|M.Sc Student||Drori Oded|
|Subject||Characterization of moving surface loads with buried|
|Department||Department of Civil and Environmental Engineering||Supervisor||Assistant Professor Eyal Levenberg|
|Full Thesis text|
The thesis was concerned with characterizing a surface traveling object by means of a cluster of inertial sensors. The concept proposed in the research was recording acceleration responses within the carrying medium generated by the passage of an object along the surface; this was done at several underground locations wherein accelerometers were buried. Recorded data from the cluster was utilized in combination with a mechanical model, simulating the load and the medium responses, in order to infer the traveling object characteristics. This is a generic problem of interest to several study disciplines, including: transportation, agriculture, border security, wildlife tracking, and crowd management.
Assuming a half-space model, virtual acceleration responses were computed for all sensors locations with a preliminary assumed set of object characteristics. These characteristics are in fact the model parameters, and were manipulated until best match was achieved between virtual and recorded accelerations. The inverse approach implemented a backcalculation scheme, wherein successive calculations were performed in an attempt to minimize the error. Once minimum discrepancy was attained, the model parameters were taken to be the sought object characteristics.
Calculations were done assuming quasi-static responses, i.e., inertia effects were neglected. Doing so allowed for simpler modeling and shorter computational times; it also lessened data acquisition demands in terms of data collection rates and quantities, and in terms of synchronization level between sensors. These features are critical when envisioning accelerometer clusters that must operate wirelessly.
As a first step, forward modeling of acceleration traces within a half-space was investigated. This stage allowed understanding of expected behavior and sensitivity with respect to object parameters. Next, a characterization framework was developed, in pursue of a feasible scheme to infer object characteristics. Facilitating the developed framework, the investigation focused on characterization of an object with a single contact point. Initially, a separate model was employed for generating synthetic “measured” responses. Several synthetic cases were tested wherein various trajectories, speeds, and loading magnitudes were included. Also studied were different cluster topologies for positioning the accelerometers. Later, realistic conditions of a traveling object were simulated, as a preparation for analysis of real (field) measurements. These synthetic-realistic conditions brought forward the necessity, and hence the development, of an improved characterization framework, which implemented model-guided smoothing for exclusion of unwanted/irrelevant acceleration data. Finally, a field test was designed and carried out - generating authentic soil acceleration response due to a moving vehicle. The characterization framework was applied to the field measurements, demonstrating applicability and capabilities.
Overall, the developed framework is deemed workable and capable of characterizing a moving surface object. Merits and limitations of the scheme were discussed throughout the thesis, along with ideas for future improvements.