The article presents a research of angular orientation based on a microelectromechanical system (MEMS) accelerometer by using machine learning (ML) and deep learning (DL) model with architectures of deep neural networks (DNNs). In the industrial environment, artificial intelligence (AI) plays a crucial role in automation which is a potential solution for better performance of inclinometer. This article was carried out to apply this intelligent model on the inertial measurement unit to accomplish the angular position. The experiment shows that the ML model correctly learns the relationship between acceleration and tracking angles via polynomial regression with an R-square of 0.98. The employed DL model with four hidden layers of ten neurons achieves an accuracy of 99.99 % and almost a nonerror performance. The acceleration acquisitions were obtained from MEMS accelerometer LSM9DS1 at a frequency of 50 Hz via microcontroller STM32F401RE. The ML and DNN models were designed based on the platform Tensorflow with high processing accuracy. The Pan-Tilt Unit was used as the angle reference for static and dynamic tests. The traditional technique is used for comparison as well as verification of the proposed models. The DL model has better precision over the ML model due to its high structure level with updating weight and error optimization from the neural network structure. Meanwhile, ML shows more stable results in dynamic circumstances.

New Artificial Intelligence Approach to Inclination Measurement Based on MEMS Accelerometer

Hoang M. L.;Pietrosanto A.
2022-01-01

Abstract

The article presents a research of angular orientation based on a microelectromechanical system (MEMS) accelerometer by using machine learning (ML) and deep learning (DL) model with architectures of deep neural networks (DNNs). In the industrial environment, artificial intelligence (AI) plays a crucial role in automation which is a potential solution for better performance of inclinometer. This article was carried out to apply this intelligent model on the inertial measurement unit to accomplish the angular position. The experiment shows that the ML model correctly learns the relationship between acceleration and tracking angles via polynomial regression with an R-square of 0.98. The employed DL model with four hidden layers of ten neurons achieves an accuracy of 99.99 % and almost a nonerror performance. The acceleration acquisitions were obtained from MEMS accelerometer LSM9DS1 at a frequency of 50 Hz via microcontroller STM32F401RE. The ML and DNN models were designed based on the platform Tensorflow with high processing accuracy. The Pan-Tilt Unit was used as the angle reference for static and dynamic tests. The traditional technique is used for comparison as well as verification of the proposed models. The DL model has better precision over the ML model due to its high structure level with updating weight and error optimization from the neural network structure. Meanwhile, ML shows more stable results in dynamic circumstances.
2022
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11386/4827812
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 7
  • ???jsp.display-item.citation.isi??? ND
social impact