Structural Damage Diagnosis Using Subspace Based Residual and Artificial Neural Networks

K. Saeed (1) , N. Mechbal (1) , Michel Vergé (1) , Gérard Coffignal (1)

1 LMSP - Laboratoire de Mécanique des Systèmes et des Procédés

An Artificial Neural Network (ANN) based approach using a new non-parametric residual is presented for damage diagnosis. The residual is associated with Observability null-space of the system and is generated by using a parity matrix obtained from covariance driven output-only Subspace Identification (SubID) algorithm. Training of ANN is established using residuals generated from Finite Element (FE) model of the structure under consideration for different defect cases. This trained ANN is in turn used to identify the predefined defect types, in semi-real time, on the actual structure. The proposed methodology is applied to an active composite beam structure and its effectiveness to identify damage is studied by testing a trained ANN with data obtained from both simulation and experimentation. Promising results were obtained showing that, it was possible to distinguish between healthy and damaged states with good accuracy and repeatability.

Domaines: Sciences de l'ingénieur [physics]

K. Saeed, N. Mechbal, Michel Vergé, Gérard Coffignal. Structural Damage Diagnosis Using Subspace Based Residual and Artificial Neural Networks. 7th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, SafeProcess, 2009, Barcelona, Spain. pp. ⟨hal-00362540⟩