On the empirical side, there might be answers for you in the national accounts. I only know about the french case : the french statistical institute (INSEE) has different depreciation data for different types of capital (e.g. buildings, machines, patents) and for different sectors. These data are supposed to reflect both physical depreciation and "normal obsolescence". The only reference I can give you is in french : Jean-François Baron, "Les comptes de patrimoine et de variation de patrimoine, base 2000", INSEE, 2008, http://www.insee.fr/fr/indicateurs/cnat_annu/base_2000/documentation/methodologie/nb10.pdf , but I suspect you can find similar data in other national accounts.
On the theoretical side, if I understand you, you mean that a capital good can become economically obsolete before it is physically out of use. I see three possible reasons (possibly overlapping) for this :
the price of inputs (e.g. the wage rate) rises so that this capital good is not profitable any more
the price of output falls but not the price of inputs, so that it is not profitable
production is demand constrained and there other more productive ("new") capital goods available, so that our capital good will remain idle (although it could be profitable if there were a higher demand)
To account for this problem, you have to drop the representation of capital as a single homogeneous magnitude (the number "K") and have different capital goods with different properties (productivity and input requirements). This is called a "vintage capital" model. I don't know if this has been done in an RBC or DSGE framework, but a quite complete neoclassical paper dealing with these issues is :
Solow, Yaari, Tobin, Weiszäcker (1966), "Neoclassical growth with fixed factor proportions"