Historically, most interest in pipe networks has focused on the development of efficient algorithms for the analysis of flow, and there are now very useful and efficient computer packages available for simulation of existing and proposed new schemes. SimNet [6], for example, is widely used in Poland for analysis and simulation of gas distribution systems. In contrast, there has been comparatively little research interest into the development of methodologies aimed at optimising the design of pipe networks. There are very few computer packages available for commercial use that help the designer to produce truly optimal pipe network designs. The current available methods for designing networks can be divided into three groups: - heuristic methods - methods which assume a continuous range of diameters is available - discrete optimisation methods
Although, these methods have developed through the years from an appreciation of what usually constitutes a good and economic design, there is no guarantee that the designs produced will in any sense be optimal. Even for a simple tree-like network there are a very large number of possible designs that provide feasible solutions. It can be seen that the chances of hitting on the best solution are very small. In PILOT method [lo] the pipes are ordered for the tree definition according to the shortest distance from the source. The program then proceeds in the usual way up to the point of balancing the flows. The pipe resistances are evaluated, then recalculated using flows obtained by Hardy-Cross loop balancing method and corresponding diameters are selected. Another Hardy-Cross to balance the network is performed determining flows and pressures through the system. The process of scaling resistances and performing Hardy-Cross is repeated until none of the pressures falls below the minimum design pressure. Once the basic solution has been found the program aims at further economic improvements. In FP6 method [9], after all pipes have been set to minimum diameter the loops are defined and initial flows assigned Hardy-Cross balancing flows. Pipes are then upgraded on maximum-lflowpath from source outwards the criterion being the cost benefit. Several flowpaths are upgraded simultaneously. Areas of influence are assigned to each governor in CONGAS method [10]. The tree system is designed for each area with minimum pressure at ends and even pressure drop loop closing pipes are sized to a minimum and Hardy-Cross flow balancing performed. The theoretical diameters are substituted by actual diameters and network is rebalanced. 3. Continuous methods One way of solving the design problem is to initially suppose that any size of diameter is possible. Doing this allows continuous optimization methods to be employed. The result of such optimization will be a set of diameters which needs to be corrected to available diameter sizes. SUMT is an algorithm for sequential unconstrained minimization technique and is a method devised and programmed by G. Boyne in.
