BEGIN:VCALENDAR VERSION:2.0 X-WR-CALNAME:EventsCalendar PRODID:-//hacksw/handcal//NONSGML v1.0//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York LAST-MODIFIED:20240422T053451Z TZURL:https://www.tzurl.org/zoneinfo-outlook/America/New_York X-LIC-LOCATION:America/New_York BEGIN:DAYLIGHT TZNAME:EDT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZNAME:EST TZOFFSETFROM:-0400 TZOFFSETTO:-0500 DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT CATEGORIES:College of Engineering,Thesis/Dissertations DESCRIPTION:Title: "Predicting Microbiologically Influenced Corrosion Sever ity from Electrochemical Impedance Spectroscopy Using Interpretable Machin e Learning" by Raksha Mohan Thesis Advisor: Maricris Mayes, Associate Prof essor, Chemistry & Biochemistry Thesis Committee:Firas Khatib, Associate P rofessor, Computer & Information ScienceDonghui Yan, Associate Professor, Mathematics Abstract: Microbiologically influenced corrosion (MIC) account s for an estimated 20-30% of global corrosion losses, yet reliable quantit ative prediction of MIC severity remains unsolved. Electrochemical measure ments combined with interpretable machine learning provide a promising fra mework for addressing this challenge. MIC depends on coupled microbial, bi ofilm, and interfacial electrochemical processes, and charge-transfer resi stance (Rct) is a useful proxy for corrosion severity, as expressed in log ₁₀(Rct) from electrochemical impedance spectroscopy (EIS). However, ph ysically interpretable predictive models for MIC remain limited. Here, we show that machine learning regressors can predict log₁₀(Rct) in MIC sy stems involving Pseudomonas and Vibrio across varying environmental condit ions. A dataset of 116 EIS and 83 potentiodynamic polarization observation s was compiled from ten peer-reviewed sources. Random forest and Gradient Boosting Machine (GBM) regressors were compared using stratified five-fold cross-validation, and Shapley additive explanations (SHAP) were used to i nterpret model behavior in physically meaningful terms. GBM outperformed R andom Forest, achieving a higher cross-validated R². SHAP analysis identi fied double-layer CPE admittance as the dominant predictor of corrosion se verity, consistent with its role as a reporter of biofilm-induced interfac ial disorder, while genomic species descriptors contributed modest but int erpretable signals consistent with known difference in metabolic versatili ty between the two organisms. This work establishes a reproducible and int erpretable baseline for quantitative MIC prediction and, to our knowledge, provides the first application of SHAP analysis to EIS-derived features i n MIC while demonstrating the value of integrating genomic descriptors wit h electrochemical features for corrosion modeling. Zoom Meeting ID: 922 35 04 5299Passcode: 941562\nEvent page: /events/cms/dat a-science-ms-thesis-defense-by-raksha-mohan.php\nEvent link: https://umass d.zoom.us/j/92235045299?pwd=gOpd6QBNGaNhrJaTwjQXCEkGjk8iSb.1 X-ALT-DESC;FMTTYPE=text/html:
Title: "Predicting Microbiologi cally Influenced Corrosion Severity from Electrochemical Impedance Spectro scopy Using Interpretable Machine Learning"
\nby Raksha Mohan
\n< p>Thesis Advisor: Maricris Mayes\, Associate Professor\, Chemistry & Bioch emistry\nThesis Committee:
Firas Khatib\, Associate Professor\
, Computer & Information Science
Donghui Yan\, Associate Professor\,
Mathematics
Abstract:
\nMicrobiologically influenced corrosi on (MIC) accounts for an estimated 20-30% of global corrosion losses\, yet reliable quantitative prediction of MIC severity remains unsolved. Electr ochemical measurements combined with interpretable machine learning provid e a promising framework for addressing this challenge. MIC depends on coup led microbial\, biofilm\, and interfacial electrochemical processes\, and charge-transfer resistance (Rct) is a useful proxy for corrosion severity\ , as expressed in log₁₀(Rct) from electrochemical impedance spectrosco py (EIS). However\, physically interpretable predictive models for MIC rem ain limited. Here\, we show that machine learning regressors can predict l og₁₀(Rct) in MIC systems involving Pseudomonas and Vibrio across varyi ng environmental conditions.
\nA dataset of 116 EIS and 83 potentiod ynamic polarization observations was compiled from ten peer-reviewed sourc es. Random forest and Gradient Boosting Machine (GBM) regressors were comp ared using stratified five-fold cross-validation\, and Shapley additive ex planations (SHAP) were used to interpret model behavior in physically mean ingful terms. GBM outperformed Random Forest\, achieving a higher cross-va lidated R². SHAP analysis identified double-layer CPE admittance as the d ominant predictor of corrosion severity\, consistent with its role as a re porter of biofilm-induced interfacial disorder\, while genomic species des criptors contributed modest but interpretable signals consistent with know n difference in metabolic versatility between the two organisms.
\nT his work establishes a reproducible and interpretable baseline for quantit ative MIC prediction and\, to our knowledge\, provides the first applicati on of SHAP analysis to EIS-derived features in MIC while demonstrating the value of integrating genomic descriptors with electrochemical features fo r corrosion modeling.
\nZoom Meeting ID: 922 3504 5299
Passcode
: 941562
Event page: /eve
nts/cms/data-science-ms-thesis-defense-by-raksha-mohan.php
Event li
nk: