First, log-linearisation using first order Taylor expansion is what was used in Uhlig's paper. It is a common way to use. Maybe I misunderstood your YY't....That should be steady state value of output times percentage output deviation to its steady state value. Similar structure on the right hand side.
Second, once you get it right. You can put the equation in Dynare. Depends on how you calibrate your model, you can figure out what are the steady state value. How do you calibrate your steady state investment to output ratio? There are loads of literature from which you can find the calibrations. Otherwise, you can estimate them with some prior conditions.
Last, I cannot see your attached file.