An adaptive distributed dipole model for enhanced dynamic inversion in magnetic field and induction sensor
Date of Issue2014
School of Mechanical and Aerospace Engineering
A magnetic-field-based sensing/imaging system, being noninvasive, has been widely implemented for anomaly detection in various applications. The sensing/imaging performance is highly dependent on the modeling approach, which decides accuracy, efficiency and implementation. Distributed Multi-pole Model, as a semi-analytical modeling method, has been developed to simulate magnetic field from permanent magnet and direct current-carrying coils, but its application is limited to static field problems due to usage of scalar-potential-based governing equations. For sensing/imaging purpose, modeling of time-varying electromagnetic field plays a more critical role because it provides abundant information in both spatial and time domains. The objective of this research is to develop a modeling approach for time-varying electromagnetic field and explore its capability for sensing/imaging purpose. The magnetic point dipole directly conceptualizes the physical field information (automatically satisfying Maxwell’s equations) and offers a compact solution to account for the field variation. Then originated from vector potential formulas, the extended distributed multi-pole model is developed to account for AC source and further exploited to describe the inductive relation between two coils. The commonly seen integral form of the induction equation can be simply discretized based on extended distributed multi-pole models. Therefore real-time and robust mutual induction computation can be realized and orientation sensing schemes being developed. As a semi-analytical modeling method utilizing discretized source terms, computation for induction can be more efficient and simply applicable for various sensor configurations. Another promising application is introduced as magnetic induction tomography (MIT), which utilizes compact measuring devices with moderate electromagnetic field, and thus could be a safer and portable imaging system for medical applications. With a number of point dipoles spanning the target domain and effectively characterizing the field from eddy current, a novel modeling approach is developed. Since there is a linear relation between the dipole moment and peripheral field variables, the forward mapping can be directly built, and the matrix for inversion is much more simplified. Compared to directly solving for conductivity, this inversion procedure for the dipole moment is proven to be fast-computing, which also provides guidance for an adaptive refining process. While induction-based sensing and tomography has been developed with an easily implementable approach that discretizes the field problems from distributed dipole models, we expect the modeling and discretization method will have broader applications in matrix regularization and signal processing.