Analytic prognostic for petrochemical pipelines

Summary of the ISO 13381-1: 2004 standard main steps. Figure 2. Estimation of the value of the RUL. used by the diagnostic module. The output of this module identifies the actual operating mode. This state is then projected in the future, by using adequate tools in order to predict the system’s future state. The intersection point between the value of each projected parameter or feature and its corresponding alarm threshold leads to what is known as RUL (remaining useful life) of the system (Figure 2). Finally, appropriate maintenance actions can be taken depending on the estimated RUL. These actions may aim at eliminating the origin of a failure which can lead the system to evolve to any critical failure mode, delaying the instant of a failure by some maintenance actions or simply stopping the system if this is judged necessary. Prognostic approaches In the engineering disciplines, fault prognosis has been approached via a variety of techniques ranging from Bayesian estimation and other probabilistic / statistical methods to artificial intelligence tools and methodologies based on notions from the computational intelligence arena. Figure 3 (Abou et al., 2010; Vachtsevanos et al., 2006) summarizes the range of possible prognosis approaches as a function of the applicability to various systems and their relative implementation cost. Prognosis technologies typically use measured or inferred features, as well as data-driven and/or physics-based models, to predict the condition of the system at some future time. Inherently probabilistic or uncertain in nature, prognosis can be applied to failure modes governed by material condition or by functional loss. Prognosis algorithms can be generic in design but specific in terms of application. Prognosis system developers have implemented various approaches and associated algorithmic libraries for customizing applications that range in fidelity from simple historical/usage models to approaches that use advanced feature analysis or physics-of-failure models. Table 1 (Vachtsevanos et al., 2006) provides an overview of the recommended models and information necessary for implementing specific approaches. Of course, the resolution of this table only illustrates three levels of algorithms, from the simplest experienced-based (reliability) methods to the most advanced physics-of-failure approaches that are calibrated by sensor data.