Test optimization selection is an important step in testability design. This paper mainly studies the test optimization selection problem under the condition of unreliable test. Firstly, the problem is reduced to a multi-objective problem for analysis. On this basis, the mathematical model of the problem is established with the false alarm rate, test cost and false alarm rate as the objectives, and the fault detection rate and isolation rate as the constraints. Then, based on the Bayesian network testability model, the e-dominant nsga-2 algorithm is used to solve the problem. Finally, the method is used to optimize the test of a certain equipment The design is selected and compared with NSGA-2 algorithm and HBPSOGA algorithm to verify the effectiveness and practicability of the method.