Thanks for your answer.
1) I wanted to start with a simple ODE because this way I could verify the confiability of my algorythm. The equation I'm really interested in solving is:
x'' + A*x + B*x*(x')^2 + C*x^3 = 0, where A,B and C are known parameters.
The scaling parameter ("e") is applied to the initial conditions, which should have the form:
x(0) = e*x(0) + e^2*x(0) + h.o.t
x'(0) = diff(x,t) | x = 0
2) You're right. x[j] depends on T,T and T; not only on T.
3) I agree with you, but, since I will need to use the MMS for a branch of equations, I thought the automatising would be a good idea. I have done it manually for a first approximation. Doing a second one, though, with terms of higer order of "e", would be easier if done autommatically. The main issue here is to rewritte the ODE, using the variable transformations suggested by the MMS.