D3P

Cohesie grant

Title: Distributed Damage Diagnosis and Prognosis for Improving Resilience of Large Scale Critical Infrastructures

Period: 2018 - 19

Budget: 60 kEur

Role: Co-PI (with Xiaoli Jiang)

Funding source: TU Delft (Cohesie 2018)

Description: Critical Infrastructures (CI) are strategic assets of our societies and range from civil engineering structures such as dams and bridges, to power, communication, water, gas and transportation networks, to large industrial complexes: they can be defined as “so vital that their incapacity or destruction would have debilitating impact on the defense or economic security” of a Country (see U.S. President Executive order 13010 ). It is then of the utmost importance to guarantee their safety and resilience to a wide range of external and internal damaging factors such as normal wear, corrosion, fatigue (aging effect), naturally occurring extreme events, human errors or intentional cyber or physical sabotages, and faults. Although such events are varying in nature, their common trait is the capability to inflict a damage to CI components, which can, if not detected and corrected, progress to a widespread and catastrophic failure of the CI. While a CI may be already endowed with a well-developed network of sensors that allow to continuously monitor them, still even, and mostly, a vast amount of data is not enough to guarantee that maintenance or emergency actions will be carried out effectively and timely. There is indeed a need for advanced analytical tools that are able to detect and diagnose the source of any damage occurring to the CI, and prognose the way it is going to develop over time and affect the performance and safety of the failing CI component as well as neighboring ones. The prognosis step, which is needed to effectively plan the deployment of any countermeasure or maintenance activity, is in particular a demanding step which is plagued by the high variability and uncertainty that characterizes the dynamics of the damage progression.

In this project we plan to investigate distributed monitoring architectures for large-scale CI. Such architectures shall be able to share data from different sensors and feed such data to dynamical models of different components of a CI. Predictions from such models will be used for

  • Detection of faults and other anomalies (such as cyber attacks)
  • Identification of detected anomalies
  • Prognosis of component residual life
  • Prognosis of the CI health status as a whole

One of the key point of the proposed approach will be the capability to include the effect of the many sources of uncertainties in a probabilistic way. Indeed, not only the sensors output will be affected by noise, but the models themselves will present unavoidable parametric or non-parametric uncertainties. Both sources of uncertainty can be effectively modeled in a probabilistic way, leading to the models being based on difference or differential stochastic equations.