Mehrdad Shafiei Dizaji, Ph.D.DT | SHM | NDE | AI | FEA

600+Citations ↗

About me

Mehrdad Shafiei Dizaji, Ph.D.

Structural Engineering · AI · NDE · Digital Twins

I am a civil and structural engineering researcher developing data-driven and physics-informed methods for structural health monitoring, nondestructive evaluation, infrastructure inspection, digital twins, and intelligent sensing.

My work connects structural mechanics, computer vision, deep learning, finite element modeling, and sensor data to create practical tools for infrastructure assessment, damage detection, and structural decision-making.

Ph.D. · University of Virginia Postdoctoral Research · FHWA / UVA Structural Health Monitoring & AI
Explore Projects Selected Publications

Transcripts

Civil engineering coursework from my undergraduate, master's, and doctoral studies, organized by degree and subject.

View PDF Download PDF

Presentations & Research Communication

My Talks

I share my research on structural health monitoring, artificial intelligence, digital twins, and full-field sensing through conference presentations, posters, and seminars.

Selected highlights

2025 · IWSHM · Stanford University · September 9–11

A Visual Question Answering Model to Automate Nondestructive Evaluation Image Analysis

Connecting visual information and natural language to automate the interpretation of NDE images.

2025 · 104th TRB Annual Meeting · Poster B587

Predicting Subsurface Abnormalities Growth using Physics-Informed Neural Networks

Physics-informed learning for subsurface condition assessment, presented in the Soil Properties and Subsurface Investigation Techniques session.

2023 · ATEM–iDICs · Awara, Fukui, Japan · October 9–12

Deep Learning-Based Digital Image Correlation to Extract Visual Measurement

Deep learning methods for extracting structural measurements from images.

2023 · Structures Congress · New Orleans, Louisiana · May 3–6

Digital Bridge Whisperer: A Cyber-Human Collaboration Framework for the Future of Bridge Inspection

A framework connecting computational tools and human expertise for bridge inspection.

2023 · Structures Congress · New Orleans, Louisiana · May 3–6

Leveraging Mixed Reality for Augmented Structural Mechanics Education

Using mixed reality to visualize structural behavior and support engineering education.

2021 · UVA Postdoc Research Symposium · November 18

Adaptive Digital Twinning – An Immersive Visualization Framework for Structural CPS

A 12-minute research talk on immersive digital twins for structural cyber-physical systems. Recognized with a Best Presentation Award.

View all 35 presentations and workshops · 2007–2025
2025

IWSHM, 2025, Society for Experimental Mechanics, “A Visual Question Answering Model to Automate Nondestructive Evaluation Image Analysis” STRUCTURAL HEALTH MONITORING 2025, 15th International Workshop on Structural Health Monitoring (IWSHM), Sep. 9-11, Stanford University, CA, USA, 2025.

2025

TRB, 2025, TRB Annual Meeting, Predicting Subsurface Abnormalities Growth using Physics-Informed Neural Networks. TRBAM-25-02241, Poster/Board: B587, Session 4033: Soil Properties and Subsurface Investigation Techniques, 104th TRB Annual Meeting 2025.

2023

ATEM-iDICs, 2023, Society for Experimental Mechanics, “Deep learning-based Digital Image Correlation to Extract Visual Measurement” Annual International ATEM-iDICs 23 Conference, Awara, Fukui, Japan, Oct. 9-12, 2023.

2023

Structure Congress, 2023, Structural Engineering Institute (SEI), “Digital Bridge Whisperer: A Cyber-Human Collaboration Framework for the Future of Bridge Inspection”, New Orleans, Louisiana, May 3-6 American Society of Civil Engineers (ASCE), 2023.

2023

Structure Congress, Structural Engineering Institute (SEI), “Leveraging Mixed Reality for Augmented Structural Mechanics Education”. New Orleans, Louisiana, May 3-6 American Society of Civil Engineers (ASCE), 2023.

2023

EMI, Engineering Mechanics Institute Conference, “A Generalized digital image correlation Using Attention-based Deep Learning Architecture to Extract Full-Field Subpixel Displacement Measurement from limited data Using Transfer Learning”, Atlanta, Georgia- June 6-9, 2023.

2022

iDICs, 2022, Society for Experimental Mechanics, “Full-Field Image-Based Sensing and Topology Optimization for Detecting and Reconstructing the 3D Geometry of Subsurface Structural Damages for Large Scale Structures”. Annual International DIC Conference, iDICs conference, Boston, Nov. 7th, 2022.

2022

SEM IMAC, 2022, Society for Experimental Mechanics, “Detecting and Reconstructing the 3D Geometry of Subsurface Structural Damages Using Full-Field Image-Based Sensing and Topology Optimization.” In Rotating Machinery, Optical Methods & Scanning LDV Methods, Volume 6 (pp. 175-180). Springer, Cham

2022

SEM IMAC, 2022, Society for Experimental Mechanics, A Hybrid-Attention-LSTM-Based Deep Convolutional Neural Network to Extract Modal Frequencies from Limited Data Using Transfer Learning. In Rotating Machinery, Optical Methods & Scanning LDV Methods, Volume 6 (pp. 167-173). Springer, Cham.

2022

SEM IMAC, Society for Experimental Mechanics, Optimal Kernel Design for the Extraction of Subtle Motions Using Convolutional Neural Network. In Rotating Machinery, Optical Methods & Scanning LDV Methods, Volume 6 (pp. 181-185). Springer, Cham.

2022

SPIE, Machine-learning to see defects: a hybrid attention-ConvLSTM-based convolutional neural network deep learning architecture for structural damage detection. In Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems 2022 (Vol. 12046, pp. 167-173). SPIE.

2021

UVA POSTDOC RESEARCH SYMPOSIUM, 2021, Research Communication, ”Adaptive Digital Twinning – An Immersive Visualization Framework for Structural CPS”, a 12-min talk, Nov 18th Postdoc Research Symposium, University of Virginia, (Oral presentation).

2021

EMI, Engineering Mechanics Institute Conference, 2021, ”Leveraging Full-field Measurement from 3D Digital Image Correlation for Topology Optimization to Detect Internal Abnormalities.”, EMI 2021/PMC 2021 conference, (Virtual presentation).

