Fabrication of triboelectric nanogenerators with multiple strain mechanisms for high-accuracy material and gesture recognition

Textile-based triboelectric nanogenerators (T-TENGs) have great potential in new-generation tactile sensors because their output signal is determined by the specific charge interaction between the target and triboelectric material. However, gesture recognition by existing T-TENG tactile sensors is very challenging because the electron affinity and polarity of the volunteer skin are similar. This study proposes a new structural sensor that integrates a Ni particle-encapsulated carbon tube (Ni@C tube) with polydimethylsiloxane (PDMS) by means of multiple strain mechanisms for improving recognition accuracy using a signal superposition technique. Interestingly, dynamic strain from multiple sources, i.e., PDMS layer strain (1.24-2.47 kPa), Ni@C tube crack (2.47-5.56 kPa), PDMS layer further strain (5.56-52.97 kPa), and Ni@C tube slip laterally (52.97-139.05 kPa), was established in the PDMS-coated Ni@C tube. In addition, the output signal of the as-fabricated T-TENG device was a combination of output signals from the PDMS/Ni@C tube and PDMS/contact object devices, which could fully reflect the stress field caused by the contact material pressing the surface and the surface characteristics, promoting the deep learning algorithm to be able to extract more hidden features of the identified materials. By integrating convolutional neural network (CNN) analysis based on the voltage signals of the T-TENG device, we explored a strategy that could enable the as-fabricated device to recognize 6 gestures with 99.3% recognition accuracy and 8 different materials with 99.8% recognition accuracy in the natural environment.

Automated summary

This study proposes a new structural sensor that integrates a Ni particle-encapsulated carbon tube (Ni@C tube) with polydimethylsiloxane (PDMS) by means of multiple strain mechanisms for improving recognition accuracy using a signal superposition technique. By integrating convolutional neural network (CNN) analysis based on the voltage signals of the T-TENG device, we explored a strategy that could enable the as-fabricated device to recognize 6 gestures with 99.3% recognition accuracy and 8 different materials with 99.8% recognition accuracy in the natural environment.