Learning‐Based Damage Recovery for Healable Soft Electronic Skins

Abstract

Natural agents display various adaptation strategies to damages, including damage assessment, localization, healing, and recalibration. This work investigates strategies by which a soft electronic skin can similarly preserve its sensitivity after multiple damages, combining material-level healing with software-level adaptation. Being manufactured entirely from selfhealing Diels-Alder matrix and composite fibers, the skin is capable of physically recovering from macroscopic damages. However, the simultaneous shifts in sensor fiber signals cannot be modelled using analytical approaches, since the materials viscoelasticity and healing processes introduce significant nonlinearities and time-variance into the skin's response. We show that machine learning of 5-layer networks after 5000 probes leads to highly sensitive models for touch localization with 2.3mm position and 95% depth accuracy. Through health monitoring via probing, damage and partial recovery are localized. Although healing is often successful, insufficient recontact leads to limited recovery or complete loss of a fiber. In these cases, complete resampling and retraining recovers the networks' full performance, regaining sensitivity and further increasing the system's robustness. Transfer learning with a single frozen layer provides the ability to rapidly adapt with fewer than 200 probes.

Effect of Secondary Particles on Self-Healing and Electromechanical Properties of Polymer Composites Based on Carbon Black and a Diels–Alder Network

“…CB360G presents a lower percolation threshold than CB260G because of its relatively higher surface area. Because of the compromise between excellent electrical conductivity and retaining decent self-healing properties, the focus in this work was on the composites using CB360G, aiming at the use of lower CB contents. ,, The increase in electrical conductivity is most significant between 5 and 7.5 wt % CB360G (Figure a), indicating that the 3D network of carbon black particles forms in the polymer network above 5 wt % CB, leading to decent electrical properties of the CB360G composites above 7.5 wt %, whereas the CB260G composites do not yet show significant electrical conductivity at this range.…”

“…This time-dependent response of the conductive network makes calibration of the sensor based on analytical models challenging. Nevertheless, machine learning techniques have proven to be a useful tool to model highly nonlinear and time-dependent self-healing flexible sensors based on Diels–Alder composites …”

“…An example is self-healing resistive strain sensors that recover their sensor behavior after multiple damage-healing cycles . In addition, the healing capacity offers the opportunity to use these sensors for damage detection as well in which both damage and recovery can be registered in the sensor signal. Therefore, their integration led to an embedded health monitoring system in devices.…”

Sensorized Skin With Biomimetic Tactility Features Based on Artificial Cross‐Talk of Bimodal Resistive Sensory Inputs

“…Changing these parameters would be expected to result in a shift in mean error, and could quickly be accounted for by retraining just the network's final layer. [ 59 ]…”

“…Changing these parameters would be expected to result in a shift in mean error, and could quickly be accounted for by retraining just the network's final layer. [59] In Section 2.1, it was hypothesized that a network's localization performance would improve with the probing depth. Thus, Figure 4c fulfilled these expectations for both the small and large skin.…”