| dc.description.abstract | The current global energy situation calls for action to be taken to reduce fossil fuel consumption and generate clean energy. One way to reduce fossil fuel consumption is by utilizing and recovering the large amounts of waste heat generated in industrial processes. Among the different technologies available used to recover waste heat, the organic Rankine cycle is an incipient technology that has operational advantages over the conventional steam Rankine cycle when the temperature of the waste heat source does not exceed 350ºC. However, the efficiency of these cycles is relatively low, as they generally employ an expander optimized exclusively for the design conditions and the conditions to which they are subjected are usually variable and far from the design point.
Of the various turbine architectures available on the market, the axial turbine fits the power ranges of an organic Rankine cycle, from a few kWel up to some MWel. With the idea of providing a powerful tool for the design of an axial turbine staging, previous work on this thesis has been aimed at developing a mean-line model for preliminary design under design and off-design conditions, as it provides an accurate first estimate of the turbine's key dimensions and parameters. This model has been used to evaluate the losses that occur in axial turbines, whereby entropy is generated in the expansion process and the efficiency of the turbine is reduced. Accurate determination of these losses is essential in order to obtain a precise model that can be used to increase the potential of these cycles. In order to evaluate them, various authors have proposed a series of empirical correlations to account for the different components of the internal losses of axial turbines. Among the different correlations considered, the Benner loss model has been chosen for this purpose as it takes into account the incidence losses when the turbine faces off-design conditions and has proven to predict turbine efficiency more accurately when comparing his work to other authors'.
In order to ensure the consistency of the model and to improve the optimization mode so that it can be carried out more efficiently, a sensitivity analysis has been conducted using one of the various design of experiments techniques, the face-centered central composite design. This technique is used to account not only for the main and interaction effects but also for the quadratic effects of the factors that are inputs to the model and have been selected for the study. With this aim, two sets of experiments have been carried out in Python within a wide design space and a local sensitivity analysis respectively, followed by a statistical analysis in order to determine the significance of the effects. From these experiments it has been concluded that, despite limitations found in the model which disabled studying the selected factors as a whole and despite the reliability of the results due to the regression model that has been used to obtain the coefficients, the mean-line model proposed as design tool seems to make sense, as it presents as significant those effects that had proven to have a considerable influence on the efficiency in a previous analysis, suggesting that none of them should be discarded when designing an ORC axial turbine. It was also found that reducing the design space in the optimization when using gradient-based algorithms does not necessarily lead to more efficient optimization. | |
