M.Sc Student | Tesler Alexander |
---|---|

Subject | Analysis of Impedance Spectroscopy Data Using Genetic Programming Methods |

Department | Department of Chemical Engineering |

Supervisors | Professor Daniel Lewin |

Professor Yoed Tsur | |

Full Thesis text |

The need to study and characterize interactions of solid-solid,
solid-liquid interfaces and physical and electrical properties of materials
results in the wide application of electrochemical methods in modern chemistry
and material science. The key attribute is the complex electric impedance -
measured as a function of frequency of small sinusoidal potential
perturbations, and referred to as *ImpedanceSpectroscopy *(IS). While the
collection of impedance data is relatively simple, their accurate analysis and
interpretation, expressed as a predictive model, is not an easy task. To
estimate the model parameters, it is necessary to solve the inverse problem
(Fredholm integral equation with complex kernel function), proceeding from
discrete measured points to a continuous model. Unfortunately, the problem is
ambiguous and ill-posed and cannot be solved directly because of the presence
of noise in the measured signal. In fact, the data can be fitted to an infinite
number of models. The goal of the research is to quantify the noise of measured
IS signals and to find the most compact models, adopted from Baltianski and
Tsur, that fit the data well enough using evolutionary programming methods. Two
complementary methods have been applied: Genetic Algorithm (GA) and Genetic
Programming (GP). The former method facilitates robust parameter estimation for
arbitrary nonlinear models, while the latter uses the adaptive GP approach of
Grosman and Lewin to create relatively non-complex models through the
penalization of unnecessarily complex models. As demonstrated, this approach
enables the most appropriate reduced-order models to be generated to match
Impedance Spectroscopy data.