Multipurpose sharable engineering knowledge repository

The chemical process industry devotes considerable time and effort on decision support systems for process engineering applications. Increasing the reusability of the industrial knowledge in these systems leverages significant resources in developing these systems. Creating a unified representational framework leverages the knowledge overlap in multiple applications and processes.; This work describes a multipurpose representation for decision support applications in continuous chemical processes involving malfunction or abnormal situations, e.g. diagnosis, HAZOP. With respect to previous efforts, this work presents an overhauled version of Functional Representation (FR) that extends its flexibility for knowledge reuse, its ability to capture complex process devices and its capacity for multiple applications. The application focus of previous efforts, while adequate for the tasks at hand, was insufficient from a multipurpose perspective.; FR is a representation formalism that models behavior, function and mode of operation in a hierarchical structure. Behavioral descriptions are represented with Causal Process Descriptions (CPDs), and an associated causal state directed graph. Function constructs classify behaviors with similar functional characteristics. Modes categorize Functions (and therefore behaviors) around modes of operation such as normal, or malfunction.; The new formalism addresses reusable representations for causal reasoning about complex devices. A complex device is modeled as an aggregation of connected component models. Knowledge reuse is enhanced by modeling every device with a single construct (from components to systems).; Causal propagation is a critical operation associated with all targeted applications. Knowledge is structured to facilitate the task of managed causal branching of automated, real-time causal propagation. Behavioral knowledge is structured from a snapshot perspective, which models the device state at an instant in time. By adopting this perspective, temporal knowledge is removed from the causal transitions, and propagation constraints can be employed. Methodologies such as causal pruning and propagation order are implemented to further limit causal branching.; To demonstrate and validate the multipurpose capacity of a knowledge repository, two different industrial processes were modeled for use in two different engineering applications. A HAZOP analysis interface was built and operated on a fines removal unit FR model, while a diagnostic system interface operated on a demethanizer unit FR model.