2021

iDICs, International Digital Image Correlation Society Conference (iDICs), Conference and Workshop, “Fusion of Computer Vision-Based Sensing and Topology Optimization Using 3D-Digital Image Correlation to Identify 3D Geometry of Structural Damage.”, Full paper submitted, Photomechanics IDICS Conference, (Virtual presentation).

2021

IMAC-IXXX, “Multi-Level Damage Detection using Octree Partitioning Algorithm”, Optical Methods & Scanning LDV Methods, Proceedings of the Society for Experimental Mechanics Series, IMAC-XXXX: Next Frontier in Structural Dynamics, (Virtual presentation).

2020

ASNT, American Society for Nondestructive Testing, ”Full-Field Non-destructive Evaluation of Mecklenburg Bridge Using Computer Vision-Based Techniques”, ASNT Research Symposium in Williamsburg, VA, April23-26, 2020, (Oral presentation)

2020

SEI, American Society of Civil Engineers, ”Computer Vision-based Tomography of Structural Elements to Detect Internal Abnormalities: A Preliminary Feasibility Study”, SEI Structure Congress in St. Louis, Missouri, April5-8, 2020, (Oral presentation).

2020

ACI, American Concrete Institute, ”Load Rating Strategies for Bridges with Limited or Missing As-Built Information”, ACI Spring Convention in Chicago, 2020, (Oral presentation).

2019

ACI, American Concrete Institute, ”Subsurface Condition Assessment and Structural Health Monitoring of Concrete Structures Using Digital Image Correlation and Topology Optimization”, ACI Fall Convention in Cincinnati, Ohio, 2019, (Oral presentation in 444 committee meeting, CRC proposal endorsement).

2019

iDICs, International Digital Image Correlation Society Conference (iDICs), Conference and Workshop, ”Mixed Finite Element-DIC Technique to Detect Unseen Abnormalities of a Structure Using an Inverse Approach”, SEM Fall Conference, October 14-17, Portland, Oregon, 2019, (Oral presentation).

2019

SPIE, Smart Structures + Nondestructive Evaluation, ”3D InspectionNet: a deep 3D convolutional neural networks-based approach for 3D crack detection on concrete columns”, Denver, Colorado, USA, 3-7 March 2019, (Oral presentation).

2018

ACI, American Concrete Institute, ”Finite Element Model Updating for Structural Evaluation of Bridges with Limited Information”, ACI Fall Convention in Las Vegas, 2018, (Oral presentation).

2018

EMI, Engineering Mechanics Institute Conference, 2018 ”Vision-Based Characterization and Identification of Structural Systems”, MIT, Boston, Massachusetts, USA, May 29 – June 1, 2018, (Oral presentation).

2018

EMI, Engineering Mechanics Institute Conference, 2018 ”Deep Learning Neuro-Membrane Neural Networks: A Training Algorithm”, MIT, Boston, Massachusetts, USA, May 29 – June 1, 2018, (Oral presentation).

2018

SPIE, Smart Structures + Nondestructive Evaluation, ”Integrating 3D Scanning within a Simulation Framework for Structural Mechanics”, Denver, Colorado, USA, 4-8 March, 2018, (Oral presentation).

2018

SPIE, Smart Structures + Nondestructive Evaluation, ”Reframing Measurement for Structural Health Monitoring: A Full-Field Strategy for Structural Identification”, Denver, Colorado, USA, 4-8 March 2018 (Oral presentation).

2018

ACI, American Concrete Institute, ”Load Rating Strategies for Bridges with Limited or Missing As-Built Information”, ACI Spring Convention in Salt Lake City, 2018, (Oral presentation).

2017

SHMII8, Structural Health Monitoring of Intelligent Infrastructure Conference, ”En-vision-ing a Novel Approach for Structural Health Monitoring – A Model for Full-Field Structural Identification using 3D-Digital Image Correlation”, December 5-8, Brisbane, Australia, 2017 (Oral presentation).

2017

Civil Engineering Department, Flash Talks, Seminar Series, “Leveraging Vision for Structural Identification: A Digital Image Correlation Based Approach” University of Virginia, Virginia, September 25, 2017, (Oral presentation).

2017

Electrical Engineering and Computer Science Department, Image Processing Seminar Series, “Integrating Imaging into a Framework for Civil Infrastructure Assessment” University of Virginia, Virginia, November 16, 2017, (Oral presentation).

2017

Graduate Research Symposium, ”Full-field Structural Identification Using 3D Digital Image Correlation (DIC)”, Civil Engineering Department, University of Virginia, Charlottesville, VA, U.S., 2017 (Poster presentation).

2017

iDICs, International Digital Image Correlation Society Conference (iDICs), Conference and Workshop, ”Finite Element Model Updating Using 3D digital image correlation”, SEM Fall Conference, November 6-9, Barcelona, Spain, 2017 (Oral presentation).

2016

iDICs, International Digital Image Correlation Society Conference (iDICs), ”Leveraging Vision for Structural Identification – A Digital Image Correlation Based Approach,” SEM Fall Conference, November 8-11, Philadelphia, PA, U.S., 2016 (Oral presentation).

2008

Master of Science Thesis, ”Simulating Nonlinear behavior of Concrete Gravity dams crest under earthquakes by Neural Networks”, Sharif University of Technology, (Supervisor: Dr. Abdolreza Joghataie, Examination Committee: Dr. Ali Kaveh, Dr. Estekanchi and Dr. Ghaemian), Tehran, Iran (Oral presentation).

2007

Master of Science Seminar, ”Nonlinear Models for Simulating of Deteriorating Hysteresis Behavior of Concrete”, Sharif University of Technology, (Supervisor: Dr. Borghei), Tehran, Iran (Oral presentation)

EXPERIMENTAL RESEARCH · 3D-DIC My Research in Action Real-Time 3D Digital Image Correlation in the Laboratory My YouTube Channel ↗

A look at my experimental work with three-dimensional digital image correlation (3D-DIC), using optical measurements to study structural deformation and response.

Watch experiment on YouTube ↗

Featured Research

Projects

Research Areas

Core expertise

Deep learning architecture with spatial attention and temporal feature fusion.
Attention-based deep learning

AI for Structural Engineering

Deep learning, engineered neural networks, transfer learning, and intelligent structural identification.

