Matlab crashing when using MoorDyn in WEC-Sim

[brittlyd]

I am trying to use MoorDyn in WEC-Sim to create a tether between two hydrodynamic bodies (a surface-piercing float and a submerged heaveplate). However, when I try to run this, my Matlab crashes. When I reported this to WEC-Sim, they mentioned that this capability should be possible, but recommended I reach out here.

Here is a drawing of the setup:

Here is a screenshot of my lines.txt file:

Here is a screenshot of my wecSimInputFile.m:

And here is a screenshot of my Simulink model:

I also attached a zip file of the full WEC-Sim folder.

Can you help me figure out why Matlab keeps crashing and if this problem is possible?

Thank you!
Brittany

MBARIWEC_v1_tether2.zip

[RyanDavies19]

Hi @brittlyd,

Can you share the error messages you are getting? Or is MATLAB not printing any errors?

[brittlyd]

Hi @RyanDavies19

Thanks for your quick reply. No errors are printed and there's not a crash report that I can see.

[RyanDavies19]

Linking to WEC-Sim/WEC-Sim#1620

[RyanDavies19]

It seems like your coupled body 2 position is being overwritten by WECSim to be 0,0,0. If you look at your MBARIWEC_v1_tether/mooring/lines_Line1.out file you can see that the pz of your end nodes is at -0.45 and 0.35 (these outputs are in the global frame). This means that in your WECSim set up, the location of Body 2 is not being initialized as 0, 0, -29.9. Hopefully that helps!

[brittlyd]

I changed the point location for body2, and Matlab still crashes, but it looks like the initial setup of the tether looks correct. I attached a screenshot of the lines.txt and the new zip folder of the results. In that folder, there's a script called plotTether.m that should help visualize the initial node positions.

MBARIWEC_v1_tether3.zip

[RyanDavies19]

Hi @brittlyd,

It seems like your MoorDyn model is unstable. It could be helpful to run it without WECSim to see if you can isolate the problem. Below is an example script that should work if you pip install moordyn that will run MoorDyn in isolation.

Modified from: https://github.com/FloatingArrayDesign/MoorDyn/blob/master/example/simpledemo.py

import moordyn
import numpy as np

### OPTIONAL ###
# import matplotlib.pyplot as plt
# import moorpy 

'''
This is a very simple example of how to use the MoorDyn Python interface. Users need to provide a moordyn input file before running.
This code is ready to run if a user provides an input file with a single coupled body. A more thorough explanation of how to drive 
MoorDyn can be found in `MoorDyn_standalone_demo.ipynb`.

The framework here is set up for a single coupled body. For other configurations, the state (x) and state derivatives (xd) need to be 
changed accordingly.
'''

# Create time array
tMax = 0.5 # max time for running time series
dt = 0.001 # coupling timestep
time = np.arange(0, tMax, dt) # time series

# Build arrays of state and state derivatives
size = (len(time), 6)
x = np.zeros(size)
xd = np.zeros(size)

# Create the MoorDyn system from the input file. 
# NOTE TO USERS:
#   Change this to your own input file. 
#   It must have one coupled body and no other coupled objects. 
system = moordyn.Create("<file name>")

# You can get the initial positions and velocities from the system itself using the MoorDyn-C API
body1 = moordyn.GetBody(system, 1)
state1 = moordyn.GetBodyState(body1) # tuple with (x, xd)
body1 = moordyn.GetBody(system, 2)
state1 = moordyn.GetBodyState(body2) # tuple with (x, xd)

# Set the first time step of the state and state derivatives to the initial values from MoorDyn
x[0,:] = np.array(state1[0], state2[0]) # x, y, z, roll, pitch, yaw, x, y, z, roll, pitch, yaw
xd[0,:] = np.array(state1[1], state2[1]) # xdot, ydot, zdot, roll_dot, pitch_dot, yaw_dot, xdot, ydot, zdot, roll_dot, pitch_dot, yaw_dot

# Setup the initial condition
moordyn.Init(system, x[0,:], xd[0,:])

# Run the simulation for five timesteps (with states held constant)
for i in range(1, len(time)):
    # call the MoorDyn step function
    f = moordyn.Step(system, x_t[0,:], xd_t[0,:], time[i], dt)    # force value on coupled DOF returned here in array

# Alright, time to finish!
moordyn.Close(system)


### Optional ###

# # Plot and animate the results using MoorPy

# # Create a MoorPy system (this loads the moordyn outputs found in <dirname>/<rootname>_<object>#.out, where <object> is either Line or Rod and # is the corresponding number)
# ms = moorpy.System(file=<path+rootname+extension>, dirname=<path>, rootname=<rootname>, qs = 0, Fortran = False) # qs tells MoorPy it is loading a MoorDyn system, Fortran = False tells it to use the MoorDyn-C output format

# # Plot the results at t = 0 in red
# ms.plot(color="red", time = 0)

# # animate the results
# animation = ms.animateLines()

# plt.show()

[RyanDavies19]

Hi @brittlyd,

Just a thought: It could be that your MoorDyn system is initializing in some unstable configuration due to the static solver (which can struggle in edge cases like yours). Try adding 1 ICgenDynamic to your MoorDyn options list to have MoorDyn use the slower (but sometimes more stable) dynamic relaxation solver.

[brittlyd]

Hi @RyanDavies19,

Thank you for your help! I don't know how to run MoorDyn without WEC-Sim, so I tried your second suggestion first and added the ICGenDynamic option to the bottom of my lines.txt file. When I did that, the WEC-Sim simulation ran and Matlab didn't crash! However, I got the following error for each time step:

ERR D:\a\MoorDyn\MoorDyn\source\MoorDyn2.cpp:751 moordyn::MoorDyn::Step(): t = 0.4 s: NaN in node positions
ERR D:\a\MoorDyn\MoorDyn\source\Line.cpp:943 moordyn::Line::getStateDeriv(): NaN detected
Line 1

Then, I tried to run it again, Matlab crashed again. I tried again, and it made it all the way through the simulation, but Matlab crashed when trying to save the WEC-Sim results. I'm not sure why it was able to run only once without crashing, but it seems like the ICgenDynamic did seem to help?

[brittlyd]

I was able to get the simulation to run again (I made the time step smaller in lines.txt) and was able to save the simulation printout with the errors (attached). It looks like the x node positions are blowing up and I'm not sure why? The body x positions are not blowing up in the WEC-Sim outputs.

simPrintout.txt

Thanks again!

[RyanDavies19]

Hi @brittlyd,

Good news, that means we are on the right track. I suspect that your initialization is still causing the system to start in an unsteady configuration. Try turning down the drag scaling (CdScaleIC) and adding 2.0 dtIC (this forces the initialization to search for when the mooring tensions are steady for > 2.0 seconds). Those changes will help make initialization more sensitive to the true steady state. I also noted that you do not have any line damping, which is likely also contributing to instability. Is this intentional? Commonly, people set the damping value to -1, which critically damps the natural frequency of the lumped mass system (which is most frequently artificially excited by numerical instabilities). More info on that is here: https://moordyn.readthedocs.io/en/latest/troubleshooting.html#model-stability-and-segment-damping

You could also try changing the number of segments you have. I would think that 15 segments would be sufficient for your system.