Convergence problem about dry cell

#Case target
In my case need to set 2 radial wells.

A radial well have 10 pipes, each pipes should extract 5,000 CMD.

Then a radial well will extract 50,000 CMD.

2 radial wells will extract 100,000 CMD.

#My model
1. Unsaturated flow (Richard's), the well zone van Genuchten parameters is: (alpha=14.5, n=2.68, sr=0.045).

2. One aquifer and one aquitard.

3. Set radial well by discrete feature with well BC

#Problem
I just set the first radial well (50,000CMD), then get non-convergence.

I have try to smooth mesh but not work.

According to the reference: https://www.waterloohydrogeologic.com/2017/10/20/experiencing-non-convergence-error-here-are-some-tips/, I consider the dry cell is the reason that result in non-convergence.

If the well location (elevation) cannot adjust(layer thickness fixed), how can i solve the convergence problem?

*The detail setting and result as shown in picture.

Your saturation data indicates that the aquifer's hydraulic conductivity is not sufficient to sustain such pumping rate without pumping well running dry (which not surprising given that 5E4 m³/day is extremely high pumping rate).
You can check how much pumping your aquifer can sustain by setting a head constraint equal to elevation for your well nodes. This would limit pumping to the rate at which wells remain in the saturated zone (the rate value can checked by looking into water budget for the well nodes). In terms of setup, I would remove well nodes at the top of the aquifer - these have very limited available head are much more likely to go dry.
I would also check some analytical solutions for partially penetrating wells - this should give you a ballpark estimate of pumping rates feasible in this setup.