_body(mesh, name, center_of_mass, mass, inertia):
body = cpy.FloatingBody(mesh=mesh, name=name, center_of_mass=center_of_mass)
body.add_all_rigid_body_dofs()
body.rotation_center = body.center_of_mass
body.inertia_matrix = make_inertia_matrix(body, mass, inertia)
body.hydrostatic_stiffness = body.immersed_part().compute_hydrostatic_stiffness()
return body
# Create floating bodies for platform and flaps and combine
platform = make_floating_body(platform_mesh, "platform", platform_cg, platform_mass, platform_inertia)
flap1 = make_floating_body(flap1_mesh, "flap1", flap1_cg, flap_mass, flap_inertia)
flap2 = make_floating_body(flap2_mesh, "flap2", flap2_cg, flap_mass, flap_inertia)
foswec_fb = platform + flap1 + flap2
Info : [ 90%] Union - Splitting solids
1.3 Hydrodynamics
Next, we run the BEM using Capytaine to obtain the hydrodynamic coefficients for the FOSWEC. Capytaine returns the hydrodynamic matrices and coefficients for the full 18 rigid body DOFs.
[4]:
# run BEM calculations
bem_data = wot.run_bem(foswec_fb, freq, depth=2)
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:25:23] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
1.4 Multibody for Everybody
After generating BEM coefficients for all rigid body DOFs, Multibody for Everybody is called to reduce from global coordinates to joint coordinates. M4E is only set up for planar DOFs, so it ignores the sway, roll, and yaw coefficients for each body and reduces the remaining coefficients (surge, heave, and pitch) to joint coordinates. Thus, the FOSWEC is reduced from 9 global planar coordinates to 5 joint coordinates. wot.utilities.setup_from_M4E() calls M4E, inputting the M4E input file,
BEM coefficients, floating bodies, and mass proerties, and sets up the reduced-coordinate impedance matrix, excitation coefficients, and hydrostatic stiffness matrix to be used in setting up the WEC object. The utilities function also returns R0, which is the transformation matrix used for coordinate conversion, and the number of M4E DOFs per body, M4E_ndof.
[5]:
from M4E_inputs import FOSWEC_M4E_inputs
# create multibody system for wavebot using M4E
m0 = [platform_mass, flap_mass, flap_mass]
J0 = [platform_inertia, flap_inertia, flap_inertia]
# return the reduced coordinate matrices
impedance_reduced, excitation_reduced, hydrostatic_stiffness_reduced, R0, M4E_ndof = wot.utilities.setup_from_M4E(FOSWEC_M4E_inputs, bem_data, [platform, flap1, flap2], m0, J0)
Problem initialization finished in: 0.000 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.052 seconds
Symbolic EOM computation finished in: 0.022 seconds
Compiling EOM expressions finished in: 0.076 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
1.5 Additional forces
Additional forces acting on the FOSWEC include the mooring stiffness and damping on the platform and linear damping on the flaps. The damping and stiffness matrices are defined in the global planar DOFs and transformed to reduced coordinates using wot.utilities.reduce_damping_stiffness_M4E() and the coordinate transformation matrix, R0. Then, the force functions are defined normally using the reduced-coordinate matrices. For the code-to-code verification, no PTO is included here.
[6]:
# set mooring stiffness and damping matrices and reduce to joint coordinates
mooring_stiffness_full = np.diag([8e3, 2e5, 2e5, 0, 0, 0, 0, 0, 0])
mooring_stiffness = wot.utilities.reduce_damping_stiffness_M4E(mooring_stiffness_full, R0)
mooring_damping_full = np.diag([8e2, 1e4, 1e4, 0, 0, 0, 0, 0, 0])
mooring_damping = wot.utilities.reduce_damping_stiffness_M4E(mooring_damping_full, R0)
def f_mooring(wec, x_wec, x_opt, wave, nsubsteps=1):
pos = wec.vec_to_dofmat(x_wec)
vel = jnp.dot(wec.derivative_mat,pos)
time_matrix = wec.time_mat_nsubsteps(nsubsteps)
mooring = -pos @ mooring_stiffness - vel @ mooring_damping
mooring_force = jnp.dot(time_matrix,mooring)
return mooring_force
# set linear damping matrix (flap linear damping) and reduce to joint coordinates
flap_linear_damping_full = np.diag([0, 0, 0, 0, 0, 10, 0, 0, 10])
flap_linear_damping = wot.utilities.reduce_damping_stiffness_M4E(flap_linear_damping_full, R0)
def f_linear_damping(wec, x_wec, x_opt, wave, nsubsteps=1):
pos = wec.vec_to_dofmat(x_wec)
vel = jnp.dot(wec.derivative_mat,pos)
time_matrix = wec.time_mat_nsubsteps(nsubsteps)
damping = - vel @ flap_linear_damping
linear_damping_force = jnp.dot(time_matrix, damping)
return linear_damping_force
# define added forces
f_add = {'f_mooring': f_mooring, 'f_linear_damping': f_linear_damping}
1.6 WEC object
The WEC object is defined using the from_impedance method with the reduced-coordinate matrices and coefficients.
[7]:
# Define the WEC object using the from_impedance method
wec = wot.WEC.from_impedance(
waves[0].freq.values,
impedance=impedance_reduced,
exc_coeff=excitation_reduced,
hydrostatic_stiffness=hydrostatic_stiffness_reduced,
constraints=None,
f_add=f_add,
)
1.7 Objective function
For the code-to-code verification, we are not applying any PTO/control forces. The objective function is therefore set to zero with no optimization variables.
