The utility model discloses a neural net generalized inverse decoupling controller for a bearingless synchronous reluctance motor which consists of a composite controlled object, a neural net generalized inverse and a linear closed-loop controller, wherein the neural net generalized inverse and the linear closed-loop controller are in series connection in front of the composite controlled object in sequence; the composite controlled object is a whole formed in a way that a first Park inverse converter, a second Park inverse converter, a first Clark inverse converter, a second Clark inverse converter, a first current tracking inverter and a second current tracking inverter that are in series connection in sequence are connected in front of the bearingless synchronous reluctance motor; the neural net generalized inverse consists of a static neural net and 5 linear links; a generalized pseudolinear system is formed in a way that the neural net generalized inverse in series connection in front of the composite controlled object; the generalized pseudolinear system comprises two position subsystems and a speed subsystem; correspondingly, the linear closed-loop controller consists of two position controllers and a speed controller; and the neural net generalized inverse decoupling controller realizes independent decoupling control among components between an electromagnetic torque and a radial suspension force as well as in two vertical directions of the radial suspension force.