Adaptive Modeling

We describe a computational technique for functional modeling of physical devices. In this technique, functions and behaviors of a given device are derived by retrieving the structure-behavior-function (SBF) model of a structurally similar device and revising the retrieved model to meet he specifications of the given device. The SBF model of a device explicitly representsits structure, its functions, and its internal causal behaviors that specify how its structure delivers its functions. The model of the known device is revised by model-revision plans, where each plan accommodates a specific type of structural difference between the new and the known devices. The model ontology gives rise to a classification of structural differences and corresponding model-revision plans. The process of model revision is focused by the organization of the known model. The revised model for the new device is stored in memory for potential reuse in future. We call this computational process adaptive modeling. Motivations, Background, and Goals Functional models of devices have proved to be quite useful for reasoning about a variety of function-related tasks. The Functional Representation (FR) scheme [Sembugamoorthy and Chandrasekaran 1986; Chandrasekaran, Goel and Iwasaki 1993], for example, has been extensively used for diagnosis (e.g., [Sticklen and Chandrasekaran 1989]), redesign (e.g., [Goel and Chandrasekaran 1989]), and, more recently, design verification (e.g., [Iwasaki and Chandrasekaran 1992)]. Since functional models explicitly represent the functions of a device and use the function representations to organize behavioral knowledge about the device, they help define problem spaces for function-related tasks, and provide access to the knowledge relevant for searching the spaces. But the origin, generation and acquisition of functional device models remain open issues. Not only are these questions fundamental, but, in addition, their answers are likely to impose additional representational constraints on the models. The research described here is motivated by both goals: exploration of the origin, generation and acquisition of functional device models, and discovery of additional representational constraints on the models. The origin of this research lies in our earlier work on the conceptual phase of functional device design. The tasks of conceptual (or preliminary) device design and (qualitative) device modeling have an "inverse relationship" with each other. The task of device design takes as input a specification of the desired output behaviors of the device, i.e., the device functions. It has the goal of giving as output a specification of the structure of the device that can deliver the desired functions. Thus, the task of device design is a function --~ structure mapping. The task of device modeling takes as input a specification of the structure of a device. It has the goal of giving as output a specification of the output behaviors of the system. Thus, the task of qualitative device modeling is a structure -+ output behavior mapping. Since device functions are a subset of the output behaviors of the device, it follows that the two tasks, though not exact inverses of each other, have an inverse relationship 1. In our work on the conceptual phase of functional design, this inverse relationship between device design and device modeling led to us to hypothesize that knowledge of device models may facilitate the adaptation of the designs of known devices to design new devices of similar functionality. If the structure --~ output behavior map of a device was known, then, we hypothesized, this map may enable adaptation of the device structure for achieving a different, though similar and related, set of device functions. In the Kritik family of systems, we have extensively investigated this hypothesis [Goel 1991, 1992; Goel and Chandrasekaran 1992]. Kritik contains a design case memory, where each case in the memory specifies a structure-behavior-function (SBF) model that explains how the structure of the de1The "output behaviors" of a device, in our terminology, include both the intended and the unintended behaviors. The "functions" of the device refer to the intended output behaviors. The "internal behaviors" of the device, in this terminology, are the causal processes that result in its output behaviors including the device functions. Goel 67 From: AAAI Technical Report WS-96-01. Compilation copyright © 1996, AAAI (www.aaai.org). All rights reserved.