Since chemical processes are often operated over a range of operating conditions and some of the system parameters are only known to a certain degree, uncertainties exist in the process model. Interval types of process models offer an attractive alternative for process description in an operating environment. In terms of fault diagnosis, an interval process model based diagnostic system is robust as compared to conventional quantitative model-based systems. In this work, an interval model is incorporated into the deep model algorithm (DMA) for fault diagnosis. A design procedure is given, and characteristics of interval DMA are also discussed. One unique property is that the interval parity equations generally give better diagnostic resolution than the crisp ones under the DMA framework. A CSTR example with interval coefficients is used to illustrate the design and effectiveness of the interval DMA. Results show that the proposed method is not only successful in handling wide range of operating conditions but also capable of identifying correct fault origins accurately.