Literature

• 1. 2nd International Workshop on Resilience (IRW) 2018, http://www.workshop-china 2018.resiltronics.org/, last access on December 27, 2018.
• 2. Workshop draft program as of October 8, 2018.
• 3. Subcommittee on Critical Infrastructure Security and Resilience, Committee on Homeland and National Security of the National Science and Technology Council. (February 28, 2018). Summary of the 2018 Critical Infrastructure Security and Resilience Stakeholder Workshop, https://www.whitehouse.gov/wp-content/uploads /2018/03/New-CISR-Stakeholder-Workshop-Summary-Formatted-FINAL.pdf, last access on December 2, 2018.
• 4. National Security Strategy of the United States of America, December 2017, The White House, Washington DC, https://www.whitehouse.goV/wp-content/uploads/2017/12/N SS-Final-12-18-2017-0905.pdf, last access on November 22, 2018.
• 5. Acatech - German National Academy of Science and Engineering. Resilience-by-design: a strategy for the technology issues of the future, acatech POSITION PAPER - Executive summary and recommendations, https://www.acatech.de/wp-content/uploads/2018/03/ acatech_POSITION_RT_KF_eng_140508.pdf. last access on November 22, 2018.
• 6. I. Haring, G. Sansavini, E. Bellini, N. Martyn, T. Kovalenko, M. Kitsak, G. Vogelbacher, K. Ross, U. Bergerhausen, K. Barker, I. Linkov. (2017). Towards a generic resilience management, quantification and development process: general definitions, requirements, methods, techniques and measures, and case studies. In: Resilience and Risk: Methods and Application in Environment, Cyber and Social Domains, Editors: I. Linkov. J. M. Palma-Oliveira. pp. 21-80, Springer, ISBN 9789402411225, http://www. springer.com/de/book/9789402411225.
• 7. Lloyd’s Register Foundation Report Series: No. 2015.2, Foresight review of resilience engineering: designing for the expected and unexpected, October 2015, http://www.lrfoundat ion.org.uk/publications/resilience-engineering.aspx, last access on November 25,2018.
• 8. K. Thoma, B. Scharte, D. Hiller, T. Leismann. (2016). Resilience engineering as part of security research: definitions, concepts and science approaches. Eur J Secur Res, Volume 1. Issue 1, pp. 3-19. doi: 10.1007/s41125-016-0002-4.

• 9. I. Haring, S. Ebenhoch, A. Stolz. (2016). Quantifying resilience for resilience engineering of socio technical systems. Ear J Secur Res, Volume 1, Issue 1, pp. 21-58. doi:10.l007/s41125-015-0001-x.
• 10. Workshop questions as transferred to and agreed by the working groups on October 30, 2018.
• 11. Final workshop questions as shared between the working groups on October 30, 2018.
• 12. Lili Xie, Gian-Paolo Cimellaro, Michel Bruneau, Zhishen Wu, Max Didier, Mohammad Noori, Ivo Haring. (2018). Frameworks, fundamentals and education for future infrastructure risk control and resilience: workshop 1 report. Proceedings of 2nd International Workshop on Resilience (IRW) 2018, Nanjing and Shanghai, China.
• 13. Aftab Mufti, Xilin Lu, Jinpin Ou, Shamim Sheikh, Ying Zhou, Marco Domaneschi, Mohammad Noori, Ivo Haring. (2018). How to improve critical infrastructure systems with emerging technologies: future critical infrastructure systems predictive simulation and emerging technologies - Report of working group 2. Proceedings of 2nd International Workshop on Resilience (IRW) 2018, Nanjing and Shanghai, China.
• 14. Teruhiko Yoda, Ertugrul Taciroglu, Ivo Haring, Anastasios Sextos, Mohammad Noori. (2018). Big data analytics, machine learning and AI for future human support against disruption events and for critical infrastructure system resilience: report of working group 3. Proceedings of 2nd International Workshop on Resilience (IRW) 2018, Nanjing and Shanghai, China.

