Real-time deformation monitoring of bridges using GPS/Accelerometers
The need for conducting real-time bridge deformation monitoring is addressed in the context of the development of bridge management system (BMS) and land transportation safety in this thesis. Current instruments used for bridge dynamic deformation monitoring are compared in terms of system productivity and reliability. An integrated sensor system of Global Positioning System (GPS) receiver and triaxial accelerometer is then proposed with the capabilities to accurately monitor long-term deformation and short-term dynamics of a bridge. Since the investigation of the bridge dynamic responses is of great importance in research and practice, the emphasis of this thesis is on the monitoring of dynamic bridge deformation. Zero baseline (ZBL) and short baseline (SBL) tests are conducted to evaluate the performance of three types of Leica GPS receivers at 10 Hz sampling rate. Statistic characteristics of positioning solutions and the achievable accuracy of each receiver type are analysed, which are employed to design optimal filters for various GPS error suppressions. By using a moving average (MA) technique, millimetre baseline accuracy can be achieved even with a single frequency receiver. It demonstrates the possibilities to conduct millimetre bridge deformation monitoring if appropriate filtering techniques are applied to the positioning solutions and integer ambiguity has been fixed. A simple but accurate triaxial accelerometer calibration technique is proposed in the thesis with a solid mathematical derivative to evaluate the precisions of estimated parameter offsets. A specially designed cage is used to house a GPS antenna with a triaxial accelerometer to avoid complex sensor alignment and simplify the coordinate transformations between different reference frames. The determination of instantaneous attitude of an accelerometer body frame is realised by three GPS stations on the deck of a bridge at a rate of 10 Hz and the sensed 3D accelerations are then transformed into a bridge coordinate system (BCS) simultaneously. BCS is the computation frame of a hybrid bridge deformation monitoring system (BDMS). Important issues in sensor integration such as local gravity determination, synchronisation of time series from different sensors are addressed. Bridge trials are briefed with the emphasis on the instrument configuration for effective error mitigation and sensor integration. A group of reference stations consisting of two reference stations closely setup near a bridge and the permanent continuous GPS stations are recommended for reducing relative tropospheric delay, multipath, and receiver noise both at reference stations and monitoring sites. GPS satellite sky distribution and its impact on propagating ranging errors in mid latitude areas such as in the UK and high latitude areas are analysed both with analytical and simulation approaches. The error propagation formulae are derived to analyse the defects of current satellite constellation on the GPS positioning solutions in each direction of a BCS. This is further exploited to improve the component accuracy of particular interest through changing the dilution of precision (DOP) values. The degree of positioning improvement is illustrated with GPS/GLONASS positioning. A simulator according to ranging error propagation is used to simulate the achievable accuracy from the best and the worst GPS constellations. Modified precise satellite ephemeris by inserting the positions of pseudolites is employed to investigate the changes of DOP values in each direction of a BCS. The summaries of this simulation have universe significance in guiding the selection of the best locations of pseudolites. Adaptive Finite Impulse Response (FIR) filtering or adaptive filtering (AF) for short and important application issues are addressed in the thesis. Autocorrelation lags of ZBL and SBL tests of each type of receiver are used to determine the filter lengths according to the fundamentals of low pass and high pass filter designs. A real-time AF algorithm is introduced and widely employed as an analytical tool in the error mitigation and real bridge deformation signal extraction. The application defects of MA technique in bridge deformation monitoring are compared with AF approach according to the component analysis of GPS positioning solutions. A recursive AF algorithm is proposed to gradually isolate actual bridge deflection signals from multipath and receiver noise both at reference stations and monitoring sites. Spectral analysis is applied to the input and output signals to investigate the efficiency of the designed filter. In order to effectively isolate actual bridge deformation, misalignment and its consequence are demonstrated with day-to-day shifted time series of bridge deflection. Cross-correlation is also used to analyse the feasibility and efficiency of the proposed AF algorithm. Acceleration aided AF approach is detailed in the thesis. A simple algorithm, based on the principles of digital signal filtering and optimal filter design, is proposed to estimate relative displacements of bridge sensed by a triaxial accelerometer in three dimensions. With the relative displacements, GPS receiver noise has been filtered out and the cleaned displacements are obtained. AS another data fusion approach, a software package based on discrete Fourier transform (DFT) to integrate GPS and accelerometer data with a position output rate up to that of a triaxial accelerometer is introduced. Relative tropospheric delay is another major error source identified in GPS-based bridge deformation monitoring. Methods applied to distinguish the impacts of multipath and tropospheric delay are presented. The cause for relative tropospheric delay is analysed and microclimate effect is recognised as the major impact factor in this particular environment. Numerical calculations also confirm the assumption. The way to effectively remove relative tropospheric delay is recommended. The research emphasis in this thesis is to develop a prototype of a hybrid BDMS to achieve centimetre level positioning accuracy at each epoch in three dimension of a BCS. The findings from this research are summarised and the future work is predicted.