Pipeline network calculations using sparse computation techniques

Abstract It is shown how computation to determine steady-state conditions in pipeline networks can be greatly facilitated using sparse computation techniques. Three network algorithms for direction assignment, node-arc ordering and construction of a minimal length cycle set have been devised for this purpose. These algorithms are readily implemented using list processing techniques, thereby making it unnecessary for program users to supply similar, but often non-optimal, information. Using these algorithms, it is possible to obtain substantial reduction in computing time and core storage. Computational accuracy may also be enhanced through reduced number of rounding errors. The reduction of computing time is particularly dramatic and estimates based on operation counts are partially verified by actual computational experience. For a given network the benefit of a single application of these algorithms is multiplied many times with iterations and different parametric values. The development demonstrates substantial advantages of applying graph-theoretic techniques directly to enhance data processing efficiency.