Research Background
Seyed Hamid Reza Sanei
I believe my research expertise and interests (as described below) will be an asset to your group. I am confident that I can
strengthen the existing research and bring new ideas with my unique background and vision.
A. Research Interests and Expertise
A.1. Optical and Electron Microscopy
To have a quantitative characterization of microstructures, high quality images using optical and electron microscopy are
acquired. Electron microscopy provides high resolution images due to short wavelength of electrons; however, it requires
intensive polishing. Alternatively, optical microscopy using epi fluorescent filter can be used. In epi-fluorescent
microscopy, an ultraviolet light passes through an excitation filter that allows only a limited range of the light spectrum to
strike the specimen. The light from the specimen forms an image that is passed through a barrier filter that blocks all light
but that which was emitted (fluoresced) from the specimen. This approach provides a robust technique for the detection of
microstructural features while requiring minimal polishing effort. During my Ph.D. I worked extensively with optical and
electron microscopy.
A.2. Image Processing
While human brain can distinguish different features of an image instantly, image segmentation is one of the most
difficult tasks in image processing. Computers require specific criteria to distinguish one feature from another. While
intensity segmentation works well for a highly contrasted image with two constituents, its application is limited for color
images with multiple constituents. Other advanced image segmentation techniques such as morphological segmentation,
edge detection, region growth and object recognition have proved to be very powerful for different applications. I have
extensive experience in image acquisition, processing and object recognition. Two of my papers are on the basis of
microstructure characterization using image analysis of actual microstructures.
A.3. Nanoindentation Testing and Analysis
Nano indentation is a robust technique to determine bulk as well as in situ properties of materials. My earlier research
focused on nano indentation testing and accurate interpretation of its results on soft materials. I served as a
nanoindentation workshop instructor responsible for training graduate students and faculty on how to use a G200
nanoindenter. I published two journal papers and three conference proceedings on nanoindentation of polymers.
A.4. Stochastic Finite Element Method
Deterministic approaches are unable to predict scatter in properties as they do not permit any microstructural variability. To
capture property scatter observed experimentally, stochastic approaches must be conducted. In stochastic finite element,
properties are randomly attributed to each element from a known distribution. I developed a python script to randomly
assign different epoxy properties from a distribution obtained by nanoindentation to elements. The study resulted in a
conference paper presented at AIAA conference. I am working on an ongoing research to use stochastic FEA for developing
stochastic failure envelopes.
A.5. Extended Finite Element Method
To determine initiation and evolution of failure, extended finite element (XFEM) can be used. Unlike standard FEA in
which the mesh has to be modified as the crack propagates to avoid high stress gradient, the crack is initiated and progressed
independently of the mesh. Using XFEM, I modeled crack initiation and evolution in synthetically generated composite
microstructure.
A.6. Multiscale Modeling
Multiscale approaches have received great attention in the past decade due to their accuracy and high computational
efficiency. During my PhD I have been working on multiscale analysis of composite materials to link microstructural
variability to macroscopic properties. I recently proposed an uncorrelated volume element (UVE) as an intermediate length
scale for multiscale analysis. I introduced UVE as an extension of the stochastic volume element to replace the commonly
used representative volume elements (hexagonal, square and diagonal configurations). The UVE defines an intermediate
intrinsic length scale for homogenization of properties such that adjacent microstructures are independent of each other.
The UVE can be considered as an intermediate length scale between constituent and lamina facilitating the stochastic
multiscale modeling. The concept of UVE permits variability from one realization to another which enables prediction of
stochastic response.
A.7. Artificial Neural Network
While linking physics of the problem to macroscopic properties is the step in the right direction, it is a very challenging
task and the progress has been incremental over the past decade. The alternative approach is to learn from examples the
same way human neurons understand everyday complex concepts. Artificial neural network (ANN) can be used to find
the pattern between microstructural features and macroscopic properties by simply analyzing a set of examples. Such
examples are used for training the network so it can learn the pattern. Experimental and simulation results are typically
used to train the network. While ANN has been employed in several disciplines, its application in composite design has
not been explored. ANN in conjunction with physics based approaches will promise more reliable structural analysis. I am
currently working on a paper to develop stochastic failure envelopes using ANN.
A.8. Wavelet Transformation for Scaling
One important challenge in failure analysis is the specimen size effect. In the design and analysis of structures, the results
from testing small coupons are used in design of actual structures. According to the homogeneity assumption, the volume
average strength is not dependent on specimen size and the results are applicable to other sizes. However, experimental
findings have shown that the strength is sample size dependent, such that larger samples have lower strength. This
phenomenon is the so-called size effect. As defects are distributed in the sample, therefore, larger samples are more likely
to have more defects. The obvious solution to the problem is to conduct experiments on life size samples which would be
infeasible given time and cost. The alternative is to scale the predicted strength based on the results of small coupons. The
two available analytical scaling methods are Weibull statistical strength theory and the fracture mechanics approach. They
have several intrinsic shortcomings limiting their applications. They are applicable for tensile loading and have poor
prediction in compression and shear loadings. Furthermore, their strength prediction is based on the assumption that the
mode of failure remains the same throughout scaling which is an unrealistic assumption in actual multiaxial applications.
A novel approach is based on employing the concept of image compression using wavelet transform technique. A
microstructure is treated as an image with distributed defects and upon scaling, microstructure is split into four subimages, approximation, horizontal, vertical and diagonal details. This unique approach is a powerful technique to capture
size effect observed experimentally.
Summary:
Working in three different research groups during my Master’s and Ph.D. was a great opportunity to experience different
research environments. Such experience give me invaluable insights in adapting to different research group dynamics. My
diverse background provide me with unique vision to tackle mechanical engineering problems differently. During my
PhD, I have been coauthor of several proposals for NSF, the US Navy, DOE and NASA.