Abstract
Accurately estimating design flow rates is essential for sizing premise plumbing components, yet prediction remains difficult because demand depends on occupancy, fixture characteristics, regional behavior, and event timing. Current design methods still rely on approaches developed in the 1940s, which studies show consistently overestimate simultaneous flow rates and lead to oversized systems. Oversizing increases capital costs, raises water age and disinfectant decay, promotes microbial growth, and reduces pump efficiency. To address these limitations, empirical, probabilistic, and stochastic methods have been explored. Stochastic models offer the greatest potential because they represent behavioral variability, but existing versions often rely on outdated assumptions, exclude key behavioral factors, or lack practical implementation pathways. This dissertation introduces a novel demand estimation framework using a stochastic event-generation model that applies events to a user-defined digital twin of a plumbing system. The work includes four components: (1) evaluation of assumptions and distributions used in existing stochastic models against regional empirical data; (2) identification of socioeconomic, demographic, and housing factors influencing water use using MGWR and OLS; (3) analysis of high-resolution datasets to determine whether consumption patterns reveal latent user populations; and (4) development of a Monte Carlo simulation that generates water-use events and routes them through the digital twin to estimate simultaneous hot- and cold-water flow rates without relying on unrealistic assumptions.