Stereo digital image correlation cameras observing a bridge girder during field testing.
Full-field bridge measurements

Structural Health Monitoring

Full-field sensing, damage detection, structural identification, computer vision, and image-based diagnostics.

Visual question answering framework for defect localization in nondestructive evaluation images.
AI-assisted defect interpretation

Nondestructive Evaluation

AI-assisted NDE interpretation, visual question answering, GPR data analysis, laser scanning, and automated inspection.

Illustrated PINN workflow linking radar observations, neural network training and electromagnetic physics constraints.
Physics-informed GPR prediction

Physics-Informed Neural Networks

Physics-constrained learning for subsurface sensing and prediction, including electromagnetic-wave-informed modeling.

Research overview connecting offline and online digital twins with mixed reality structural visualization.
Digital twins and mixed reality

Digital Twins & Mixed Reality

Adaptive digital twins, immersive structural visualization, AR/VR, and cyber-physical infrastructure systems.

Solid and wireframe finite element representations of four study bridges.
Bridge finite element models

Finite Element & Bridge Engineering

Finite element model updating, bridge load rating, structural optimization, testing, and sensor-data integration.

Selected projects

01

Physics-Informed Neural Networks for GPR-NDE Data

Developed a PINN framework that incorporates electromagnetic wave propagation into neural-network training for subsurface-data prediction.

PINNsGPRNDEFHWA
PHYSICS-INFORMED GPR FORECASTING
Physics-informed forecasting of GPR dataA radar scans a concrete bridge deck. Six historical B-scan frames from BEAST surveys feed CNN spatial features and SFCA, ConvLSTM and TFFA. Data fitting and electromagnetic wave residual jointly train the network, which predicts a future GPR map. All signals and the highlighted evolving anomaly are schematic. 01 / OBSERVE THE SUBSURFACE 02 / LEARN FROM REPEATED SURVEYS GPR pulse → reflections from rebar and anomalies t − 5t − 4t − 3t − 2t − 1t − 0BEAST dataset · six-frame input sequence 03 / EXTRACT SPATIAL AND TEMPORAL PATTERNSCNNInception featuresSFCAChannel attentionConvLSTMSequence memoryTFFAFrame attention 04 / PHYSICS-INFORMED TRAINING Observed-data error + weighted EM-wave residual ∇²E − με ∂²E/∂τ² = 0 (σ = 0) Automatic differentiation · trainable ε and loss weight · Adam 05 / FORECASTFuture B-scan · evolving subsurface patterns Survey time indexes repeated scans; τ denotes electromagnetic wave travel time.

Based on “Physics-Informed Neural Networks to Predict Subsurface Abnormalities Growth in GPR-NDE Data Using the BEAST Dataset.” Conceptual animation; signals and anomaly evolution are illustrative, not experimental results.

02

NDEChat: AI-Based Nondestructive Evaluation

Developed a visual question answering approach for NDE image analysis using deep learning and natural-language generation.

VQAComputer VisionNLPResNet
ChatNDE / FROM NDE IMAGES TO LANGUAGE
01 / INSPECTION IMAGE
Impact Echo · illustrative response mapGPR · illustrative attenuation map

BEAST NDE imagery connects visual inspection patterns with expert-written descriptions and question–answer pairs.

DOMAIN-EXPERT ANNOTATIONSImage + caption → VLP training
Image + question + answer → VQA fine-tuning

02 / Describe the image · Figure-to-Caption

ResNet-502,048 visual featuresLinear projectionVisual → language spaceGPT-2Caption generation
ILLUSTRATIVE CAPTION

03 / Ask about the image · Visual Question Answering

ViT image encoderImage patchesBERT-like encoderQuestion tokensAnswer decoderMultimodal attention
ILLUSTRATIVE QUESTION & ANSWER
04 / EVALUATE · Compare generated text with expert references using BLEUExpert review → refine data and model

Overview of the paper’s two model pathways. The report found stronger results for focused VQA responses than for long-form captions. This animation uses illustrative maps and example text; it is not a live diagnostic model. Map colors depend on the NDE modality and scale.

03

Subsurface Damage Reconstruction with DIC & Topology Optimization

Reconstructing hidden defects in steel coupons from full-field surface measurements. A finite element model updates element stiffness through a regularized SIMP formulation and the method of moving asymptotes (MMA), minimizing the mismatch between computed response and experimental 3D digital image correlation data.

Topology Optimization3D-DICInverse IdentificationSubsurface Damage
3D SPECIMENS / INVERSE TOPOLOGY OPTIMIZATIONINITIAL STIFFNESS FIELD
01 / MEASURE THE SURFACE3D surface · schematic strain contours
02 / UPDATE THE MODEL3D voxel mesh · evolving stiffness
03 / RECONSTRUCT DAMAGE3D backside recesses · estimated geometry
0 / 350
Low stiffness / possible defectIntact

Begin with a uniform stiffness field. Play or drag the iteration slider to reveal the hidden damage.

Illustrative reconstruction inspired by the paper; animated fields and iterations are schematic, not a live FE/MMA solve or recorded experimental results. Specimens have backside recesses, not through-holes. The threshold controls classification of the illustrative stiffness field.

Figure 15: experimental DIC and updated numerical strain and displacement fields, with absolute difference maps, for the two-defect specimen.
Figure 15 · Measured vs. reconstructed responseExperimental DIC fields, updated model fields, and absolute differences for configuration 2. Strain components εxx, εxy, εyy and displacement components U, V, W.Click to expand · Published results
Figure 18: simulated topology optimization for centered-defect configuration 4, showing initial values, target, estimated stiffness, binary detections, true positives, false positives, false negatives and true negatives.
Figure 18 · Reconstructing the centered defectSimulated configuration 4: (a) initial values, (b) target, (c) element stiffness, (d) binary detections, (e) true positives, (f) false positives, (g) false negatives, and (h) true negatives.Click to expand · Published results

Read the study ↗ · Dizaji, Alipour & Harris · Engineering Structures 230 (2021), 111712

04

Computer Vision-Based Structural Health Monitoring

Developed full-field image-based approaches using 3D digital image correlation, topology optimization, and laser scanning for structural identification and damage assessment.

