Strategic and Analytics-Driven Inspection Operations for Infrastructure Resilience

**Abstract**:

This talk addresses the design of optimal inspection and response strategies to improve the resilience of critical infrastructure networks to failure events resulting from cybersecurity attacks and/or natural events. For strategic planning, we consider the problem of finding an inspection strategy utilizing the minimum number of smart detectors to ensure a target detection performance against multiple adversarial disruptions. This problem can be formulated as a mathematical program with constraints involving Nash equilibria of a large strategic game. We develop a scalable approach that computes randomized strategies based on solutions of a minimum set cover problem and a maximum set packing problem, along with optimality guarantees. For operational response in the aftermath of a natural disaster, we consider a stochastic orienteering problem with probing constraints to localize network failures. We first develop a predictive failure model using data from San Francisco Bay area’s gas pipeline inspection operations. Then, we exploit the problem structure to design a scalable non-adaptive algorithm based on integer programming that can be solved for large-scale instances. Our results lead to practical strategies for response operations, with small optimality gaps. We demonstrate the value of utilizing real-world failure data and network properties for improving inspection and response operations.

**Bio**:

Mathieu Dahan is a PhD candidate in the Center for Computational Engineering at MIT. Mathieu earned a M.S. in Computation for Design and Optimization from MIT, an M.Eng. and B.Eng. in Applied Mathematics from École Centrale Paris, and a B.S. in Mathematics from the University of Paris Sud. His research interests are in combinatorial optimization, game theory, and predictive analytics, with applications to service operations and disaster logistics. His work in strategic network inspection is aimed at detecting cybersecurity threats to critical infrastructure systems. He is collaborating with utility companies on the use of mobile sensors and predictive analytics for improving disaster response operations. Previously, he worked as a research scientist intern at Amazon.

]]>This talk addresses the design of optimal inspection and response strategies to improve the resilience of critical infrastructure networks to failure events resulting from cybersecurity attacks and/or natural events. For strategic planning, we consider the problem of finding an inspection strategy utilizing the minimum number of smart detectors to ensure a target detection performance against multiple adversarial disruptions. This problem can be formulated as a mathematical program with constraints involving Nash equilibria of a large strategic game. We develop a scalable approach that computes randomized strategies based on solutions of a minimum set cover problem and a maximum set packing problem, along with optimality guarantees. For operational response in the aftermath of a natural disaster, we consider a stochastic orienteering problem with probing constraints to localize network failures. We first develop a predictive failure model using data from San Francisco Bay area’s gas pipeline inspection operations. Then, we exploit the problem structure to design a scalable non-adaptive algorithm based on integer programming that can be solved for large-scale instances. Our results lead to practical strategies for response operations, with small optimality gaps. We demonstrate the value of utilizing real-world failure data and network properties for improving inspection and response operations.

]]>