Pouya Tavousi

Dr. Pouya Tavousi


Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 February 2021 in Journal of Nondestructive Evaluation
Verified
Journal Article
Open access
PDF file
MKMatthew Konnik
Matthew Konnik
BABahar Ahmadi
Bahar Ahmadi
NMNicholas May
Nicholas May
JFJoseph Favata
Joseph Favata
ZSZahra Shahbazi
Zahra Shahbazi
PTPouya Tavousi
Pouya Tavousi
X-ray computed tomography (CT) is a powerful technique for non-destructive volumetric inspection of objects and is widely used for studying internal structures of a large variety of sample types. The raw data obtained through an X-ray CT practice is a gray-scale 3D array of voxels. This data must undergo a geometric feature extraction process before it can be used for interpretation purposes. Such feature extraction process is conventionally done manually, but with the ever-increasing trend of image data sizes and the interest in identifying more miniature features, automated feature extraction methods are sought. Given the fact that conventional computer-vision-based methods, which attempt to segment images into partitions using techniques such as thresholding, are often only useful for aiding the manual feature extraction process, machine-learning based algorithms are becoming popular to develop fully automated feature extraction processes. Nevertheless, the machine-learning algorithms require a huge pool of labeled data for proper training, which is often unavailable. We propose to address this shortage, through a data synthesis procedure. We will do so by fabricating miniature features, with known geometry, position and orientation on thin silicon wafer layers using a femtosecond laser machining system, followed by stacking these layers to construct a 3D object with internal features, and finally obtaining the X-ray CT image of the resulting 3D object. Given that the exact geometry, position and orientation of the fabricated features are known, the X-ray CT image is inherently labeled and is ready to be used for training the machine learning algorithms for automated feature extraction. Through several examples, we will showcase: (1) the capability of synthesizing features of arbitrary geometries and their corresponding labeled images; and (2) use of the synthesized data for training machine-learning based shape classifiers and features parameter extractors.

3 May 2023 in PLOS ONE
Verified
Journal Article
Open access
PDF file
NMNicholas May
Nicholas May
HCHongbin Choi
Hongbin Choi
APAdrian Phoulady
Adrian Phoulady
SAShahram Amini
Shahram Amini
PTPouya Tavousi
Pouya Tavousi
SSSina Shahbazmohamadi
Sina Shahbazmohamadi
Cross sectioning is a critical sample preparation technique used in a wide range of applications, that enables investigation of buried layers and subsurface features or defects. State-of-the-art cross-sectioning methods, each have their own pros and cons, but generally suffer from a tradeoff between throughput and accuracy. Mechanical methods are fast but lack accuracy. On the other hand, ion-based methods, such as focused ion beam (FIB), offer high resolutions but are slow. Lasers, which can potentially improve this tradeoff, face multiple challenges that include creation of heat affected zones (HAZs), undesirably large spot size as well as material redeposition. In this work, we utilized, for the first time, a femtosecond pulsed laser, which has been shown to cause minimal to zero HAZ, for rapid creation of large cross sections that are comparable with FIB cross sections in quality. The laser was integrated with a targeted CO2 gas delivery system for redeposition control and beam tail curtailing, and a hard mask for top surface protection and further shrinkage of the effective spot size. The performance of the proposed system is showcased through real world examples that compare the throughput and quality resulted from the laser and FIB cross sectioning techniques.

19 July 2022 in Scientific Reports
Verified
Journal Article
Open access
PDF file
APAdrian Phoulady
Adrian Phoulady
NMNicholas May
Nicholas May
HCHongbin Choi
Hongbin Choi
YSYara Suleiman
Yara Suleiman
SSSina Shahbazmohamadi
Sina Shahbazmohamadi
PTPouya Tavousi
Pouya Tavousi
Acquiring detailed 3D images of samples is needed for conducting thorough investigations in a wide range of applications. Doing so using nondestructive methods such as X-ray computed tomography (X-ray CT) has resolution limitations. Destructive methods, which work based on consecutive delayering and imaging of the sample, face a tradeoff between throughput and resolution. Using focused ion beam (FIB) for delayering, although high precision, is low throughput. On the other hand, mechanical methods that can offer fast delayering, are low precision and may put the sample integrity at risk. Herein, we propose to use femtosecond laser ablation as a delayering method in combination with optical and confocal microscopy as the imaging technique for performing rapid 3D imaging. The use of confocal microscopy provides several advantages. First, it eliminates the 3D image distortion resulting from non-flat layers, caused by the difference in laser ablation rate of different materials. It further allows layer height variations to be maintained within a small range. Finally, it enables material characterization based on the processing of material ablation rate at different locations. The proposed method is applied on a printed circuit board (PCB), and the results are validated and compared with the X-ray CT image of the PCB part.

