An important step of multi-sources of prior information fusion is to get a reasonable weight for each prior distribution in reliability assessment. There are some different shortcomings of existing methods which calculate the weights for prior distribution, in order to determine the weights easily and rationally, it presents a new method which based on test data’s confidence distribution. The method is directly linked to the targets which we want to assess, so it can reduce the assessment error from the headstream. An example shows that the method is simple and has obvious application values in the reliability analysis and other project fields.