Digital Twin for Health Monitoring of a Cantilever Beam Using Support Vector Machine

PurposeDigital twinning has emerged as a notable technology trend in recent years, primarily due to its widespread applicability in the industrial domain. However, practical adoptions of this technology have been slower. The combination of big data analytics and artificial intelligence/machine learning (AI-ML) methods with digital twinning strengthens its significance and research possibilities. Structural elements can undergo changes over time due to degradation mechanisms or other environmental factors, affecting its condition and performance. In some cases, the behavior of a degraded yet acceptable structure may resemble that of a damaged one, potentially leading to misinterpretation and misguided maintenance decisions. The effectiveness of a classical damage detection system largely relies on the selection of features and the classifier, often neglecting the ongoing changes in the structure and the evolution of damage.MethodsThis study addresses this issue by proposing a novel approach to develop a Digital Twin framework utilizing the Support Vector Machines(SVM) for the Structural Health Monitoring(SHM) of a cantilever beam. The virtual model of the physical system is formulated as a computational model using differential equations involving two distinct time scales. The slow time notion is employed to separate the gradual evolution of the system stiffness from the instantaneous time. The computational model is used to investigate different damage scenarios and stiffness-degraded conditions. The SVM classifier is trained with data in the form of the Frequency Response Function(FRF) taken from the computational model. Various parameters, such as damage location, intensity, and stiffness degradation, are considered when constructing the training datasets. Different kernels for the SVM classifier and ratios for the training to-testing data sets are tested for better accuracy of the proposed Digital Twin framework.ResultsThe Digital Twin is found to be robust and yields reasonably accurate results for the Support Vector Machine classifier with second order polynomial kernel.ConclusionThis study presents a Digital Twin (DT) framework for Structural Health Monitoring (SHM) of a cantilever beam using AI-ML techniques. Damage identification and localization were evaluated under various conditions, including hold-out ratios. The DT framework, employing SVM with polynomial kernels (2nd and 3rd order), achieved high accuracy for distinguishing healthy and damaged states. For damage localization, the 2nd order polynomial kernel reliably classified all damage locations, outperforming linear and RBF kernels. This underscores the suitability of polynomial kernels within the DT framework for SHM applications.