3D-DICSHMTopology Optimization
IMAGE-BASED TOMOGRAPHY / SURFACE RESPONSE TO HIDDEN DAMAGE
Initial intact model
Surface measurements reveal unseen structural damageStereo cameras observe a speckled steel beam under a midspan load. Full-field surface strain and displacement measurements are mapped to a common grid with a finite-element model. A regularized SIMP stiffness representation and MMA topology optimization reduce response discrepancies to reconstruct a hidden damaged region. Geometry and iteration progress are illustrative.01 / MEASURE THE SURFACESteel beam · midspan loading · stereo 3D-DICStereo images → surface strain and displacement04 / RECONSTRUCT HIDDEN DAMAGEUpdated element stiffness reveals the damaged regionInitial intact model3D geometry is inferred from surface responseCutaway and defect shape are schematic3D-DIC + FEMMeasured and computed fieldsCOMMON GRIDInterpolate strain and displacementSIMP + MMAUpdate element stiffnessDAMAGE GEOMETRYInfer location, shape and extentIteratively minimize the discrepancy between full-field responsesConceptual response mismatchBoundary conditions are calibrated using intact specimens before damage identification.Illustrative progression

Based on Dizaji, Harris and Alipour (2022), “Integrating visual sensing and structural identification using 3D-digital image correlation and topology optimization to detect and reconstruct the 3D geometry of structural damage.” The study uses full-field optical measurements and an inverse finite-element problem on steel beams. Animation, defect geometry, and mismatch progression are illustrative, not experimental results.

05

Bridge Load Rating with Incomplete As-Built Information

Developed finite-element-model-driven strategies combining field measurements, bridge testing, and sensor data for load rating.

BridgesFEMLoad RatingVDOT
FEMU-BASED BRIDGE LOAD RATING / INCOMPLETE AS-BUILT INFORMATION
Bridge load rating through finite-element model updatingAn instrumented concrete bridge provides static strain, deflection and rotation data and dynamic acceleration measurements. An initial model with uncertain reinforcement, concrete properties and support restraint is updated by a genetic and gradient optimization scheme linking MATLAB and ABAQUS. Identified properties support capacity calculation and analytical load rating. Visual signals and progress are conceptual.01 / FIELD TESTINGLive-load testing and ambient / impact vibration ● Strain · displacement · rotation■ Accelerometers → EFDD → first three frequencies02 / COMPARE MEASURED AND MODEL RESPONSESMeasured responseFE model responseAbstract response comparison · normalized schematicUnknowns: As · Ec · f′c · support restraintUpdate uncertain properties to improve agreementINITIAL FE MODELObserved geometry + assumed detailsABAQUS ↔ MATLABCompute and compare responseGA + GRADIENTIteratively identify unknown propertiesMinimize response discrepancies03 / CAPACITY AND LOAD RATINGIdentified section / material properties → nominal capacity → analytical LRFR rating factorRF

Based on the supplied paper, “Automated Finite Element Model Updating-based Load Rating of Bridges with Incomplete As-Built Information by Integrating Numerical data with Sensor-Based measurement.” The study evaluated two concrete slab and two T-beam bridges, treating available details as unknown and using plans as a comparison baseline. The hybrid method combined local static and global vibration response. The moving truck, response curves, and updating progress are illustrative; this preview does not compute a bridge rating factor.

07

SMA Recentering Dampers for Seismic Protection

Investigated a passive recentering damper combining NiTiHfPd shape memory alloy bars and friction springs, using material tests and nonlinear seismic analyses of six- and nine-story steel frames.

Shape Memory AlloysSeismic ProtectionPassive ControlRecentering
SMARD / ENERGY DISSIPATION AND RECENTERING
Cyclic shaking
SMA recentering damper for seismic protectionA simplified SMARD cutaway shows NiTiHfPd alloy bars, friction springs and a moving piston. A protected and unprotected steel frame illustrate cyclic response and unloading. Geometry and motion are schematic.01 / SMA + FRICTION SPRING DAMPERHybrid passive device · restoring force and energy dissipationGold: NiTiHfPd barsTeal: friction springsPiston motion compresses the internal SMA–spring assemblies.02 / STEEL FRAME RESPONSEQualitative comparison under shared excitationUnprotected frameWith SMARDShake · dissipate · unload · recenter03 / MATERIAL CHARACTERIZATIONNiTiHfPd tests: −35°C, 5°C and 25°CLoading frequencies: 0.05–1 HzSuperelastic recovery and friction-based energy dissipation04 / BUILDING-LEVEL EVALUATION6- and 9-story steel special moment frames44 ground motions · DBE and MCE hazard levelsNonlinear response-history analysis of drift and accelerationIllustrative motion and geometry; no quantitative seismic-response simulation is performed.

Based on Farzad Shafiei Dizaji and Mehrdad Shafiei Dizaji, “A Novel Smart Memory Alloy Recentering Damper for Passive Protection of Structures Subjected to Seismic Excitations Using High Performance NiTiHfPd Material.” Material and spring tests informed the proposed device and numerical building studies. The paper calls for further experimental validation of the complete device. This animation illustrates the mechanism and residual-drift concept; it does not reproduce quantitative test or building-analysis results.

08

Hysteresis Identification with EPNN

Developed an Extended Preisach Neural Network that combines deteriorating-stop memory neurons and sigmoid neurons to identify nonlinear hysteresis, including asymmetric, non-congruent, and rate-dependent behavior.

HysteresisNeural NetworksSystem IdentificationSmart Materials
EXTENDED PREISACH NEURAL NETWORK / MEMORY + RATE EFFECTS
Extended Preisach Neural Network for hysteresis identificationInput x and its rate of change enter a first hidden layer with normalized deteriorating-stop neurons and a second hidden layer of sigmoid neurons, followed by a linear output. A moving point traces illustrative hysteresis loops. Hybrid training combines genetic search and subgradient optimization with space dilatation.01 / INPUT AND HYSTERETIC MEMORYCurrent input and rate preserve information about the loading path.x(t)ẋ(t)DSDSDSDSσσσσŷ(t)Input layerNDS memorySigmoid layerLinear outputNDS = normalized deteriorating stop · schematic neuron countx = 0.00, ẋ = 0.00 (normalized illustration)02 / HYSTERESIS LOOPIllustrative normalized input–output relationshipxy03 / HYBRID TRAININGGENETIC ALGORITHMSearch for a suitable parameter regionSUBGRADIENT METHODRefine with space dilatationIDENTIFIED RESPONSEMinimize mean squared output errorAdaptive parameters: connection weights and deterioration parameters βStudy examples include smart actuators and material hysteresis across different engineering domains.Memory supports non-congruent loops; sigmoid neurons support smooth asymmetric / non-Masing behavior.Adding ẋ(t) enables rate-dependent identification. The examples here illustrate capabilities, not a trained EPNN.

