Bio-inspired Models for Infrastructure Network Optimization, Deployment and Adaptability

Wednesday, January 18, 2017

Abstract: Astonishing similarities were noted between the geometry of networks formed by living organisms (e.g. roots and slime mold) and that of infrastructure facilities (e. g. railway systems). The underlying assumption of this research work is that where classical optimization fails, alternative designs can be found in natural processes. In an attempt to optimize the efficiency, resiliency and versatility of subsurface networks, we propose to gain fundamental understanding of accommodation, adaptation and competition of biological networks. In this presentation, we model the topological evolution plant root systems and leaf venation networks in the presence of obstacles (accommodation) and under environmental fluctuations (adaptation). First, we present root growth experiments undertaken by our collaborators at Oak Ridge National Laboratory (ORNL), and we perform image analyses to calibrate a Root System Architecture (RSA) model. Second, we improve this RSA model to predict root growth around obstacles. Third, we explain the limitations of leaf venation models to predict the dynamic accommodation of ducts around obstacles and we propose an alternative based on a Steiner Tree (ST) algorithm. Then, we compare the efficiency and cost of networks obtained by RSA, leaf venation and ST models during deployment (topology change) and service (flow without topological change). Lastly, we formulate an analog problem of engineering design to test the suitability of biological models to optimize the design of a subsurface network of water lines. We present root growth experiments done at ORNL in in agar, on plates that contain obstacles and food spots. The repartition of obstacles and food spots corresponds to that of Georgia Tech buildings and fire hydrants, respectively. For this purpose, a simplified map of Georgia Tech campus was 3D printed. We compare the current water line network to the root system observed in the experiment and to the RSA and ST model predictions. We also discuss potential design improvements. We will use a similar modeling approach to study the design of the Google Fiber network in the city of Atlanta.

Bio: Dr. Arson is a theoretical and numerical expert in damage and healing rock mechanics, thermo-chemo-poromechanics, and underground storage. Her group designs and formulates models that link microscopic damage and healing processes to macroscopic rock behavior. For example, they explain how salt grain sliding mechanisms can result in crack propagation and how diffusive mass transfer at grain interfaces can actually heal these cracks. They study crack propagation at multiple scales, in shale for instance, and how crack patterns affect rock strength, stiffness and permeability. They also use principles of micro-mechanics and thermodynamics to understand fragmentation processes in granular assemblies – ballast for example.