Current increase in induction motor

I have an induction motor which is fed by a three phase inverter. I am trying to use vector control with d-axis aligned with the rotor flux vector. I have a load which increases its torque as the speed increases. I use a fixed value for direct component of current (Id) reference.

I start to test at low speed and everything goes fine, voltage amplitude increases with frequency and current is almost constant. But when I try greater speeds, at some point, the voltage amplitude reduces notoriously (instead of keep increasing), and quadrature component of current (Iq) also increases notoriously.

I also observed that Iq has a very low value before the problem occurs. Almost all current is contributed by Id. Is this low value of Iq a normal thing? or should it be increasing with speed? Is there some kind of relationship betbeen Id and Iq, that I should follow? I suspect this is related to saturation. How can I avoid this?

UPDATE: I tested with a larger value of field generating current (which I call Id) and the problem reduces, but still happens at higher speeds. So I believe may be is not saturation, but could it be like the motor looses its field?

is it normal that my controller tries keeps Iq at low values, even at high torque?

Id and Iq are perpendicular to each other. We can say that Id is the excitation current, while Iq is torque current in the same fashion as brush DC motor with separate excitation, where Id is the field and Iq is the armature current.

There is a certain relationship between voltage and current, V/f characteristics. The voltage increases with speed and at certain point it can't go further (supply voltage). Beyond this speed it is possible to increase the speed by field weakening method - reducing the Id current, which reduces the flux.

The output torque is the product of both currents Id and Iq. within the nominal speed operation the Id is constant, while the Iq is has a mirrored value of the load torque, when accelerating is positive, when deccelerating is negative - generator mode, when the motor rotates unloaded, steady speed it should have a minimal value - bearing friction torque.

Without more details is difficult to find where the problem in your case is. The Id should be constant, the Iq small if you spin it with low acceleration. Did you made this vector control by yourself? What is the hardware used : eval board, FOC library,...? Did you placed current transformers correctly?