Horizon 2020 Marie Skłodowska-Curie Innovative Training Network

Siyuan Chen

BE, MSc
Home/Siyuan Chen
Siyuan Chen 2018-11-08T10:44:43+00:00

Research Interests:

CAD/CAM; Additive manufacturing; Unmanned aerial vehicle application

Biography:

Siyuan holds a MSc degree in Mechanical and Aerospace Engineering from Syracuse University. After receiving his Masters, he worked as a research assistant at Columbia University, in Numerical Modeling, and prior to, at Tsinghua University, developing and supporting interdisciplinary courses in Learning Process Design, and Workshop Management.

He received his bachelor degree from the Beijing Institute of Petrochemical Technology in Mechanical Engineering.

His previous research experience has included under-sea pipeline repair system design (as part of the China National Research Program), rescue equipment design and 3D printer design. He joined TRUSS ITN in September 2015. A summary of his research highlights and training, dissemination and outreach activities in TRUSS  other than network-wide events, is provided in the pdf below, followed by more detailed info on his research outputs.

ESR14_Summary

Research Outputs:

  • Dou, Y., Cai, X. and Chen, S. (2012), “Design of an Amphibious Rescuing Car Based on Fischertechnik Model”, Laboratory Research and Exploration, 31(6).
  • Chen, S., Cai, X. and Li, H. (2012), P.R. China Patent ZL201120186204.0: Electromagnetically Locking Device for Automatic Window.
  • Han, L., Liu, Z. and Chen, S. (2010), “Optimization of Automobile Suspension System Based on SUMT”, Science Technology and Engineering, 110(25).

Publications in TRUSS

Journal papers

This paper reports on patents worldwide related to both hardware and software for the contraction and deployment of UAVs and is intended to provide a snapshot of currently available unmanned aerial vehicles (UAVs) technologies, as well as to identify recent trends and future opportunities in affiliated hardware and software. Basic components related to self-designed units are explained (e.g. platform selection, autopilot control comparison and sensor selection). Current applications and research areas of UAVs are discussed. The research and product development trends focus on extending the flight time, enhancing the water and wind resistant capabilities, improving autonomous navigation abilities, and enriching the payload capacity. Since autonomous navigation is a key technology in UAVs applications, concepts about this are also explained. [DOI] -> Link to full text in repository

Conference contributions
Using an Unmanned Aerial Vehicle (UAV) for documentation and inspection of civil infrastructures has become increasingly popular. One area of interest is in bridge inspection as it holds the potential of being safer, more economical, and less disruptive, with respect to traffic flow. With 3D reconstruction method, structural deficiencies and 3D models can be obtained from a 3D point cloud generated from UAV imagery data. However, shadows and water reflectivity may affect the quality of the point cloud generated from images, which causes difficulty in data processing. This paper presents a detailed workflow of removing outlier data points through the statistical filter and geometric-based filter. The experimental results showed that the statistical filter gives the best performance.
Imagery-based, three-dimensional (3D) reconstructions from Unmanned Aerial Vehicles (UAVs) hold the potential to provide a safer, more economical, and less disruptive approach for bridge inspection. This paper describes a methodology using a low-cost UAV to generate an imagery-based, dense point cloud for bridge deck inspection. Structure from motion (SfM) is employed to create a three-dimensional (3D) point cloud. Outlier data are removed through a density-based filtering method. Next,  the unsupervised learning algorithm k-means and an object-based region growing algorithm are compared for accuracy with respect to bridge deck extraction. Last, an automatic pavement evaluation method is proposed to estimate the deck’s pavement condition. The procedure is demonstrated through an actual case study, in which a 3D point cloud of 16 million valid points was generated from 212 images. With that data set, the region growing method successfully extracted the deck area with an F-score close to 95%, while the unsupervised learning approach only achieved 76%. In the last, to evaluate the surface condition of the extracted pavement, a polynomial surface fitting method was designed to evaluate and visualise the damages.

This paper reports on recent contributions by the Marie Skłodowska-Curie Innovative Training Network titled TRUSS (Training in Reducing Uncertainty of Structural Safety) to the field of structural safety in rail and road bridges (http://trussitn.eu). In TRUSS, uncertainty in bridge safety is addressed via cost efficient structural performance monitoring and fault diagnostics methods including: (1) the use of the rotation response due to the traffic traversing a bridge and weigh-in-motion concepts as damage indicator, (2) the combination of design parameters in probabilistic context for geometrical and material properties, traffic data and assumption on level of deterioration to evaluate bridge safety (via Bayesian updating and a damage indicator based on real time measurement), (3) the application of a fuzzy classification technique via feature selection extracted using empirical mode decomposition to detect failure, and (4) the testing of alternative vibration based damage sensitive features other than modal parameters. Progress has also been made in improving modern technologies based on optical fiber distributed sensing, and sensors mounted on instrumented terrestrial and on aerial vehicles, in order to gather more accurate and efficient info about the structure. More specifically, the following aspects have been covered: (a) the spatial resolution and strain accuracy obtained with optical distributed fiber when applied to concrete elements as well as the ideal adhesive, and the potential for detecting crack or abnormal deflections without failure or debonding, (b) the possibility of using the high-resolution measurement capabilities of the Traffic Speed Deflectometer for bridge monitoring purposes and, (c) the acquisition of bridge details and defects via unmanned aerial vehicles. -> Link to full text in academic repository

This paper introduces a three-dimensional reconstruction experiment based on a physical laboratory-based experiment on a brick wall. Using controlled shooting distances and angles, different images sets were captured and processed with a structure from motion-based technique, which can reconstruct 3D models based on multi-view, Two-Dimensional (2D) images. Those 2D geometries are shown to generate significant deformations within the resulting point cloud, especially where there were large angles (with respect the camera position and the wall’s normal direction) and at close distances to the wall’s surface. This paper demonstrates that by overlapping different flawed image sets, the deformation problem can be minimised.  -> Link to full text in repository

Selected presentations from TRUSS dissemination events

1st TRUSS Symposium (Portoroz (Slovenia), 21st June 2017)

TRUSS Workshop (Dublin (Ireland), 29th August 2018)