Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List

Full text release has been delayed at the author's request until May 05, 2026

ETD Abstract Container

Abstract Header

Novel Forward-Inverse Estimation and Hypothesis Testing Methods to Support Pipeline and Brain Image Analyses.

Yazbeck, Maha
http://orcid.org/0000-0003-0657-3099
http://rave.ohiolink.edu/etdc/view?acc_num=osu1704410077295309

Abstract Details

2024, Doctor of Philosophy, Ohio State University, Industrial and Systems Engineering.
This dissertation addresses two applied problems relating to images. The first relates to images of pipeline corrosion and the second relates to images of the human brain and individuals with Attention-Deficit/Hyperactivity Disorder (ADHD). The corrosion of oil and gas pipelines is important because there are thousands of leaks every year costing billions of dollars for cleanups. ADHD is important because a substantial fraction of the world population has the disorder causing significant suffering and hundreds of billions of dollars of losses to the world economy. To address both image analysis problems, novel statistical and operations research techniques are proposed which have potentially wide applicability. Relating to pipeline corrosion, an established simulation method is called the “voxel” method which permits predictions about how images and pipelines or other media will change as corrosion evolves. In most realistic cases, we find that the parameter values or “inputs” (Xs) needed to run the simulation are unknown. We only have the images which are essentially outputs (Ys) which can be generated by real world experiments or simulations. The phenomenon of having incomplete inputs for simulation is common in many engineering and science situations and a critical challenge for both people and artificial intelligence. We and others have called this important subject, “empirical forward-inverse estimation” since we can gather data (empirically) in the forward manner progressing from assumed inputs (Xs) to measured outputs (Ys) and then generate inverse predictions from Ys to Xs. With (hopefully) accurately estimated X values, the experimental setup or simulation can then predict the future corrosion evolution and whether repair in critically needed. Relating to forward-inverse analyses, 24 variants of an established two stage method or framework are studied in relation to enhanced inverse prediction accuracy for two test cases including pipeline corrosion modeling and the Susceptible-Infected-Removed (SIR) model. Novel proposed sub-methods for use in the two-stage framework include approximate space filling designs, artificial neural networks, and a conservative “nearest point-based” inverse estimation method. Several proposed combinations offer promising accuracy benefits including the use of Latin hypercube sampling and our nearest point-based inverse prediction. To address the computational challenges, efficient optimization methods are proposed to support the large-scale deployment of the forward inverse methods needed for cases with thousands of simulations runs and 9 Xs and 9 Ys. Dramatic improvements in accuracy and computational efficiency are demonstrated using both SIR and pipeline cases using the proposed methods. Additionally, metrics for general “empirical invertibility” and so-called “focused invertibility” characterize in percentage units the typical accuracy that could be expected for random and more relevant point inversions respectively. The focused invertibility is based on the root mean squared error value on test set points selected to be nearby points of interest and scaled to account for numbers of inputs and ranges. We argue that focused invertibility values of less than 10% imply that the outputs are likely sufficient to predict inputs accurately, so long as the proposed empirical model-building methods are applied. To further support pipeline modeling, novel extensions of existing voxel 3D grid stochastic methods are proposed to improve the ability to match outputs with existing images of pipelines. Such types of simulation are at mesoscopic scale and are considered gray box models, allowing to capture atomic sources of corrosion, and to set point attributes and conditions of probability according to physics but also functioning without governing partial differential equations like the case of white box models. Also relating to pipeline image data, a generative artificial intelligence (AI) method is presented to reconstruct images of corrosion and conditioning their generation on specific attributes. Image and its characteristics are entangled in a latent representation and the decoder is conditioned on an attribute vector before outputting the image. Relating to ADHD, Electroencephalogram (EEG) data is analyzed for youths known to have the condition and presented in terms of deviations in expression from a large normal population. The deviations are given in terms of the standard normal or “z-score” values of paired expressions. These expressions are sorted to identify important candidates for reductions in therapies. Yet, the significance of the sorted values, which are essentially “order statistics” cannot properly be evaluated in terms of individual z-score values. Therefore, a hypothesis testing method is proposed for the sorted expressions based on Monte Carlo estimation of the order statistic quantiles. The method is illustrated in the context of real data and implications for the design of therapies are presented.
Theodore T. Allen (Advisor)
William (Bill) Notz (Committee Member)
Samantha Krening (Committee Member)
Marat Khafizov (Committee Member)
138 p.
Engineering; Industrial Engineering; Materials Science; Statistics
Statistical Quality Improvement of Simulations; Forward-Inverse Functions; Parameter Estimation; Semi-physics Models of Corrosion; Invertibility Index; Monte Carlo Simulations; Parametric Bootstrap; Hypothesis Testing; Brain Connectivity Analysis; Image Analysis; Image Generation; Variational Auto-encoders.

Recommended Citations

Citations

  • Yazbeck, M. (2024). Novel Forward-Inverse Estimation and Hypothesis Testing Methods to Support Pipeline and Brain Image Analyses. [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1704410077295309

    APA Style (7th edition)

  • Yazbeck, Maha. Novel Forward-Inverse Estimation and Hypothesis Testing Methods to Support Pipeline and Brain Image Analyses. 2024. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1704410077295309.

    MLA Style (8th edition)

  • Yazbeck, Maha. "Novel Forward-Inverse Estimation and Hypothesis Testing Methods to Support Pipeline and Brain Image Analyses." Doctoral dissertation, Ohio State University, 2024. http://rave.ohiolink.edu/etdc/view?acc_num=osu1704410077295309

    Chicago Manual of Style (17th edition)

osu1704410077295309
© 2024, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.