根据现有相似理论推导出非完全相似定律，并提出满足模型材料本构关系下降段的非完全相似要求的处理技巧。根据已有模型材料的特性，数值重构地震动力破坏模型试验，将非完全相似模型与原型及完全相似模型计算结果进行比较发现，在坝体振型及损伤的分布方面，非完全相似模型与原型结果十分一致，此外，非完全相似模型与原型的频率及最大拉应力都是符合非完全相似定律的。当非完全模型材料泊松比由0.17变为0.2时，由模态分析可知模型结构频率及振型的变化十分微小。结构的Rayleigh阻尼与其各阶频率及振型的阻尼比有关，因此计算时采用原、模型各自的Rayleigh阻尼解决了模型试验中阻尼相似的问题。总之，通过数值重构地震动模型试验可知，在试验设备及条件、模型材料不完全满足相似要求的情况下，根据非完全相似定律设计模型，并采取一定的相似技巧是可以得到稳定可靠的试验结果的。

The non-full similarity laws are deducted on the basis of the existing similar theory. The disposal skill is proposed to meet non-full similitude requirements for descending stage in constitutive relationship of model material. According to the properties of the model material, the numerical reconstruction of dynamic failure modal test is conducted to compare the results of prototype, full and non-full similarity models. It is found that the non-full similarity model has a good agreement with prototype in the first four modes and distribution of maximum tensile stress and damage in dam body, and the relation between the natural frequencies and maximum tensile stresses of prototype and non-full similarity model meet the similarity requirements. When the poison’s ratio of the material model is changed from 0.17 to 0.2 for non-full similarity model, there are no significant changes for the natural frequencies and models. The Rayleigh damping is relevant to the frequencies and models damping ratio of structure, so the Rayleigh damping of numerical prototype and model is applied to solve the problem of damping in model test. In summation, the analysis of numerical reconstruction of model test implicates that model test results are reliable with model designed according to non-full similarity laws and the treatment skills adopted for similarity requirements under the cases which the experiment equipment and condition, model material don’t meet all of the similitude requirements.