The reconstructions of 3D temperature fields by acoustic tomography are simulated studied. Four planes along the axial of the space to be measured are selected as typical planes. Eight acoustic sensors are amounted on each plane’s periphery. Using the reconstruction algorithm based on radial basis function approximation and Tikhonov regularization we proposed, the 2D temperature fields in the four planes are reconstructed, then the 3D temperature field is reconstructed by interpolation of the 2D images. Reconstructions of three temperature field models by using exact and corrupted time-of-flight data demonstrate that the method proposed in this paper has good ability in complex 3D temperature field reconstruction.