For the problem that single testing method is difficult to achieve comprehensive assessment of the object，the comprehensive testing using a variety of nondestructive testing methods is suggested to be effective，in which the credibility assessment of comprehensive test results is critical． In the investigation，firstly，the cloud nondestructive testing system architecture for large-scale steel structures was designed based on Hadoop，and the information flow of large-scale steel structures cloud nondestructive testing Hadoop architecture was analyzed． Then，the D-S evidence theory joint operator of credibility fusion of large-scale steel structures cloud nondestructive testing data was defined，and the multi-source data credibility MapＲeduce fusion algorithm based on D-S evidence theory was proposed． Finally，a large steel tube tower structure testing experimental platform was established，and the credibility of visual endoscopic inspection，eddy current testing and conventional ultrasonic inspection multisource data was tested by fusion evaluation． The results show that the detection rate of defects in each single testing method is improved by the MapReduce fusion algorithm for the credibility of cloud testing data． It meets the cloud nondestructive testing requirements of large-scale steel structures