[8]:
# Objective function - none for now
obj_fun = lambda wec, x_wec, x_opt, waves : 0
nstate_opt = 0
2. Model verification
2.1 Solve
With the objective function set to zero, WecOptTool solves for a feasible periodic response by enforcing the dynamic residual constraint, effectively solving the multibody dynamics problem. The problem is solved for each verification wave condition. For post-processing, wot.utilities.post_process_M4E() formats the results and converts them back to the global planar coordinates.
[9]:
# Solve
scale_x_wec = 1e1
scale_x_opt = 1e-3
scale_obj = 1e-2
results_all = []
wec_fdom_full_all = []
wec_tdom_full_all = []
nsubsteps = 10
for wave in waves:
# Solve
results = wec.solve(
wave,
obj_fun,
nstate_opt,
scale_x_wec=scale_x_wec,
scale_x_opt=scale_x_opt,
scale_obj=scale_obj,
)
# Post-process results
wec_fdom_full, wec_tdom_full = wot.utilities.post_process_M4E(
wec, results, wave, nsubsteps, R0
)
# Store outputs
results_all.append(results)
wec_fdom_full_all.append(wec_fdom_full)
wec_tdom_full_all.append(wec_tdom_full)
Optimization terminated successfully (Exit mode 0)
Current function value: 0.0
Iterations: 1
Function evaluations: 2
Gradient evaluations: 1
Optimization terminated successfully (Exit mode 0)
Current function value: 0.0
Iterations: 1
Function evaluations: 2
Gradient evaluations: 1
Optimization terminated successfully (Exit mode 0)
Current function value: 0.0
Iterations: 1
Function evaluations: 2
Gradient evaluations: 1
2.2 Code-to-code comparison
To verify the WecOptTool+M4E model, the results are compared to an equivalent WEC-Sim model. The small waves ensure the system remains entirely within the linear range of operation since M4E linearizes the dynamics while WEC-Sim models the nonlinear dynamics. Within the small wave conditions, a near perfect match is achieved between the WecOptTool+M4E and WEC-Sim models. As the wave height is increased, the difference between the WecOptTool and WEC-Sim results increases. Still, the relatively low error across the tested wave conditions verifies the WecOptTool+M4E model and supporting its use in co-design and optimization studies.
[10]:
# Load combined WEC-Sim results
ws_data = xr.open_dataset("data/FOSWEC_WS_data.nc")
for i in range(ws_data.sizes["case"]):
ws_output = ws_data.isel(case=i)
wave_height = ws_output.wave_height.values
wec_tdom_full = wec_tdom_full_all[i]
fig, axs = plt.subplots(2, 1, sharex=True, figsize=(6, 7),gridspec_kw={"hspace": 0.05})
ws_time_period = ws_output.time
wec_time_period = wec_tdom_full["time"]
# Platform response
axs[0].plot(wec_time_period,wec_tdom_full.sel(realization=0).pos[0, :],"C0-",linewidth=1,label="surge, WecOptTool",)
axs[0].plot(ws_time_period,ws_output.platform_response[:, 0],"C0--",linewidth=2,label="surge, WEC-Sim",)
axs[0].plot(wec_time_period,wec_tdom_full.sel(realization=0).pos[1, :],"C1-",linewidth=1,label="heave, WecOptTool",)
axs[0].plot(ws_time_period,ws_output.platform_response[:, 2] - platform_cg[2],"C1--",linewidth=2,label="heave, WEC-Sim",)
axs[0].plot(wec_time_period,wec_tdom_full.sel(realization=0).pos[2, :],"C2-",linewidth=1,label="pitch, WecOptTool",)
axs[0].plot(ws_time_period,ws_output.platform_response[:, 4],"C2--",linewidth=2,label="pitch, WEC-Sim",)
axs[0].set_ylabel("Platform response [m, rad]")
axs[0].set_title(f"H = {wave_height:.3f} m")
axs[0].legend(fontsize=10, handlelength=1.5, labelspacing=0.3)
# Flap response
axs[1].plot(wec_time_period,wec_tdom_full.sel(realization=0).pos[5, :],color="C0",linewidth=1,label="WecOptTool - flap 1",)
axs[1].plot(wec_time_period,wec_tdom_full.sel(realization=0).pos[8, :],color="C1",linewidth=1,label="WecOptTool - flap 2",)
axs[1].plot(ws_time_period,ws_output.pto1_response[:, 4],"--",color="C0",linewidth=2,label="WEC-Sim - flap 1",)
axs[1].plot(ws_time_period,ws_output.pto2_response[:, 4],"--",color="C1",linewidth=2,label="WEC-Sim - flap 2",)
axs[1].set_xlabel("Time [s]")
axs[1].set_ylabel("Flap pitch [rad]")
axs[1].legend(fontsize=10, handlelength=1.5, labelspacing=0.3)
plt.tight_layout()
/tmp/ipykernel_4237/252279905.py:34: UserWarning: This figure includes Axes that are not compatible with tight_layout, so results might be incorrect.
plt.tight_layout()
3. Top-mounted ballast study
3.1 Top-mounted ballast impact on hydrostatic stiffness
Adding ballast to the top of the flaps moves their center of gravity upward (increase center of gravity height). As shown below, this reduces the distance between the center of gravity and center of buoyancy, creating a destabilizing effect that reduces the hydrostatic stiffness. The reduced hydrostatic stiffness may increase the WEC response and improve power capture, particularly if the flap rotation constraints are not already active.
