How to protect the endangered languages is a typical and important social problem around the world. Now, the existing social computational models of language evolution are generally short of language endangerment warning and language recovery. These two parts are very important for language evolution. In this paper, an artificial society is cultivated based on multi-agents with purpose of observing language endangerment and its recovery phenomenon. In view of endangerment factor of population language, resilience inner protective factor and resilience composite protective factor, corresponding social computational model is given. From experimental results we can see that the computational model draws the endangerment trend of the weaker one in the language competition and the recovery process of weaker endangered language under the effect of resilience composite protective factor. To break through research bottleneck for high cost of language endangerment experiment in real society, the proposed computational model is working in an artificial society. The introduction of artificial society plays an important and practical role in the design and evaluation of endangered language protection scheme.