Multi-Hazard Supply Chain Network Management: Assessment of Modeling Elements and Data Requirements for Preand Post- Disaster Restoration Dr. Suzanna Long, Missouri S&T Dr. Tom Shoberg, U.S. Geological Survey Dr. Steven Corns, Missouri S&T Dr. Héctor Carlo, University of Puerto Rico at Mayaguez Project #G13AC00028 June 26, 2013 Students Involved • Varun Ramachandran, PhD Student, Engineering Management, Missouri S&T • Wilson Alvarez, Graduate Student, Industrial Engineering, University of Puerto Rico at Mayaguez • Alejandro Vigo, Undergraduate Student, Industrial Engineering, University of Puerto Rico at Mayaguez • Victor David, Undergraduate Student, Industrial Engineering, University of Puerto Rico at Mayaguez • Lizzette Pérez, PhD Student, Engineering Management, Missouri S&T 2 Aim: To approach SCSI restoration as a complex adaptive systems problem and define the necessary model elements, data needs/element, component interdependencies and metrics for success. Resiliency and scalability are then incorporated into a multi-hazard management decision-making tool. 3 Emergency Response • Short Term: Focus on people – Rescue & Recovery (FEMA) – Command & Control (DHS) • Long Term: Focus on infrastructure – Scalability – Resiliency – Sustainability 4 Gap Analysis • Medium- to long-term Supply Chain Strategic Infrastructure (SCSI) recovery after extreme event disruption has yet to be adequately modeled. • Interdependencies between SCSI elements are not well mapped. • Much infrastructure data are proprietary and, as such are difficult to acquire. • Decision making and handoffs between private and public entities lead to restoration bottlenecks. • No single approach exists to model SCSI recovery from planning stage to restoration stage. 5 Uniqueness of Approach • Integration of geospatial and supply chain data. • Model Based Systems Engineering (MBSE) and Complex Adaptive Systems (CAS) to account for the different elements of the supply chain network. • Model considers: – Scalability – Resiliency – Sustainability 6 Previous Results • Used Combinatorial Graph Theory. • Feasibility analysis of the model against an actual EF-5 tornado. • Development of Priority Restoration Matrix. Infrastructure Model Joplin Difference Communication 10 days 10 days Same as Joplin Electricity 15.5 days 14 days +1.5 days Local Transportation 10 days total major 90% major roads +1 day Network roads 5-6 days cleared. Water Pipelines 8 days 10 days Resiliency 22.5 +/- 2.5 days 24.5 +/- 3.5 days +/- 2 days Lines -2 days 7 Project Goals • Integration of information to make a comprehensive MBSE model. • Develop a framework of how sub-system level resilience metric can be used to calculate resiliency time and cost to get sub-systems back to pre-event levels. • Make scalable models that span multiple elements of the supply chain network across several regions in response to a variety of possible extreme events. • Improve decision-making algorithms to approach optimal restoration associated with the destruction caused by an extreme event. 8 Tools to Approach the Project • Model Based Systems Engineering (MBSE) • Complex Adaptive Systems (CAS) • Agent Based Modeling 9 What is Model Based Systems Engineering? • Away of doing systems engineering using models • The shift from a document centric systems engineering paradigm to a model centric (Friedenthal) • New topic and still evolving 10 Source: INCOSE MBSE Initiative – Mark Sampson Vision: Integrated systems-oriented decision support… Power Rating: 18 Amps Hydraulic Fluid: SAE 1340 notcompliant Sensor MTBF: 3000 hrs Minimum Turn Radius: 24 ft. Dry Pavement Braking Distance at 60 MPH : 110 ft. ft. 90 ft at 60 MPH : 110 Thermal/Heat Dissipation: 780° Ergonomic/Pe dal Feedback: 34 ERGS Hydraulic Pressure: 350 PSI Automatic Cruise Control <FAULT> Complex Adaptive Systems • Complex Adaptive Systems are dynamic systems that may represent cells, species, individuals, nations, constantly acting and reacting to what the other entities around them are doing. • Because extreme event SCN restoration involves a large number of coupled, dynamic sub-systems, the reconstruction effort must be approached as a complex adaptive system. 12 Complex Adaptive Systems Changing external environment System Feedback Changing external environment Changing external environment Feedback Emergent Behavior Changing external 13 environment Combining Information from Multiple Models Freight capacity Data Infrastructure Data Geospatial Data Disaster Disaster type: type: Tornado Earthquake Restore supply Restore supply chain chain to to 80% 80% prepreevent capacity. event capacity. Restoration Data MBSE Model Location Data Transportation Data 14 Agent-based Modeling ABM is defined as decentralized, non- system level approach to model design. The agents or active components need to identify component behavior and an environment must be defined which establishes connections for the simulation. Characteristics of Agents: • Identifiable and discrete. • Rules for interactions with environment • Goal-directed. • Flexible. • Learns from surroundings. 