@nathanmtom and @jleonqu can you review this and run it in 2020b?

Aside from a few small typos in the bodyClass, all changes are in one function in the Simulink library:
WECSim_Lib/Body Elements/Rigid Body/Hydrodynamic Body/Wave Radiation Forces Calculation/SS CI and Constant-Damping-CoeVariant Subsystem/Convolution Integral Calculation/ConvolutionIntegral_interp

Block location:

Original CI function:

New CI function:

I was able to check this pull request locally in my computer. I run the case using Matlab 2020b and it works.

I have discussed this issue with Mathworks support. They confirmed that R2020a contained a bug allowing Simulink to compile without an error message. This allowed bsxfun() to function incorrectly and did not give the intended results. The bug was fixed in R2020b.

My update permutes the velocity and IRKB dimensions within the convolution integral function so that the dimensions match correctly for multiplication. 8/25 continuous integration tests still fail, because of the previous version's inaccurate bug.

@nathanmtom When you have time, can you please review the updated convolution integral function and check that it is calculating the integral correctly according to the theory documentation?

Mathworks support also confirmed to me that this bug first appeared in R2017b (released September 2017). In 2017b, the requirements for dimension expansion were relaxed. It became possible for MATLAB to miss a runtime error for incompatible dimension operations, such as bsxfun().

I checked releases v2.2 and v3.0 which were before/after R2017b released. The convolution integration function does change between these versions, but I have not confirmed if v2.2 should have had this same error. It is likely the error in the changed function was missed because it overlapped with the simulink bug release which allowed it.

I am currently testing the accuracy of this change using the B2B application. This app compares regular wave, regular+B2B, regularCIC, regularCIC+B2B, regularCIC+SS, regularCIC+B2B+SS cases. The new CI update should match the state space response very closely. However there are some significant differences now between the CI and SS cases as shown in the radiation damping force plots:

Radiation damping in Surge:

Radiation damping in Heave:

Radiation damping in Pitch:

I solved the last problems with this bug. In addition to the permutations in the CI block, there was an issue with the selector block that chose the velocity values to pass into the CI function. I have solved both issues and the function now works correctly. All continuous integration tests are passing.

The below figures show that the CI calculations (black lines) are now more closely aligned with the state space calculations (light blue) and likely more accurate than previously.

Surge:

Heave:

Pitch: