Semi-automated impact device based on human behaviour recognition model for in-service modal analysis

Modal analysis is a reliable method for the study of structural behaviour. A novel modal analysis technique called impact synchronous modal analysis (ISMA) was developed using which modal analysis can be performed in the presence of ambient forces. However, studies determined that the manual operation of this technique is laborious, time intensive and has limited practicality due to the lack of control and knowledge of the impact with respect to the phase angle of the disturbances using conventional impact hammer. A fully automated impact device called automated phase controlled impact device (APCID) was developed to perform in-service modal analysis with minimum number of impacts. However, large size and heavy weight of this device made it unsuitable for real world applications. In this paper, a portable semi-automated impact device is used to perform in-service modal analysis. The device uses the conventional manual impact hammer and is equipped with inertial measurement unit (IMU). It is operated manually and uses human behaviour recognition along with control of APCID which gives indication to impart impact based on human's physical behaviour. This physical behaviour is recognized by classifying different impact types and predicting impact times using machine learning technique from the inertial sensor data. The cyclic load components at 20 Hz and 30 Hz are reduced by 91.2% and 92.5%, respectively, using the proposed ISMA with IMU. The extracted modal parameters are also in good correlation with the benchmark, experimental modal analysis data as well as the previous work using APCID. All the modes are identified with less than 3% difference in natural frequencies, less than 10% difference in damping values and modal assurance criterion values greater than 0.9 for all modes at running frequencies of 20 Hz and 30 Hz.