Semiconductor production fabs regularly encounter faults which result in unscheduled tool downtime. Among these are real-time tool faults, preventative maintenance recovery problems and tool mis-matching at start-up and process transfer. This downtime can be reduced by applying a Fault Detection and Classification scheme where the core problem is identified as rapidly as possible, replacing the usual "trial-and-error" approach to fault identification. Scientific Systems have developed a non-intrusive, high-resolution impedance sensor which is designed to aid fault identification. The sensor is used to characterize a baseline process, operating within control limits, by measuring the Fourier components of RF voltage, current and phase. This results in a unique impedance fingerprint of the chamber. When a fault condition occurs, the impedance fingerprint varies in a predictable pattern. By comparing the fault fingerprint to the baseline, it is possible to classify faults through a diagnostic methodology. Using the system, hardware problems can be separated from process issues and changes in individual process inputs can be identified. We report a number of case studies where the system has been successfully deployed.