The research project describes the development of an assistance system for the injection molding process of large-format plastic components such as car bumpers. For process data acquisition, 30 specified sensors were integrated into an injection mold for passenger car bumpers. Part of the sensor technology is integrated into the cavity, while another part is located in the vicinity of the cavity. For example, the displacement sensors positioned around the cavity detect the deformation of the injection mold during the injection process. For recording and time-synchronous storage of all process data, hardware specially adapted to the high data input was developed and put into operation. The investigations provide information on the extent to which displacement measurement signals, which are interpreted using machine learning methods, can be used for indirect inline quality assessment of the components. The database obtained from the investigations is the basis for the use of machine learning methods from the low-data domain. With the aid of the support vector regression algorithms used, it is possible to derive the different quality criteria from the sensor signals. In this way, the component quality can be evaluated in the process cycle, even before the mold opens and a new cycle produces another component. It is also possible to derive statements from the algorithms that can be used to make recommendations for correcting the process parameters. These are recommended to the machine operator via the display according to the "man-in-the-loop" principle. The aim of the research project is to provide plastics processing companies with a mold-based assistance system to actively support the injection molding process. Furthermore, the investigations are to show which type of sensors and their positions in the mold are decisive in order to be able to make a binding prediction about the produced component quality insitu of the process.