Editorial: Robust Monitoring, Diagnostic Methods and Tools for Engineered Systems

Manolis N. Chatzis, Eleni Chatzi, Costas PapadimitriouFrontiers in Built Environment 2020 · DOI · OpenAlex

Abstract

This editorial synthesizes a collection of research aimed at making Structural Health Monitoring (SHM) more resilient to real-world uncertainties. By integrating advanced Bayesian frameworks, machine learning, and Building Information Modeling (BIM), the featured works move beyond idealized simulations to provide robust, data-driven diagnostic tools for complex engineered systems like bridges, wind turbines, and buildings. (ISSN: 2297-3362)

Introduction

Engineered systems—from skyscrapers to gas turbines—operate in environments defined by significant uncertainty. Engineers often struggle with limited information regarding material laws, unpredictable environmental loads (such as wind or temperature), and the complex influence of non-structural components.

Traditional simulation models frequently fail to account for these variables, leading to "model bias" and inaccurate predictions. Furthermore, real-world data collection is often problematic: sensors may have low sampling rates to conserve power, data points might go missing, and systems often exhibit "non-smooth" behaviors like sliding or impacts that standard filters cannot track. This special issue is dedicated to bridging the gap between theoretical monitoring and practical, reliable asset management.

Methodologies for Robust Diagnostics

The research featured in this issue introduces several novel methodologies designed to enhance the reliability of SHM:

  • Hierarchical Bayesian Frameworks: These are used to identify not just structural parameters, but also the stochastic properties of modeling errors, effectively neutralizing model bias.
  • Operational Modal Analysis (OMA): This method identifies the dynamic properties of buildings using only ambient vibration (like wind or traffic) rather than controlled shaking.
  • Discontinuous Unscented Kalman Filter (UKF): A modified filter designed specifically for "non-smooth" systems where phenomena like friction, sliding, or plasticity occur.
  • Gaussian Processes (GPs): Employed to update auto-regressive models in real-time to account for varying environmental conditions, such as fluctuating wind speeds on turbines.
  • Error-Domain Model Falsification (EDMF): A framework used to update Finite Element models by identifying which model assumptions are incompatible with observed data.
  • BIM-SHM Integration: A web-based framework that links real-time sensor data directly into Building Information Models for easier visualization and decision-making.

Experimental Validation and Case Studies

The editorial references a diverse array of experimental and field-operational datasets used to validate these methodologies:

  • The "Albina Yard": An experimental dataset from a three-story timber building, made available to the scientific community.
  • Venoge Bridge: Long-term strain data from a bridge monitored since 1995.
  • London, Ontario Steel Bridge: A case study for the integration of BIM and SHM.
  • Gas Turbine Vibration Data: Used for training novelty-detection machine learning models.
  • Wind Turbine Data: Both on-shore field data and simulated datasets for fatigue life estimation.
  • Two-bay Steel Frame: Used to test methods for handling missing or sub-Nyquist sampled data.

Key Insights and Results

  • The "Secondary Element" Effect: Research indicates that non-structural elements (like exterior walls) have a surprisingly significant impact on a building's modal properties. Ignoring these in a model can lead to major inaccuracies in dynamic assessment.
  • Embracing Bias: Rather than trying to build a "perfect" model, researchers are now using Bayesian tools to quantify and account for the inherent bias in simplified engineering models.
  • Data Reconstruction: New methods can robustly reconstruct signals even when data is missing or sampled at frequencies below the Nyquist rate—a critical advancement for low-power wireless sensor networks.
  • Fatigue and Damage Detection: Methods using Gaussian Processes successfully estimated the fatigue life of wind turbines, while ARMAX models coupled with sensor clustering proved capable of detecting structural damage in simulated multi-story buildings.

Conclusion and Future Directions

While these tools are maturing, significant hurdles remain. Distinguishing between actual structural damage and changes caused by environmental factors (like temperature swings) remains a primary challenge. Furthermore, there is a need for a "roadmap" to help managing authorities transition from traditional inspection methods to these high-tech sensorial approaches. The transition to data-driven asset management requires appropriate training for the engineers and authorities who will use these tools in the field.

Summary updated

Abstract

ISSN:2297-3362
BibTeX
@inproceedings{chatzis2020editorial,
  author = {Manolis N. Chatzis and Eleni Chatzi and Costas Papadimitriou},
  title = {{Editorial: Robust Monitoring, Diagnostic Methods and Tools for Engineered Systems}},
  booktitle = {Frontiers in Built Environment},
  year = {2020},
  doi = {10.3389/fbuil.2020.00125},
}
Figure