Data Driven Assembly of Procedurally Modeled Facilities

Abstract

We present a method to arrange components of industrial facilities in a constrained site footprint. We use a probabilistic graphical model of industrial sites and existing procedural modeling methods to automate the assembly and 3D modeling of wastewater treatment plants. A knowledge engineered approach produces a combination of components that inherently contains domain specific information like process dependencies and facility size. The inferred combination is laid out using mathematical optimization or via a physics-based simulation resulting in an arrangement that respects the industrial process and design plausibility.