[11]:
# plot mass on top effects
top_mounted_ballast_vec = np.linspace(0, 5, 11)
# Define mesh for single centered flap
with pygmsh.geo.Geometry() as geom:
flap = geom.add_polygon(
[[-flap_thickness_bottom / 2, -flap_width / 2, -flap_draft],
[flap_thickness_bottom / 2, -flap_width / 2, -flap_draft],
[flap_thickness_top / 2, -flap_width / 2, flap_height - flap_draft],
[-flap_thickness_top / 2, -flap_width / 2, flap_height - flap_draft],],mesh_size=0.1,)
geom.extrude(flap, [0, flap_width, 0])
flap_mesh = geom.generate_mesh()
# determine new center of gravity height and hydrostatic stiffness for different top-mounted ballast values
cg_height_vec = []
hydrostatic_stiffness_vec = []
for top_mounted_ballast in top_mounted_ballast_vec:
# Define FloatingBody
cg_height_above_hinge_new = (cg_height_above_hinge*flap_mass + top_mounted_ballast*flap_height)/(flap_mass + top_mounted_ballast)
flap_cg = [0, 0, -flap_draft + cg_height_above_hinge_new]
flap = cpy.FloatingBody(mesh=flap_mesh, name="flap", center_of_mass=flap_cg)
# Define rotation dof
flap.rotation_center = (0, 0, -flap_draft)
flap.add_rotation_dof(name='Pitch')
# Set FloatingBody inertia matrix
flap.mass = flap_mass + top_mounted_ballast
flap_inertia_hinge = flap_inertia + (flap_mass + top_mounted_ballast)*cg_height_above_hinge_new**2
rigid_inertia_matrix_xr = xr.DataArray(data=np.asarray((np.diag([flap_inertia_hinge]))),
dims=['influenced_dof', 'radiating_dof'],
coords={'influenced_dof': list(flap.dofs),
'radiating_dof': list(flap.dofs)},
name="inertia_matrix")
flap.inertia_matrix = rigid_inertia_matrix_xr
# record cg height and hydrostatic stiffness
flap.hydrostatic_stiffness = flap.immersed_part().compute_hydrostatic_stiffness()
cg_height_vec.append(cg_height_above_hinge_new)
hydrostatic_stiffness_vec.append(flap.hydrostatic_stiffness)
fig, axs = plt.subplots(2, 1, sharex=True, figsize=(6, 5))
axs[0].plot(top_mounted_ballast_vec, cg_height_vec)
axs[0].set_ylabel("Center of gravity height [m]")
axs[0].grid(True)
axs[1].plot(top_mounted_ballast_vec, np.squeeze(hydrostatic_stiffness_vec))
axs[1].set_xlabel("Top-mounted ballast [kg]")
axs[1].set_ylabel("Hydrostatic stiffness [N·m/rad]")
axs[1].grid(True)
3.2 Top-mounted ballast study loop
The top-mounted ballast study loops through ballast mass, wave period, and wave amplitude. Capytaine is rerun for each ballast mass and wave period to generate the hydrodynamic coefficients using the updated mass properties and required frequencies. The PTO, additional forces, and constraints are defined for each case, and the optimization is run to determine the optimal controller gains and added stiffness coefficients.
Sometimes the optimization converges to local optima. To reduce sensitivity to the initial guess, each case is optimized multiple times using different initial conditions, and the solution with the highest generated power is selected.
Note that a few details differ from the corresponding paper, including the case discretization and controller formulation: this tutorial uses a proportional controller, while the paper uses an unstructured controller.
3.2.1 PTO
Since no PTO was included in the verification setup, the PTO dynamics, kinematics, and constraints are now included here. Although there are two PTO units in reality, they can be represented together by one kinematics matrix and one impedance matrix. The PTOs are connected between the flap and the platform and resist the relative pitch motion between them.
The PTO kinematics matrix maps the global planar coordinates to the two PTO coordinates: flap 1 pitch relative to the platform and flap 2 pitch relative to the platform. This kinematics matrix is reduced to joint coordinates using wot.utilities.reduce_PTO_kinematics_M4E().
The PTO dynamics are defined using the two-port impedance model based on the drivetrain and generator parameters. Because there are two PTOs, the combined PTO impedance matrix is 4x4: two mechanical ports and two electrical ports. In this formulation, rows/columns 1 and 3 correspond to the mechanical and electrical ports of the first PTO, while rows/columns 2 and 4 correspond to the mechanical and electrical ports of the second PTO.
3.2.2 Optimal drivetrain stiffness
Along with optimizing the PTO controller gains, an optimized stiffness coefficient is added to each flap to represent an added torsional spring. The added stiffness coefficients give the optimizer an additional design variable that can compensate for changes in hydrostatic stiffness while still satisfying the maximum flap rotation constraints.
3.2.3 Controller
A proportional controller is used here for efficient optimization, providing linear damping control. There are two proportional gains: one for each PTO.
3.2.4 Constraints
The constraints include rotational constraints for each flap and peak and rms generator torque limits.
