Prediction of damage parameters in an inclined curved cracked beam based on vibration signatures using an Adaptive Neuro-Fuzzy Inference System (ANFIS)
摘要
Curved beams are extensively used in engineering applications when weight-saving requirements are paramount. Therefore, it is essential to thoroughly research their dynamic behavior. In this research study, Differential Quadrature Method (DQM), Finite Element Method (FEM) using ANSYS, and experimental analysis are conducted to analyse the free in-plane vibration of curved cracked beams, with the aim of determining vibration signatures such as natural frequencies and mode shapes. Adaptive Neuro-Fuzzy Inference System (ANFIS) a reverse model is developed to predict the damage parameters using over three thousand datasets obtained from the aforementioned theoretical, numerical and experimental analysis. Boundary conditions, radius of curvature (R), and the first three natural frequencies are the input parameters considered for the ANFIS model, while Relative Crack Depth (RCD) and Relative Crack Location (RCL) are the output parameters of the model. The effectiveness of the input parameters is assessed using various statistical indicators, including the Root Mean Square Error (RMSE) and the coefficient of determination (R2). Sensitivity analysis is also used to identify and prioritize the input parameters that significantly influence the natural frequencies. Furthermore, the ANFIS model demonstrated superior effectiveness, boasting a higher coefficient of determination (≥ 0.99) and lower root mean square error (≤ 0.0076). Experimental validation further confirmed the accuracy of the predictive models in comparison with experimental findings.