Ultra-Small Damage Detection via EIT in Soft Nanocomposite Sensors

Abstract

Electrical impedance tomography (EIT) has received much attention as a potential sensing modality in robotic, biomedical, human-interfacing, and even structural applications. To date, however, work in this area has focused on sensing relatively large artifacts via EIT---spatially large pressure distributions, sizable damages in composites, large cracks in conductive coatings, etc. This is important because it is often much more desirable to detect early stage damage before it grows. EIT tends to struggle with small damage detection because the method is predicated on the damage causing a measurable voltage change at the domain's boundary. Small damages cause only small boundary voltage perturbations, which can easily be masked by noise. Additionally, commonly used regularizers such as the Laplace operator tend to "smooth over" small damages further masking them. With these limitations in mind, this manuscript makes three contributions: First, a mixed regularization scheme is implemented in the EIT inverse problem. Prior work has shown that this regularizer excels at finding small, highly localized damage. Second, the effect of injection-measurement scheme is explored. And third, a novel EIT test fixture is used that overcomes many of the challenges associated with experimental EIT. Experimental validation is conducted on a representative piezoresistive nanocomposite sensor, carbon nanofiber (CNF)-modified polyurethane (PU), which can be envisioned as a distributed pressure sensor, an artificial or robotic skin, or a sensing coating applied to a structural component. Our results show that with appropriate regularization and injection-measurement schemes, EIT can indeed find very small defects that are on the order of 0.0027% of the sensor's area. These results represent an order-of-magnitude improvement over the current state of the art.