System under large stress: Prediction and management of catastrophic failures

Abstract

T he tensile strength of a chain is determined by its weakest link. Does this idea apply to more complex systems too? For instance, does the weakest thread of a spider web initiate cascading failure, when a strong wind gust is stretching the web to its limit? What happens to a computer when both the supply voltage and the ambient temperature are more than 20% outside its normal range of operations?Climate change, an increasingly more densely populated world and the rapid change of technology seem to put more systems under large stress. Engineering sustainable systems with a more favorable response to large stress appears to be an urgent societal need. Emergency evacuations of hospitals after hurricane Katharina and Sandy, and the May 22, 2011 tornado in Joplin illustrate the urgent need for modeling the adaptive capacity of hospitals during an extended loss of infrastructure [1]. Presidential Policy Directive 21 [2] and the U.S. Department of Homeland Security National Infrastructure Protection Plan (NIPP) [3] call for increasing resilience of the nation's critical infrastructure.The resilience is the ability to withstand, adapt to, and rapidly recover from a disruptive change [4][5][6][7][8][9]. Complex systems researchers use a holistic interdisciplinary approach to predict emergent behaviors in systems under large stress with the intent to better manage them. The long term goal is to engineer and design sustainable systems with more favorable behaviors under large stress. Typically the following 5-step approach is used: (1) First one uses cloud data, knowledge field experts and practitioners to construct a flow chart with many components. One keeps track of all relationships including nonlinearities. Figure 1 shows a schematic of a partial flow chart for a university. (2) Linear response theory is used to verify the response to small stress. The linearized model is considered complicated due to its many components, but is not called complex. (3) Agent based models on super computers such as the UIUC Blue Water system and NCSA's computer clusters are used to simulate the flow chart dynamics under large stress. (4) Key concepts of complex systems theory are used to describe the computed response to large stress. For instance the theory of catastrophes (including the failure of the weakest link and nonlinear resonances), deterministic chaos, synchronization, fractals, genetic algorithm, self-organized criticality, cellular automata, neural nets, rare events, and multiscale networks help to choose reproducible meaningful observables to analyze the resulting large data sets and to discover relationships between the control parameters and the observables.