Mechanical properties are the features that measure the ability of a metal to withstand several loads and tensions. Specifically, ultimate tensile strength is the force a material can resist until it finally breaks. This property is one of the variables controlled during the foundry process. The only way to examine this feature is to apply destructive inspections that make the casting invalid with the subsequent cost increment. Modelling the foundry process using machine learning allows algorithms to foresee the value of a certain variable, in this case, the probability of a certain value of ultimate tensile strength for a foundry casting. However, this approach needs to label every instance to generate the model that will classify the castings. In this paper, we present a new approach for detecting faulty castings through collective classification to reduce the labelling requirements of completely supervised approaches. Collective classification is a type of semi-supervised learning that optimises the classification of partially-labelled data. We perform an empirical validation demonstrating that the system maintains a high accuracy rate while the labelling efforts are lower than when using supervised learning.