15 SCSI Modeling A system is comprised of: • Nodes (with a variety of “types”) • Links or “connections” to other nodes (with a variety of “modes”) • Local rules for Nodal and Link behavior • Local Adaptation of Behavioral Rules • “Global” forcing from Policy Connect nodes appropriately to form a system (network) Critical Infrastructures are to be modeled as Complex systems because they are composed of many parts whose interaction yields emergent structure (networks) and behavior (cascades), they grow and adapt in response to policy and contain people which makes the behavior unpredictable 16 Model Framework 17 Model Layout 18 Sub-System Model • A critical infrastructure model is constructed from three key data sources. – The National Map (TNM) of the U.S. Geological Survey – Geospatial data – Infrastructure elements derived from these data. – Missouri and Illinois departments of transportation. • These are integrated with transportation capacity data from the U. S. Department of Commerce to model the flow of goods and services through the urban center 19 Data Requirements • Lack of centralized data repository for comprehensive analysis of critical infrastructures. • Data needed for this research: Geospatial Data with geographic and Transportation data (road, air, rail, elevation data and sea) Hydrographic models of water basins Infrastructure interdependency data Regional nodes of critical infrastructure Real-time data and not static data Hazard Data e.g. geo-seismic data Social, administrative, economical data of a the region under consideration 20 Data Acquisition - Transportation Category Commodity Freight Manufactured Goods Raw Materials Data Ownership Difficulties with data Type Freight Data Data Description Food, Agriculture, Paper etc. Electronics, Machinery, Vehicles etc. Tons Tons Coal, Fuel, Chemicals etc. Tons Public Private/ Public Private/ Public 1)Static data; 2)Generalized data; 3) Proprietary data Private/ Public Private Private Private/ Public Private/ Public 1)Inconsistency; 2)Estimation maybe required; 3)Private-Public ownership Freight Flow Data Road Transportation Rail Transportation Air Transportation Water Transportation Pipeline Transportation Goods Transported by Road Tons Goods Transported by Rail Goods Transported by Air Goods Transported by Water Tons Tons Tons Goods Transported by Pipeline Tons Infrastructure Capacity Data Road- Hub Rail-Hub Water-Hub Bulk, General Cargo, Containers Bulk, Break Bulk, Intermodal, Shunting, etc. Rail Car Storage, Dry Storage, Liquid Storage etc. Tons Tons Tons/ Bushels Private Private Private 1) Varied amount of data required; 2) Different capabilities of hubs; 3) Interdependency of data 21 Data Acquisition - Transportation Category Data Difficulties with Ownership Type data Infrastructure Location Data/Geospatial Data Data Description Hub Location Number of hubs in the area Number Private Utility Location Location of all utilities that aid in freight flow Number Private/Public Road & Bridge Location Location of infrastructure that aids in road transportation Number Public Airport Location Location of infrastructure that aids in air transportation Number Private Pipeline Location Location of infrastructure that aids in pipeline transportation Number Private Number Private Number Private River Location Rail Location Location of infrastructure that aids in river transportation Location of infrastructure that aids in rail transportation 1)Ever increasing data set; 2) Use of software; 3) Static data Restoration Data # of People Number of people available and required to work on restoration Number Private/Public Travel Time Time required for different teams to arrive at the damage area Hours/Days Private/Public Skill Set Skilled people required to work on different aspects - Private/Public Mode Substitution If possible, substitution of mode to allow freight flow - Private/Public Task Management Assignment and management of different tasks - Private/Public Equipment Required Goods required for restoration to take place Tons/Pieces Private/Public 1)Different time dependence factors; 2)Vast amount of data; 3) Scalability; 4)Ownership of data 22 Data Size Estimation Data Set Elevation Hydrography Orthoimagery Roads Rail Location U.S. Geological Survey (USGS) – The National Map (TNM) USGS – National Hydrography Data (NHD) from TNM USGS – TNM USGS – TNM Missouri Department of Transportation (MoDoT) - Illinois Department of Transportation (IDoT) Size (MB) 2000 93.6 674,000 354 85.2 Airports USGS – Center for Excellence in Geospatial Information Science (CEGIS) 0.310 Electric Grid Bridges