Pressure Dependent Analysis (PDA) is preferred over Demand Dependent Analysis (DDA) to analyse any water distribution network (WDN) under pressure deficient condition. In PDA, outflow at any node is considered dependent on the available nodal pressure. The demand at a node is considered to be satisfied: (1) fully, if the available pressure is more than the desirable pressure; (2) partially, if the available pressure is between the minimum required and desirable pressure; and not at all, if the pressure less than the minimum required pressure. The outflow rate (q) is considered to increase non-linearly with the increase in available pressures (h) in partial supply conditions. Different node head-flow relationships (NHFR) have been suggested by different researchers for partial supply conditions. The most common NHFR is q=K hn, in which K and n are the constants. As the demands of several consumers are lumped at a node in a large WDN, the nature of NHFR at a lumped node depends on the relative elevations of different consumer connections and the head loss in the secondary network originated from the lumped nodes. The EPANET 2.0 is largely used as a network solver and carry out DDA. Several methodologies have been suggested to make use of EPANET 2.0 for PDA in which partial flow condition is simulated by adding some fictitious components, like emitter or a pipe. These fictitious elements can consider different values of constant K at different nodes in EPANET 2.0, however a common value of n is to be adopted. Recently released EPANET 2.2 has a direct PDA facility, however, it even does not allow considering different minimum and desirable pressure values at different nodes, therefore requiring use of DDA option with fictitious components. In this study, two sets of NHFR at different lumped nodes of a real-life network are developed. In the first set, best fit NHFRs are obtained in which both K and n are considered different at different nodes. In the second set, best fit NHFRs are obtained by fixing the same value of n at different nodes. The PDA with the first set of values is carried out using a self-developed code based on the Hardy Cross method. The PDA with the second set of NHFRS is carried out using EPANET 2.0 with addition of fictitious elements. The comparison of results shows that the NHFR with different K and n value provide better results as compared to that obtained by fixing the same value of n at all nodes. This indicates the importance of different n values at different nodes to exhibit proper NHFR.
