<?xml version="1.0" encoding="UTF-8"?><?xml-stylesheet type="text/xsl" href="static/style.xsl"?><OAI-PMH xmlns="http://www.openarchives.org/OAI/2.0/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.openarchives.org/OAI/2.0/ http://www.openarchives.org/OAI/2.0/OAI-PMH.xsd"><responseDate>2026-11-21T05:11:25Z</responseDate><request verb="GetRecord" identifier="oai:macsphere.mcmaster.ca:11375/33314" metadataPrefix="dim">https://prod-ms-be.lib.mcmaster.ca/server/oai/request</request><GetRecord><record><header><identifier>oai:macsphere.mcmaster.ca:11375/33314</identifier><datestamp>2026-07-01T07:06:41Z</datestamp><setSpec>com_11375_271</setSpec><setSpec>col_11375_272</setSpec></header><metadata><dim:dim xmlns:dim="http://www.dspace.org/xmlns/dspace/dim" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:doc="http://www.lyncode.com/xoai" xsi:schemaLocation="http://www.dspace.org/xmlns/dspace/dim http://www.dspace.org/schema/dim.xsd">
   <dim:field mdschema="dc" element="contributor" qualifier="advisor">Swartz, Christopher L.E.</dim:field>
   <dim:field mdschema="dc" element="contributor" qualifier="author">Aarabi Dahej, Sepehr</dim:field>
   <dim:field mdschema="dc" element="contributor" qualifier="department">Chemical Engineering</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="accessioned">2026-06-30T19:49:52Z</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="issued">2026</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="uri">https://hdl.handle.net/11375/33314</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="uri">https://doi.org/10.71548/176</dim:field>
   <dim:field mdschema="dc" element="description" qualifier="abstract">Partial differential equation (PDE) models are foundational tools for the design, optimization, and control of many chemical processes, yet their high computational cost may pose a critical barrier in scenarios requiring rapid or repeated evaluations. This burden is further compounded in parameter estimation, where identifying unknown model parameters demands solving large-scale systems within an optimization problem at costs that scale poorly with spatial resolution. This work addresses both bottlenecks through two complementary contributions: the development of Physics-Informed Neural Network (PINN)-based surrogate modeling frameworks, and their application to parameter estimation in PDE-constrained systems. 

Two surrogate models are developed, each tailored to distinct applications. The first, the Simulation-PINN, is conditioned on static boundary conditions and PDE parameters, allowing it to generate the full spatiotemporal solution in a single forward neural network pass and making it particularly well-suited for offline applications. The second, the PCA-PINN, targets real-time applications by encoding the current system states using Principal Component Analysis (PCA) and predicting the solution at the next time step, allowing the system states to be propagated forward in time while accommodating changes in inputs at each time step under a zero-order hold. Both frameworks are validated on isothermal and non-isothermal axial-dispersion plug-flow reactor (PFR) case studies, where they achieve computational speedups exceeding three orders of magnitude relative to baseline numerical solvers while maintaining high accuracy across concentration and temperature fields. Importantly, both models also generalize robustly to operating conditions outside their training domains. 

Building on these frameworks, parameter estimation for an isothermal PFR system with two unknown parameters is then investigated using three strategies: classical simultaneous optimization using finite differences, sequential optimization leveraging the Simulation-PINN surrogate, and a simultaneous forward-inverse PINN. Among these, the Simulation-PINN surrogate proves particularly effective, enabling rapid estimation with cheap gradient evaluations and substantially lower computational cost.

Taken together, these results show that PINN-based models can provide practical and computationally efficient alternatives to repeated numerical solution of first-principles PDE models for simulation, online prediction, and parameter estimation in chemical engineering. They also indicate that input design, architecture selection, and the way PINN surrogates are embedded in simulation and estimation workflows are important in determining whether such methods remain accurate and efficient in distributed process applications.</dim:field>
   <dim:field mdschema="dc" element="description" qualifier="sponsorship">Support from the Natural Sciences and Engineering Research Council of Canada (NSERC) and the McMaster Advanced Control Consortium (MACC) is gratefully acknowledged.</dim:field>
   <dim:field mdschema="dc" element="description" qualifier="degreetype">Thesis</dim:field>
   <dim:field mdschema="dc" element="description" qualifier="degree">Master of Applied Science (MASc)</dim:field>
   <dim:field mdschema="dc" element="language" qualifier="iso">en</dim:field>
   <dim:field mdschema="dc" element="subject">PINNs</dim:field>
   <dim:field mdschema="dc" element="subject">Surrogate Modeling</dim:field>
   <dim:field mdschema="dc" element="subject">Partial Differential Equations</dim:field>
   <dim:field mdschema="dc" element="subject">Parameter Estimation</dim:field>
   <dim:field mdschema="dc" element="subject">Physics-Informed Neural Networks</dim:field>
   <dim:field mdschema="dc" element="subject">Process Simulation</dim:field>
   <dim:field mdschema="dc" element="title">Physics-Informed Neural Networks (PINNs) for Chemical Process Simulation, Real-Time Operation, and Parameter Estimation</dim:field>
   <dim:field mdschema="dc" element="type" lang="en">Thesis</dim:field>
   <dim:field mdschema="others" element="access-status">embargo</dim:field>
</dim:dim></metadata></record></GetRecord></OAI-PMH>