No abstract

No abstract

The fundamental concepts and methods of fault detection and diagnosis are reviewed. Faults are defined and classified as additive or multiplicative. The model-free approach of alarm systems is described and critiqued. Residual generation, using the mathematical model of the plant, is introduced. The propagation of additive and multiplicative faults to the residuals is discussed, followed by a review of the effect of disturbances, noise, and model errors. Enhanced residuals (structured and directional) are introduced. The main residual generation techniques are briefly described, including direct consistency relations, parity space, and diagnostic observers. Principal component analysis and its application to fault detection and diagnosis are outlined. The article closes with some thoughts about future directions. measurements to estimates obtained, from other measurements, by the model; any discrepancy may be an indication of faults. Another class of techniques (generally but incorrectly called "data driven"), most notably principal component analysis (PCA), include the estimation of an implicit model, from empirical plant data, and then use this in ways similar to the model-based methods. These approaches will be described in more detail in the sequel.

No abstract

No abstract