An Approach to Consistency-based Process Diagnosis

The Model-Based Diagnosis (MBD) approach [24] relies on deep knowledge of how a system works. It employs circumscription [14, 15] or default logic [23] as the underlying formalism. This is very elegant but requires many restrictive assumptions. MBD only caters for static domains and component-oriented artifacts. Thus, for real-life systems, there is a need for a dynamic MBD framework, through some explicit representation of time and events. In this paper, which builds on the work in [18], we present a Temporal First Order Nonmonotonic Logic (Temporal FONL) based on partial models in order to allow for a natural representation of temporal knowledge using time points and intervals. We show how dynamic behaviour of physical systems can be expressed in the language of Temporal FONL. The system can be extended to allow the representation of events. We show how the qualification problem and frame problem can be circumvented.

Artificial Intelligence, 2002

In this paper we propose a new characterization of model-based diagnosis based on process algebras, a framework which is widely used in several areas of computer science. We show that process algebras provide a powerful modelling language which allows us to capture, in an uniform way, different types of models of physical systems, including models of time-varying and dynamic behavior. Then we provide a characterization of diagnosis which is equivalent to the "classical" abductive one. This suggests new interesting opportunities for research on relations between modelbased reasoning and process algebras. (L. Console), picardi@di.unito.it (C. Picardi), marina@di.unito.it (M. Ribaudo). 0004-3702/02/$ -see front matter  2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 0 4 -3 7 0 2 ( 0 2 ) 0 0 2 9 2 -8 20 L. Console et al. / Artificial Intelligence 142 (2002) 19-51 (e.g., computational complexity [13,57] or diagnosability ). Moreover, they provided conceptual tools for analyzing application problems and domains and for relating them to the different approaches for modelling and problem solving; as a result, they have been used for defining frameworks which provide guidelines for studying which approaches to modelling and diagnosis are suitable for a given application problem or domain (e.g., see ). Together with the applications, the foundational works contributed to singling out new open problems and opportunities for research. Last, but not least, they contributed to the creation of bridges between model-based reasoning and other areas of artificial intelligence and computer science such as logical and non-monotonic reasoning, probabilistic reasoning, machine learning, control theory, to mention only some of them.

1993

We show how qualitative simulation can be used to monitor and diagnose faults in the behaviour of dynamic physical systems. Our system synchronously tracks the evolution of the real system and iteratively searches for faults based on a characterisation of the modelling space. A particular innovation is the use of priority measures to rank the multiple behaviours predicted by the simulation and to use these in establishing the most likely fault. The operation of the system is illustrated by application to an experimental process rig that consists of continuous and dynamic thermal and flow processes.

Given the increasing use of digital control systems in manufacturing plants, the theoretical potential for error monitoring and diagnosis has significantly increased. However, conventional monitoring systems often still suffer from the problems of past decades where automated error recognition is nonexistent and the monitoring system restricts itself mainly to communications and signal processing. As a result, the diagnostic load in case of malfunctions is left to the operator. We describe a model-based diagnosis system that is integrated into the control of a manufacturing plant by providing an additional layer atop the existing monitoring system. The diagnosis system provides flexible, high-level and resilient diagnosis capability and can directly use the monitoring system's reports as observations for the diagnosis process. We also discuss general principles of developing models for this domain and show a simple model used in our current diagnosis/monitoring prototype system.

Model-based diagnosis provides a well founded theory and a set of algorithms for finding and fixing a misbehavior caused by components. Actually applying model-based diagnosis effectively requires a flexible implementation which is capable of handling the differing requirements of multiple application domains. The diagnosis framework described in this paper has been developed for the purpose of being used in an industrial setting. It derives a significant part of its effectiveness from being integrated with a component oriented language for describing diagnosis models. The framework itself contains a class library comprising several different diagnosis engines having a standardized interface and allows rapid prototyping of diagnosis applications. The paper describes the framework, shows application domains, where the framework was applied, and gives an overview of the capability of the system description language AD2L.