[12]:
top_mounted_ballast_vec = np.linspace(0, 2, 3)
wave_periods = [1, 3, 6]
wave_amplitudes = [0.05, 0.15, 0.25]
case_outputs = []
for top_mounted_ballast in top_mounted_ballast_vec:
# redefine floating bodies with top-mounted ballast
platform = make_floating_body(platform_mesh, "platform", platform_cg, platform_mass, platform_inertia)
total_flap_mass = flap_mass + top_mounted_ballast
cg_height_above_hinge_new = (flap_mass * cg_height_above_hinge + top_mounted_ballast * flap_height) / total_flap_mass
flap_inertia_new = flap_inertia + (flap_mass*top_mounted_ballast/total_flap_mass)*(flap_height - cg_height_above_hinge)**2
flap1_cg = [-flap_center_distance_apart / 2, 0, -flap_draft + cg_height_above_hinge_new]
flap1 = make_floating_body(flap1_mesh, "flap1", flap1_cg, total_flap_mass, flap_inertia_new)
flap2_cg = [flap_center_distance_apart / 2, 0, -flap_draft + cg_height_above_hinge_new]
flap2 = make_floating_body(flap2_mesh, "flap2", flap2_cg, total_flap_mass, flap_inertia_new)
foswec_fb = platform + flap1 + flap2
for wave_period in wave_periods:
# define frequency object
wavefreq = 1/wave_period # Hz
f1 = wavefreq
nfreq = 2
freq = wot.frequency(f1, nfreq, False) # False -> no zero frequency
# run BEM calculations for wave period
bem_data = wot.run_bem(foswec_fb, freq, depth=2)
# create multibody system for wavebot using M4E
m0 = [platform_mass, total_flap_mass, total_flap_mass]
J0 = [platform_inertia, flap_inertia_new, flap_inertia_new]
# return the reduced coordinate matrices
impedance_reduced, excitation_reduced, hydrostatic_stiffness_reduced, R0, M4E_ndof = wot.utilities.setup_from_M4E(FOSWEC_M4E_inputs, bem_data, [platform, flap1, flap2], m0, J0)
for wave_amplitude in wave_amplitudes:
# define wave case
amplitude = wave_amplitude
phase = 0 # degrees
wavedir = 0 # degrees
waves = wot.waves.regular_wave(f1, nfreq, wavefreq, amplitude, phase, wavedir)
## PTO impedance definition
omega = bem_data.omega.values
gear_ratio = 3.75
torque_constant = 1.021
winding_resistance = 1.028
winding_inductance = 0.0
drivetrain_inertia = 0.05
drivetrain_friction_aft = 3.27 / 3.75**2
drivetrain_friction_bow = 2.7 / 3.75**2
drivetrain_stiffness = 0.0
drivetrain_impedance_aft = (
1j * omega * drivetrain_inertia
+ drivetrain_friction_aft
+ 1 / (1j * omega) * drivetrain_stiffness
)
drivetrain_impedance_bow = (
1j * omega * drivetrain_inertia
+ drivetrain_friction_bow
+ 1 / (1j * omega) * drivetrain_stiffness
)
winding_impedance = winding_resistance + 1j * omega * winding_inductance
pto_impedance_11_aft = -1 * gear_ratio**2 * drivetrain_impedance_aft
pto_impedance_11_bow = -1 * gear_ratio**2 * drivetrain_impedance_bow
off_diag = np.sqrt(3.0 / 2.0) * torque_constant * gear_ratio
pto_impedance_12 = -1 * (off_diag + 0j) * np.ones(omega.shape)
pto_impedance_21 = -1 * (off_diag + 0j) * np.ones(omega.shape)
pto_impedance_22 = winding_impedance
pto_impedance = np.zeros((4, 4, len(omega)), dtype=complex)
pto_impedance[0, 0, :] = pto_impedance_11_aft
pto_impedance[1, 1, :] = pto_impedance_11_bow
pto_impedance[0, 2, :] = pto_impedance_12
pto_impedance[1, 3, :] = pto_impedance_12
pto_impedance[2, 0, :] = pto_impedance_21
pto_impedance[3, 1, :] = pto_impedance_21
pto_impedance[2, 2, :] = pto_impedance_22
pto_impedance[3, 3, :] = pto_impedance_22
# PTO settings
pto_ndof = 2
controller = wot.controllers.pid_controller(ndof_pto=pto_ndof, proportional=True, integral=False, derivative=False)
loss = None
names = ["flap 1 PTO","flap 2 PTO"]
# PTO kinematics
pto_kinematics_full = np.array(
[[0, 0, -1, 0, 0, 1, 0, 0, 0],
[0, 0, -1, 0, 0, 0, 0, 0, 1],]
)
pto_kinematics_reduced = wot.utilities.reduce_PTO_kinematics_M4E(pto_kinematics_full, R0)
pto = wot.pto.PTO(pto_ndof, pto_kinematics_reduced, controller, pto_impedance, loss, names)
mooring_stiffness = wot.utilities.reduce_damping_stiffness_M4E(mooring_stiffness_full, R0)
mooring_damping = wot.utilities.reduce_damping_stiffness_M4E(mooring_damping_full, R0)
flap_linear_damping = wot.utilities.reduce_damping_stiffness_M4E(flap_linear_damping_full, R0)
def f_mooring(wec, x_wec, x_opt, wave, nsubsteps=1):
pos = wec.vec_to_dofmat(x_wec)
vel = jnp.dot(wec.derivative_mat, pos)
time_matrix = wec.time_mat_nsubsteps(nsubsteps)
force = -pos @ mooring_stiffness - vel @ mooring_damping
return jnp.dot(time_matrix, force)
def f_linear_damping(wec, x_wec, x_opt, wave, nsubsteps=1):
pos = wec.vec_to_dofmat(x_wec)
vel = jnp.dot(wec.derivative_mat, pos)
time_matrix = wec.time_mat_nsubsteps(nsubsteps)
force = -vel @ flap_linear_damping
return jnp.dot(time_matrix, force)
def f_drivetrain_stiffness_flap1(wec, x_wec, x_opt, wave, nsubsteps=1):
pto_pos = pto.position(wec, x_wec, x_opt, wave, nsubsteps)
flap1_stiffness = x_opt[-2]
spring_force_pto = jnp.zeros_like(pto_pos)
spring_force_pto = spring_force_pto.at[:, 0].set(-flap1_stiffness * pto_pos[:, 0])
spring_force_wec = jnp.dot(spring_force_pto, pto_kinematics_reduced)
return spring_force_wec
