You can use xparse:

\documentclass{article}
\usepackage{amsmath}
\usepackage{xparse}

\ExplSyntaxOn
\NewDocumentCommand{\repsum}{O{3}mmm}
 {% #1 = optional number of starting summands
  % #2 = final number
  % #3 = first symbol
  % #4 = second symbol
  \int_step_inline:nn { #1 } { #3\sb{##1}#4\sb{##1} + }
  \dotsb
  \int_step_inline:nnn { #2 - 1} { #2 } { + #3\sb{##1}#4\sb{##1} }
 }
\ExplSyntaxOff  

\begin{document}

First test: $\repsum{9}{F}{u}$

Second test: $\repsum[2]{6}{F}{u}$

The CUF Refined theory expands the summation as
\begin{equation}
u=\repsum{9}{F}{u}=F_\tau u_\tau
\end{equation}
where the last expression exploits the Einstein notation. 

\end{document}

The idea is to make a cycle from 1 to 3 (or the number specified in the optional argument), printing the summands with their subscripts followed by +; then print the dots and + followed by the summands from #2-1 (#2 is the final number of summands) to #2.

With this implementation, you are responsible for ensuring no overlap. So you can do \repsum[1]{4}{F}{u}, but with less than four summands it won't work.