The furnace exit gas temperature (FEGT) is one of the fundamental parameters necessary to determine the energy balance of the boiler in a coal-fired power plant, and is thus beneficial to the production of reliable thermo-fluid models of its operation and the operation of the systems down and upstream. The continuous measurement of the FEGT would also be a useful indicator to predict, prevent and diagnose faults, optimize boiler operation and aid the design of heat transfer surfaces. Acoustic pyrometry, a technique that measures temperature based on the travel time of an acoustic wave in a gas, is investigated as a viable solution for continuous direct measurement of the FEGT. This study focuses specifically on using acoustic pyrometry to reconstruct the temperature profile at the furnace exit including methods for accurately determining the time of flight (TOF) of acoustic waves. An improved reconstruction technique using radial basis functions (RBF) for interpolation and a least squares algorithm is simulated and its performance was compared to cubic spline interpolation, regression and Lagrange interpolation by evaluating its reconstruction accuracy in terms of mean and root-mean-squared (RMS) error when reconstructing set temperature profiles. Various parameters including transceiver positions, grid divisions and time of flight error, are investigated in terms of how they inform acoustic pyrometry implementation. The improved RBF interpolation function managed to reconstruct complex temperature profiles and had a greater reconstruction accuracy than compared interpolation methods, improving on the accuracy of previous work done. Random acoustic path error was found to not be additive with reconstruction error however repeating acoustic TOF readings improved reconstruction accuracy to mitigate this effect. In general, it was also found that symmetrical transmitter/receiver positions produced more accurate reconstructions as well as positioning receivers/transceivers and grid lines closer to the furnace walls, where the greatest temperature change occurs. In addition to testing reconstruction methods, a low-cost experimental set-up was constructed to measure the time of flight. The focus of this study was on using various signal processing methods to determine the time of flight and evaluating their accuracy in the presence of noise. Methods such as threshold detection with bandpass filtering, cross correlation, generalized cross-correlation (GCC) and a new method developed employing variable notch filters with locations and widths based on repetitive frequencies identified in the noise with cross correlation. The performance of methods was experimentally tested under varying signal to noise ratios (SNR) and noise conditions. These SNR tests showed that cross-correlation methods produced more reliable TOF readings under lower SNRs than threshold detection methods. Under white noise the smooth coherent transform (SCOT) GCC variation proved to produce the most accurate results producing an average TOF error of 0.84 % up until a SNR of 1.4 before reducing in accuracy. In coloured noise (generated based on previous boiler recordings) the variable notch filter method with crosscorrelation was able to identify repetitive noise frequencies filter them out and ultimately produced results with an average TOF error of 1.99 % up until a SNR of 0.67, where the noise level exceeds that of the signal.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:uct/oai:localhost:11427/29516 |
| Date | 14 February 2019 |
| Creators | Raikes, Geoff |
| Contributors | Mouton, Hennie, Fuls, Wim |
| Publisher | University of Cape Town, Faculty of Engineering and the Built Environment, Department of Mechanical Engineering |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Master Thesis, Masters, MSc (Eng) |
| Format | application/pdf |
Page generated in 0.0019 seconds