Farrokh, F. S. Dizaji and M. S. Dizaji (2022), “Hysteresis Identification Using Extended Preisach Neural Network,” Neural Processing Letters. The loop examples and animated signals are schematic, normalized illustrations of the capabilities discussed in the paper. This preview does not run a trained EPNN or reproduce experimental data.

09

Neuro-Skins: Neural Intelligence in a Continuous Membrane

Co-developed a neuro-membrane model coupling neurons with nonlinear dynamic finite elements. Each finite neuro-element converts local displacements into distributed traction, connecting neural activity with mechanical response and plasticity. The study examines neuron type and cell size.

Finite Neuro-ElementsNeural NetworksNonlinear DynamicsPlasticitySmart Materials

Open interactive demonstration in full screen ↗ · Read the paper ↗

After training: from excitation to a learned response

Explore the trained-model demonstration in full screen ↗

10

Neuroplasticity: Learning Through an Adaptive Base Plate

Co-developed a learning approach for Dynamic Plastic Continuous Neural Networks (DPCNNs), combining synaptic-weight optimization with selected material-property updates in a plastic base plate. The medium transmits information through wave propagation and contributes time delay and history dependence. The interactive extension below connects 200 neuro cells one-to-one with a 20 × 10 mesh displayed as a single layer of 3D elements.

NeuroplasticityContinuous Neural NetworksWave PropagationNonlinear FEMSensitivity-Based Learning

Open interactive demonstration in full screen ↗ · Read the paper ↗

11

Hybrid Training of Continuum Neuro-Skin Networks

Proposed a hybrid learning strategy combining particle swarm optimization (PSO) with L-BFGS-B refinement for continuum neuro-skin networks. The technical manuscript demonstrates four regional design parameters; this interactive extension gives every element in a 200-cell mesh its own independent parameter. A single layer of 3D elements includes a 20-cell damaged area with reduced stiffness, and visualizes the full learning process.

Hybrid OptimizationPSOL-BFGS-BNeuro-SkinsElement-Wise Learning

Open all-element learning demonstration in full screen ↗

Applied engineering · Field testing · Bridge assessment

Industry Projects

VDOT / VTRC · University of Virginia · Final report · May 2020

Load Rating Strategies for Bridges With Limited or Missing As-Built Information

Assessing bridge load-carrying capacity when drawings and structural details are incomplete. This collaborative project combined field measurements, finite element model updating, and vibration-based identification to estimate unknown structural properties and support analytical load ratings.

Bridge load ratingFull-scale field testingFinite element model updatingStructural identification
4 study bridgesTwo slab bridges and two T-beam bridges in Virginia
My roleGraduate Research Assistant, UVA · Report co-author
Engineering applicationMeasurement-informed assessment for bridges with limited records
Figure 8 · report p. 28

Four Virginia bridge case studies

War Branch, Smacks Creek, Flat Creek, and Brattons Creek bridges: two reinforced concrete slab bridges and two T-beam bridges.

Figure 9 · report p. 29

Field instrumentation & load testing

War Branch Bridge instrumentation layout for live-load testing, showing strain gauges, string potentiometers, and tiltmeters.

Figure 19 · report p. 37

Finite element models & internal geometry

Solid and wireframe representations of the four study bridges, used to represent structural geometry and investigate unknown details.

Figure 2 · report p. 8

From field measurements to load rating

Static, vibration, and hybrid model-updating approaches connect measured bridge response with calibrated analytical load ratings.

Project findings

Across the four study bridges, ratings from finite element model updating were equal to or up to 17% lower than conventional calculations. The simplified vibration-based method produced differences from −16% to +16%. The study also demonstrated the feasibility of using a limited number of sensors for these methods.

Source: Harris, Ozbulut, Bagheri, Shafiei Dizaji, Ndong & Alipour, Load Rating Strategies for Bridges With Limited or Missing As-Built Information, FHWA/VTRC 20-R27 (May 2020). Figures reproduced from the report; © 2020 Commonwealth of Virginia.

VDOT Richmond District · VTRC · University of Virginia

Evaluation of Mecklenburg Bridge Following Repair Using 3D Digital Image Correlation

A field investigation of a repaired steel girder on the Mecklenburg Bridge, following earlier web buckling and crippling. Stereo-camera 3D digital image correlation captured full-field deformation during the lowering of the beam onto its bearing, supported by handheld laser scanning and conventional strain gauges.

Steel girder repair3D digital image correlationLaser scanningFull-field diagnostics
Mecklenburg County, VirginiaNorthbound I-85 over U.S. Route 1
My roleFirst author · Collaborative UVA / VTRC investigation
Engineering objectiveEvaluate repaired beam-end behavior during load transfer
Figure 1 · p. 1416

Mecklenburg Bridge & evaluated girder

I-85 over U.S. Route 1 in Mecklenburg County, Virginia, showing the bridge and the location of the girder under investigation.

Figure 5c · p. 1420 · selected panel

Stereo-camera measurement in the field

Two 3D-DIC systems monitor the repaired web panel and tail-end region as the girder is lowered onto its bearing.

Figure 8a · p. 1424 · selected panel

Full-field out-of-plane deformation

Measured out-of-plane displacement over the web panel (Region 0) at the end of the lowering sequence, revealing spatial response across the observed surface.

Figure 10 · p. 1426

Finite element model & vertical strain results

The ABAQUS model shows the girder geometry, mesh, loading, and support conditions (a), followed by vertical strain contours near the repaired beam end (b). The analysis supports the order of magnitude of the measured strains; differences from 3D-DIC results reflect model approximations and uncertainty in boundary conditions.

