Dual-Modal Material Identification Method via MTEG-TENG Synergistic Sensing and Machine Learning Optimization in Multiple Environments

Material identification sensors, as the core components
that endow
robots with intelligent perception capabilities, are crucial for their
development and innovation. However, the complexity and diversity
of environmental conditions pose greater challenges to the accuracy
of sensors in identifying materials. This paper proposes a dual-modal
collaborative material identification method based on microthermoelectric
generator (MTEG) and triboelectric nanogenerator (TENG). A prototype
is fabricated which consists of a thermal tactile material identification
unit (TT-IU) based on MTEG and a contact electrification material
identification unit (CE-IU) based on TENG. The TT-IU measures voltage
induced by the difference in temperature between its two ends, reflecting
the material’s thermal diffusivity. The CE-IU measures voltage
produced when materials contact with the unit, indicating the electron
affinity of materials. Since individual material has distinct thermal
diffusivity and electron affinity, the classification of materials
can be achieved by correlating and analyzing these two independent
voltage data. To verify the material identification capability of
this method, a MTEG-TENG dual-modal collaborative characteristic material
identification performance validation experiment system is set up.
Furthermore, this paper delves into the impact of external conditions
and contact conditions such as contact pressure, material surface
roughness, ambient temperature and humidity on recognition performance.
The experiment results indicate that under open conditions, the material
identification method can significantly distinguish between materials.
Integrated with machine learning techniques, the material identification
method achieves identification of eight characteristic materials under
various external conditions with an overall identification accuracy
of 93.54%.