This paper introduces methods to evaluate the wear and to estimate the remaining useful life of non-return valves based on process and machine data of injection molding machines. In industry, non-return valves of injection molding machines are often replaced either too early due to maintenance intervals or after failure risking to damage adjacent components, since the current state of wear is not known without elaborate disassembly. A reliable predictive maintenance system helps to avoid unscheduled shutdowns due to component failures, to utilize the entire components service life and, consequently, to increase the overall equipment effectiveness. This paper shows an automated predictive maintenance approach for non-return valves by using available process and machine data. First, wear-specific key figures are created from process data using process monitoring tools, which can be used to classify the type of wear. Based on this, an estimation of the remaining service life can be predicted by regression methods.