Inverse Heat Conduction problem of the Quenching a Rotary Cylinder by Multiple Water Impinging Jets

Sammanfattning: The thesis deals with solving the time dependent inverse heat conduction and heat transfer problem of the quenching process of a rotary solid cylinder by multiple impinging water jets. The development of such investigation consists of two parts that complement each other. As is the case of any scientific experiment, first of all, an initial hypothesis will be set to be demonstrated theoretically. The numerical validation is carried out with a series of artificial cooling curve data and sensitivity analyses in the inverse solution. Then, a series of recorded temperature data were implemented into the inverse solution to predict the surface heat transfer during the quenching process.The numerical study consists of the solution of a two-dimensional linear time dependent inverse heat conduction problem based on the Generalized Minimal Residual Method (GMRES). The inverse solution method is based on the solution of an iterative problem, validated by a set of artificial temperature data. Such solution allows the prediction of the surface temperature and heat flux distribution in the quenching process, making use of recorded internal temperatures of the specimen. In order to solve the problem, the Matlab and Comsol Multiphysics programs were used. The GMRES algorithm was written as Matlab code, while the computational domain was defined in Comsol Multiphysics. Moreover, both programs collaborated in the solution of the inverse problem. Once the problem was solved, a sensitivity analysis was carried out in order to study the dependence of the numerical result on various parameters and optimize the inverse solution setup for application of recorded experimental data.The validated inverse solution setup examined by the sensitivity analyses was used on a set of experimental data, allowing the demonstration of the initially proposed hypothesis. This sensitivity analyses were performed consecutively for different key parameters regarding the numerical definition of the problem. The values for the parameters were considered optimal when minimum values for the error of the predicted surface temperature were recorded. In this case, the analyzed parameters were the m-value, mesh cell size, effect of noise, initial quenching temperature and quenching cooling rate. The connection between the experimental and numerical studies is obvious, as the first oneprovides the latter with input data of the inner temperature data of the specimen for the solving of the inverse problem, as is the case of the practical application of the code developed in the present thesis, and the inverse solution is essential in order to predict thesurface temperature and heat flux that are key information in studying quenching systems.