In the development stage and use stage, reliability demonstration test is important in determining products' reliability at a certain time. However, it is usually difficult to get the distribution of products with little testing data. In this paper, four methods for determining distribution parameters are proposed and compared. They are based on optimization with complex boundary and D-S evidence theory. These method can decrease the uncertainty of maximum entropy method by fusing several types of expert judgment and other prior information. By using bound search and visualization, the principle of decreasing the uncertainty is found to narrow the feasible region by changing the bound determined by prior information. And based on the result, an example is calculated by using different methods. The results indicates that by fusing prior information can effectively decrease risks of reliability demonstration test plan and optimization method will give lower risks for the same test plan than D-S evidence method.
 

D-S evidence theory,reliability demonstration,maximum entropy,boundary search,risk calculation