Structural Health Monitoring and Impact Detection Using Neural Networks for Damage Characterization

This report discusses the development of an embeddable impact detection system utilizing an array of piezoelectric wafer active sensors (PWAS) and a microcontroller. Embeddable systems are a critical component to successfully implement a complete and robust structural health monitoring system. System capabilities include impact detection, impact location determination and digitization of the impact waveform. A custom algorithm was developed to locate the site of the impact.. The embedded system has the potential for additional capabilities including advanced signal processing and the integration of wireless functionality. For structural health monitoring applications it is essential to determine the extent of damage done to the structure. In an attempt to determine these parameters a series of impact tests were conducted using a ball drop tower on a square aluminum plate. The response of the plate to the impact event was recorded using a piezoelectric wafer sensor network attached to the surface of the plate. From this testing it was determined that several of the impact parameters are directly correlated with the features recorded by the sensor network.

Download Full-text

A variational Bayesian neural network for structural health monitoring and cost-informed decision-making in miter gates

Structural Health Monitoring ◽

10.1177/1475921720904543 ◽

2020 ◽

pp. 147592172090454 ◽

Cited By ~ 2

Author(s):

Manuel A Vega ◽

Michael D Todd

Keyword(s):

Neural Network ◽

Neural Networks ◽

Decision Making ◽

Structural Health Monitoring ◽

Health Monitoring ◽

Informed Decision ◽

Informed Decision Making ◽

Bayesian Neural Network ◽

Bayesian Neural Networks ◽

Structural Health

Many physics-based and surrogate models used in structural health monitoring are affected by different sources of uncertainty such as model approximations and simplified assumptions. Optimal structural health monitoring and prognostics are only possible with uncertainty quantification that leads to an informed course of action. In this article, a Bayesian neural network using variational inference is applied to learn a damage feature from a high-fidelity finite element model. Bayesian neural networks can learn from small and noisy data sets and are more robust to overfitting than artificial neural networks, which make it very suitable for applications such as structural health monitoring. Also, uncertainty estimates obtained from a trained Bayesian neural network model are used to build a cost-informed decision-making process. To demonstrate the applicability of Bayesian neural networks, an example of this approach applied to miter gates is presented. In this example, a degradation model based on real inspection data is used to simulate the damage evolution.

Download Full-text

Artificial Neural Networks for Structural Health Monitoring of Helicopter Harsh Landings

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.390.192 ◽

2013 ◽

Vol 390 ◽

pp. 192-197

Author(s):

Giorgio Vallone ◽

Claudio Sbarufatti ◽

Andrea Manes ◽

Marco Giglio

Keyword(s):

Neural Networks ◽

Artificial Neural Networks ◽

Structural Health Monitoring ◽

Health Monitoring ◽

Structural Damage ◽

Numerical Models ◽

Damage State ◽

Operational Conditions ◽

Structural Health ◽

Artificial Neural

The aim of the current paper is to explore fuselage monitoring possibilities trough the usage of Artificial Neural Networks (ANNs), trained by the use of numerical models, during harsh landing events. A harsh landing condition is delimited between the usual operational conditions and a crash event. Helicopter structural damage due to harsh landings is generally less severe than damage caused by a crash but may lead to unscheduled maintenance events, involving costs and idle times. Structural Health Monitoring technologies, currently used in many application fields, aim at the continuous detection of damage that may arise, thereby improving safety and reducing maintenance idle times by the disposal of a ready diagnosis. A landing damage database can be obtained with relatively little effort by the usage of a numerical model. Simulated data are used to train various ANNs considering the landing parameter values as input. The influence of both the input and output noise on the system performances were taken into account. Obtained outputs are a general classification between damaged and undamaged conditions, based on a critical damage threshold, and the reconstruction of the fuselage damage state.

Download Full-text