Abstract

This PhD work presents novel methods of mechanical property identification for soft nonlinear materials and methods of recreating and modeling the deformation behavior of these nonlinear materials for tactile feedback systems.

For the material property identification, inverse modeling method is employed for the identification of hyperelastic and hyper-viscoelastic (HV) materials by use of the spherical indentation test.
Identification experiments are performed on soft foam materials and fresh harvested bovine liver tissue.
It is shown that reliability and accuracy of the identified material parameters are directly related to size of the indenter and depth of the indentation. Results show that inverse FE modeling based on MultiStart optimization algorithm and the spherical indentation, is a reliable and scalable method of identification for biological tissues based on HV constitutive models.

The inverse modeling method based on the spherical indentation is adopted for realtime applications using variation and Kalman filter methods.
Both the methods are evaluated on hyperelastic foams and biological tissues on experiments which are analogous to the robot assisted surgery.
Results of the experiments are compared and discussed for the proposed methods.
It is shown that increasing the indentation rate eliminates time dependency in material behavior, thus increases the successful recognition rate. The deviation of an identified parameter at indentation rates of V=1, 2 and 4 mm/s was found as 28%, 21.3% and 7.3%.
It is found that although the Kalman filter method yields less dispersion in identified parameters compared to the variance method, it requires almost 900 times more computation power compared to the variance method, which is a limiting factor for increasing the indentation rate.
Three bounding methods are proposed and implemented for the Kalman filter estimation.
It was found that the Projection and Penalty bounding methods yield relatively accurate results without failure. However, the Nearest Neighbor method found with a high chance of non-convergence.

The second part of the thesis is focused on the development of tactile displays for modeling the mechanical behavior of the nonlinear materials for human tactile perception.
An accurate finite element (FE) model of human finger pad is constructed and validated in experiments of finger pad contact with soft and relatively rigid materials. Hyperfoam material parameters of the identified elastomers from the previous section are used for validation of the finger pad model.
A magneto-rheological fluid (MRF) based tactile display is proposed and its magnetic FE model is constructed and validated in Gauss meter measurements.
FE models of the human finger pad and the proposed tactile display are used in a model based control algorithm for the proposed display. FE models of the identified elastomers are used for calculation of control curves for these elastomers.
An experiment is set up for evaluation of the proposed display.
Experiments are performed on biological tissue and soft nonlinear foams.
Comparison between curves of desired and recreated reaction force from subject's finger pad contact with the display showed above 84% accuracy.

As a complementary work, new modeling and controlling approaches are proposed and tested for tactile displays based on linear actuators.
Hertzian model of contact between the human finger pad and actuator cap is derived and curves of material deformation are obtained and improved based on this model. A PID controller is designed for controlling the linear actuators.
Optimization based controller tuning approach is explained in detail and robust stability of the system is also investigated.
Results showed maximum tracking error of 16.6% for the actuator controlled by the PID controller. Human subject tests of recreated softness perception show 100% successful recognition rate for group of materials with high difference in their softness.