Hybrid 3D printed three-axis force sensor aided by machine learning decoupling

Abstract

Identification of magnitude and orientation for spatially applied loading is highly desired in the fields of not only the machinery components but also human-machine interaction. Despite the fact that the 3-axis force sensor with different structures has been proposed to measure the spatial force, there are still some common limitations including the multi-step manufacturing-assembly processes and complicated testing of decoupling calibration. Here, we propose a rapid fabrication strategy with low-cost to achieve high-precision 3-axis force sensors. The sensor is designed to compose of structural Maltese cross base and sensing units. It is directly fabricated within one step by a hybrid 3D printing technology combining deposition modeling (FDM) with direct-ink-writing (DIW). In particular, a machine learning (ML) model is used to convert the strain signal to the force components. Instead of a mount of calibration tests, this ML model is trained by sufficient simulation data based on programmed batch finite element modeling. This sensor is capable of continuously identifying a spatial force with varying magnitude and orientation, which successfully quantify the applied force of traditional Chinese medicine physiotherapy including Gua Sha and massage. This work provides insight for design and rapid fabrication of multi-axis force sensors, as well as potential applications.

Keywords

• Hybrid 3D printing
• 3-axis force sensor
• machine learning
• nonlinear decoupling