Field investigation · Laser scanning

Handheld laser scanning at the repaired beam end

A handheld laser scanner captures the girder web surface while the connected laptop displays the scan data. The reconstructed geometry complements 3D-DIC measurements in evaluating the repaired region.

Figure 11 · p. 1427

Before-and-after laser-scan assessment

Surface scans collected before lowering and two weeks afterward, with a deviation analysis comparing the reconstructed beam-end geometry.

Project findings

The monitored beam end showed no evidence of renewed local buckling following the lowering sequence. Full-field measurements and scan comparisons supported evaluation of the repaired region. The 3D-DIC results also enabled back-calculation of dead-load sharing, with conclusions limited to the single beam location studied.

View published paper ↗

Source: Dizaji, M. S., Harris, D. K., Kassner, B., & Hill, J. C. (2021). Full-field non-destructive image-based diagnostics of a structure using 3D digital image correlation and laser scanner techniques. Journal of Civil Structural Health Monitoring, 11, 1415–1428. Figures reproduced from the paper; selected panels are identified above. © Springer-Verlag GmbH Germany, part of Springer Nature 2021.

Hampton Roads, Virginia · Full-scale field load testing

Evaluation of Five Spans of the Hampton Roads Bridge Tunnel

A field evaluation of five prestressed concrete girder spans using digital image correlation (DIC) and a cable-suspended reference (CSR) system. The testing addressed selected spans that did not pass load rating, with six VDOT tandem dual dump trucks filled with gravel providing controlled loading.

Work took place approximately one mile offshore, where an open-water setting offered no land-based fixed reference. The team installed instrumentation during the day and tested at night to minimize traffic disruption. Modified GoPro cameras mounted at pier caps observed midspan targets for DIC displacement measurements.

Bridge load testingDigital image correlationOffshore instrumentationDeflection measurement
5 evaluated spansWestbound 31, 32, 34 · Eastbound 56, 57
Two measurement systemsDIC and cable-suspended reference measurements
Field contributionCollaborative bridge load testing and sensor-data analysis
Slide 1

Hampton Roads Bridge Tunnel

The I-64 crossing between Norfolk and Hampton, providing the setting for the five-span field evaluation.

Slide 2

Instrumentation beneath the bridge

The field team accesses the girders over open water to install instrumentation before nighttime load testing.

Slide 3

Camera-based displacement measurement

Modified low-cost GoPro cameras mount at the pier cap and observe targets at midspan for DIC measurements.

Slide 4

Controlled truck loading

Truck Configuration 2 illustrates the transverse loading position used for the example load-test results.

Slide 4

Westbound Span 34 · DIC results

Example Load Case 2 deflection profiles across the bridge cross section for repeated runs measured using DIC.

Slide 4

Westbound Span 31 · CSR results

Example Load Case 2 deflection profiles measured with the cable-suspended reference system on a different tested span.

Example measured response

The presentation documents cross-section deflection profiles under Truck Configuration 2, including DIC results for Westbound Span 34 and CSR results for Westbound Span 31. These plots illustrate measured response across repeated runs on the respective spans.

Source: supplied project presentation, Case Study 1 – Load Testing Hampton Roads Bridge Tunnel, slides 1–4. Photographs, loading diagram, and result plots selected from the original slides.

Varina-Enon Bridge · Field monitoring · 3D digital image correlation

Continued Monitoring of the Varina-Enon Bridge Response to Special Permit Loads

Continued monitoring of bridge response to special permit loads using 3D digital image correlation. The field work connects camera-based measurements with structural response assessment, using patterned surfaces to track deformation over an observed area.

The presentation documents access to the bridge interior, measurement regions at different positions along the span, and the camera and lighting setup used in the monitoring campaign.

Special permit loads3D digital image correlationFull-field measurementBridge response monitoring
Monitoring objectiveAssess bridge response to special permit loads
Measurement approachNon-contact, full-field 3D digital image correlation
Field deploymentPatterned surfaces with camera and lighting instrumentation
Project presentation · Slide 1

Varina-Enon Bridge overview

Project overview showing the bridge and illustrative full-field response maps included in the presentation.

Project presentation · Slide 1

Bridge access & measurement locations

Field photographs identify the access point and patterned measurement surfaces near the beginning and middle of the monitored span.

Project presentation · Slide 1

DIC measurement setup

Camera and lighting equipment face a speckled surface prepared for digital image correlation measurements inside the bridge.

Project presentation · Slide 1

Instrumentation inside the bridge

An interior field photograph documents the working space and instrumentation arrangement for the monitoring campaign.

Source: supplied project presentation, Case Study 3 – Continued Monitoring of the Varina-Enon Bridge, slide 1. All four pictures selected from the original slide.

Research Development & Funding

Funding

Research Proposals

Selected funded and submitted proposals, alongside a PhD qualifying-examination proposal, spanning AI-based damage prediction, digital twins, and infrastructure assessment.

2023Awarded

Data Fusion Using Multi-Model Deep Learning Strategy to Predict the Spatial-temporal Implications of Damage on the Life Cycle of Structural Systems

NRC Research Associateship Program

Role: Principal Investigator (PI)

$73,000 annuallyAnnual fellowship funding
2022Awarded

Engaging the Two-Way Interaction for a Digital Twin for Large Scale Infrastructure

CCI–Smart Cities Program, University of Virginia

Role: Supporting contributor (partial assistance)

$30,000Award amount
2021Awarded

Digital Twin Framework of Cyber-Physical Structural Systems

NSF EAGER

Role: Supporting contributor (partial assistance)

$300,000Award amount
2019Qualifying examination

Deep Learning-Based Automated 3D Crack Detection for Post-Disaster Building Assessment

Proposal in partial fulfillment of the PhD qualifying examination

Role: Principal Investigator (PI)

2019Submitted

Image-based Tomography for Bridge Damage Identification

Submitted to BRAC

Role: Co-Principal Investigator (Co-PI)

Selected Honors · Fellowships & Awards

Fellowships & Awards

NRC · $73,000/yearNational Research Council Research Associateship Fellowship · 2023 (approximately two years)
ASCE / SEI · $3,000Presenter Scholarship ($1,500 · 2023) and Young Professional Scholarship ($1,500 · 2021)
Best Paper · $1,500IMAC Best Paper Nominee ($500 · 2021) and SPIE Best Paper Nominee ($1,000 · 2020); amounts listed in CV
UVA · $14,500Graduate Teaching Internship Fellowship ($9,500 · 2019) and Virginia Engineering Foundation Graduate Fellowship ($5,000 · 2019–20)

Additional selected honors: ASCE Student Fellowship ($1,000), ASNT travel grant ($1,000), ACI travel stipend ($1,000), UVA CEE travel award ($1,500), UVA Postdoc Symposium Best Presentation ($100), UVA Alumni “Hoos Coming Back” fund ($500), and iDICs travel fellowship ($400). Amounts are as recorded in the CV; nominations and travel stipends are not necessarily cash prizes.