1 August 2023 in Scientific Reports
Verified
Journal Article
Open access
PDF file
PHPouria Hoveida
Pouria Hoveida
APAdrian Phoulady
Adrian Phoulady
HCHongbin Choi
Hongbin Choi
NMNicholas May
Nicholas May
SSSina Shahbazmohamadi
Sina Shahbazmohamadi
PTPouya Tavousi
Pouya Tavousi
Counterfeit products pose significant economic, security, and health risks. One approach to mitigate these risks involves establishing product provenance by tracing them back to their manufacturing origins. However, current identification methods, such as barcodes and RFIDs, have limitations that make them vulnerable to counterfeiting. Similarly, nonvolatile memories, physically unclonable functions, and emerging techniques like Diamond Unclonable Security Tag and DNA fingerprinting also have their own limitations and challenges. For a traceability solution to gain widespread adoption, it must meet certain criteria, including being inexpensive, unique, immutable, easily readable, standardized, and unclonable. In this paper, we propose a solution that utilizes ultrashort pulsed lasers to create unique, unclonable, and immutable physical tags. These tags can then be read nondestructively using far-field Terahertz (THz) spectroscopy. The primary objective of this paper is to investigate the feasibility of our proposed approach. We aim to assess the ability to distinguish laser marks with varying depths, evaluate the sensitivity of THz reading to laser engraving parameters, examine the capacity to capture high-information-density marks, and explore the ability to capture subsurface tags. By addressing these aspects, our method holds the potential to serve as a universal solution for a wide range of traceability applications.

17 January 2025 in Scientific Reports
Verified
Journal Article
Open access
PDF file
APAdrian Phoulady
Adrian Phoulady
YSYara Suleiman
Yara Suleiman
HCHongbin Choi
Hongbin Choi
NMNicholas May
Nicholas May
SSSina Shahbazmohamadi
Sina Shahbazmohamadi
PTPouya Tavousi
Pouya Tavousi
Printed Circuit Board (PCB) design reconstruction is essential for addressing part obsolescence, intellectual property recovery, compliance, quality assurance, and enhancing national capabilities. Traditional methods for PCB design extraction, both non-geometry-based and geometry-based, have limitations in accuracy, efficiency, and scalability. This paper presents an automated approach, combining image processing and machine learning, to achieve 3D semantic segmentation of PCB X-ray Computed Tomography (X-ray CT) images and subsequent netlist extraction. By employing a 3D U-Net architecture with a ResNet-18 backbone and training on synthetic data, we introduce a first-of-its-kind method for direct 3D semantic segmentation, significantly improving over previous efforts. Our approach eliminates the need for extensive labeled datasets by using inherently labeled synthetic data. Further, this method enhances ease of segmentation by significantly reducing or eliminating the preprocessing effort required for 2D image stacks. It also improves universality by expanding the scope of application beyond images with specific 2D stack criteria, segmenting the 3D image in its entirety. Additionally, this method enables the processing of images of PCBs that have undergone bending, which is common among PCBs with a thickness below a certain threshold. The implications of this approach extend beyond PCBs, finding applications in various physical and biological sciences where 3D image segmentation is crucial. This methodology includes high-resolution 3D imaging, watershed segmentation, machine learning-based semantic segmentation, and netlist extraction. Validation with both synthetic and real-world PCB datasets shows high accuracy and robustness, offering a scalable solution for PCB design reconstruction.

1 June 2021 in Journal of Micro and Nano-Manufacturing
Verified
Journal Article
Closed access
PDF file
MCMeysam T. Chorsi
Meysam T. Chorsi
PTPouya Tavousi
Pouya Tavousi
CMCaitlyn Mundrane
Caitlyn Mundrane
VGVitaliy Gorbatyuk
Vitaliy Gorbatyuk
Kazem KazerounianKazem Kazerounian
Kazem Kazerounian
HIHorea T. Ilies
Horea T. Ilies
Natural nanomechanisms such as capillaries, neurotransmitters, and ion channels play a vital role in the living systems. But the design principles developed by nature through evolution are not well understood and, hence, not applicable to engineered nanomachines. Thus, the design of nanoscale mechanisms with prescribed functions remains a challenge. Here, we present a systematic approach based on established kinematics techniques to designing, analyzing, and controlling manufacturable nanomachines with prescribed mobility and function built from a finite but extendable number of available “molecular primitives.” Our framework allows the systematic exploration of the design space of irreducibly simple nanomachines, built with prescribed motion specification by combining available nanocomponents into systems having constrained, and consequently controllable motions. We show that the proposed framework has allowed us to discover and verify a molecule in the form of a seven link, seven revolute (7R) closed-loop spatial linkage with mobility (degree-of-freedom (DOF)) of one. Furthermore, our experiments exhibit the type and range of motion predicted by our simulations. Enhancing such a structure into functional nanomechanisms by exploiting and controlling their motions individually or as part of an ensemble could galvanize development of the multitude of engineering, scientific, medical, and consumer applications that can benefit from engineered nanomachines.