Search

This paper presented an experimental study on simultaneously reconstruction of three-dimensional temperature and inhomogeneous radiative properties distribution by radiation image processing in a 600 MWe supercritical arch-fired boiler. Two independent monochromatic radiation intensities extracted from the flame image were used to solve the temperature and radiative properties respectively by iteration, and provided a reasonable option for the radiative properties. Finally the convergent radiative properties were fixed for the on-line detection of temperature. The visualized temperatures obtained by the present method for different cases agreed well with those measured by infrared pyrometer at different locations, while the largest relative error was less than 4.60%. The three-dimensional temperature measurement results indicated that, the undershoot depth of flame was increased and the flame scoured the cold hopper when the F secondary air was downward sloping in the boiler. The present on-line temperature measurement method based on inhomogeneous radiative properties was reliable for the combustion diagnostic in the combustion industry, and would help to improve the awareness level of combustion in the supercritical arch-fired boiler.

In order to improve the quality of tomography reconstruction for ill-posed problems with limited projections, a new idea of fabricating virtual lines of sight which are consistent with, but not lineally dependent on, the measured lines of sight is proposed in this paper. Using a scheme of equilateral triangle plane meshes whose edge length equals twice the width of the lines of sight, based on the spatial continuity of the source term distribution, some independent virtual lines of sight are fabricated by an interpolation method described in this paper. Non-negative linear least squares method is used to solve the reconstruction problem. As the times of interpolation for virtual lines approach a critical level (it is 6 for the case in this paper) the condition number of the matrix of the forward problem drops obviously, indicating that the independent interpolation method may effectively overcome the `ill-posed-ness' of the problem. Accordingly, if the times of interpolation for virtual lines are less than the critical level, the shape of the assumed source distribution, obviously could not be reconstructed; as the times of interpolation become larger than the critical level, the shape of the assumed source distribution has been recovered visually. With 10 times of interpolation for virtual lines, most of the relative errors near the four peaks are below 10 %, thus the effectiveness of the interpolation method is reasonably verified. In addition, the reconstruction method has good noise immunity.

The distribution of ratios of energy scattered by the medium or reflected by the boundary surface (DRESOR) method can provide radiative intensity with high directional resolution, but also suffers the common drawbacks of the Monte Carlo method (MCM), i.e., it is time-consuming and produces unavoidable statistical errors. In order to overcome the drawbacks of the MCM, the so-called equation-solving DRESOR (ES-DRESOR) method, an equation-solving method to calculate the DRESOR values differently from the MCM used before, was proposed previously. In this method, a unit blackbody emission is supposed within a small zone around a specified point, while there is no emission elsewhere in a plane-parallel, emitting, absorbing, and isotropically scattering medium with transparent boundaries. The set of equations for the DRESOR values based on two expressions for the incident radiation was set up and solved successfully. In this paper, the ES-DRESOR method is extended to a one-dimensional system with diffusely reflecting boundaries. The principle and formulas are given. Several examples with different parameters are taken to examine the performance of the proposed method. The results showed that all the DRESOR values obtained using the ES-DRESOR method agree well with those got using MCM. The average relative error for the intensity obtained by the ES-DRESOR method is 9.446 × 10−6, lower by over 1 order of magnitude than the 2.638 × 10−4 obtained by the MCM under the same conditions. More importantly, the CPU time for computing the DRESOR values, which ranges from several hundred seconds to several thousand seconds using the MCM, is reduced to 0.167 s using the ES-DRESOR method. The computation time is shortened by about 3 orders of magnitude. The overall performance of the ES-DRESOR method is excellent.