FLOATING POINT ERROR - OVERFLOW in inversion to obtain unsaturated parameter

[Yesser]

I am trying to obtain the parameters of the soil-water retention curve through the inversion of infiltration test data using a tension disc infiltrometer with a 20 cm diameter. The tests consisted of four stages with suctions varying at -20, -15, -10, and -3 cm, each lasting 50 minutes. Readings were taken every 5 minutes during the first three stages due to minimal differences between consecutive readings, and every 2 minutes during the last stage.

To achieve this, I created an axisymmetric model with a height of 30 cm and a width of 40 cm. The infiltrometer radius (10 cm) is represented by a "variable head" boundary condition, while the model's bottom is set to "free drainage," and the rest of the domain has a "no flux" condition. The initial condition was based on the soil moisture content, and the final moisture content is an inversion parameter. The model uses a finite element mesh with a 2 cm grid size and a refinement of 0.5 cm near the disc location.

It is worth noting that the soil exhibits bimodal behavior in the retention curve, as determined by other methods. However, the modeling does not converge and displays a "Floating point error - Overflow." I would like to know if there are any recommendations to help the model converge for all my tests, which involve a total of nine soils.

[Jirka]

You should look at the Disc example, which is distributed with the software on how to correctly set up this problem. You have a wrong definition of the objective function. First, the flux values are negative for inflow. Second, the Position should be 3 (a time-variable head BC), not 1 (which is a constant head BC). J.

[Yesser]

Thank you very much for your response, Professor. I made the suggested changes, but the model still does not converge satisfactorily. I reviewed the example in DISC and noticed that the fluxes entered in the software are positive, while in the manual they are negative. I tested both approaches in the example, but it still does not converge. I am attaching the file again for any recommendations, which I would greatly appreciate.

[Jirka]

This is clearly not true. The cumulative fluxes entered in the software to define the objective function must be negative. I do not see why you would want to argue this with me. I’m not planning to look at your project again. J.

[Yesser]

I'm Sorry, that was not my intention. I truly appreciate your advice and help, as it has cleared up my doubts. I will keep working to improve my model, and if necessary, I will try to ask again.

[Jirka]

Set constraints on n, so that it cannot get lower than 1.1. J.