Frameworks, Fundamentals, and Education for Future Infrastructure Risk Control and Resilience

This section covers the identification of future fundamental resilience research needs and related academic educational challenges, as found at the 2nd International Workshop on Resilience 2018. Fundamental research needs were agreed to consist of the development of flexible and generally accepted frameworks, i.e., assessment process models, resilience improvement, development, implementation, and optimization models. Such models should also take the cultural, societal context, and expectations of operators, users, and citizens into account, in particular, how to translate them into acceptance and evaluation criteria. Much more w'ork is believed to be necessary to understand and simulate local loadings (e.g., on building level in case of earthquakes and flooding), especially of combined multiple threats, such as physical impact, flood loading, or cyber attacks combined with physical and natural hazards. Measuring and metrics for resilience are believed to remain ongoing future tasks, in particular, on a system level. Specific challenges identified comprise megacities, legacy infrastructure, fast simulation of large-scale urban built environments as well as increasingly interlinked infrastructure systems, and the use of (unspecific and dedicated) data, data analytics up to self-learning approaches (machine learning, KI). Modern semantic digital building and infrastructure formats are expected to enable novel developments regarding (semi) automatic assessments. The education focus for long-term scientific and applied capacity improvement is believed to build on strong (multiple) MINT subject domain experts, who should also take advanced courses in system science modeling approaches as well as data-driven sciences. In all cases, students are proposed to be taught within broad real-world application projects to learn how to involve users and decision makers and respective participatory science approaches for enhancing future resilience research. Along with the main argumentative lines, this section provides lists of key questions and challenges and identifies similarly most prominent approaches and methods. It outlines a tentative timeline for the implementation of the future fundamental and education resilience research roadmap.

Background and Introduction

As threat modalities, variability, and demand levels to infrastructure systems of today’s worldwide societies and communities are increasing, the need for advanced overall and sustainable risk control proves to be a fundamental prerequisite for thriving societal, individual, economic, and environmental well-being and development.

At the 2nd International Resilience Workshop (IRW) 2018 held at Nanjing and Shanghai [1], the experts present at the workshop committed to outlining future resilience research goals from a mainly engineering-technical perspective. To arrive at sufficient specific recommendations, as discussed in more detail in the introduction to the three workshop reports [2], the focus of working group 1 was to roadmap research needs regarding the enhanced understanding of fundamental processes underlying natural and extreme events on various spatial and temporal scales. This comprised of taking account of the variability inherent in such hazards and events, and on curricular changes necessary to better prepare the future generation of civil engineers for critical infrastructure resilience (CIS) research. Working Group 2 addresses advanced predictive modeling for critical infrastructure and related technology innovations, in particular, data gathering [3]. Workshop 3 covered the leverage of data analytics, machine learning up to AI approaches for better infrastructures, and the support of humans in case of disruption events [4].

Fundamental processes to be addressed within the research needs road-mapping were specified to cover advanced risk control and resilience management processes, risk, and resilience analysis, mechanical, technical, cyber, etc. properties, and behaviors of infrastructures at risk, as well as organizational issues, societal agendas, and even worldwide contexts framing the understanding of modern socio cyber-physical infrastructure systems including users, operators, and decision makers, from a (mainly technical and engineering) science and fundamental perspective, respectively.

The following report first defines fundamental infrastructure resilience research and education top-level goals (Section 1.2.2), from which key challenges (Section 1.2.3) are derived by contrasting with a research landscape review of the state of the art (Section 1.2.4). To ensure comprehensiveness, Section 1.2.5 provides a boundary condition aware framework for future fundamental and educational infrastructure resilience research needs. This sets the stage for concepts, methods, and technologies to be advanced in future resilience research (Section 1.2.6). Section 1.2.7 provides a tentative timing schedule for the identified research endeavors.

