.png){kind=icon}
.png){kind=icon}
Return
Instantaneous power prediction for industrial robots using tree-based machine learning methods
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2025.112339.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This article presents a tree-based machine learning methodology for instantaneous power prediction designed, tested and validated using data from an articulated industrial robot. The proposed methodology for instantaneous power prediction materializes through a generic system architecture with functionalities consisting of data acquisition, time alignment of data samples, storage, model learning, instantaneous power prediction and integration in time to evaluate energy consumption at robot operation level. This methodology is designed to evaluate offsite energy consumption of robotized workstations for different layouts characterized by relative position of the robot with respect to the serviced and fly-by points. This is important both for offline virtual commissioning of robotized workstations (determine layout) and for online operation for maintenance purposes (determine energy spikes different from normal model). The analyzed operation is the linear motion of the robot Tool Control Point in Cartesian space, characterized by the complexity of the kinematic model: each joint operates in coordinated motion, adjusting its velocity and acceleration continuously to ensure a straight path with constant speed. A custom Internet of things (IoT) device enables synchronized energy and motion data logging for robots, ensuring consistent values for sampled trajectories. Justification for the usage of tree-based methods and experimental results are provided.
Journal
IF:
8
Papers:
5.4K
Citations:
3.5W