Abstract: Based on similarity theory and dimensional analysis, a similarity index was deduced as a criterion to judge whether the test result is similar to the reality or not. Taking the mixture of sand, gypsum and diatomite as imitative specimens, a similarity simulation test was performed. The primary experiment was designed orthogonally where the level of similarity was controlled by a compression similarity index. Through a level-difference analysis, the importance of four influencing factors on mechanical properties were revealed as similarity indices. Then, the secondary experiment was performed with specimens' mixture ratio readjusted in a small range, thus finding out an optimal mixture ratio in conformity to the similarity indices based on required mechanical properties. In this way the orthogonally designed experiments were used instead of overall ones so as to reduce greatly the number of experiments with time saved. Experiment results showed the similar materials' mixture ratio and how the mechanical indices change, which can offer a reference to the project of which the materials to be used are similar to those in the simulation.