The traditional reliability evaluation is based on the failure data,which needs a large amount of statistical data and samples,and comprises the characteristics of long test cycle and high cost. To solve this problem,without failure data,a method of reliability evaluation was proposed which was based on the performance degradation data. The performance degradation model was built in the method which was based on the least square support machine. With the help of NSGA-II,the parameters optimization of the method above had been done,and the process of parameters optimization and reliability evaluation was given.Last but not least,the availability of the method is validated.