The primary goal of water distribution systems is to ensure the delivery of high-quality drinking water at adequate pressure. However, the operational efficiency could be impacted by aleatory uncertainties stemming from random variables like water demand and pipe roughness, as well as epistemic uncertainties due to incomplete data and complex system interactions. Water demand at nodes is influenced by these uncertainties, being random, due to variable customer behavior, and fuzzy, because of the difficulty in specifying exact demand values. This research aims to represent water demand as a fuzzy random variable to quantify both uncertainties in a unified modeling framework. Previous works that utilized a fuzzy random approach assumed a normal distribution with a fuzzy mean and a standard deviation set as a fixed percentage of demand, using triangular or trapezoidal functions to represent uncertainty. Such a fixed percentage for variation assumes uniform uncertainty across all nodes and conditions, which is not realistic in practice. This research intends to avoid simplifying assumptions and implement unsupervised soft clustering methods to define fuzzy membership functions. Uncertainty in water demand limits the value and use of current distribution system modeling. Due to this uncertainty, models are too inaccurate and rigidly defined to use for setting day-to-day operational thresholds. This research presents a water demand modeling methodology that better captures uncertainty and can be used to identify realistic ranges of conditions expected in water distribution systems.