MULTI-MODAL PARAMETER BASED DELAMINATION DETECTION IN COMPOSITE STRUCTURES USING METHODS OF ARTIFICIAL INTELLIGENCE

Vibration based delamination detection methods in composites, currently employed for structural health monitoring solutions uses Natural frequency as their basic damage indicator. Other indicators such as the mode shapes and damping parameters if used could enhance their capability of damage detection. Degradation of the structures due to delamination will reduce the strength of the material, i.e., its flexural stiffness, which will alter not only the natural frequencies but the mode shapes and damping parameters which could detect the presence of delamination, assessing its size, location and the interface. The problem is basically divided into two parts, solving the forward problem of identifying the changes in natural frequencies and solving the inverse problem from the forward data. Artificial Neural networks like the Multi-Layer Perceptron has the ability to incorporate multiple damage indicators for producing the outputs. FE Models have been used along with the experiments to compare the data generated and validated against the simulated ones. FE models by themselves are computationally expensive, so surrogate models have been used to reduce the computational expense. This method is called Surrogate Assisted Optimization (SAO). In order to carry out SAO, Response Surface Models (RSM) has been developed to reduce the number of training data sets required for solving the inverse problem. It has been seen that the algorithms are highly efficient in detection of damage in Composite beams and plates despite of the limited number of training datasets provided and the introduction of artificial errors and noise. Moreover the robustness of the algorithm were clearly evident when the errors where quantified using the experimental and simulated data. Thus the algorithms provide an efficient means of detecting the delamination parameters accurately from the vibration data.