Decomposition of a bimodal distribution function into two normal components by the maximum likelihood estimation method