| dc.description.abstract | Magnetic manipulation has the potential to recast the medical field both from an operational and drug delivery point of view as it can provide wireless controlled navigation over surgical devices and drug containers inside a human body. The accuracy and precision of controlled navigation will provide access to delicate organs and decrease the rehabilitation time. The advantages of achieving such a task have absorbed engineers' and researchers' attention and effort in the electromagnetic, imaging, mechanical, and robotic fields to implement the principles and make a functional magnetic manipulator.
The main idea behind magnetic manipulators is to regulate electrical currents fed to the coils to precisely position and orient an agent- also known as a robot or a magnetic tool- inside a working space. The presented system in this research is formed with nine coils, also known as electromagnets, placed normal to the spherical volume. The radius of this space is directly correlated with the dimensions and the number of coils, which can be utilized to parameterize the spatial constraints.
Extending the number of coils forming a spherical volume, also known as spherical workspace, has led to developing a unique geometrical constraint to optimize the coil placement. The determination of the constraints resulted in a specific outer diameter for each coil. In order to design a coil that produces the maximum axial force with the least power combustion with a given outer radius, Fabry Factor equation and Finite Element Method (FEM) were adopted. Fabry Factor relates the dimensions of the coils to each other such that the power consumption is minimized. Therefore, various iron-core coils were simulated using this method, and then the axial force of each coil at the furthest operational point in the working space was measured using FEM. The optimization result led to a cylindrical iron-core coil with an inner diameter of 20.5 mm, an outer diameter of 66 mm, and a length of 124 mm.
The FEM results in 3D for a complex system is mostly associated with errors between actual and simulated values of the magnetic field, around 17 percent less than the actual values in this project. In order to eliminate this error, the magnetic field of the manufactured coil had been predicted using Artificial Intelligence (AI) techniques for experimental purposes. Regression models of Artificial Neural Network (ANN), a hybrid method called Artificial Neural Network with Simulated Annealing (ANN/SA), and Gene Expression Programming (GEP) had been built individually. ANN/SA has shown outstanding performance with an R-squared equal to 0.99 and root mean square error of 0.0028; hence, it has been used in the actuation process for magnetic field prediction.
Finally, to indicate the functionality of the system, a simple 1D PI actuation logic with 𝑘𝑝= 3.25 and 𝑘𝑖=0.01 using a laser sensor had been successfully investigated. It first predicts the magnetic field using ANN/SA at the agent's current position provided by the laser sensor; then, regulates the current flowing through each coil till the agent settles at the final destination. | en |
