Since your question is related to modeling in OPL you should ask it on the OPL Forum.
In any case, the square root function is neither linear nor quadratic, hence it cannot be used. What you can try is to reformulate. For example,
dexpr float d= sqrt(d1+d2+d3);
minimize d;
is equivalent to
dvar float+ sqrtd;
minimize sqrtd;
subject to {
sqrtd^2 >= d1 + d2 + d3;
...
}
However, this turns out to be a non-convex quadratic constraint in your case, which CPLEX cannot handle either.
Maybe you can find another way to reformulate your model?
#CPLEXOptimizers#DecisionOptimization