On-line, model-based fault diagnosis of production lines is accomplished by examining the output times of the system, i.e. the output event-timing signature of the system. Specifically, the detection of a disturbance in the input arrival times is demonstrated. A timing residual is formed using an event-timing model based on the max-plus algebra of the production line. The residual is then analyzed using standard statistical process control (SPC) techniques to detect failures. Simulation results for a simple two-machine production line illustrate the approach. The novelty of this approach is the use of a max-plus event-timing model with an additive noise model and the application of an event-timing model to the fault diagnosis problem.