An efficient method of constructing $textsc{k}$-minimal path sets to evaluate the reliability of a flow network is presented. The network is considered to be in a functioning state if it can transmit a maximum flow which is greater than or equal to a specified amount of flow, $textsc{k}$say, and a $textsc{k}$-minimal path set is a minimal set of branches that satisfies the given flow constraint. In this paper, under the assumption that minimal path sets of the network are known, we generate composite paths by adding only a minimal set of branches at each iteration to get $textsc{k}$-minimal path sets after possibly the fewest composition, and compute maximum flow of composite paths using only minimal path sets. Thereby we greatly reduce the possible occurrence of redundant composite paths throughout the process and efficiently compute the maximum flow of composite paths generated. Numerical examples illustrate the method.