Problem with b.c. of k-e model
 

I have jast made 3d k-e model for channel flow.

My program is checking the errors of u,v,w,k,e while it is running.

I got a problem of convergence of error of epsilion.

the others are ok except epsilon.

So, I forced to converge epsilon then check the results. I have no idea why the value of epsilon has suddenly big value at the free stream boundary condition. Also, the value of k is reducing value with respect to height of channel but, it is suddenly incresing near the free stream. I did a lot of times check the code. but i coudn't find whats wrong with this code. Is there any idear for me?

Thank you