The invention relates to a dynamic reliability model updating method based on Bayes factor optimization. The dynamic reliability model updating method comprises the following steps that 1, the failure physics of a product is analyzed, a failure physics reliability model of the product is built, and reliability degree R0 (ti) at some given time points is estimated; 2, the relevant reliability information of the product is collected in sites, the reliability information is subjected to statistics and physical property analysis, a statistics reliability model is built, and reliability degree at some given time points is estimated; and 3, the failure physics reliability model is used as a selection model, in-site reliability data and reliability test data are combined, and a Bayes factor expression at different time intervals is built. The dynamic reliability model updating method has the beneficial effects that the dynamic reliability model updating at small samples is realized, and meanwhile, higher practical values and higher engineering significance are realized on the reliability design optimization of new generation products of new technology and new materials.