Neural networks are a form of Artificial Intelligence based on the architecture of the human brain. They allow complicated non-linear relationships to be learnt from example data, and for further test data to be identified according to the relationship previously learnt. This allows the construction of control systems and diagnostic systems of geotechnical processes which were previously not possible due to their complicated non-linear nature. The main topic of research is the application of neural networks to the diagnosis of the condition of ground anchorages. Ground anchorages are in use in many engineering structures such as tunnels, retaining walls and dams and it has been reported that only 5-10% are routinely monitored during service. The conventional method of testing is load lift-off testing, which is expensive and time consuming. The patented technique, GRANIT, makes use of neural networks to learn the complicated relationship between the vibrational response of an anchorage to an applied axial impulse and its post-tension level. Research has been conducted into the parameters of the system which affect the diagnostic ability of the neural network. Further research into the application of the GRANIT technique to the identification of other faults in the anchorage has been conducted, such as change in free length, or gaps in the grouting. An automated procedure for the identification of the frequencies of interest in the response signatures of the GRANIT system has been investigated, and an example is given of an application of this automated procedure in the area of vibro-impact ground moling, a patented technique which uses both vibration and impact to maximise its penetration depth. Further research into the use of neural networks in an automated process has also been undertaken, and the development of a new technique is presented. This new technique has the potential of returning parameters of interest from any given group of signals, and has potential of application outwith geotechnical data. A patent application for this new technique has now been filed by the author.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:369120 |
Date | January 2001 |
Creators | Starkey, Andrew J. |
Publisher | University of Aberdeen |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | http://digitool.abdn.ac.uk:80/webclient/DeliveryManager?pid=217212 |
Page generated in 0.0156 seconds