Research

Research

Academic & research appointments

2023–2025

Postdoctoral Research Associate

Turner-Fairbank Highway Research Center · Federal Highway Administration

2021–2023

Postdoctoral Research Associate

University of Virginia · Engineering Systems and Environment

2020–2021

Postdoctoral Scholar

University of Massachusetts Lowell · Mechanical Engineering

2015–2020

Research Assistant

University of Virginia · Engineering Systems and Environment

Experimental research · Instrumentation

Lab experience

Explore the instruments and systems featured in my laboratory presentation, from structural testing and optical sensing to data acquisition and immersive visualization.

1 / 22

Digital image correlation

DIC workstation

Workstation and analysis display for digital image correlation.

Source: laboratory presentation · Slide 1

Browse the thumbnails or choose a category. Use the arrow keys while browsing the gallery.

Education

Degrees

University of Virginia university emblem
Ph.D., Civil Engineering
University of Virginia
Sharif University of Technology university emblem
M.Sc., Civil Engineering
Sharif University of Technology
University of Tabriz university emblem
B.Sc., Civil Engineering
University of Tabriz

Technical Skills

Tools

Python MATLAB TensorFlow PyTorch Scikit-learn ANSYS ABAQUS SAP2000 ARTeMIS

Graduate Research

Dissertations

Doctoral Research · University of Virginia

Ph.D. Dissertation

FROM SURFACE MOTION TO HIDDEN GEOMETRYInteractive inverse topology optimization

Four connected element layers with no gaps. Two rectangular recesses on opposite edges follow the reference figure; only the top two layers lose elements.

01 Measure surface motion

Stereo 3D-DIC · full-field observations

02 Update element stiffness

Inverse model · topology refinement

03 Reveal defect geometry

Top two layers optimized · bottom two preserved

Low stiffness / defectIntact materialDrag a model or use ← ↑ ↓ → to rotate

Start with a uniform specimen; play or scrub to reveal the hidden region.

Conceptual animation of the dissertation framework. Illustrative fields and iterations; not a live finite element solver or measured dissertation results.

Ph.D. Dissertation · University of Virginia

A Computer Vision-Based Structural Health Monitoring Framework: Feature-Mining of Damage for Predictive Numerical Simulations

Mehrdad Shafiei Dizaji · Civil Engineering · 2020
Advisor: Devin K. Harris

Can surface motion reveal damage hidden inside a structure? My dissertation combines stereo 3D digital image correlation, finite element model updating, and topology optimization to identify subsurface defects and reconstruct their three-dimensional geometry from full-field surface measurements.

3D-DICInverse MechanicsTopology OptimizationHidden Damage
Read Ph.D. Dissertation ↗

M.Sc. research · Neural networks & structural dynamics

M.Sc. thesis

Designing High-Precision Fast Nonlinear Dam Neuro-Modelers and Comparison with Finite-Element Analysis

Abdolreza Joghataie & Mehrdad Shafiei Dizaji · Journal of Engineering Mechanics, 139(10), 1311–1324 · 2013

How can a neural network learn a dam’s response to an earthquake? This research develops a dam-specific neuro-modeler trained on finite element simulations. A moving window of ground acceleration and feedback from the previous crest response allow a feed-forward network to represent nonlinear, history-dependent behavior. The paper investigates the Koyna, Pine Flat, and Sefid-Rud concrete gravity dams.

Earthquake EngineeringNeural NetworksNonlinear DynamicsConcrete Gravity Dams
INSIDE THE NEURO-MODELERSimulation time 0.00 s · Δt = 0.02 s
GRAVITY DAM · CREST RESPONSE u(t) Schematic geometry · motion exaggerated
0.000u · normalized displacement0.000v · normalized velocity
1 · Input history2 · Hidden layer3 · Next response4 · Feedback

Inspect a hidden neuron

Choose a neuron to inspect its weighted sum and activation.

H1 · linear activation

H1–H2: linear · H3–H8: tanh

Synthetic ground acceleration · normalized
Illustrative network crest displacement · normalized
About this interactive model: the 15–8–2 architecture (including one input bias), two linear and six tanh hidden neurons, feedback arrangement, and 0.02 s time step follow the paper. This browser demo uses chosen illustrative weights and synthetic signals; it does not reproduce the paper’s trained models, earthquake records, finite element results, or dam-specific predictions. All response values are normalized. Read the published paper ↗
01 / Generate response data

Analyze each dam with a nonlinear finite element model using a smeared crack formulation. Collect crest displacement and velocity under training excitation.

02 / Train the neuro-modeler

Adjust connection weights with back-propagation. Use acceleration history and previous response as inputs to learn the next response.

03 / Test on unseen earthquakes

Compare the trained network with finite element analyses under earthquake excitations excluded from training, including near- and far-field events.

Publications

Selected peer-reviewed work

2023
A hybrid-attention-ConvLSTM-based deep learning architecture to extract modal frequencies from limited data using transfer learning.
Mechanical Systems and Signal Processing, 187, 109949. Read publication ↗
2022
Integrating visual sensing and structural identification using 3D-digital image correlation and topology optimization to detect and reconstruct the 3D geometry of structural damage.
Structural Health Monitoring. Read paper ↗
2021
Subsurface damage detection and structural health monitoring using digital image correlation and topology optimization.
Engineering Structures, 230, 111712. Read published paper ↗
2021
Image-Based Tomography of Structures to Detect Internal Abnormalities Using Inverse Approach.
Experimental Techniques. Read publication ↗
2018
Leveraging full-field measurement from 3D digital image correlation for structural identification.
Experimental Mechanics, 58(7), 1049–1066. Read publication ↗

Refer to Google Scholar to access to all papers.

