Characterization and Size Optimization of Additively Manufactured Flexible Piezoresistive Sensors for Vibration Strain Sensing

Abstract

Flexible piezoresistive strain sensors have promising applications in areas such as wearables and soft robotics. For sensing dynamic strains, such as a runner’s gait or a slipping object held by a robotic gripper, these sensors must capably measure strain over a range of amplitudes and frequencies. This thesis presents the characterization and optimization of a flexible piezoresistive sensor with triply periodic minimal surface (TPMS) structures for vibration strain sensing. Sensors are fabricated using an additive manufacturing (AM) process to subsurface coat a silicone rubber (SR) matrix with graphene nanoplatelets (GNP). These sensors are then characterized under uniaxial compressive strain amplitudes from 0-10% and frequencies of 10-110 Hz. Frequency and time domain analyses are used to demonstrate sensor performance and explain unique deformation mechanisms of the TPMS structure. Low sensor delays of less than 6.3 ms, and 0.420 ms on average, demonstrate its capability for high-frequency sensing. Frequency independence of the sensor is also demonstrated, as the mean error due to its sensitivity changing with frequency is only ±3.89%. A second-degree polynomial calibration of the sensor is shown to predict the relationship between strain amplitude and sensor resistance change well, with a mean error of 3.56% for 2-10% strain amplitudes. Sensor durability is proven by testing ten sensors over 15×10^6 cycles and 80 hours without breaking. In addition, a multi-objective size optimization is performed for the TPMS sensor design, with the goal of improving its frequency independence and strain sensitivity. The optimization is solved using a multi-objective firefly algorithm (FA) and accounts for several fabrication constraints when finding a feasible sensor design. The first objective is to maximize the sensor’s first natural frequency, which results in a reduced mean frequency dependence error of ±2.18% during testing. To also attempt improving the sensor’s strain sensitivity, given a negative (compressive) applied strain, the average principal strain at the sensor surface (where the GNP coating is located) is minimized. The implemented algorithm converged within 1618 unique function evaluations, a reduction of 85.5% compared to the entire set of feasible solutions. This design optimization is the first for a flexible piezoresistive sensor in literature.