147 research outputs found
A particle filter-based model selection algorithm for fatigue damage identification on aeronautical structures
No full textThe early diagnosis of cracks in aeronautical structures is a fundamental task for the safe system operation and the optimization of maintenance policies, in view of the increasing interest in life extension programs of several high-investment industries. In principle, these tasks could be fulfilled within a condition-based framework, where direct or indirect observations of the degradation evolution are processed, possibly in real time, by proper diagnostic computational tools. In the past, several attempts have been made to build real-time monitoring systems collecting strain signals acquired from sensor networks. However, in real applications, some issues remain unresolved, for example, the large number of observations available to be handled within a unique diagnostic framework, their relationship with the underlying crack size, and their typical large uncertainties. In this paper, we provide a practical solution by innovatively combining a particle filtering-based model identification algorithm with a measurement system exploiting real-time observations of the crack length reconstructed by a committee of artificial neural networks. The artificial neural networks are trained by simulated strain fields generated by a finite element model. The resulting tool allows to perform an automatic, simultaneous, and real-time (a) selection of the model more properly describing the system state evolution, so as to detect the crack propagation onset time, (b) estimation of the model parameters, and (c) estimation of the crack length, within a unique probabilistic framework based on particle filtering. The methodology is demonstrated with reference to a real helicopter panel subject to fatigue and equipped with a fiber Bragg grating sensor network
Real-time prognosis of random loaded structures via Bayesian filtering: A preliminary discussion
No full textParticle filters are effective tools for the monitoring of damage propagation phenomena. However, a common hypothesis of particle filters for damage prognosis is the constant-amplitude fatigue loading affecting the damage growth. This work constitutes a preliminary analysis of the performance of particle filtering in case of random load, relaxing the hypothesis of constant-amplitude fatigue. Two case studies referring to stationary random loads are introduced: the first concerning a narrow-band stress history, while the second focusing on a wide-band stress spectrum. A solution for each case study is provided and validated using numerical simulations of fatigue cracks in a metallic plate
Real- Time sequential monte carlo sampling based on a committee of artificial neural networks for residual lifetime prediction of a component subjected to fatigue crack growth
No full textAbstractMost of the studies available in the literature about sequential Monte-Carlo sampling algorithm assume that a sufficient number of process observations is available, to guarantee the convergence of the algorithm on the target process evolution. This requirement is remotely met if the process of fatigue crack growth is concerned, due to the costs of maintenance procedures, especially within the aeronautical field. A real-time diagnostic system is the enabler of the prognostic health monitoring methodology.This work is about the application of sequential Monte-Carlo sampling to estimate the probabilistic residual lifetime of a monitored structural component, subjected to fatigue crack propagation. A real-time diagnostic unit for crack detection and damage assessment, trained with Finite Element simulations of damage evolution, generates the information as input to the prognostic unit. A crack propagation model provides the knowledge of the residual lifetime prior to the application of fatigue loads. The prognostic unit updates in real-time the probability density functions of the component residual lifetime at each discrete time during fatigue crack evolution. The methodology is preliminarily applied in simulated environment to an aeronautical metallic panel and the overall performance of a fully autonomous prognostic health monitoring system based upon simulated strain measures is evaluated.
Continuous crack growth monitoring and residual life prediction under variable- Amplitude loading conditions
No full textAbstractThe paper deals with the problem of fatigue crack growth under variable-amplitude loading from a lifetime prediction viewpoint. A sequential Monte Carlo technique is employed to monitor crack propagation in presence of several uncertainties related to the material properties, measurement systems and environmental variability. The algorithm is able to estimate the most probable parameters describing crack growth data focusing on the most probable crack growth trajectories and enhancing the prediction of the residual life of the structure. Monte Carlo sampling allows accounting for the variable amplitude loading condition, simulating several crack growth evolutions using different loads and selecting the more appropriate for the actual crack evolution. The outcome of the algorithm that is the residual life prediction is used to appreciate the performances of the method. The end of the paper discusses the application of the method within structural health monitoring systems and lifetime predictor frameworks
Evaluation of multiple damage-mode models for prognostics of carbon fiber-reinforced polymers
No full textDamage growth models are the linchpin of model-based prognostics for aging or damaged structures. Fatigue damage growth in composites can be characterized through estimating strain energy release rates and observed through stiffness reduction in the structure. The work reported herein investigates several existing models for the estimation of strain energy release rate and reduction in effective stiffness of cross-ply laminates under multiple damage modes. These models account for coexistence of matrix cracks and delamination, and have been examined using publicaly available fatigue data on carbon fiber-reinforced polymers coupons. This study is driven by the desire to identify suitable models in a prognostic context, i.e., to predict the remaining useful life of composite laminates subject to fatigue loads. Therefore, the capability of models in estimating the strain energy release rate and in accurately describing progressive degradation of the material has been analyzed. Selected models will then be used in conjunction with modified Paris' law to predict the evolution of damage progression in cross-ply laminates
A Bayesian framework for fatigue life prediction of composite laminates under co-existing matrix cracks and delamination
No full textThis paper proposes a particle filter-based Bayesian framework for damage prognosis of composite laminates exhibiting concurrent matrix cracks and delamination. Literature shows a number of applications of particle filtering for real-time prognosis of metallic structures and, recently, matrix crack density evolution in composites. The work presented here enhances the methodology proposed in previous papers by extending the Bayesian framework to multiple damage mechanisms, and validates the approach using damage progression data from notched cross-ply CFRP coupons subject to tension-tension fatigue. A multiple damage-mode model for the estimation of the strain energy release rate and the remaining stiffness of damaged laminates constitutes the core of the particle filtering algorithm, thus allowing the prognostic framework to extend for monitoring of simultaneous, coexisting damages. Also, the damage state can be evolved into the future enabling simulation of damage progression and prediction of remaining useful life of the composite material. The proposed prognostic unit successfully predicts damage growth and fatigue life of the laminate, and the results are critically discussed with respect to filtered estimation of damage progression and remaining life prediction
Sequential Monte Carlo sampling for crack growth prediction providing for several uncertainties
No full textThe problem of fatigue crack growth monitoring and residual lifetime prediction is faced by means of sequential Monte Carlo methods commonly defined as sequential importance sampling/resampling or particle filtering techniques. The algorithm purpose is the estimation of the fatigue crack evolution in metallic structures, considering uncertainties coming from phenomenological aspects and material properties affecting the process. These multiple uncertainties become a series of unknown parameters within the framework of the dynamic state-space model describing the crack propagation. These parameters, if correctly estimated within the particle filtering algorithm, will cover the uncertainties coming from the real environment, improving the prognostic performances. The standard particle filter formulation needs additional methods to augment the state vector and to correctly estimate the parameters. The prognostic system composed by the sequential Monte Carlo algorithm able to account for different uncertainties is tested through several crack growth simulations. The applicability of the method to real structures and the employment in presence of real environmental conditions (i.e. variable loading conditions) is also discussed at the end of the paper. 1
- …