Objective Details

Working group 1 identified the following objectives regarding fundamental frameworks, research needs, and methodological gaps for future resilient infrastructures:

• • As acceptable overall risk control and resilience strongly depends on the societal context and consensus, it needs to be better clarified which overall frameworks, process models, stepw'ise-iterative approaches, and framing methods are necessary and sufficient for the
• • Contextualization,
• • Organizational and technical assessment,
• • Risk and resilience evaluation, decision making, and
• • Design and/or improvement of critical infrastructure systems.
• • Better understanding of threats, their combinations (threat vectors), param- etrization, and loading description, on high-resolution scales, especially combined classical, malicious, cyber, and intelligent threats.
• • Better modeling and simulation of socio-cyber-physical infrastructure systems at risk, exposed to threats, and tested with respect to their resilience, taking account of inter and intra dependencies, and the blurry boundaries of systems.
• • Provision of risk control and resilience quantities accepted by end-users and academia.
• • Improve risk and resilience evaluation criteria and consensus, sufficient for processing engineering-technical risk and resilience quantities.
• • Provision of fundamental principles, approaches, methods, and (structural and dynamic) solutions for better risk control and resilience, including fast but predictive models.
• • Definition of curricula guidelines for future resilience research taking account of the high variability of subject domains and the need for specific knowledge to allow for progress.
• • Addressing the need for continuous academic education.

Key Challenges for Fundamental Resilience Research and Education

To illustrate the overall objective of future-proof resilience research ambitions, the following specific challenges were identified, in part also in addition to the questions posed in Section 1.2.3:

• • Resilience frameworks and processes need to address in a systematic way boundary contexts, e.g., decision making competences and resources.
• • Especially technical science-driven resilience frameworks need to better take account of and extend resilience quantification beyond civil engineering or purely technical aspects to encapsulate social and economic impacts and dimensions, considering entire communities holistically.
• • It needs to be made more explicit how types of country, cultures, communities, and local contexts determine the boundary conditions of resilience assessments, e.g., what is considered as resilient or safe versus not resilient/ unsafe in different cultures.
• • The advantages and disadvantages of the different functions of infrastructure versus system structural risk control and resilience approaches (i.e., whether to follow a more functional, or system dynamic versus a more static, or system structural, design approach) need to be better understood.
• • There is a lack of risk control and resilience, overall chance and options qualitative, semi-quantitative, quantitative measures and quantities, metrics and aggregated evaluation options, in particular, such quantities that cover several scales of resolution from local components (structural members), to structures, buildings, quarters urban regions, countries, and worldwide regions.
• • It needs to be identified at which scale resilience should be assessed, in particular, whether most relevant resolution levels are at the level of individual decision makers (e.g., house owners), at the community level (e.g., quarter major) or at higher political or social levels.
• • Scaling and normalization of resilience metrics are open questions, e.g., how to compare regions with high hazard levels with low hazard regions.
• • There is a lack of systematic and generally accepted approaches of resilience management, e.g., based on well-defined risks of not being resilient, are missing, in particular, how, if at all, to embed them in current risk assessment schemes.
• • The specific challenges of megacities are not yet understood; for instance, spatial scales need to be covered and resolved. Positive as well as negative scaling effects are expected, e.g., when comparing the resilience with respect to local threats with the resilience with respect to non-localized threats, e.g., local terror events versus large-scale weather calamities. In the former, rescue forces will not be limited at all, whereas in the latter accumulation effects need to be considered.
• • Account for legacy infrastructures, anticipate long-term use, allow for the replacement of sensitive aging (digital) parts of infrastructures.

So far, research on critical infrastructures in a civil context has not yet been canonized by any means. It is observed that true progress in this domain often depends on a broad knowledge in several disciplines. Typically advancing the domain requires the application of fundamental knowledge from related disciplines, e.g., civil engineering, mechanical engineering, computer science, physics, or other MINT subjects. Balancing the subject-specific fundamentals and the emerging interdisciplinary science is considered to be a challenging endeavor. It becomes even more challenging when considering the employability of developed models or acquired knowledge in non-academic domains.

Major Research Gaps

Resilience frameworks are often not ambitious enough regarding their scope and generalizability, but are, on the contrary, too generic to be applicable to different domains while keeping all their advantages. This occurs in particular if such frameworks and processes are de facto adapted to very specific applications, e.g., earthquake engineering or explosive terroristic threats only.

Current frameworks do not systematically aim at the maximum separation possible between resilience assessment and improvement process steps (see, e.g., the discussion in [5] within the initial sections), e.g., consider the use of the term scenario. On the other hand, schemes tend to miss interdependencies, e.g., consider the often much too implicit selection of resilience assessment quantities. For instance, it is not distinguished between management objectives, quantities that are computed for assessment, reference and comparison quantities, and resilience evaluation and acceptance steps.

The advantage of deductive and inverse methods in resilience event propagation root cause reduction and possible event identification has not yet been taken into account. Also, a systematic resilience dimensional analysis and a reduction to sufficient and necessary resilience dimensions are missing. How to relate risk and resilience management, especially in operational contexts, is still an often-discussed issue. Possible starting points and concepts of the questions discussed in this text section are given, for example, in the respective sections of [6].

Existing frameworks are typically tailored to well-known hazards, in particular for earthquake engineering. Examples include the MCEER framework of the Multidisciplinary Center for Earthquake Engineering Research of the University at Buffalo [7] and to a lesser extent the PEOPLES [8] framework. For instance, the schematic step-by-step procedure of the MCEER methodology proposes the following sequence: (1) Define extreme event scenarios (e.g., probabilistic seismic hazard analysis (PSHA), ground motion selection); (2) Define the system model; (3) Evaluate the response of the model; (4) Compute different performance measures (e.g., losses, recovery time, functionality, resilience); (5) Identify remedial mitigation actions (e.g., advanced technologies) and/or resilience actions (e.g., resourcefulness, redundancy, etc.); (6) Redesign the system. This approach strongly focuses on the engineering steps and does not provide explicit support of where to place context, evaluation, and decision making considerations.

Current damage models at various scales, from structures to buildings and infrastructures, are still mainly focusing on initial damage effects, rather than on recovery and even less post-event improvement options. It is expected that advanced modeling options will allow to better foresee how to rebuild faster or even rebuild better.

Current approaches too often fail to explain the application context, non-technical constraints, and the expected impact of the research results on improving risk control, resilience, and the decision making process.

Several Universities are already setting up or already offering curricula in the domain of critical infrastructure protection and resilience research and engineering. Examples include: (i) University at Buffalo, MCEER, a national multidisciplinary and multi-hazard earthquake engineering research center [9]; (ii) University of Freiburg, Department of Sustainable Systems Engineering (INATECH) with its thematic topic resilience engineering [10] focusing on transfer of research to industrial innovation [11].

Framework to Address the Challenges

Frameworks of improved risk control and resilience for critical infrastructure systems will need to better take into account the following topics:

• • Existing standardizations and their gaps
• • Seamless assessment of standard and non-standard operations

• • Include societal and individual expectations more explicitly, as well as decision options and available resources
• • Much more explicitness regarding overall risk acceptance criteria and societal priorities
• • Allowance for participative and informed decision making of individuals, decision makers, and the representatives of the society
• • Frameworks should take advantage of the available access to digitalized spatial data and semantic infrastructure data
• • Frameworks should be modular and sufficiently specific to predict effects on the level of individual buildings
• • Modeling and simulation approaches should be scalable, i.e., cover a broad range of modeling options from empirical-statistical, via engineering-analytical to simulative approaches in order to be able to handle the variability of the amount and quality of available data
• • Modeling and simulation approaches should deliver their output adapted to the intended use of the results, e.g., detailed structural data versus traffic- light-assessments for volunteer rescue forces
• • The framework should take account of social media data and computational resources, of increasing interconnectedness of devices and of local computing options
• • Segregation and diversification of communication channels need to be taken into account
• • Frameworks should address the level of education and experience required to conduct the necessary assessments and decisions

