Loose Bolt Identification via Machine Learning

Project Background: Loosening of bolts in large constructions is a common occurrence which may arise from a variety of causes such as structural vibrations due to fluctuating loads, improper installation procedures, bolt fracture, etc. In recent years many structural health monitoring techniques have been developed and the knowledge aquired can be applied for this problem [1, 2]. Bolt loosening may cause a broad spectrum of faults ranging from overstressing to reduced fatigue life of the structure, thus requiring consistent and systematic maintenance over the life cycle of a structure [3]. This is an extremely costly and often dangerous operation which requires a one-by-one check of each bolted connection at regular time-intervals. Current technologies to perform bolt monitoring range from classical technique executed via acoustic emission of hammer strikes, to a variety of more advanced methods such as ultrasonic testing. These techniques require trained technicians as well as significant support facilities and preparation [4]. This is an extremely repetitive, intensive task requiring specialized personnel and which makes the inspection operation burdened with high associated costs as well as significant risk for the human operators. A new solution, which seeks to minimise the need for human intervention during bolt inspections entails the use of a remote monitoring system based on Laser Ablation for Impulse Vibration Excitation [4, 5, 6]. Laser ablation technique is highly reproducible and is able to selectively affect the high frequency portion of the vibrational response, which is the one better suited for segregating the effect of loose bolts from other sources of vibration.
 
Project Summary: The use of laser ablation technology for bolt inspection is still in its infancy [4] and needs to be complemented with strong data-processing tool [6] in order to enable its upscaling, automation and, ultimately, its commercialization. The characterization of the high frequency response of a large-scale structure is a strongly non-linear and high-dimensional problem, which lends itself to treatment via machine learning techniques thanks to the occurrence of underlying dominant patterns. The PhD candidate will design a machine learning approach to automate the process of loose bolt identification in vast sections of largescale structures. This is a complex problem which will require access to new, high-quality data from laser ablation testing and a high proficiency in data-driven identification techniques for non-linear systems.
 


Requirements and Eligibility: 
Well qualified and strongly motivated PhD candidates are invited to join an expanding research group in Structural Health Monitoring. Applicants will be selected using the following criteria:

  • Possess, or be about to obtain, an upper second (2.1) UK BEng Hons or equivalent, MEng or postgraduate MSc degree in a relevant engineering or physics related subject.
  • Applicants are expected to have excellent analytical skills and a solid background in structural analysis/mechanics, engineering vibration and advanced numerical modelling of structures.
  • The ideal candidate should have knowledge or be familiar in signal processing and in system identification methods; familiarity with model updating methodologies; possible familiarity with the topic of wind turbine facilities; (ideally previous) experience in measurement instrumentation and data acquisition techniques.
  • Applicants with a deep understanding of statistical inference and/or machine learning techniques will be preferred.
  • Relevant professional experience is very welcomed
  • Have skills and understanding of MATLAB/Python and its use in data analysis

Research Environment: The candidate will work jointly between the Institute of Infrastructure and the Environment (IIE), the Institute of Integrated Micro and Nano Systems (IMNS) and the Mechanical Dynamics Lab at Shibaura Institute of Technology in Tokyo, Japan. Development of the Data Driven tools will be developed at the University of Edinburgh, while experimental testing will be performed in Japan.  

Supervision Team:

  1. Dr David García Cava (david.garcia@ed.ac.uk), Institute of Infrastructure and the Environment, University of Edinburgh, UK
  2. Dr Francesco Giorgio-Serchi (f.giorgio-serchi@ed.ac.uk), Institute of Integrated Micro and Nano Systems, University of Edinburgh, UK
  3. Prof. Naoki Hosoya, Shibaura Institute of Technology, Japan
  4. Prof. Shingo Maeda, Shibaura Institute of Technology, Japan

Suggested Reading:  

[1] Farrar CR, Worden K. An introduction to structural health monitoring. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences. 2007 Feb 15;365(1851):303-15.
[2] García D, Tcherniak D. An experimental study on the data-driven structural health monitoring of large wind turbine blades using a single accelerometer and actuator. Mechanical Systems and Signal Processing. 2019 Jul 15;127:102-19.
[3] HSE Information Sheet Guidance on management of ageing and thorough reviews of ageing installations Offshore Information Sheet No. 4/2009.  
[4] Kajiwara I., Miyamoto, D., Hosoya N. and Nishidome C., Loose Bolt Detection by High Frequency Vibration Measurement with Non-Contact Laser Excitation, Journal of System Design and Dynamics, Vol. 5, No. 8, 2011.  
[5] Kajiwara I. and Hosoya N., Method for measurement of vibration property of structure, and vibration property measurement device, EUROPEAN PATENT SPECIFICATION, EP2584335B1, PCT/JP2011/003412, WO 2011/158503
[6] N Hosoya, T Niikura, S Hashimura, I Kajiwara, F Giorgio-Serchi, Axial force measurement of the bolt/nut assemblies based on the bending mode shape frequency of the protruding thread part using ultrasonic modal analysis, Measurements, 2020, DOI: 10.1016/j.measurement.2020.107914.
 

Further Information: 

The University of Edinburgh is committed to equality of opportunity for all its staff and students, and promotes a culture of inclusivity. Please see details here: https://www.ed.ac.uk/equality-diversity

Closing Date: 

Saturday, May 1, 2021
Bolted connections on a Wind Turbine Monopile
Bolted connections on a Wind Turbine Monopile

Principal Supervisor: 

Assistant Supervisor: 

Eligibility: 

Minimum entry qualification - an Honours degree at 2:1 or above (or International equivalent) in a relevant science or engineering discipline, possibly supported by an MSc Degree. Further information on English language requirements for EU/Overseas applicants.

Funding: