capillary tube, isobutane, cycle simulation, non-adiabatic
Capillary tubes including a suction line heat exchanger are typical expansion devices used in nowadays domestic refrigeration appliances. To account for their functionality in transient cycle simulations, including the highly transient start-up and shut-down operations, different capillary tube models and their implementation in such a cycle are investigated. These non-adiabatic capillary tube models comprise dimensionless correlations, neural network methods and one-dimensional homogeneous models which stem partly from open literature and from previous work of the authors. The difficulty in application of capillary models during off-design conditions is when two phase flow or even superheated vapor which enters the capillary tube. These conditions are not covered by most of the schemes. In this work a transient cycle simulation including a 1d formulation of the heat exchangers and a semi-empirical compressor model serves as virtual test bench for several capillary tube models. The range of parameters is chosen according to the need of domestic applications using R600a - mass flow rates range between 0 and 5 kg/h at inlet pressures up to 10 bar. The comparison of different implementation strategies is carried out in terms of speed, accuracy and stability in off-design. The predictability of the models is evaluated by steady state experimental data from literature and own experiments as well as an in-house 1d model for the regions where no measurements exist. It is concluded that the direct implementation of the 1d code bears the disadvantage of low speed, whereas common dimensionless correlations lose accuracy off its design point. Neural networks turn out to be a good trade-off between speed, reliability and accuracy.