Concepts, Methods, and Technologies to Be Further Developed

Concepts, methods, and technologies expected to be most relevant and to be further developed in the future to improve risk control and resilience for critical infrastructure systems include:

• • Large-scale simulations at the city and regional level
• • Parallel computing, e.g., use of simplified access to graphic process units (GPUs) parallel computations via the CUDA API for fast real-time simulations [12]
• • Use of spatial distribution of ground motion at the regional level and in spatially distributed infrastructures (water, power, gas distribution networks, etc.) with a high local resolution
• • Soil-structure interaction effects at the regional scale and city scale. How different foundations of skyscrapers might interfere with the soil and modify the dynamic response
• • Use and automatic analysis of digital pictures/images to collect model input data, to support the selection of models, to identify the damage, or the propagation of hazards, such as fire
• • Use of Bayesian networks to model critical infrastructure system’s service demand and recovery

• • Development of dynamic Bayesian probabilistic networks (BPNs) or agent-based models (ABMs) to model the recovery processes, for instance, sequences of such processes, repair times, while considering resource constraints
• • Use of Weighted Bayesian updating, e.g., to quantify building stock fragility
• • Use and analysis of big data as obtained, e.g., from twitter, smart grids, sensors in buildings, for a broad spectrum of tasks including the detection of threats, damaging events, health monitoring (pre- and post-event), anomaly detection, etc.
• • Leverage of machine learning and artificial intelligence algorithms, e.g., random forests or neural networks to assess repair costs or to predict continuous degradation. Anticipated levels include
• • Informed application
• • Tailoring
• • Further development of methods

Roadmaps and Strategies Proposed for Future Implementation

It is expected that mainly existing frameworks and approaches will be further extended, interlinked, and enriched with new technological approaches and methods. Furthermore, standardization is expected to increase on different levels.

At the same time, dominating approaches are expected (platform effect, “the winner takes all”) as standardization and digitalization of infrastructures increase as well as the interchange between formats becoming more and more automatized for structured data and more and more practicable even for unstructured data.

Especially such international standards as CityGML [13] on the semantic digital city, infrastructure, and building levels and building information models [14] on a single building level are candidate formats, not only for the exchange of digital data, but also for assessment procedures using such data. Another example are extended GIS formats as well as, for example, OpenStreetMap or similar proprietary formats.

The expectations of the prediction accuracy of models, along with associated uncertainty prediction, will further rise. At the same time, the computation time will be further increased, increasing the demand for advanced computing methods. As advanced computation requires more consideration and more event trajectories, computation resources are expected to remain a strong limiting factor for future decades.

Regarding time scales, reliable data gathering is expected to remain a limiting factor as well, due to large legacy infrastructure system fractions.

Summary and Conclusions

In summary, working group 1 was challenged to identify the future fundamentals of technical science-driven resilience research for improving critical infrastructure systems.

• It was identified that comprehensive frameworks and processes need to be further developed. To obtain stable technical assessments, quantifications, and solutions that adequately take up the true needs of all actors without anticipating solutions from a purely technical point of view are required.

Several approaches were identified as being of main interest for future resilient and resource-effective infrastructure solutions:

• • Seamless reliability, failure, and disruption handling capability.
• • Large-area simulation capability for a range of natural and man-made catastrophes, development with high resolution on local scales, e.g., for local building loading, taking into account known geophysics.
• • Leverage of disciplinary methods for quantitative and probabilistic resilience research requires fundamental disciplinary research guided by true user needs.
• • Need for the development of fast computation capabilities for multiple scenario analysis, using advanced computational approaches that also can be employed in case of emergencies.
• • Advanced computational and/or engineering approaches include technical solutions, hierarchical models, abstract tailored models, and real-time modeling based on data.
• • Overall life-cycle considerations including possible major disruptions as well as aging.
• • The development of curricula for resilience engineering needs to be rooted in dedicated domain-specific anchor subjects, as well as generic capabilities such as complex systems modeling, graphical models, all methods of (classical) systems analysis and engineering.