do not know what your thermoset matrix is composed of. I suppose this is very viscose?
In that case sonication will have only poor effect. Ball milling or similar with high shear forces, e.g. extrusion, is said to be more effective in this case. Of course your nanoparticles should be compatible with the matrix, otherwise after dispersion they will reagglomerate. For using sonication you may use first a volatile solvent compatible with both particles and matrix and sonicate. Then mix with matrix and evaporate.