Damage Diagnosis of Single-Layer Latticed Shell Based on Temperature-Induced Strain under Bayesian Framework

Jie Xu, Zhengyang Zhao, Qian Ma, Ming Liu, Giuseppe Lacidogna2022 · DOI

TL;DR

This paper introduces a novel probabilistic approach for diagnosing damage in large-span latticed shell structures by analyzing temperature-induced strain. By employing Bayesian theory and Markov Chain Monte Carlo (MCMC) methods, the approach effectively accounts for uncertainties in monitoring data, enabling more reliable identification of structural damage and support failures.

What Problem Does This Paper Address?

Large-span latticed shells, commonly found in public buildings like stadiums and airports, are critical infrastructure. Ensuring their structural integrity through Structural Health Monitoring (SHM) is paramount. Traditional vibration-based SHM methods can be sensitive to environmental factors like temperature, and static-based methods often require controlled loading conditions that are difficult to achieve in practice for these structures. While temperature-induced strain has been explored for damage diagnosis, existing methods are often deterministic and fail to adequately account for uncertainties arising from environmental fluctuations, sensor noise, and wind loads. This paper addresses the need for a robust, probabilistic method that can reliably diagnose damage in latticed shells by specifically leveraging temperature-induced strain while explicitly managing uncertainties.

Methods

The core of this paper lies in a probabilistic damage diagnosis framework built upon Bayesian theory. The authors propose a new damage diagnosis index ($I_T$) that quantifies the relationship between temperature-induced strain and structural parameters. This index is then analyzed using Markov Chain Monte Carlo (MCMC) methods, specifically the Metropolis-Hastings algorithm, to estimate the posterior probability distribution of the damage diagnosis index. This probabilistic approach allows for the determination of confidence intervals for the initial (baseline) state and the calculation of damage probabilities for unknown states.

Key methodological steps include:

  • Damage Diagnosis Index Formulation: A new index ($I_T$) is derived based on the correlation between temperature-induced strain and structural parameters like rod stiffness and constraint stiffness.
  • Bayesian Framework: The problem is framed within a Bayesian context, where prior knowledge (assumed to be a normal distribution for the index) is updated with observed data (measured strain and temperature) to obtain a posterior probability distribution.
  • Markov Chain Monte Carlo (MCMC): The Metropolis-Hastings algorithm is employed to sample from the posterior distribution, which is often complex and non-standard. This allows for the estimation of the probability distribution of the damage diagnosis index, accounting for uncertainties.
  • Importance and Sensitivity Analysis: To efficiently monitor a large structure with limited sensors, the authors perform importance analysis (based on strain energy change upon member removal) and temperature sensitivity analysis to identify critical members for monitoring.

Datasets

The study utilizes a finite element model of a single-layer spherical mesh shell structure, based on the original roof design of the Beijing Laoshan Velodrome. This model is used to simulate various damage scenarios (rod damage and bearing restraint failure) and generate synthetic data for evaluating the proposed method. No real-world monitoring data is explicitly used for training or validation in this paper; the evaluation is performed on simulated data derived from the FE model.

Key Insights and Ideas

  • Temperature-Induced Strain as a Reliable Indicator: The research demonstrates that temperature-induced strain, when analyzed probabilistically, can be a powerful indicator of damage in latticed shells, overcoming limitations of purely vibration-based or deterministic static methods.
  • Uncertainty Quantification is Crucial: Explicitly accounting for uncertainties in the monitoring process (e.g., noise, environmental fluctuations) through Bayesian inference significantly enhances the reliability of damage diagnosis.
  • Targeted Monitoring is Effective: Importance and temperature sensitivity analyses are vital for selecting key monitoring locations in complex structures, optimizing sensor placement and maximizing diagnostic effectiveness.
  • Damage Manifests as Changes in the Diagnosis Index: The proposed damage diagnosis index ($I_T$) shows a clear trend of decreasing absolute values as damage severity increases, providing a quantifiable measure of structural integrity.
  • Differential Sensitivity to Damage Types: The method shows varying effectiveness for different types of damage and structural elements. It is particularly adept at diagnosing damage in radial and circumferential members and horizontal support failures, while less sensitive to oblique spars and z-directional support failures.

Key Results

The proposed method was validated using a finite element model of a latticed shell.

  • Rod Damage Diagnosis: The study successfully diagnosed mild, moderate, and severe damage in selected rods (members 1, 146, and 469). For instance, damage to member 1 and 146 resulted in significant changes in the damage probability, reaching 100% for severe damage. Member 469, an oblique spar, showed less pronounced changes for minor damage, highlighting its lower sensitivity.
  • Bearing Restraint Failure Diagnosis: The method effectively diagnosed bearing restraint failures, particularly in the x-direction, with damage probabilities reaching 100% for members 1, 146, and 469. Failures in the y-direction also showed high abnormal probabilities, while z-direction failures were less effectively diagnosed.
  • Qualitative Diagnosis: The mean values of the diagnostic index showed clear trends with increasing damage, allowing for qualitative judgment of damage severity.

Limitations and Future Work

  • Simulation-Based Validation: The current validation relies solely on numerical simulations. The authors acknowledge that real-world monitoring data is more complex and suggest that future work should focus on verifying the method's effectiveness in actual monitoring environments.
  • Sensitivity to Damage Types: The method's effectiveness varies for different structural elements and damage types. It is less effective for oblique spars and z-directional support restraints. Future research could explore modifications or complementary methods to address these limitations.
  • Complexity of Real-World Data: The paper mentions that real-world data is more complicated than simulated data. Further investigation into handling diverse and complex real-world monitoring data is warranted.
  • Sensor Placement Optimization: While importance and sensitivity analyses guide sensor selection, further research into optimal sensor network design for comprehensive coverage and robustness could be beneficial.

Summary updated

BibTeX
@misc{xu2022damage,
  author = {Jie Xu and Zhengyang Zhao and Qian Ma and Ming Liu and Giuseppe Lacidogna},
  title = {{Damage Diagnosis of Single-Layer Latticed Shell Based on Temperature-Induced Strain under Bayesian Framework}},
  year = {2022},
  doi = {10.3390/s22114251},
}
Figure