For reconstructing three-dimensional (3D) temperature field of small-scale and impenetrable solid, such as printed circuit board (PCB), a new method based on fuzzy relationships between temperatures of different points was proposed and software was developed for solving the issue. Firstly, fuzzy incomplete relationships between temperatures of inside points and surface points were introduced. Secondly, from the fuzzy incomplete relationships complete relationships could be obtained using inference process of Markov logic networks. Finally, based on constraint least-squares support vector machine (CT_LSSVM) and the temperature field on the surface of the solid the 3-D temperature field inside the impenetrable solid could be constructed. The experiment results showed that the proposed algorithm not only has excellent generalization performance and prediction accuracy but also offers explanation for the predict results, which can provide some decision support for temperature controls on PCB and even real-time decision support.