def f_drivetrain_stiffness_flap2(wec, x_wec, x_opt, wave, nsubsteps=1):
pto_pos = pto.position(wec, x_wec, x_opt, wave, nsubsteps)
flap2_stiffness = x_opt[-1]
spring_force_pto = jnp.zeros_like(pto_pos)
spring_force_pto = spring_force_pto.at[:, 1].set(-flap2_stiffness * pto_pos[:, 1])
spring_force_wec = jnp.dot(spring_force_pto, pto_kinematics_reduced)
return spring_force_wec
f_add = {
"f_mooring": f_mooring,
"f_linear_damping": f_linear_damping,
"f_drivetrain_stiffness_flap1": f_drivetrain_stiffness_flap1,
"f_drivetrain_stiffness_flap2": f_drivetrain_stiffness_flap2,
"PTO": pto.force_on_wec
}
# Constraints
max_flap_pitch = np.deg2rad(30.0)
max_pto_torque = 40.0
max_rms_pto_torque = 20.0
def const_flap1_rotation(wec, x_wec, x_opt, wave, nsubsteps=nsubsteps):
pto_pos = pto.position(wec, x_wec, x_opt, wave, nsubsteps)
flap1_pos = pto_pos[:, 0]
return max_flap_pitch - jnp.abs(flap1_pos.flatten())
def const_flap2_rotation(wec, x_wec, x_opt, wave, nsubsteps=nsubsteps):
pto_pos = pto.position(wec, x_wec, x_opt, wave, nsubsteps)
flap2_pos = pto_pos[:, 1]
return max_flap_pitch - jnp.abs(flap2_pos.flatten())
def const_motor_torque(wec, x_wec, x_opt, wave):
pto_force = pto.force(wec, x_wec, x_opt, wave, nsubsteps)
motor_torque = pto_force / gear_ratio
return max_pto_torque - jnp.abs(motor_torque.flatten())
def const_motor_rms_torque(wec, x_wec, x_opt, wave):
pto_force = pto.force(wec, x_wec, x_opt, wave, nsubsteps)
motor_torque = pto_force / gear_ratio
rms_by_flap = jnp.sqrt(jnp.mean(motor_torque**2, axis=0) + 1e-12)
return max_rms_pto_torque - rms_by_flap
constraints = [
{"type": "ineq", "fun": const_flap1_rotation},
{"type": "ineq", "fun": const_flap2_rotation},
{"type": "ineq", "fun": const_motor_torque},
{"type": "ineq", "fun": const_motor_rms_torque},
]
# define optimization states and bounds
n_force_states = 1 * 2 # 1 P gain * 2 PTO DOFs
nstate_opt = n_force_states + 2 # 2 optimal stiffnesses for the two flaps
bounds_opt = (((-1e8, 1e8),) * n_force_states + ((0.0, 1e4),) + ((0.0, 1e4),))
# create WEC object
wec = wot.WEC.from_impedance(
waves.freq.values,
impedance=impedance_reduced,
exc_coeff=excitation_reduced,
hydrostatic_stiffness=hydrostatic_stiffness_reduced,
constraints=constraints,
f_add=f_add,
)
obj_fun = pto.average_power
wave_coeffs = waves.sel(realization=0, wave_direction=0).values
x_wec_mat = np.zeros((2 * len(wave_coeffs), 5))
x_wec_mat[0::2, :] = np.real(wave_coeffs)[:, None]
x_wec_mat[1::2, :] = np.imag(wave_coeffs)[:, None]
x_wec_0 = x_wec_mat.flatten() # Set x_wec_0 based on wave coefficients for initial guess of WEC states
x_opt_0_list = [[0, 0, 250, 250],[0, 0, 500, 500]] # Set last 2 elements of x_opt_0 to 250 for initial guess of optimal stiffnesses
# Run the optimization mutliple times with different initial conditions and select best power
best_case = None
nsubsteps_post = 10
for x_opt_0 in x_opt_0_list:
results = wec.solve(
waves,
obj_fun,
nstate_opt,
scale_x_wec=1e1,
scale_x_opt=1e-2,
scale_obj=1.0,
x_wec_0=x_wec_0,
x_opt_0=x_opt_0,
bounds_opt=bounds_opt,
optim_options={"maxiter": 500},
)
wec_fdom_full, wec_tdom_full = wot.utilities.post_process_M4E(wec, results, waves, nsubsteps_post, R0)
pto_fdom, pto_tdom = pto.post_process(wec, results, waves, nsubsteps=nsubsteps_post)
avg_power = np.mean(pto_tdom["power"][0, 1, :, 0].values + pto_tdom["power"][0, 1, :, 1].values)
electrical_power = -avg_power
if best_case is None or electrical_power > best_case["electrical_power"]:
best_case = {
"results": results,
"wec_fdom_full": wec_fdom_full,
"wec_tdom_full": wec_tdom_full,
"pto_fdom": pto_fdom,
"pto_tdom": pto_tdom,
"electrical_power": electrical_power,
}
x_full = np.asarray(best_case["results"][0].x)
nstate_wec = int(wec.nstate_wec)
x_opt_sol = np.array(x_full[nstate_wec:], copy=True)
opt_stiffness_flap1 = x_opt_sol[-2]
opt_stiffness_flap2 = x_opt_sol[-1]
pos = np.squeeze(best_case["pto_tdom"]["pos"].sel(realization=0).values)
max_abs_pos = np.max(np.abs(pos), axis=0)
flap1_peak_rotation_utilization = max_abs_pos[0] / max_flap_pitch
flap2_peak_rotation_utilization = max_abs_pos[1] / max_flap_pitch
case_outputs.append(
{
"top_mounted_ballast": top_mounted_ballast,
"wave_period": wave_period,
"wave_amplitude": wave_amplitude,
"electrical_power": best_case["electrical_power"],
"flap1_peak_rotation_utilization": flap1_peak_rotation_utilization,
"flap2_peak_rotation_utilization": flap2_peak_rotation_utilization,
"opt_stiffness_flap1": opt_stiffness_flap1,
"opt_stiffness_flap2": opt_stiffness_flap2,
"results": best_case["results"],
"wec_fdom_full": best_case["wec_fdom_full"],
"wec_tdom_full": best_case["wec_tdom_full"],
"pto_fdom": best_case["pto_fdom"],
"pto_tdom": best_case["pto_tdom"],
}
)
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:25:46] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING:capytaine.bem.solver:Mesh resolution for 19 problems:
The resolution of the mesh might be insufficient for omega ranging from 12.566 to 12.566.
