.png){kind=icon}
.png){kind=icon}
Return
Enhancing data-driven input reconstruction via optimized sensor balancing
.png){kind=icon}
PRE
AI
DOI：10.1016/j.ymssp.2024.111184.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The utilization of data -driven techniques is gaining popularity in modeling the behavior of complex mechanical systems. Specifically, non -parametric estimators based on deconvolution operations of measured quantities have been extensively utilized to reconstruct input signals that influence these systems. This study introduces a sensor balancing approach that leverages two commonly measured quantities, strain and acceleration, within the context of input force estimation using the impulse response filter. An optimization methodology is proposed, based on hammer impact measurements used as training data, to determine the optimal sensor scaling values that improve force estimates. The proposed method provides a customized balancing scheme for non -parametric estimators, eliminating estimate bias caused by noisy sensors and the different scaling units associated with the tracked quantities. To illustrate the benefits of the sensor fusion approach, an industrial validation case involving a vehicle rear differential sub -frame is presented. Additionally, a comparison is made in terms of computational time and performance for various optimization alternatives in both the time and frequency domains.
Keywords:
Impulse response filter
Sensor fusion
Deconvolution
Strain and acceleration
Non-parametric model
Weighted least squares
Journal
IF:
8.9
Papers:
1.3W
Citations:
6.6W