A novel scanline algorithm for generating ordered variables

Rapid and timely network reliability assessment is one of the effective means for the reliable operation of infrastructure systems or other systems. Previous studies have found that binary decision diagram (BDD) -based network reliability algorithms are more efficient than other algorithms due to their advantage of avoiding repeated computation. The computational complexity of the BDD-based network reliability algorithm depends heavily on the maximum number of partitions in the network decomposition, notated as Fmax; the higher the Fmax is, the lower the computational efficiency is, and vice versa. However, previous studies did not consider the Fmax in generating ordered variables by depth-first search (DFS), breadth-first search (BFS) or network-driven search (NDS) algorithms, therefore, the efficiency of the network decomposition is largely a matter of luck by the previous algorithms. In this paper, a novel scanline algorithm was proposed aiming to obtain ordered variables with as small Fmax as possible. Experiments on the selected networks results show that the Fmax of the ordered variables generated by the proposed algorithm is smaller and more robust than that of the previous algorithms or their improved versions.