In this paper, we have estimated the reliability of a system composed of elements connected in series and subject to the same stress, by the nonparametric kernel estimation method using lognormal kernel. This kernel is chosen for its advantageous properties, including an optimal convergence rate for the mean integrated squared error, non-negativity, free from boundary bias, and naturally varying in shape. Asymptotic properties such as bias, variance, and mean squared error have been established for the proposed estimator. Furthermore, we address the selection of the optimal bandwidth parameter an essential aspect of kernel estimation using both the rule of thumb and the unbiased cross validation techniques. Finally, a simulation study is carried out to highlight the performance of the system reliability estimator, based on lognormal kernel and to compare the two bandwidth selection techniques to determine the best one.