Modeling and Inference with Relational Dynamic Bayesian Networks

Present manufacturing systems are equipped with sensors that provide a basis for real-time monitoring and diagnosis; however, placement of these sensors is constrained by the the functions that they perform and the system's structure. Moreover, sensors cannot be placed across all components in the equipment due to significant data challenges. This results in non-observable components, which limit our ability to support effective real-time monitoring and fault diagnosis initiatives. Consequently, product quality drifts found during inspection often result in unscheduled breakdown of all equipment involved in respective production operation. This situation becomes more complex for automated manufacturing lines, where success depends on our ability to capitalize maximum production capacities. This paper proposes a methodology that exploits historical data over unobserved equipment components to reduce the search space of potential faulty components, followed by a more accurate diagnosis of failures and causes. The purpose is to improve the effectiveness and efficiency of both the real-time monitoring of potential faulty components and the diagnosis of causes. In the proposed approach, we use a logical diagnosis approach to reduce the search space of suspected equipment in the production flow, which is then formulated as a Bayesian network. The proposed approach computes the risk priority for suspected equipment with corresponding factors (such as human factor and recipe), using joint and conditional probabilities. The objective is to quickly and accurately localize the possible fault origins in real-time and support effective corrective maintenance decisions. The key advantages offered by this approach are: (i) reduced unscheduled equipment breakdown duration, and (ii) stable production capacities, required for success in highly competitive and automated manufacturing systems. Moreover, this is a generic methodology and can be deployed on fully or semi-automated manufacturing systems.

show abstract

“…The conditional probabilities are computed based on principal theories reported in the literature. 43 The BN model is developed as follows:…”

Section: Proposed Diagnosis Approachmentioning

confidence: 99%

Fault diagnosis for the complex manufacturing system

Nguyen

Duong

Zamaï

et al. 2016

Proceedings of the Institution of Mechanical Engineers, Part O:

View full text Add to dashboard Cite

show abstract

“…In contrast to standard HMM, the logical representation allows the model to represent compactly probability distributions over sequences of logical atoms, rather than propositional symbols. Similarly, DBNs have been extended using first-order logic [Manfredotti 2009;Manfredotti et al 2010]. A tree structure is used, where each node corre-sponds to a first-order logic expression, e.g.…”

Section: Related Workmentioning

confidence: 99%

Probabilistic Event Calculus for Event Recognition

Skarlatidis

Παλιούρας

Artikis

et al. 2015

ACM Trans. Comput. Logic

View full text Add to dashboard Cite

Symbolic event recognition systems have been successfully applied to a variety of application domains, extracting useful information in the form of events, allowing experts or other systems to monitor and respond when significant events are recognised. In a typical event recognition application, however, these systems often have to deal with a significant amount of uncertainty. In this article, we address the issue of uncertainty in logic-based event recognition by extending the Event Calculus with probabilistic reasoning. Markov logic networks are a natural candidate for our logic-based formalism. However, the temporal semantics of the Event Calculus introduce a number of challenges for the proposed model. We show how and under what assumptions we can overcome these problems. Additionally, we study how probabilistic modelling changes the behaviour of the formalism, affecting its key property-the inertia of fluents. Furthermore, we demonstrate the advantages of the probabilistic Event Calculus through examples and experiments in the domain of activity recognition, using a publicly available dataset for video surveillance.

show abstract

“…• In this paper, in order to compute the conditional probabilities, we need to extract the distribution over some subset of variables or a single variable, and thus, we need to marginalize or sum out the variables other than the variables of interest as explained in [21]. The marginalization rule for any sets of variable X and Y is given by :…”

Section: Diagnosis Modelmentioning

confidence: 99%

Dynamic structure identification of Bayesian network model for fault diagnosis of FMS

Nguyen

Duong

Zamaï

et al. 2014

IECON 2014 - 40th Annual Conference of the IEEE Industrial Electronics Society

View full text Add to dashboard Cite

This paper proposes an approach to accurately localize the origin of product quality drifts, in a flexible manufacturing system (FMS). The logical diagnosis model is used to reduce the search space of suspected equipment in the production flow; however, it does not help in accurately localizing the faulty equipment. In the proposed approach, we model this reduced search space as a Bayesian network that uses historical data to compute conditional probabilities for each suspected equipment. This approach helps in making accurate decisions on localizing the cause for product quality drifts as either one of the equipment in production flow or product itself.

show abstract

Modeling and Inference with Relational Dynamic Bayesian Networks

Cited by 30 publications

References 53 publications

Fault diagnosis for the complex manufacturing system

Fault diagnosis for the complex manufacturing system

Probabilistic Event Calculus for Event Recognition

Dynamic structure identification of Bayesian network model for fault diagnosis of FMS

Contact Info

Product

Resources

About