I'm trying to solve an MINLP problem where the following division term is appearing in the objective: $$z_r = \frac{x_{ry}}{\sum_r d_r x_{ry}}, \forall r, y,$$ where $x_{ry}$ is a 2D binary variable and $d_r$ is a non-zero real number. In addition, there is a constraint $\sum_r x_{ry} \leq 1$. Is there a suitable way to linearize this division?
I tried to use a new variable $M_r = z_r \times \sum_r d_r x_{ry}$, but the situation is still the same for the commercial solvers.