Model reference Neural Network-based methodology for vibration control in a five-story steel structure.
The recent natural hazards that have taken place around the world led to the development of structures that can truly adapt to the ever-changing operational conditions. One of the main goals that an adaptive structure is to mitigate the vibrations that a civil structure is being subjected to, where artificial intelligence-based proposals are one of the active research topics, since they have been demonstrated noticeable results in vibration mitigation in civil engineering. Bearing this information in mid, this work proposes a reference-based vibration controller using a neural network implementation to reduce the negative impact of such conditions and, consequently, avoid damages into the structure. In this regard, a recurrent topology network is employed to implement the model reference and the controller, respectively. A magnetorheological damper is chosen to provide the actuator device, as it offers the best compromise between the required power and has the flexibility to still offer a protection when it is not powered. A simulation scheme is developed in this work, where the obtained results show a 18%-reduction in the vibrations measured at the second story of a five-story steel building, demonstrating the capabilities that the proposal could have for vibration control in real-life scenarios.
Paper
Full text
Model reference Neural Network-based methodology for vibration control in a five-story steel structure.
Semantic Scholar · Engineering · 2020
Abstract
The recent natural hazards that have taken place around the world led to the development of structures that can truly adapt to the ever-changing operational conditions. One of the main goals that an adaptive structure is to mitigate the vibrations that a civil structure is being subjected to, where artificial intelligence-based proposals are one of the active research topics, since they have been demonstrated noticeable results in vibration mitigation in civil engineering. Bearing this information in mid, this work proposes a reference-based vibration controller using a neural network implementation to reduce the negative impact of such conditions and, consequently, avoid damages into the structure. In this regard, a recurrent topology network is employed to implement the model reference and the controller, respectively. A magnetorheological damper is chosen to provide the actuator device, as it offers the best compromise between the required power and has the flexibility to still offer a protection when it is not powered. A simulation scheme is developed in this work, where the obtained results show a 18%-reduction in the vibrations measured at the second story of a five-story steel building, demonstrating the capabilities that the proposal could have for vibration control in real-life scenarios.