and Overpasses Tunnels and Culverts Water Plants and Stations Dams and Locks Power plants Railroad yards and stations River Docks/Ports Communication Towers Restoration Rates USGS – CEGIS USGS – CEGIS USGS – CEGIS USGS – CEGIS USGS – CEGIS USGS – CEGIS USGS – CEGIS USGS – CEGIS U. S. Federal Communications Commission (FCC) Missouri University of Science and Technology (S&T) – Department of Engineering Management and Systems Engineering (DEMSE) (Compiled from literature and interviews) U.S. Department of Commerce (Commerce), U.S. Department of Transportation (USDoT) and Private business (Compiled and housed at S&T – DEMSE) 0.980 0.436 0.223 0.063 0.053 0.007 0.031 0.053 0.020 0.250 Simulated data from disaster scenario model, compiled and executed by USGS and S&T, housed at S&T 100.6 Supply Chain Flow Rates Disaster Damage/Hazard Scenario Total 50.56 676,686 23 Model work flow 24 Types of Interdependencies • Physical – one system depends on another for operation (ex, wastewater depends on power) • Geographic – co-located systems • Cyber – linked electronically or through information-sharing • Logical – other, such as shared financial market Types of Interdependent Failures • Cascading – direct disruption • Escalating – exacerbates already-existing disruption, increasing severity or prolonging • Restoration – impacts the restoration of another system • Compound damage propagation – leads to disruption that causes serious damage • Substitutive – disruption due to excessive demands placed on a system to substitute for failed system Algorithm for Interdependency 1: Load data for each element. 2: while data exists do 3: for i = 1 to number_of_infrastructure_elements 4: Select one element 5: Make directed graph with nodes and edges 6: while intersection do 7: Planarize edge and use weight as distance between the line segment 8: end while 9: Use Dkijstras algorithm find nearest edge 10: while not reached end point of agents destination do 11: Calculate list of edges that make the route 12: for j = 1 to number_of_coordinates_to_pass 13: Map according to rules 14: end for 15: end while 11: Update database 10: end for 11: end while 27 Combining Sub-Systems Infrastructure Data Freight capacity Data Restoration Data Geospatial Data Transportation Data CAS Restoration Optimization GUI Decision Framework 28 Summary • An inventory of the necessary data is presented along with information on how well these data can be estimated and integrated from public access sources. • Creating a SCSI model using public data is a daunting task, but is possible. • ABM can be used for mapping interdependencies and creating simulations. Summary • Use SCN availability function to estimate the expected economic impact of a given extreme event. • Restoration management routing decisions during extreme events • Facility location in terms SCN resiliency • Using resilience as a metric of SCN reliability during and after extreme events. 30 Summary • Economic Recovery following extreme events dependent on robustness of supply chain networks. • SCN highly dependent on geospatial data for facility location but this element is significantly underdesigned • Improving the SCN resilience and scalability after Extreme Events 31 Thank you • Questions? • For more information contact: Dr. Suzanna Long, EMSE Telephone: 1-573-341-7621 Email: longsuz@mst.edu 32 References • Guha, S., Moss, A., Naor, J., Schieber, B., 1999. Efficient recovery from power outage. In: Vitter, J., Larmore, L., Leighton, F. (Eds.), Proceedings of the Symposium on Theory of Computing (STOC). Atlanta, GA, USA • Ang, C., 2006. Optimized recovery of damaged electrical power grids. Unpublished master’s thesis, Naval Postgraduate School. • Xu, N., Guikema, S., Davidson, R., Nozick, L., Cagnan, Z., Vaziri, K., 2007. Optimizing scheduling of post-earthquake electric power restoration tasks. Earthquake Engineering and Structural Dynamics 36 (2), 265–284. • E.E. Lee, J.E. Mitchell, and W.A. Wallace. Restoration of services in interdependent infrastructure systems: A network flows approach. IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews, 37(6):1303{1317, 2007. • J. Gong, E.E. Lee, J.E. Mitchell, and W.A. Wallace. Logic-based multi-objective optimization for restoration planning. In W. Chaovalitwongse, K.C. Furman, and P.M. Pardalos, editors, Optimization and Logistics Challenges in the Enterprise, chapter 11. Springer, 2009. • Cavdaroglu, B., Hammel, E., Mitchell, J., Sharkey, T., Wallace, W.: Integrating restoration and scheduling decisions for disrupted interdependent infrastructure systems. Annals OR 203(1 ): 279-294 (2013) • Nurre, S., Cavdaroglu, B., Mitchell, J., Sharkey, T., Wallace, W.: Restoring infrastructure systems: An integrated network design and scheduling (INDS) problem. European Journal of Operational Research 223(3): 794-806 (2012) 33