Analysis of sampling and multi-vehicle separation for BWIM systems
dc.contributor.author | Grabau, Mathew A.C. | |
dc.contributor.examiningcommittee | Cai, Jun (Electrical and Computer Engineering) Issa, Mohamed (Civil Engineering) | en_US |
dc.contributor.supervisor | McNeill, Dean (Electrical and Computer Engineering) | en_US |
dc.date.accessioned | 2015-09-14T15:05:53Z | |
dc.date.available | 2015-09-14T15:05:53Z | |
dc.date.issued | 2015 | |
dc.degree.discipline | Electrical and Computer Engineering | en_US |
dc.degree.level | Master of Science (M.Sc.) | en_US |
dc.description.abstract | Structural Health Monitoring provides ample opportunity for deep analysis of our infrastructure, including using the bridge as a scale through a process called (Bridge) Weigh in Motion (BWIM). Many variables impact the BWIM’s capabilities, accuracy, and by extension, overall usefulness. The overall goal of the research conducted was to identify methods of improving the accuracy and performance of BWIM. That goal was narrowed down to two specific objectives: 1) assess if a higher sampling rate leads to increased BWIM performance as postulated by some sources [1]; and 2) attempt to develop a means of analyzing complex multi-vehicle events where the distributed strain profile cannot be processed by the standard BWIM algorithms. The first objective was accomplished using a Matlab simulation to generate sampled strain signals, with different sampling nonidealities such as offsets in the start of captured events. The testing demonstrated that the sampling rate is sufficient at 100 Hz, with minute peak detection errors manifesting only when running at unreasonable levels of accuracy for the Matlab analysis. Given that information, there is potentially room for reducing the sampling rate which benefits BWIM installations by saving on data storage requirements. Beyond that, no further testing is recommended in the area of sampling rate. The second objective was accomplished by using higher-order signal processing techniques such as Independent Component Analysis (ICA). These techniques aim to, at a minimum, ensure that heavy-vehicle events are detected and recorded. A total of four tests were performed on Independent Component Analysis — two on simulated strain mixtures and two on signal samples collected from bridge data. The results of the tests demonstrate ICA may potentially be introduced into a BWIM implementation pending further refinements. The most likely targets for improvement are through analyzing the independence of truck signals using correlation, taking measures to decorrelate the mixtures, and also testing whether post-separation filtering of the strain readings impacts the result. Notwithstanding those areas of improvement, the overall verdict is that a clear recommendation for using ICA in active BWIM analysis is not currently feasible. | en_US |
dc.description.note | October 2015 | en_US |
dc.identifier.uri | http://hdl.handle.net/1993/30784 | |
dc.language.iso | eng | en_US |
dc.rights | open access | en_US |
dc.subject | ICA | en_US |
dc.subject | BWIM | en_US |
dc.title | Analysis of sampling and multi-vehicle separation for BWIM systems | en_US |
dc.type | master thesis | en_US |
local.subject.manitoba | yes | en_US |