This warning appears when the largest panel of this mesh has radius > wavelength/8.
WARNING Mesh resolution for 19 problems: The resolution of the mesh might be insufficient for omega ranging from 12.566 to 12.566. This warning appears when the largest panel of this mesh has radius > wavelength/8.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.021 seconds
Symbolic EOM computation finished in: 0.003 seconds
Compiling EOM expressions finished in: 0.066 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -0.4722878874610539
Iterations: 25
Function evaluations: 26
Gradient evaluations: 25
Optimization terminated successfully (Exit mode 0)
Current function value: -0.4722878891864572
Iterations: 22
Function evaluations: 25
Gradient evaluations: 22
Optimization terminated successfully (Exit mode 0)
Current function value: -4.250590985587285
Iterations: 17
Function evaluations: 20
Gradient evaluations: 17
Optimization terminated successfully (Exit mode 0)
Current function value: -4.2505909905573676
Iterations: 16
Function evaluations: 19
Gradient evaluations: 16
Optimization terminated successfully (Exit mode 0)
Current function value: -11.807197105638778
Iterations: 14
Function evaluations: 16
Gradient evaluations: 14
Optimization terminated successfully (Exit mode 0)
Current function value: -11.807197182031025
Iterations: 15
Function evaluations: 18
Gradient evaluations: 15
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:26:17] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.020 seconds
Symbolic EOM computation finished in: 0.002 seconds
Compiling EOM expressions finished in: 0.066 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -4.985812060491261
Iterations: 20
Function evaluations: 24
Gradient evaluations: 20
Optimization terminated successfully (Exit mode 0)
Current function value: -4.985812082765188
Iterations: 20
Function evaluations: 24
Gradient evaluations: 20
Optimization terminated successfully (Exit mode 0)
Current function value: -6.316251609737131
Iterations: 27
Function evaluations: 31
Gradient evaluations: 27
Optimization terminated successfully (Exit mode 0)
Current function value: -6.316251609728284
Iterations: 15
Function evaluations: 21
Gradient evaluations: 15
Optimization terminated successfully (Exit mode 0)
Current function value: -6.310441763356556
Iterations: 24
Function evaluations: 31
Gradient evaluations: 24
Optimization terminated successfully (Exit mode 0)
Current function value: -6.310441762815542
Iterations: 10
Function evaluations: 15
Gradient evaluations: 10
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:26:47] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.021 seconds
Symbolic EOM computation finished in: 0.003 seconds
Compiling EOM expressions finished in: 0.067 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -0.29746914782847783
Iterations: 21
Function evaluations: 22
Gradient evaluations: 21
Optimization terminated successfully (Exit mode 0)
Current function value: -0.0019202054189914894
Iterations: 5
Function evaluations: 6
Gradient evaluations: 5
Optimization terminated successfully (Exit mode 0)
Current function value: -1.600522110284948
Iterations: 16
Function evaluations: 19
Gradient evaluations: 16
Optimization terminated successfully (Exit mode 0)
Current function value: -1.600522110291004
Iterations: 19
Function evaluations: 21
Gradient evaluations: 19
Optimization terminated successfully (Exit mode 0)
Current function value: -1.6015540616052333
Iterations: 15
Function evaluations: 20
Gradient evaluations: 15
Optimization terminated successfully (Exit mode 0)
Current function value: -1.6015540697792665
Iterations: 16
Function evaluations: 19
Gradient evaluations: 16
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:27:17] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING:capytaine.bem.solver:Mesh resolution for 19 problems:
The resolution of the mesh might be insufficient for omega ranging from 12.566 to 12.566.
This warning appears when the largest panel of this mesh has radius > wavelength/8.
WARNING Mesh resolution for 19 problems: The resolution of the mesh might be insufficient for omega ranging from 12.566 to 12.566. This warning appears when the largest panel of this mesh has radius > wavelength/8.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.021 seconds
Symbolic EOM computation finished in: 0.003 seconds
Compiling EOM expressions finished in: 0.068 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -0.49450957839986254
Iterations: 23
Function evaluations: 25
Gradient evaluations: 23
Optimization terminated successfully (Exit mode 0)
Current function value: -0.4945095767343216
Iterations: 23
Function evaluations: 26
Gradient evaluations: 23
Optimization terminated successfully (Exit mode 0)
Current function value: -4.45058619090312
Iterations: 17
Function evaluations: 20
Gradient evaluations: 17
Optimization terminated successfully (Exit mode 0)
Current function value: -4.4505861880724975
Iterations: 17
Function evaluations: 20
Gradient evaluations: 17
Optimization terminated successfully (Exit mode 0)
Current function value: -12.362739418669578
Iterations: 15
Function evaluations: 19
Gradient evaluations: 15
Optimization terminated successfully (Exit mode 0)
Current function value: -12.362739418493595
Iterations: 16
Function evaluations: 19
Gradient evaluations: 16
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:27:45] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.022 seconds
Symbolic EOM computation finished in: 0.003 seconds
Compiling EOM expressions finished in: 0.070 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -4.836629438436521
Iterations: 20
Function evaluations: 25
Gradient evaluations: 20
Optimization terminated successfully (Exit mode 0)
Current function value: -4.836629448003479
Iterations: 29
Function evaluations: 34
Gradient evaluations: 29
Optimization terminated successfully (Exit mode 0)
Current function value: -6.320837968770919
Iterations: 8
Function evaluations: 10
Gradient evaluations: 8
Optimization terminated successfully (Exit mode 0)
Current function value: -6.32083796877009
Iterations: 14
Function evaluations: 20
Gradient evaluations: 14
WARNING:wecopttool.core:Iteration limit reached (Exit mode 9)
Iteration limit reached (Exit mode 9)
Current function value: 49.6279462938092
Iterations: 500
Function evaluations: 1997
Gradient evaluations: 501
[19:28:17] WARNING Iteration limit reached (Exit mode 9)
Optimization terminated successfully (Exit mode 0)
Current function value: -6.312186964577563
Iterations: 13
Function evaluations: 19
Gradient evaluations: 13
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:28:18] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.021 seconds
Symbolic EOM computation finished in: 0.003 seconds
Compiling EOM expressions finished in: 0.067 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -0.3223896950922931
Iterations: 21
Function evaluations: 24
Gradient evaluations: 21
Optimization terminated successfully (Exit mode 0)
Current function value: -0.001754466563553272
Iterations: 5
Function evaluations: 6
Gradient evaluations: 5
Optimization terminated successfully (Exit mode 0)
Current function value: -1.602846438773955
Iterations: 20
Function evaluations: 24
Gradient evaluations: 20
Optimization terminated successfully (Exit mode 0)
Current function value: -1.602846438408145
Iterations: 22
Function evaluations: 25
Gradient evaluations: 22
Optimization terminated successfully (Exit mode 0)
Current function value: -1.6006607084109574
Iterations: 15
Function evaluations: 20
Gradient evaluations: 15
Optimization terminated successfully (Exit mode 0)
Current function value: -1.600660708410552
Iterations: 19
Function evaluations: 24
Gradient evaluations: 19
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:28:48] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING:capytaine.bem.solver:Mesh resolution for 19 problems:
The resolution of the mesh might be insufficient for omega ranging from 12.566 to 12.566.
This warning appears when the largest panel of this mesh has radius > wavelength/8.
WARNING Mesh resolution for 19 problems: The resolution of the mesh might be insufficient for omega ranging from 12.566 to 12.566. This warning appears when the largest panel of this mesh has radius > wavelength/8.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.020 seconds
Symbolic EOM computation finished in: 0.002 seconds
Compiling EOM expressions finished in: 0.066 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -0.5155149335643173
Iterations: 23
Function evaluations: 26
Gradient evaluations: 23
Optimization terminated successfully (Exit mode 0)
Current function value: -0.5155149364967618
Iterations: 22
Function evaluations: 24
Gradient evaluations: 22
Optimization terminated successfully (Exit mode 0)
Current function value: -4.639634409950427
Iterations: 19
Function evaluations: 20
Gradient evaluations: 19
Optimization terminated successfully (Exit mode 0)
Current function value: -4.639634399477887
Iterations: 17
Function evaluations: 20
Gradient evaluations: 17
Optimization terminated successfully (Exit mode 0)
Current function value: -12.88787333975127
Iterations: 15
Function evaluations: 18
Gradient evaluations: 15
Optimization terminated successfully (Exit mode 0)
Current function value: -12.887873323859989
Iterations: 14
Function evaluations: 17
Gradient evaluations: 14
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:29:17] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.020 seconds
Symbolic EOM computation finished in: 0.003 seconds
Compiling EOM expressions finished in: 0.065 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -4.692228335300319
Iterations: 23
Function evaluations: 28
Gradient evaluations: 23
Optimization terminated successfully (Exit mode 0)
Current function value: -4.692228336889064
Iterations: 34
Function evaluations: 39
Gradient evaluations: 34
Optimization terminated successfully (Exit mode 0)
Current function value: -6.32800815770614
Iterations: 7
Function evaluations: 8
Gradient evaluations: 7
Optimization terminated successfully (Exit mode 0)
Current function value: -6.328008156999417
Iterations: 14
Function evaluations: 18
Gradient evaluations: 14
WARNING:wecopttool.core:Iteration limit reached (Exit mode 9)
Iteration limit reached (Exit mode 9)
Current function value: 16.160995310365873
Iterations: 500
Function evaluations: 2170
Gradient evaluations: 501
[19:29:47] WARNING Iteration limit reached (Exit mode 9)
Optimization terminated successfully (Exit mode 0)
Current function value: -6.314767175089055
Iterations: 11
Function evaluations: 15
Gradient evaluations: 11
WARNING:wecopttool.core:Using the geometric centroid as the center of gravity (COG).
[19:29:49] WARNING Using the geometric centroid as the center of gravity (COG).
WARNING:wecopttool.core:Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
WARNING Using the center of gravity (COG) as the rotation center for hydrostatics. Note that the hydrostatics do not use the axes defined by the FloatingBody degrees of freedom, and the rotation center should be set manually when using Capytaine to calculate hydrostatics about an axis other than the COG.
Problem initialization finished in: 0.001 seconds
All branches in the graph:
Branch 1: [1, 2]
Branch 2: [1, 3]
R and RD matrices calculation finished in: 0.021 seconds
Symbolic EOM computation finished in: 0.003 seconds
Compiling EOM expressions finished in: 0.067 seconds
[Info] Updating float body 1 IC from (x=0.0, z=0.0) → (x=0.0, z=-0.8) relative to Reference_frame_Origin: (x=0, z=0) .
