AbstractInspired by the rationale of Robust Design, a novel methodology is presented for the design of diagnostic systems for fault detection and isolation (FDI). Detection/isolation capability and robustness, i.e. sensitivity to faults and insensitivity to noise, are addressed in an integrated way within a fully stochastic framework. Although FDI is a classical problem in control engineering, this new approach improves the current state of the art both in terms of general applicability and optimality of the design solution. It demonstrates the potential of robust design in fostering innovation in a variety of technical sectors. For illustrative purposes, the methodology is applied to the design of a FDI system for a fluid mixer. Copyright © 2006 John Wiley & Sons, Ltd.