Wireless sensor network can provide long-term and real-time data support to earthen sites preventive conservation. The knowledge obtained from the monitoring data could provide decision supports for earthen sites preventive conservation. How to solve missing, faults and systematic errors has been an urgent problem for the improvement of data quality. And it is also imperative to obtain the scientific laws between existing environment and survival conditions of earthen sites. At first, the methods of data cleaning are studied with the comparison analysis among average method, Bayes method and regression model via defining the evaluation criterion of data quality in this paper. Then the common norms between environment conditions and survival conditions of earthen sites are analyzed and summarized. And the danger theory of immune systems is used as immune factors to build survival model. From the processing of monitoring data of Daming Palace soil sample, results show that regression model method has the highest completion accuracy rate among three methods. Cracks growth conditions and stress tolerance also verify the certain accuracy rate of survival condition model.