Semi-Supervised Support Vector Machines algorithms as Classification Methods in Structural Health Monitoring

One of the fields in data-based structural health monitoring (SHM) that has not been widely considered is the data classification step. Applications of the semi-supervised methods in data classification is getting more attention nowadays. In this study, an efficient semi-supervised support vector machine (S3VM) algorithm is used to for classifying between healthy and unhealthy stages. For this reason, a combined model-based and data-based approach is taken to determine the damage sensitive features. A hybrid approach has been utilized to generate the feature vectors. Using the vibrational data of the structure, the dynamic properties is obtained by system identification methods. Modal strain energy used as damage sensitive features (DSF). Different states of healthy and unhealthy conditions of the structure is used to evaluate the effectiveness of the proposed algorithm. Also, the Support Vector Machines (SVM) algorithm is utilized to compare the results. Since the semi-supervised support vector machines algorithm is based on support vector machines formulation, it is a suitable algorithm to compare the result with. It can be seen that the use of unlabeled data will enhance the effectiveness of the classification methods especially in the lack labeled data. When the labeled dataset is large enough, the result for both supervised and semi-supervised support vector machines is almost the same.