Development Of ARMA Algorithm For System Identification In Structure Health Monitoring
Download| Volume 3 Issue 2 2022 | |
|---|---|
| Author(s): |
Faizal Afzal M. Irfan Anis Marium Shakeel |
| Abstract | Amplitudes, peaks shift and waveforms altered the signal response when structural damage severity and position change, thus having a strong association between damage cases and signal response shapes. The damage detection and building structural state evaluation are tough to comprehend. As the number of structural failures has increased, the development of methods for recognizing the degradation become increasingly important. Structural health of buildings and critical infrastructure should be monitor for signs of deterioration or impending failure. Time series modeling and forecasting was an effective tool for structural analysis. To emulate via artificial neural networks the behavior of time series model, the Autoregressive with Moving Average (ARMA) represents one of the most effective methods for structural characterization in operative environments. The novelty of this work is to attain using the unique TinyML approach, that allowed for the embodiment of the developed models into a resource-constrained device. The proposed models were trained and tested using data collected by sensors strategically placed on engineering structures. The identical models were then converted to the TFLite format, that are designed for devices with limited resources. The converted models were tested using the same dataset on Arduino and STM32 boards to assess how they perform in real-world scenarios. For univariate time series forecasting an Artificial Neural Network (ANN) method reduce computational effort and reduced severity outperformed Recurrent Neural Networks (RNN) and Convolutional Neural Networks (CNN). |
| Keywords | Degree of Freedoms, Structural Health Monitoring, Tiny Machine Learning, TensorFlow Lite, Neural Network |
| Year | 2022 |
| Volume | 3 |
| Issue | 2 |
| Type | Research paper, manuscript, article |
| Recognized by | Higher Education Commission of Pakistan, HEC | Category | Journal Name | ILMA Journal of Technology & Software Management | Publisher Name | ILMA University | Jel Classification | -- | DOI | - | ISSN no (E, Electronic) | 2709-2240 | ISSN no (P, Print) | Country | Pakistan | City | Karachi | Institution Type | University | Journal Type | Open Access | Manuscript Processing | Blind Peer Reviewed | Format | Paper Link | https://ijtsm.ilmauniversity.edu.pk/arc/Vol3/i2/1.pdf | Page |