Research Software & Educational Technology

Apps & UI/UX

Mobile mixed-reality structural mechanics applications developed collaboratively at the University of Virginia, as described in my ASEE conference publications. These animated demonstrations illustrate the published concepts rather than reproducing original app screenshots.

UI/UX · Augmented Reality

Structural analysis brought into the real world

Explore a bridge model through a mobile augmented reality interface that brings structural analysis results into the surrounding physical space.

From numerical results to spatial understanding

This recording demonstrates my augmented reality structural visualization work. The mobile interface offers element, member, and system analysis views. In the bridge example, a three-dimensional model appears in the camera view, allowing the user to inspect it from different viewpoints.

The demonstration shows a finite element mesh, load and support indicators, and color-coded structural response contours. On-screen controls switch between response components, including strain and displacement, while node information connects the visualization to locations in the model. This approach helps make structural behavior easier to explore and explain in engineering education and design discussions.

Augmented RealityFinite Element VisualizationStructural MechanicsEngineering Education
Open Demo Video
IMMERSIVE ENGINEERING · INTERACTIVE LEARNINGEngineering meets augmented reality.

Explore research-driven mobile experiences that bring structural behavior, computational models, and interactive visualization into the physical world.

◈ 3D visualization◈ Structural mechanics◈ AR education
STRUCT-AR · CONCEPT SIMULATION AR TRACKING ACTIVE
↓APPLIED LOAD
STRESS CONTOUR
LOWHIGH
Illustrative simulation · not measured results
FINITE ELEMENT MESH · LOAD RESPONSE · AUGMENTED REALITY
ASEE 2022AR · Structural Design

AR Visualization for Structural Design

An educational augmented-reality approach for visualizing structural behavior and exploring design concepts through interactive three-dimensional models.

STRUCTURAL DESIGN PREVIEW3D VISUALIZATION
3D AR DESIGN STUDIO ● MODEL ACTIVE
3D STRUCTURAL FRAMEAPPLIED LOADS
Original geometry Simulated response

Related publication: Yehia, Chao, Lyons, Dizaji & Harris (2022), “A Case Study on Leveraging Augmented Reality for Visualization in Structural Design,” ASEE Annual Conference & Exposition.

The STRUCT-AR store links above open the released application. The rotating 3D website animations are illustrative browser-rendered demonstrations, not live instances of the mobile app.

Academic Teaching

Teaching Experience & Student Mentorship

My teaching connects structural engineering fundamentals with numerical modeling, experimental mechanics, computer vision, and hands-on design. I have supported graduate and undergraduate instruction, laboratory education, and interdisciplinary research mentorship.

FALL 2019 · UNIVERSITY OF VIRGINIA

Graduate Teaching Internship Fellow

CE 5340 · Advanced Topics in Structural Engineering

Co-instructor with Professor Jose Gomez. Developed advanced course modules, organized the syllabus and course project, and delivered lectures on numerical analysis and finite element concepts. Graduate-level instruction included in-person and web-based delivery.

SPRING 2019 & 2020 · UNIVERSITY OF VIRGINIA

Teaching Fellow

CE 3330 · Introduction to Design of Structural Systems

Co-developed the syllabus and course project, organized instructional modules on computer vision for structural analysis, and supported student project planning and progress.

SPRING 2019 & 2020 · UNIVERSITY OF VIRGINIA

Laboratory Instructor

CE 3330 · Structural Systems Laboratory

Coordinated experiments and project activities using MTS testing equipment, 3D printing, digital cameras, and digital image correlation. Guided problem-solving sessions, exercises, and quizzes.

FALL 2008 · SHARIF UNIVERSITY OF TECHNOLOGY

Teaching Assistant

Introduction to Structural Optimization

Contributed to syllabus and course-project development, taught MATLAB-based computational modules, and guided problem-solving sessions, quizzes, and exercises.

Course Syllabus · Fall 2019 · University of Virginia

CE 5340 · Advanced Topics in Structural Engineering

Matrix structural analysis through the direct stiffness method, virtual work, and computational implementation. Instructors: Jose P. Gomez, P.E., Ph.D., and Mehrdad S. Dizaji.

Matrix analysisDirect stiffness methodMATLAB & MathcadFinite elements
Explore syllabus highlights

Prerequisite: CE 3300 Structural Mechanics or equivalent.

Course focus: Relate external loads to member forces, deformations, and global displacements using equilibrium, force–deformation relationships, and compatibility. Formulate element stiffness matrices, assemble the global stiffness matrix, and solve structural systems with MATLAB, Mathcad, and Excel.

Topics: Statics and matrix algebra; virtual forces and displacements; force and displacement methods; planar trusses, beams, and frames; symmetry, support settlements, releases, and temperature effects. Time permitting: three-dimensional systems, nonlinear behavior, and an introduction to finite element analysis using ABAQUS.

Assessment: Homework 10%; Test 1 20%; Test 2 20%; Computer Project I 25%; Computer Project II 25%.

Recommended texts: Matrix Analysis of Structures (Kassimali); Matrix Structural Analysis (McGuire et al.); Matrix Analysis of Structures (Sennett).

Download CE 5340 syllabus (.doc)

Research Mentorship

Mentored and collaborated with undergraduate, graduate, and visiting researchers at UVA on adaptive digital twins, mixed-reality structural mechanics education, computer vision, and structural identification.

Contact

Research collaboration, academic opportunities & engineering innovation

I am interested in research and professional opportunities involving structural engineering, AI, digital twins, structural health monitoring, computer vision, and nondestructive evaluation.

Facebook X Email Google Scholar LinkedIn GitHub YouTube ResearchGate Instagram

Beyond Engineering

Fitness & Strength Training

Outside engineering, I enjoy strength training and the discipline of working toward personal goals.

Incline Dumbbell Press

300 lb22 sec

A moment from my upper-body strength training.

Barbell Squat

405 lb14 sec

Building lower-body strength through consistent training.

Dips

180 lb34 sec

Bodyweight training as part of my gym routine.

Barbell Bench Press

405 lb6 sec

Working on upper-body strength with a spotter.