2003 IEEE Aerospace Conference Proceedings (Cat. No.03TH8652), 2003

Over the past decade, the number of Earth orbiters and deep space probes has grown dramatically and is expected to continue in the future as miniaturization technologies drive spacecraft to become more numerous and more complex. This rate of growth has brought a new focus on autonomous and self-preserving systems that depend on fault diagnosis. Although diagnosis is needed for any autonomous system, current approaches are almost uniformly "ad-hoc," inefficient, and incomplete. Systematic methods of general diagnosis exist in literature, but they all suffer from two major drawbacks that severely limit their practical applications. First, they tend to be large and complex and hence difficult to apply. Second and more importantly, in order to find the minimal diagnosis set, i.e., the minimal set of faulty components, they rely on algorithms with exponential computational cost and hence are highly impractical for application to many systems of interest.

2003

Fault detection and diagnosis is an important problem in process engineering. It is the central component of abnormal event management (AEM) which has attracted a lot of attention recently. AEM deals with the timely detection, diagnosis and correction of abnormal conditions of faults in a process. Early detection and diagnosis of process faults while the plant is still operating in a controllable region can help avoid abnormal event progression and reduce productivity loss. Since the petrochemical industries lose an estimated 20 billion dollars every year, they have rated AEM as their number one problem that needs to be solved. Hence, there is considerable interest in this field now from industrial practitioners as well as academic researchers, as opposed to a decade or so ago. There is an abundance of literature on process fault diagnosis ranging from analytical methods to artificial intelligence and statistical approaches. From a modelling perspective, there are methods that require accurate process models, semi-quantitative models, or qualitative models. At the other end of the spectrum, there are methods that do not assume any form of model information and rely only on historic process data. In addition, given the process knowledge, there are different search techniques that can be applied to perform diagnosis. Such a collection of bewildering array of methodologies and alternatives often poses a difficult challenge to any aspirant who is not a specialist in these techniques. Some of these ideas seem so far apart from one another that a non-expert researcher or practitioner is often left wondering about the suitability of a method for his or her diagnostic situation. While there have been some excellent reviews in this field in the past, they often focused on a particular branch, such as analytical models, of this broad discipline. The basic aim of this three part series of papers is to provide a systematic and comparative study of various diagnostic methods from different perspectives. We broadly classify fault diagnosis methods into three general categories and review them in three parts. They are quantitative model-based methods, qualitative model-based methods, and process history based methods. In the first part of the series, the problem of fault diagnosis is introduced and approaches based on quantitative models are reviewed. In the remaining two parts, methods based on qualitative models and process history data are reviewed. Furthermore, these disparate methods will be compared and evaluated based on a common set of criteria introduced in the first part of the series. We conclude the series with a discussion on the relationship of fault diagnosis to other process operations and on emerging trends such as hybrid blackboard-based frameworks for fault diagnosis.

Annals of Mathematics and Artificial Intelligence, 1994

Diagnosis as a real human activity of problem solving, involves many actions and reasoning steps that seem to conflict with the use of exact, formal, and complete models of physical systems in model-based diagnosis. This paper focuses on demonstrafing that some features occurring in practical diagnostic tasks, namely coping with intermittent faults, inaccurate models and observations, and testing, can be incorporated in consistency-based diagnostic systems. The basis is established by an extension to consistency-based diagnosis that enables the use of multiple models and hypothetical reasoning. It provides a sound theoretical foundation for building systems that can reflect simplifications and plausible inferences in multiple models, rather than using implicit, hard-wired heuristics, and relate the various models by logical links which clearly determine the underlying assumptions and the impact of a model shift on the diagnostic result.

2013

The existing theory of consistency-based diagnosis and its implementations have proven successful in a number of technical applications. However, they turn out to be inherently limited to a very specific class of systems to be diagnosed: They are tailored for artifacts consisting of components in a fixed structure, and they are aimed at a particular kind of diagnosis and repair, namely failing components and their replacement. In order to cover systems that comprise processes and change their structure dynamically, faults that are due to unanticipated objects and interactions, and therapy tasks that involve complex interference with the system, we need an extended theory of diagnosis and therapy. In this paper, we present theoretical and practical results of our work on process-oriented consistency-based diagnosis. For this purpose, a logical reconstruction of a processoriented modeling paradigm is needed, which forms the foundation for both the automated composition and the revisio...

In this paper we propose the use of the Timed Observation theory as a powerful frameworks for model-based diagnosis. In fact, they provide a global formalism for modeling a dynamic system (TOM4D), for characterizing and computing di-agnoses of the system under investigation