At the inlet, did you set the volume fraction of feed to "1" and all others to 0?
Otherwise you will pipe phase 1 as default in, which might be different to feed...

As I said before, you can use patch to set initial volume fraction of each phase. or use UDF.

I did set the volume fraction of feed by 1, but same problem occur too.

I don't see how that helps here. He does not seem to have a problem with initial values, but with his settings. The pictures at t=0 look like they should...

Some setting is wrong. Which multiphase model do you use? Please post values / settings you changed in Fluent?

here is the setting part 1

here is the setting part2
Thanks

Hi,

1)
Are you sure about your time step size?
With 20m/s at the velocity inlet and dt=60s the fluid will move 1200m each timestep! I am afraid that these settings can't converge... Can you show your residual history?

2)
Why don't you set the velocity normal to the surface at the velocity inet, but instead you use a direction vector?

EDIT: I just realized that you did some curious things in your calculation settings... Do you know, what all these values mean? Timestep, Number of Iterations, ...

Yes you can!

This is the first time I'm contact with CFd, I still fresh with it and still got a big room for improvement. I'll spend more to time on it.
Btw, in order to maintain filling rate at 20m/s, what time step and interaction I should use?

I would: Estimate how far I want to travel the feed during one timestep. Let's say 0.5m.
Now, your time step has to be 0.5m / 20m/s = 0.025s.
Then, you set the number of iterations per timestep to 1. This can't work. Since the set of equations is nonlinear, you have to iterate each time step for convergence. I would set the number of iterations to 25 and see how the residuals behave.

You could do 10 time steps with these settings and post a picture of the residuals here. Good luck!

here is the residual result by using time step of 0.05, Number of timestep=10, and interations/time step=25.

This looks bad...
How did you initialize your domain?

How you differentiate between good and bad residual from the graph?
Sorry, I can't understand your question.. Can u further explain to me?

It looks like some of the residuals nearly don't change at all during these iterations. This should not happen.

What do you use for initialization? There is a button "initialize" at "Solution->Solution Initialization->" Do you use "hybrid"? I would go for "standard" and set every text box to zero. Then, hit "Initialize".

Can you post a picture of your mesh?

i set all the value as 0 except the temperature=113.

Yes you can, except that you'll spend your time on forums to find clues. When it comes to complex cases, you'll be completely lost and won't be able to tell if your results make sense especially if you do not have validation data.

I still believe that to get the most out of CFD, you need to study it. Everyone can tick boxes but understanding what you are doing is different story.

Send me your case file and I'll tell you what you are doing wrong

Can you give me your personal email?

Now, what was the problem?