Thermal transients study for the ESS target helium cooling system

Abstract: The European Spallation Source (ESS) is going to be the worlds most powerful spallation source once in full operation. The neutrons used by scientists and engineers are spallated from a rotating wheel of wolfram, being bombarded by accellerated protons. The wheel, also called the target, develops heat in the process, and is therefore cooled by a helium cooling system. The target primary cooling system, named system 1010, is modelled in the language Modelica but is not fully verified, nor has any practical testing been conducted. ETHEL is a test rig located at LTH, similar in function to that of system 1010. It can be used to conduct experiments, and the results can to some degree be extrapolated to system 1010. A model of ETHEL was constructed. The model was intended to both be faithful to the physical contraints of ETHEL, but also constructed similarly enough to the system 1010 model to allow for extrapolation of the conclusions made. Transient experiments were formulated, based on expected cases of failure for system 1010. The experiments were conducted on the ETHEL rig, the ETHEL model, and the system 1010 model. Based on the experiment results a model evaluation of the ETHEL model was conducted, both in regards to time-response behaviour, and in regard to data error. System identification on both the system 1010 model data, and the ETHEL data was conducted, with the purpose to identify significance of sensor location in the former case, and to determine the relative speed of the sensors in the latter case. Three sensor types were examined for their speed, however the analysis lacked the tools to evaluate their accuracy. The location analysis revealed that the exact location of the sensor in the system 1010 model did not significantly impact the time-response. The model evaluation showed that the ETHEL model performance was qualitatively good, demonstrating a good time-response compare to the real system, however significant quantitative errors were found. The results were extrapolated to the system 1010 model and its real counterpart. This implies that the location analysis is fair to the real system, and further conclusions about this fact was made.