[HydroAdapter] Prepared bodies:
[1] name='platform', CoM=[ 0. 0. -0.8]
[2] name='flap1', CoM=[-0.72 0. -0.42]
[3] name='flap2', CoM=[ 0.72 0. -0.42]
Optimization terminated successfully (Exit mode 0)
Current function value: -0.3523215870764252
Iterations: 23
Function evaluations: 25
Gradient evaluations: 23
Optimization terminated successfully (Exit mode 0)
Current function value: -0.0016077541350345028
Iterations: 5
Function evaluations: 6
Gradient evaluations: 5
Optimization terminated successfully (Exit mode 0)
Current function value: -1.607822692726673
Iterations: 18
Function evaluations: 22
Gradient evaluations: 18
Optimization terminated successfully (Exit mode 0)
Current function value: -1.6078226938545082
Iterations: 20
Function evaluations: 22
Gradient evaluations: 20
Optimization terminated successfully (Exit mode 0)
Current function value: -1.6005326595019904
Iterations: 13
Function evaluations: 17
Gradient evaluations: 13
Optimization terminated successfully (Exit mode 0)
Current function value: -1.600532664114216
Iterations: 17
Function evaluations: 21
Gradient evaluations: 17
3.3 Top-mounted ballast study results
The results of the top-mounted ballast study are summarized in the plots below. Adding ballast can improve average power by reducing the flap hydrostatic stiffness and increasing flap motion. However, the benefits are limited by the peak flap pitch contraints. As a result, top-mounted ballast is most beneficial in smaller-amplitude waves, while larger-amplitude waves may already reach the rotation constraint and therefore see little additional benefit from destabilization.
[13]:
fig, axs = plt.subplots(
4,
len(wave_periods),
sharex="col",
sharey="row",
figsize=(2.6 * len(wave_periods), 9),
gridspec_kw={"hspace": 0.15},
constrained_layout=True,
squeeze=False,
)
colors = ["#9ecae1", "#4292c6", "#08306b"]
markers = ["o", "s", "^"]
for j, period in enumerate(wave_periods):
for amplitude, color, marker in zip(wave_amplitudes, colors, markers):
case_list = sorted(
[
c for c in case_outputs
if np.isclose(c["wave_period"], period)
and np.isclose(c["wave_amplitude"], amplitude)
],
key=lambda c: c["top_mounted_ballast"],
)
if len(case_list) == 0:
continue
ballast = np.array([c["top_mounted_ballast"] for c in case_list])
mean_elec_power = np.array([c["electrical_power"] for c in case_list])
flap1_util = np.array([c["flap1_peak_rotation_utilization"] for c in case_list])
flap2_util = np.array([c["flap2_peak_rotation_utilization"] for c in case_list])
stiffness_flap1 = np.array([c["opt_stiffness_flap1"] for c in case_list])
stiffness_flap2 = np.array([c["opt_stiffness_flap2"] for c in case_list])
baseline_power = mean_elec_power[np.where(np.isclose(ballast, 0))[0][0]]
power_improvement = 100 * (mean_elec_power - baseline_power) / baseline_power
axs[0, j].plot(ballast, mean_elec_power, marker=marker, color=color, alpha=0.85)
axs[1, j].plot(ballast, power_improvement, marker=marker, color=color, alpha=0.85)
axs[2, j].plot(ballast, flap1_util, marker=marker, linestyle="-", color=color, alpha=0.85)
axs[2, j].plot(ballast, flap2_util, marker=marker, linestyle="--", linewidth=2, color=color, markerfacecolor="none", alpha=0.85,)
axs[3, j].plot(ballast, stiffness_flap1, marker=marker, linestyle="-", color=color, alpha=0.85)
axs[3, j].plot(ballast,stiffness_flap2, marker=marker, linestyle="--", linewidth=2, color=color, markerfacecolor="none", alpha=0.85,)
axs[0, j].set_title(f"T = {period:g} s")
axs[3, j].set_xlabel("Top-mounted ballast [kg]")
for ax in axs.flatten():
ax.grid(True, alpha=0.3)
axs[0, 0].set_ylabel("Electrical power [W]")
axs[1, 0].set_ylabel("Power improvement [%]")
axs[2, 0].set_ylabel("Peak rotation utilization [-]")
axs[3, 0].set_ylabel("Optimal stiffness [N·m/rad]")
amplitude_handles = [
Line2D([0], [0], color=color, marker=marker, linestyle="-", label=f"A = {amp:.2f} m")
for amp, color, marker in zip(wave_amplitudes, colors, markers)
]
axs[0, -1].legend(handles=amplitude_handles, fontsize=8, framealpha=0.9, loc="upper right")
axs[1, -1].legend(handles=amplitude_handles, fontsize=8, framealpha=0.9, loc="upper left")
flap_handles = [
Line2D([0], [0], color="k", linestyle="-", label="Flap 1"),
Line2D([0], [0], color="k", linestyle="--", label="Flap 2"),
]
amp_legend = axs[2, -1].legend(handles=amplitude_handles, fontsize=8, framealpha=0.9, loc="lower left")
axs[2, -1].add_artist(amp_legend)
axs[2, -1].legend(handles=flap_handles, fontsize=8, framealpha=0.9, loc="lower right")
amp_legend = axs[3, -1].legend(handles=amplitude_handles, fontsize=8, framealpha=0.9, loc="upper left")
axs[3, -1].add_artist(amp_legend)
axs[3, -1].legend(handles=flap_handles, fontsize=8, framealpha=0.9, loc="upper right")
[13]:
<matplotlib.legend.Legend at 0x7f3464d1c8c0>
